Evaluation of ACMG/AMP guidelines for classifying APC variants in Korean patients with suspected familial adenomatous polyposis

Abstract

Background

APC is the most frequently mutated gene in familial adenomatous polyposis (FAP). Recent advancements in sequencing technologies and the introduction of the 2024 APC-specific ACMG/AMP variant classification guidelines have reshaped the diagnostic landscape for hereditary polyposis syndromes.

Methods

We retrospectively reviewed APC variant data from 425 Korean patients clinically suspected of FAP between 2006 and 2020. Genetic testing included direct sequencing, targeted next-generation sequencing (NGS), multiplex ligation-dependent probe amplification, and mutant enrichment with 3′-modified oligonucleotide PCR. Variants were reclassified using the 2024 APC-specific ACMG/AMP criteria.

Results

Pathogenic or likely pathogenic APC variants were detected in 166 of 425 patients (39.1%). Twenty-one novel truncating mutations were identified. The application of the 2024 APC-specific ACMG/AMP criteria led to the reclassification of 13 variants, with 2 pathogenic or likely pathogenic and 7 likely benign. Nine patients showed evidence of APC mosaicism.

Conclusion

This is the first large-scale study to evaluate the clinical utility of the 2024 APC-specific ACMG/AMP variant interpretation guidelines, enabling the reclassification of previously designated variants of uncertain significance into more definitive categories. The detection of mosaic variants emphasizes the need for higher sequencing depth. Our results support the integration of NGS and further diagnostic evaluation into APC variant testing.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12935-026-04405-0.

Introduction

The APC (Adenomatous Polyposis Coli, NM_000038.6) gene, located on chromosome 5q22.2, is the most commonly mutated gene in familial adenomatous polyposis (FAP) [1]. A deleterious mutation in APC, a known tumor suppressor gene, impairs the suppression of cellular growth, leading to the formation of adenomas, and eventually conferring malignant potential [2]. The gene comprises 16 exons, starting with non-coding exon 1. Exon 16 is the longest and the most mutation-prone region [3]. Most of the pathogenic mutations in APC are frameshift, nonsense, or splicing site mutations, ultimately resulting in truncated protein products. Uniquely, the APC gene shows a relatively high rate of mosaicism, estimated to reach up to 20% in some cohorts [4].

Given these complexities, genetic testing strategies for APC have evolved significantly over time. Direct sequencing was used in initial studies, which facilitated early insights into the mutational landscape of the gene. However, this method could not reliably detect mosaic mutations, had difficulty identifying large-scale variants, and required complementary techniques such as multiplex ligation-dependent probe amplification (MLPA) to detect exon-level deletions or duplications. The advent of next-generation sequencing (NGS) has addressed many of these limitations. NGS allows for simultaneous evaluation of APC and other hereditary colonic polyposis-related genes, detection of low variant allele frequency (VAF) variants, and identification of exon-level copy number changes [5]. However, the relatively shallow sequencing depth typically used in germline testing can hinder the detection of mosaic variants, especially in genes like APC where such events are not uncommon. Techniques such as mutant enrichment with 3′-modified oligonucleotide (MEMO)-PCR have been introduced to improve the detection of low-level variants [6].

Although the original ACMG/AMP variant interpretation guidelines offered a general framework for classifying germline variants, the emergence of gene-specific adaptations has improved accuracy [7]. In 2024, a set of APC-specific ACMG/AMP guidelines was introduced [3]. This revised framework removed ambiguous or gene-irrelevant criteria such as PM1, PM3, PM4, PP2, PP4, and PP5. The criterion PM2, related to absence in controls, was downgraded to PM2_Supporting. To classify variants as benign, precise allele frequency thresholds were defined (0.1% for BA1 and 0.001% for BS1). This allowed many variants of uncertain significance (VUS) to be reclassified as benign or likely benign.

Under the ACMG/AMP framework, evidence codes are categorized according to pathogenic or benign strength levels, including PVS1 (pathogenic very strong, predicted loss-of-function variants in genes where loss of function is a known disease mechanism), PS4 (pathogenic strong, increased prevalence in affected individuals), PM2_Supporting (pathogenic moderate, low population frequency), BA1 (benign stand alone, high allele frequency consistent with benignity), and BS1 (benign strong, allele frequency greater than expected for the disorder) (Supplementary Table S1). The 2024 APC-specific recommendations further refined the thresholds and application strengths of these criteria for APC variant interpretation, particularly for truncating, splicing, and population frequency-based variants.

Despite the clinical importance of these revisions, the APC-specific ACMG/AMP guidelines remain largely overlooked and have only been applied in retrospective reclassification efforts [8]. Large-scale data from real-world clinical laboratories are still lacking. Therefore, this study aims to review the spectrum of APC variants identified in patients with suspected FAP from multiple laboratories over the past 15 years. Additionally, we evaluate the impact of the new APC-specific ACMG/AMP criteria on variant classification and analyze the prevalence and detection of mosaicism using various testing modalities including NGS, MLPA, and MEMO-PCR.

Materials and methods

Study population

Between May 2006 and June 2020, 425 patients clinically suspected of FAP were referred for APC mutation testing. Non-Korean patients were excluded from the study. Among the tested samples, 267 were referred to a large commercial laboratory from various locations throughout Korea, and 176 were submitted through a single institutional healthcare system. A retrospective chart review including clinical diagnosis, family history, colonoscopy findings, and pathological results was conducted for 233 patients who visited two tertiary hospitals. This study was approved by the Institutional Review Board of Yonsei University Severance Hospital (IRB No. 4-2020-0776), and all procedures conformed to the Declaration of Helsinki.

Direct sequencing of APC

Genomic DNA was extracted from peripheral blood samples. Before 2010, APC mutation analysis was performed at a commercial laboratory using denaturing high-performance liquid chromatography (WAVE Maker System, Transgenomic Inc., San Jose, CA, USA), followed by Sanger sequencing for heterozygous PCR products. After 2010, direct sequencing of APC was conducted using PCR amplification and AccuPower HotStart PCR PreMix (Bioneer, Daejeon, Korea), with primers designed to target all of the coding exons and flanking introns. Sequencing was performed using the ABI 3730 Genetic Analyzer (Thermo Fisher Scientific, Waltham, MA, USA) and a BigDye Terminator v3.1 Cycle Sequencing Kit (Thermo Fisher Scientific).

Targeted next-generation sequencing and confirmatory analysis

Targeted NGS was performed using a customized panel (Dxome Co. Ltd., Gyeonggi-do, Korea) composed of 60 genes, including those associated with hereditary colonic polyposis or other hereditary cancer syndromes (Supplementary Table S2). Capture probes were designed to target all of the coding exons and flanking introns of the selected genes. Double-stranded DNA was quantified and adjusted to 5 ng/µL using a Qubit dsDNA BR Assay Kit on a Qubit 2.0 Fluorometer (Thermo Fisher Scientific). DNA was fragmented, purified, and prepared for hybrid capture using standard protocols. Libraries were sequenced on the Illumina NextSeq 550Dx platform (Illumina, San Diego, CA, USA). Copy number analysis by base-level normalization was used to determine copy number variants (CNVs) [9]. All small nucleotide variants were visually confirmed using the Integrated Genomics Viewer [10]. CNVs in the APC gene were validated by MLPA (P043 Probemix; MRC Holland, Amsterdam, The Netherlands), and low-level variants were confirmed by MEMO-PCR using 3′-modified blocking primers targeting mutation sites [6].

Mutation identification and classification

Variants were annotated according to the GRCh37 (hg19) reference genome assembly. Variants were classified as “pathogenic,” “likely pathogenic,” “VUS,” “likely benign,” or “benign” based on APC-specific recommendations by the ACMG/AMP [7]. PS4 evidence was conservatively assigned according to the APC-specific phenotype point-based framework. Patients clinically suspected or documented as having FAP/attenuated FAP were considered eligible only for the minimum phenotype category, in order to avoid potential overclassification in the setting of incomplete phenotypic information. For annotation and classification, ClinVar and gnomAD were used [11, 12]. Mutations were mapped onto protein structures using the ProteinPaint tool [13].

Results

In total, 425 patients participated in the study, including 235 males and 185 females (5 unknown, Table 1). Direct sequencing was performed in 279 patients and targeted NGS in 164 patients. Eighteen patients underwent both assays based on strong clinical suspicion. Of these, six exhibited either APC mosaicism or copy‑number variants with pathogenic mutations (Supplementary Table S3). Counting those six as APC mutation-positive yielded 166 mutation‑positive patients (39.1%) and 259 mutation‑negative patients (60.9%). Additionally, eight patients harbored pathogenic or likely pathogenic variants in other genes on the targeted NGS panel (Supplementary Table S4).

Table 1.

Demographic features of clinically suspected familial adenomatous polyposis

Category		Median (25th–75th percentile) or n (%)
Age, years		41 (29–54)
Sex	Male	235 (55.3%)
	Female	185 (43.5%)
Method	Direct sequencing	279 (65.6%)
	Next-generation sequencing	164 (38.6%)
APC mutation status	P/LP/VUS	177 (41.6%)
	LB/B/not detected	248 (58.4%)
Total patients		425

P, pathogenic; LP, likely pathogenic; VUS, variants of uncertain significance; LB, likely benign; B, benign

Pathogenic or likely pathogenic APC variants were detected by direct sequencing in 107 of 279 patients (38.4%) and by targeted NGS in 59 of 164 patients (36.0%) (Supplementary Figure S1A). Across all APC variants—including pathogenic, likely pathogenic, and VUS—the combined proportion of pathogenic/likely pathogenic variants was 93.8% (166 out of 177). Nonsense, frameshift, splice-site, and missense mutations accounted for 42.4%, 41.2%, 7.9%, and 5.1% (75, 73, 14, 9 respectively), respectively (Supplementary Figure S1B). Suspected copy‑number variants detected by NGS were further evaluated by MLPA in patients with NGS‑suggested deletions of exons 1–2. A loss from upstream of exon 1 through exon 2 (Supplementary Figure S2A) and whole‑exon deletions (Supplementary Figure S2B) called by NGS were validated by MLPA. Variant distribution analysis revealed that exon 16 harbored the highest mutation frequency, with c.3927_3931del being the most recurrent variant (Fig. 1).

Fig. 1.

ProteinPaint analysis of detected mutations in APC

APC variants were classified according to APC‑specific ACMG/AMP criteria (Supplementary Data 1). Twenty‑one novel truncating mutations (six nonsense and fifteen frameshifts) were identified (Table 2); all met PVS1 and PM2_Supporting and were classified as likely pathogenic. One recurrent variant (c.5819_5820del, p.(Ile1940ThrfsTer8)) observed in two individuals also met PS4_Supporting and was classified as pathogenic. Thirteen variants previously reported in ClinVar were reclassified under the APC‑specific framework (Table 3). Three partial gene deletions were downgraded to VUS according to the PVS1-specific criteria (PVS1_Moderate or PVS1_N/A). The splice‑site variant c.1959‑2 A > C was reclassified to VUS according to PVS1-specific criteria (PVS1_N/A) despite likely pathogenic reports. The intronic variant c.423‑8 A > G received PS3_VeryStrong support from a functional RNA study and was reclassified as pathogenic. Seven variants met the BS1 threshold (gnomAD GroupMax Filtering Allele Frequency ≥ 0.001%), enabling classification as likely benign (n = 6) or benign (n = 1). No significant difference in age at diagnosis was observed between patients with reclassified benign/likely benign variants and those with pathogenic/likely pathogenic variants; however, detailed polyp burden information was not consistently available in this retrospective referral-based cohort. The missense variant c.933G > C—initially reported likely pathogenic in ClinVar—was re‑evaluated as PVS1_Supporting by PVS1-specific criteria, PS3_Moderate by transcript‑truncation functional data, PS4_Supporting by prior evidence, and PM2_Supporting by low population frequency, resulting in a final classification of likely pathogenic. A Sankey plot was generated to illustrate the reclassification of APC variants previously categorized as pathogenic, likely pathogenic, or VUS, excluding those that were already classified as likely benign or benign in ClinVar or were evidently non-pathogenic based on previous interpretations (Fig. 2).

Table 2.

Novel APC mutations detected

Pt #	Sex	Age	Nucleotide change	Amino acid change	Zyg.	Pathogenicity
133	N/A	N/A	c.141dup	p.(Leu48ThrfsTer2)	Het.	LP
217	F	29	c.946del	p.(Tyr316IlefsTer20)	Het.	LP
414	M	63	c.1257del	p.(Cys420ValfsTer34)	Het.	LP
419	M	30	c.1587del	p.(Val530TrpfsTer4)	Het.	LP
125	M	46	c.1708 A > T	p.(Lys570Ter)	Het.	LP
89	M	28	c.1772_1778del	p.(Ala591GlyfsTer17)	Het.	LP
45	M	14	c.1928_1958 + 14del		Het.	LP
102	F	20	c.2319del	p.(Phe773LeufsTer4)	Het.	LP
151	M	37	c.2327_2328dup	p.(Asp777Ter)	Het.	LP
178	M	32	c.2492T > A	p.(Leu831Ter)	Het.	LP
99	M	33	c.2923 A > T	p.(Lys975Ter)	Het.	LP
49	M	27	c.3169del	p.(Glu1057LysfsTer4)	Het.	LP
101	M	32	c.3263dup	p.(Phe1089ValfsTer30)	Het.	LP
105	F	20	c.3300_3301del	p.(Pro1101IlefsTer17)	Het.	LP
25	M	24	c.3833_3834del	p.(Ser1278PhefsTer6)	Het.	LP
35	F	48	c.3833_3834del	p.(Ser1278PhefsTer6)	Het.	LP
291	F	44	c.3857_3859A[5]	p.(Ile1287LysfsTer2)	Het.	LP
180	F	55	c.4148del	p.(Met1383SerfsTer32)	Het.	LP
179	F	54	c.4429dup	p.(Gln1477ProfsTer10)	Het.	LP
43	M	23	c.4666dup	p.(Thr1556AsnfsTer3)	Het.	LP
192	M	60	c.5208_5209del	p.(His1738GlnfsTer30)	Het.	LP
98	M	50	c.5819_5820del	p.(Ile1940ThrfsTer8)	Het.	P
112	M	47	c.5819_5820del	p.(Ile1940ThrfsTer8)	Het.	P

M, male; F, female; Het., hetero; SD, standard deviation; P, pathogenic; Pt, patient, LP, likely pathogenic, Zyg.; Zygosity

Table 3.

Pathogenicity changed or confirmed by new criteria

Pt #	Sex	Age	Nucleotide change	Amino acid change	Zyg.	Clinvar	New criteria	New classification
362	M	55	c.1959–2 A > C		Het.	Not evaluated	PVS1_N/A, PS4_Supporting, PM2_Supporting	VUS
330	F	22	exon1 deletion		Het.	Not evaluated	PVS1_N/A (Exon 1 is a non-coding region)	VUS
358	M	36	exon 6–8 deletion		Het.	Not evaluated	PVS1_Moderate, PM2_Supporting	VUS
385	F	24	exon 1–2 deletion		Het.	Not evaluated	PVS1_N/A	VUS
416	F	56	c.423–8 A > G		Het.	Conflicting	PS3_VeryStrong, PS4_Supporting, PM2_Supporting	P
310	F	52	c.3378 C > G	p.(Ser1126Arg)	Het.	Conflicting	BS1, BS2_Supporting	LB
4	F	37	c.3964G > A	p.(Glu1322Lys)	Het.	Conflicting	BS1	LB
309	F	70	c.5257G > C	p.(Ala1753Pro)	Het.	Conflicting	BS1	LB
317	M	36	c.5378 C > G	p.(Ala1793Gly)	Het.	Conflicting	BS1	LB
89	M	28	c.5708 A > G	p.(Asn1903Ser)	Het.	Conflicting	BS1	LB
269	M	35	c.6896 C > T	p.(Pro2299Leu)	Het.	Conflicting	BS1	LB
313	F	58	c.7150T > A	p.(Leu2384Ile)	Het.	Conflicting	BS1	LB
58	M	41	c.933G > C	p.(Lys311Asn)	Het.	LP (Low evidence)	PS3, PVS1_Supporting, PM2_Supporting, PS4_Supporting	LP

M, male; F, female; Het., Hetero; P, pathogenic; Pt, patient, VUS, variants of uncertain significance; LP, likely pathogenic; LB, likely benign; Zyg.; Zygosity

Fig. 2.

Sankey plot of reclassification of APC variants previously categorized as pathogenic, likely pathogenic, or variants of uncertain significance in ClinVar

Nine patients with suspected APC mosaicism (VAF < 20%) were detected by targeted NGS (Table 4). All of these patients harbored frameshift or nonsense mutations. Confirmation by MEMO‑PCR was performed on available samples. One representative case (Supplementary Figure S3) showed that MEMO‑PCR of peripheral blood identified the c.3295_3296del (p.(Val1099PhefsTer19)) variant and an identical mutation was detected in neoplastic polyp tissue. Tissue NGS demonstrated a normalized copy‑number depth near − 0.5, consistent with whole‑gene deletion, whereas peripheral blood NGS and MLPA did not reveal CNVs.

Table 4.

Characteristics of patients with suspected somatic mutations in the APC gene (n = 9)

No	Sex/Age	Exon	Mutation	Type	VAF	Clinical finding	Confirmation test
329	F/47	6	c.694 C > T, p.(Arg232Ter)	NS	3.40%	Colonoscopy: polyps	Sequencing
						Total colectomy: Numerous tubular adenomas, consistent with FAP (> 70)	(adenoma)
397	F/44	8	c.902del, p.(Pro301LeufsTer4)	FS	5.10%	Colonoscopy: numerous adenomas in entire colon	MEMO-PCR
							(PB)
275	F/52	14	c.1754del, p.(Leu585ProfsTer5)	FS	2.00%	Colonoscopy: adenomas	-
						Anterior resection of sigmoid colon: Adenocarcinoma, MD	(sample unavailable)
305	M/36	15	c.2626 C > T, p.(Arg876Ter)	NS	11.80%	Colonoscopy: numerous adenomas in entire colon	-
324	M/36	15	c.3211_3238dup, p.(Glu1080AlafsTer10)	FS	19.50%	Colonoscopy: rectal cancer	Sequencing
							(PB)
134	F/45	15	c.3295_3296del, p.(Val1099PhefsTer19)	FS	6.80%	Colonoscopy: numerous adenomas in entire colon	MEMO-PCR
					(adenoma 20.6%)		(PB and adenoma)
333	F/55	15	c.3566 C > G, p.(Ser1189Ter)	NS	11.40%	Colonoscopy: numerous adenomas in entire colon	Sequencing
							(PB)
291	F/44	15	c.3857_3859A[5], p.(Ile1287LysfsTer2)	NS	9.40%	Total proctocolectomy: Adenocarcinoma, MD + multiple tubular and tubulovillous adenomas (> 100)	MEMO-PCR
							(PB)
37	F/44	15	c.4348 C > T, p.(Arg1450Ter)	NS	6.40%	Colonoscopy: numerous adenomas in entire colon	-

F, female; FAP, familial adenomatous polyposis; FS, frameshift mutation; M, male; MD, moderately differentiated; MEMO-PCR, mutant enrichment with 3′-modified oligonucleotide polymerase chain reaction; NS, nonsense mutation; PB, peripheral blood; VAF, variant allele frequency

Discussion

This 15-year retrospective study provides a comprehensive framework for interpreting APC gene variants, encompassing the application of a novel classification system, assessment of CNVs, identification of novel mutations, and detection of mosaicism. Our analysis reflects the evolution of genetic testing strategies in Korea, from the early use of direct sequencing in 2006 to the adoption of targeted NGS for CNV analysis and MEMO-PCR for mosaicism detection in the 2020s. Pathogenic or likely pathogenic APC variants were identified in 166 of 425 patients (39.1%), highlighting the clinical utility of comprehensive APC testing strategies and updated variant interpretation frameworks. Notably, among 18 patients with a high clinical suspicion for APC-related polyposis who tested negative using conventional approaches, six were found to harbor pathogenic variants through NGS-based analysis. This study represents the first large-scale implementation of the 2024 APC-specific ACMG/AMP classification criteria and underscores the clinical and technical value of an updated and integrated variant interpretation strategy.

Several key considerations emerged during the application of the APC-specific ACMG/AMP classification criteria in this study [3]. First, in defining PVS1, the adjustments made for nonsense or frameshift variants were relatively minor; variants located between codons 49 and 2645 were considered applicable for PVS1, whereas those outside this region were not. In contrast, substantial modifications were observed in the classification of splice-site mutations. According to the guideline, known or predicted splicing mutations are stratified into five tiers, ranging from PVS1 (List A), which applies to variants expected to disrupt biologically relevant transcripts or protein domains, to not applicable (List E) for those expected to result in small in-frame deletions. The classification provides a relatively straightforward system in which the appropriate PVS1 strength is assigned based on predefined transcript and exon-level information. Exon deletions are also stratified based on their impact, ranging from PVS1 to not applicable, depending on the region deleted. Notably, full gene deletions are considered sufficient for classification as pathogenic under a standalone PVS1 criterion. Among the 58 pilot variants curated during the development of the APC-specific ACMG/AMP recommendations by Spier et al. [3], the number of pathogenic or likely pathogenic variants was similar to that in ClinVar. However, some variants previously classified as pathogenic or likely pathogenic, such as c.32dup (p.(Gln12fs)) and c.8514 C > A (p.(Tyr2838Ter)), were reclassified as VUS. The c.32dup (p.(Gln12fs)) variant was reclassified because truncating variants located upstream of codon 49 are considered not applicable for PVS1 according to the APC-specific recommendations. Similarly, the c.1959–2 A > C variant was reclassified from likely pathogenic in ClinVar to VUS in our study, as it did not meet the updated PVS1 criteria.

Second, in evaluating PS4 (increased prevalence of a variant in affected individuals compared with controls), the guideline adopts a point-based system in which 1 point is assigned for a phenotype highly specific to APC and 0.5 points for a phenotype consistent with, but not highly specific to, APC. In our study, APC testing was performed exclusively in individuals suspected of having FAP, and literature-based variant scoring was also limited to FAP-suspected cases. We thus applied 0.5 points or higher based on the presence of documented diagnoses such as FAP or attenuated FAP. For polyposis, the guideline provides concrete criteria, including adenoma counts, allowing future studies to assign 1 or 0.5 points more consistently. However, for extracolonic manifestations such as desmoid tumors, medulloblastoma, or hepatoblastoma, assessment requires CTNNB1 status, necessitating tumor sequencing, which presents financial and logistical challenges.

Finally, the allele frequency thresholds for benign classification were revised. The previous BA1 rule, defined as an allele frequency > 5%, and BS1, defined as an allele frequency greater than expected for the disorder, were updated to thresholds of ≥ 0.1% and ≥ 0.001%, respectively, using the gnomAD Popmax Filtering Allele Frequency [12]. In the current study, we used the GRCh38-based gnomAD v4.1.0 dataset, which reports GroupMax filtering allele frequencies, and applied these thresholds accordingly. Previous studies have reported that the revised BA1 and BS1 criteria increase the proportion of variants classified as benign or likely benign. Among 58 pilot variants curated during the development of the APC-specific ACMG/AMP recommendations by Spier et al. [3], the number of benign or likely benign variants increased from 15 to 24 under the updated rules. Additionally, in our full cohort, 7 variants previously reported in ClinVar as having conflicting classifications of pathogenicity were reclassified as likely benign or benign based on the application of BS1. These clearer classification thresholds reduced the number of VUS, enabling a more confident interpretation of variants and removing the need for further functional studies in cases consistent with benign allele frequencies in the general population. Nevertheless, caution is warranted when applying population frequency-based benign criteria such as BA1 and BS1, particularly in underrepresented ethnic populations. Founder variants or population-specific risk alleles may demonstrate relatively elevated allele frequencies despite clinical relevance. In addition, limited representation of East Asian populations in public databases may affect the accuracy of allele frequency-based interpretation. These changes highlight the utility of disease-specific guidelines in refining variant interpretation and ultimately improving clinical decision-making in FAP.

ClinVar is one of the most widely used public databases for clinical variant interpretation and includes submissions from multiple clinical laboratories, research groups, and expert panels. However, inter-laboratory variability and conflicting interpretations remain important limitations, particularly for rare germline variants. Other interpretation platforms, such as VarSome, may provide additional aggregated evidence and automated ACMG/AMP-based interpretation support, but these platforms often rely on overlapping resources including ClinVar and population databases. In this context, the application of APC-specific ACMG/AMP criteria may help improve the consistency and reproducibility of APC variant interpretation across different laboratories and interpretation frameworks.

Mosaicism in the APC gene has long been a topic of discussion. However, definitive conclusions have yet to be reached. The debate primarily revolves around whether APC mosaicism is somatic or germline in origin [14]. Some studies have reported that somatic mosaicism accounts for approximately 20% of simplex cases [4]. However, depending on whether the mosaic event occurred before or after the separation of embryonic germ layers—specifically, endoderm (colon), mesoderm (blood), or ectoderm (buccal mucosa)—the possibility of transmission to offspring arises if both mesodermal and ectodermal tissues are involved [15]. In such cases, although mosaicism is de novo, it cannot be strictly considered somatic due to its potential for inheritance [16]. Certain studies have suggested performing familial testing only in select cases, such as when a germline APC mosaic variant is identified in a de novo polyposis patient or when the same APC variant is repeatedly detected in multiple adenomas and confirmed in additional tissue [14]. However, performing separate variant testing for each adenoma is often impractical due to financial and logistical constraints. In our study, for two individuals suspected of harboring mosaic variants, testing was conducted on only one tissue and one blood sample, and multiple tissue types were not analyzed. Given these limitations, we propose an alternate strategy of increasing sequencing depth in blood samples, which are relatively easy to obtain and relevant for assessing heritability, to enhance the detection of mosaicism in clinical settings.

Previous studies have highlighted APC mosaicism as an underrecognized cause of otherwise unexplained adenomatous polyposis. A systematic review identified 57 patients with APC mosaicism across 19 publications, emphasizing its clinical relevance in individuals with de novo or unexplained polyposis [14]. In addition, tissue-based studies have shown that mosaic APC variants may be variably detectable across leukocyte DNA, non-neoplastic intestinal mucosa, and colorectal adenomas, and that analysis of multiple adenomas can improve detection sensitivity [15]. In our cohort, patients with suspected APC mosaicism also showed clinically significant phenotypes, including numerous colorectal adenomas or colectomy-level polyposis in several cases, while colorectal cancer was documented in three patients. These findings further support the clinical importance of detecting low-level APC mosaic variants in patients with suspected APC-related polyposis, particularly when conventional germline testing is unrevealing.

During this study, we identified several additional aspects that could enhance the diagnostic utility of APC testing. The most frequently observed APC variants were c.3927_3931del (15 cases), c.3183_3187del (9 cases), c.646 C > T (6 cases), and c.694 C > T (5 cases). In addition, splice-site variants near c.1744, including c.1744-1G > A and c.1744–2 A > G, were found in six individuals. Furthermore, when applying ACMG/AMP criteria to classify variant pathogenicity, we found that the BS1 criterion was largely dependent on population databases. However, allele frequency data remain limited for underrepresented populations, such as East Asians. This highlights the need for more comprehensive, population-specific frequency data to support accurate variant interpretation in diverse populations. In our cohort, 8 patients harbored pathogenic or likely pathogenic variants in genes other than APC, as identified through panel-based NGS. These included BMPR1A (n = 4), MUTYH (n = 2), POLE (n = 1), and PMS1 (n = 1), all of which are associated with polyposis syndromes. Additional variants identified in non-APC hereditary cancer-related genes should be interpreted cautiously in the context of polyposis predisposition. Monoallelic MUTYH variants, truncating POLE variants, and PMS1 variants may not represent established causative mechanisms for classical adenomatous polyposis syndromes. Therefore, these findings should be considered exploratory or potentially contributory rather than definitive explanations for the clinical phenotype.

This study has several limitations. First, it was a retrospective analysis, and the study population was limited to individuals clinically suspected of having FAP. As such, selection bias may have been introduced. A broader, prospective study that includes healthy individuals from the East Asian population is warranted in future investigations. Second, a portion of the dataset was based on earlier diagnostic results using direct sequencing. Although all participating laboratories now use NGS for APC, the previous use of conventional sequencing methods may have resulted in differences in detection sensitivity. Third, comprehensive genotype–phenotype correlation analysis could not be robustly performed because standardized phenotypic information, including detailed polyp burden, extracolonic manifestations, and longitudinal clinical follow-up, was not consistently available in this retrospective referral-based cohort. In addition, formal co-segregation analysis could not be performed because family relationship information was not systematically available. Robust genotype–phenotype correlation analysis according to APC variant location could not be reliably performed because standardized phenotypic information, including detailed polyp burden and extracolonic manifestations, was not consistently available in this retrospective referral-based cohort. Finally, the study lacked comprehensive tissue-based testing in patients suspected of having mosaicism. Previous studies suggest that confirming mosaicism requires testing across multiple tissue types beyond peripheral blood. Therefore, for individuals with suspected mosaic APC variants, we recommend parallel testing using biopsy-derived tissues or other accessible samples to increase diagnostic accuracy.

Conclusion

In this study, we applied the newly proposed APC-specific ACMG/AMP guidelines to a large cohort of Korean patients clinically suspected of FAP, providing real-world evidence for their clinical utility. The gene-specific framework enabled more precise classification of variants, including the reclassification of previously ambiguous variants into benign, likely benign, or likely pathogenic categories. Moreover, our data emphasize the importance of advanced molecular techniques such as targeted NGS, MLPA, and MEMO-PCR in detecting complex genetic events, including mosaicism and copy number variations, which are particularly relevant to APC. Notably, 9 APC mutation-positive cases were attributed to mosaicism, emphasizing the necessity of high-sensitivity methods for accurate diagnosis. Our findings underscore the value of integrating updated classification criteria with comprehensive genetic testing strategies to optimize variant interpretation and clinical management of FAP. Future studies in larger multi-institutional cohorts are warranted to further validate the APC-specific guidelines and to refine mosaicism detection protocols.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Conceptualization: Seo-Jin Park, Kyung-A Lee; Methodology: Dongju Won, Saeam Shin; Formal analysis and investigation: Seo-Jin Park, Namsoo Kim; Writing - original draft preparation: Seo-Jin Park, Namsoo Kim; Writing - review and editing: Namsoo Kim, Saeam Shin, Kyung-A Lee; Supervision: Kyung-A Lee.

Funding

This research received no external funding.

Data availability

All relevant data generated or analyzed during this study are included in the article and its supplementary materials.

Declarations

Ethics approval and consent to participate

This study was approved by the Institutional Review Board of Yonsei University Severance Hospital (IRB No. 4-2020-0776), and all procedures conformed to the Declaration of Helsinki

Competing interests

The authors declare no competing interests.