Gene-specific somatic epigenetic mosaicism of FDFT1 underlies a non-hereditary localized form of porokeratosis

Summary

Porokeratosis is a clonal keratinization disorder characterized by solitary, linearly arranged, or generally distributed multiple skin lesions. Previous studies showed that genetic alterations in MVK, PMVK, MVD, or FDPS—genes in the mevalonate pathway—cause hereditary porokeratosis, with skin lesions harboring germline and lesion-specific somatic variants on opposite alleles. Here, we identified non-hereditary porokeratosis associated with epigenetic silencing of FDFT1, another gene in the mevalonate pathway. Skin lesions of the generalized form had germline and lesion-specific somatic variants on opposite alleles in FDFT1, representing FDFT1-associated hereditary porokeratosis identified in this study. Conversely, lesions of the solitary or linearly arranged localized form had somatic bi-allelic promoter hypermethylation or mono-allelic promoter hypermethylation with somatic genetic alterations on opposite alleles in FDFT1, indicating non-hereditary porokeratosis. FDFT1 localization was uniformly diminished within the lesions, and lesion-derived keratinocytes showed cholesterol dependence for cell growth and altered expression of genes related to cell-cycle and epidermal development, confirming that lesions form by clonal expansion of FDFT1-deficient keratinocytes. In some individuals with the localized form, gene-specific promoter hypermethylation of FDFT1 was detected in morphologically normal epidermis adjacent to methylation-related lesions but not distal to these lesions, suggesting that asymptomatic somatic epigenetic mosaicism of FDFT1 predisposes certain skin areas to the disease. Finally, consistent with its genetic etiology, topical statin treatment ameliorated lesions in FDFT1-deficient porokeratosis. In conclusion, we identified bi-allelic genetic and/or epigenetic alterations of FDFT1 as a cause of porokeratosis and shed light on the pathogenesis of skin mosaicism involving clonal expansion of epigenetically altered cells.

Graphical abstract

We identified a skin disease associated with gene-specific epigenetic mosaicism. Biallelic genetic alterations and/or promoter hypermethylation of FDFT1 cause porokeratosis, a clonal keratinization disorder. While germline variants underlie hereditary porokeratosis, epigenetic silencing in a mosaic manner causes non-hereditary porokeratosis, predisposing specific skin areas to the disease.

Introduction

Genetic mosaicism refers to the presence of two or more genetically distinct cell populations within an individual.¹ It mainly results from postzygotic somatic variants acquired during embryogenesis. Genetic mosaicism is associated with several diseases, including monogenic disorders such as segmental neurofibromatosis type 1 (MIM: 162200), and has been well recognized in cutaneous manifestations.² The distribution of cutaneous manifestations in mosaic disorders has been categorized into five distinct types, including those following lines of Blaschko, which represent the lines of epidermal cell migration and proliferation during development.^1,3

Porokeratosis (MIM: 175900, 175800, 614714, 616631) is a clonal keratinization disorder of the epidermis exhibiting annular or circular skin lesions.⁴ The distribution of the skin lesions includes generalized forms known as disseminated porokeratosis, as well as localized forms with solitary, few, or linearly arranged multiple skin lesions.⁵ Among the localized forms, porokeratosis with large plaque skin lesions, either solitary or occurring in small numbers, is referred to as porokeratosis of Mibelli, as this specific variant of porokeratosis was initially delineated by Mibelli in 1893.^6,7,8

Porokeratosis is distinguished by a ridge-like raised border termed a cornoid lamella, which allows precise demarcation of the clonal skin lesion; porokeratosis is, therefore, an excellent model for understanding the clonal expansion of non-cancerous cells. Malignant transformation of porokeratosis occurs in a relatively small subset of individuals,^9,10 and thus, the skin lesions of porokeratosis are considered premalignant clonal lesions.

Porokeratosis has an autosomal-dominant mode of inheritance, but individuals often have no family history of porokeratosis, especially in porokeratosis of Mibelli.¹¹ Recent genetic studies have identified pathogenic variants in MVK (MIM: 251170), PMVK (MIM: 607622), MVD (MIM: 603236), and FDPS (MIM: 134629), which are related to the mevalonate pathway, in both familial and simplex cases.^12,13 Most affected individuals have variants in only one of these genes^13,14,15; thus, porokeratosis is considered a monogenic disorder. Individuals with heterozygous germline variants develop porokeratosis skin lesions by acquiring lesion-specific secondary somatic variants in the same gene,^16,17 suggesting that bi-allelic deficiencies of these genes are required for skin lesion formation. However, previous studies included individuals without pathogenic variants detected in reported genes,¹³ indicating the possibility of other genes related to porokeratosis.

In this study, we conducted a comprehensive genetic and epigenetic analysis of 21 lesion and 7 non-lesion samples from 8 individuals with porokeratosis who had no pathogenic variants detected in previously reported genes and had no family history. Our study not only identified germline and somatic pathogenic variants in FDFT1 (MIM: 184420) as the cause of porokeratosis but also discovered that somatic gene-specific epigenetic silencing of FDFT1 underlies the non-hereditary localized form of porokeratosis.

Material and methods

Individuals included in this study

Eight individuals with a diagnosis of porokeratosis, who had no pathogenic variants detected in previously reported genes (MVK, PMVK, MVD, and FDPS) and had no family history, were included in this study. This study was approved by the Ethics Committee of Keio University School of Medicine (20120226, 20170394, 20221206, and 20236014), Kobe University (B230011), and National Center for Child Health and Development (926, 2020-326 and 2023-027) in accordance with the Declaration of Helsinki. Written informed consent was obtained from all individuals. Written informed consent for the publication of clinical images was also obtained from all individuals. The skin was biopsied under local anesthesia to obtain full-thickness specimens. The epidermis was separated from the dermis by dispase treatment as we described previously.¹⁸ Peripheral blood was also obtained from all individuals as germline control samples except for individual 7, for whom a normal skin sample (non-lesion 1) was used as a germline control, because the individual had undergone bone marrow transplantation.

Isolation of genomic DNA and RNA

Genomic DNA (gDNA) was extracted from peripheral blood leukocytes and skin epidermis using a Maxwell RSC Instrument and Maxwell RSC Blood DNA Kit (Promega). RNA was extracted from skin epidermis and cultured keratinocytes using the Maxwell RSC simplyRNA Tissue Kit (Promega).

Whole-genome sequencing and whole-exome sequencing

Whole-genome sequencing (WGS) libraries were prepared using the TruSeq Nano DNA Sample Preparation Kit (Illumina) and sequenced using DNBSEQ-T7 (BGI), generating standard 150 bp paired-end data at BGI. The average depth was between 25× and 39× (median 30), and the genome coverage at 20× reads ranged from 73% to 95%. Whole-exome sequencing (WES) libraries were prepared using the SureSelect Human All Exon Kit version 6 (Agilent) and sequenced using NovaSeq6000 (Illumina), generating standard 150 bp paired-end data at Macrogen. The average depth was 92×–173× (median 140), with coverage of 92%–98% of the exonic regions at 30× reads.

Sequencing reads were mapped to the human reference genome (GRCh37) using BWA-MEM (v.0.7.17). Duplicates were marked using biobambam (v.0.0.148). Germline and somatic variants (substitutions and short indels) were identified using VarScan2 (v.2.4.2)¹⁹ and annotated using Annovar (v.2017-07-17). Variants only present in unidirectional reads were excluded. For WES, we focused on coding and splice site variants.

To identify causative germline variants, we excluded common single-nucleotide polymorphisms (SNPs). Specifically, we eliminated common SNPs that were documented in the NCBI dbSNP (build 150) or exhibited a prevalence of 0.0001 or more in any of the following datasets: National Heart, Lung, and Blood Institute Exome Sequencing Project 6500, the 1000 Genomes Project October 2014 release, gnomAD exome collection v.2.1.1, and the Tohoku Medical Megabank Organization 3.5KJPNv2 v20181105open.

For somatic calls, EBfilter (v.0.2.2) was used to filter out false-positive variants.²⁰ Variants were considered a true positive if they were classified as “high confidence” by VarScan2, had a p value of VarScan2 < 0.01, and had a p value of EBfilter <0.001. Common SNPs were excluded, as was the case with germline calls. Using the pileup data generated by Python package pysam (v.0.15.1), the presence of identical variants (with variant allele frequency >0.03) was examined in other samples from the same individual in which the variant was found.

Amplicon deep sequencing

Custom probes targeting 36 genes or loci known to be or suspected to be associated with porokeratosis and other mosaic skin disorders were generated using the SureDesign online tool by Agilent (Table S1). Sequencing libraries were constructed using the Haloplex target enrichment system (Agilent) and were sequenced using MiSeq (Illumina) generating 250 bp paired-end data. The average depth was between 512× and 1,166× (median: 982), with a coverage of 90%–95% of the targeted regions at 100× reads.

Sequencing reads were mapped to the human reference genome (GRCh37) using BWA-MEM. Somatic and germline variants were identified using SureCall (Agilent; using default parameters but minimum allele frequency 0.01) and manually curated by visual inspection with Integrative Genomics Viewer (v.2.10.2).²¹

SNP array

DNA amplification, labeling, and hybridization were performed according to the manufacturer’s instructions using the Infinium Asian Screening Array-24 v.1.0 BeadChip (Illumina). The array slides were scanned on an iScan system (Illumina), and log R ratios and B-allele frequencies were calculated and visualized using GenomeStudio (v.2.0; Illumina).

Evaluation of FDFT1 variants

We used GenBank: NM_004462.5 for variant annotation of FDFT1. MutationTaster implemented in ANNOVAR (v.2017-07-17) was used for function prediction. For FDFT1 conservation analysis, multiple sequence alignment was performed using CLUSTALW (https://www.genome.jp/tools-bin/clustalw) and displayed using Jalview (v.2.11.1.7).

Determination of allelic configurations between germline and somatic variants and between somatic variants and heterozygous SNPs

Allelic configurations were evaluated when the genomic distance between germline and somatic variants was <10 kb. For individual 2 lesion 2 and individual 8 lesions 3–4, the allelic configuration was determined by CisChecker (https://github.com/nccmo/CisChecker)²² using next-generation sequencing reads. For individual 1 lesions 1–2 and 5 and individual 8 lesion 2, the allelic relations were determined using Sanger sequencing of the subcloned gDNA. Specifically, gDNA spanning both germline and somatic variants was amplified by PCR using KOD-FX Neo (TOYOBO) with the primer pairs shown in Table S2. The amplified DNA was subcloned into the pCR-Blunt vector (Invitrogen), and each clone was sequenced using the primers listed in Table S2, as previously described.¹⁶

Bisulfite sequencing data from the Roadmap Epigenomics Consortium

The fractional methylated values, obtained from whole-genome bisulfite sequencing data of the Roadmap Epigenomics Project,²³ were downloaded from the data portal (https://egg2.wustl.edu/roadmap/data/byDataType/dnamethylation/WGBS/FractionalMethylation.tar.gz). The median fractional methylation values across all CpGs were calculated for the CpG islands within the promoter regions of FDFT1 (chr8:11,659,676–11,660,795), H19 (MIM: 103280, chr11:2,019,565–2,019,863), and IGF2 (MIM: 147470, chr11:2,154,033–2,154,387).

Methylation array

gDNA was bisulfite converted using the EpiTect plus DNA bisulfite kit (Qiagen) following the standard protocol. Genome-wide methylation was measured using the Illumina Infinium MethylationEPIC BeadChip array. The raw data were processed using Illumina GenomeStudio software (v.2011.1) to generate methylation β values and signal intensities. The quality of the data was assessed, and no samples were filtered because of a high mean detection p value (i.e., mean > 0.05). Probes were excluded if any of the following criteria were met: (1) probes did not perform well (detection p value of a probe >0.01 for at most one sample), (2) probes were located on the sex chromosomes, or (3) they were probes with known SNP sites or with cross-reactivity.²⁴ This filtering approach excluded 117,581 probes, leaving 748,337 probes for further analysis. CpG island annotations were downloaded from the UCSC genome browser.²⁵

Methylation array data of epidermis from unaffected individuals

Methylation data of 38 epidermal samples from 19 unaffected individuals were obtained from the Gene Expression Omnibus (GSE51954).²⁶ The median fractional methylation values across all CpGs were calculated for the CpG islands within the promoter regions of MVK (chr12:110,010,994–110,011,599), PMVK (chr1:154,908,962–154,909,931), MVD (chr16:88,729,108–88,730,410), FDPS (chr1:155,278,407–155,278,743), and FDFT1 (chr8:11,659,676–11,660,795).

Long-read WGS and haplotype phasing

Sequencing libraries were constructed following the manufacturer’s protocol using a SMRTbell Express Template Prep Kit 3.0 (PacBio). The libraries were sequenced on the PacBio Sequel II System with 2 8M SMRTcells under HiFi mode at Macrogen. Sequencing reads were mapped to the human reference genome (GRCh37) using pbmm2 (v.1.10.0). Allelic configurations of heterozygous SNPs were determined using CisChecker (https://github.com/nccmo/CisChecker).²²

Amplicon deep sequencing of enzymatically converted gDNA

gDNA was fragmented using Covaris sonicator (Covaris). The sheared DNA was treated with the NEBNext Enzymatic Methyl-seq Conversion Module (New England Biolabs) to enzymatically convert unmethylated cytosine to uracil while leaving 5-methylcytosine unaltered. The enzymatically converted gDNA was then subjected to PCR amplification using primers listed in Table S2. The PCR products were used to prepare sequencing libraries using the NEBNext Ultra DNA Library Prep Kit for Illumina (New England Biolabs) and finally sequenced on a MiSeq (Illumina). Fastq files were trimmed using trim_galore (v.0.6.6), and alignment was performed with Bismark (v.0.22.3) with the parameter non_directional. Allelic configurations of the methylation status of the CpG site (chr8:11,660,733) and the heterozygous SNP (chr8:11,660,764) were determined using CisChecker (https://github.com/nccmo/CisChecker).²²

Keratinocyte isolation and in vitro culture

Keratinocytes were isolated from skin biopsies at Japan Tissue Engineering Co. as previously described.²⁷ Isolated keratinocytes were cultured in KGMTM-2 Growth Medium (Lonza) supplemented with or without 5 μg/mL cholesterol (Synthechol, Sigma-Aldrich).

RNA-sequencing

RNA sequencing (RNA-seq) libraries were prepared using the TruSeq Stranded Total RNA Library Prep kit with Ribo-Zero Human/Mouse/Rat (Illumina) and sequenced using NovaSeq6000 (Illumina), generating 50 bp paired-end data. RNA-seq reads were aligned to the human reference genome (GRCh37) using STAR (v.2.6.1c).²⁸ Genes were quantified using RSEM (v.1.3.3).²⁹ Differentially expressed genes (DEGs) were calculated using DESeq2 (v.1.26.0).³⁰ Multiple-testing correction was performed using the Benjamini-Hochberg method. Significant DEGs were selected at the thresholds of absolute log₂ fold change >1 and q value <1.0 × 10⁻²⁰. We conducted a functional enrichment analysis of the DEGs using gProfiler2 (v.0.2.0) with default parameters. The gene sets from Reactome and Gene Ontology Biological Process were assessed.

Immunohistochemistry of FDFT1

Slides of formalin-fixed, paraffin-embedded tissues were deparaffinized in xylene and rehydrated through graded alcohols. The slides were blocked for endogenous peroxidase activity using PBS containing 0.3% hydrogen peroxide and 0.1% sodium azide. Slides were then placed in 100 mM Tris-HCl and 1 mM EDTA buffer (pH 9.0) and boiled in a microwave oven (500 W, 2 min, followed by 200 W, 20 min) for antigen retrieval. Slides were then incubated for 1 h in a blocking buffer solution (1× PBS containing 10% fetal bovine serum and 5% goat serum) and were incubated overnight at 4°C with an anti-FDFT1 antibody (Abcam; see Table S3 for details). The primary antibody was detected using the ImmPRESS polymerized reporter enzyme staining system (ImmPRESS reagent kit; Vector Laboratories) according to the manufacturer’s protocol. Counterstaining was performed using Mayer’s hematoxylin solution (Muto Pure Chemicals).

Protein extraction and immunoblotting

Cells were harvested and lysed with 1× radioimmunoprecipitation assay (RIPA) buffer (50 mM Tris, 150 mM NaCl, 1% Triton X-100, 0.5% sodium deoxycholate, and 0.1% SDS) supplemented with a complete EDTA-free protease inhibitor cocktail (Roche) on ice for 10 min. Lysates were then centrifuged at 12,000 × g for 3 min before protein concentration was quantified using the Pierce BCA Protein Assay kit (Thermo Fisher Scientific). Proteins were separated on a 10% separating gel and transferred to an Immobilon-P membrane (Millipore). The blots were incubated with the antibodies listed in Table S3 and visualized using Western Lightning Plus-ECL (PerkinElmer).

Quantitative reverse-transcription PCR

Quantitative reverse-transcription PCR (RT-qPCR) was performed using a StepOnePlus Real-Time PCR system (Applied Biosystems) with 2× KAPA SYBR FAST qPCR Kit Master Mix (Kapa Biosystems) following the manufacturer’s instructions. All the primers used are listed in Table S2. Transcript quantities were calculated relative to standard curves and normalized to GAPDH mRNA levels.

Cell proliferation assay

Lesion- and non-lesion-derived keratinocytes were plated in 6-well plates at 50,000 cells per well and incubated with or without 5 μg/mL cholesterol (Synthechol, Sigma-Aldrich). Cells were collected 120 h later and counted on a Countess 3 Automated Cell Counter (Thermo Fisher Scientific) using trypan blue. Cell images were acquired using a Leica DMI 6000 B microscope (bright field; Leica Microsystems).

Topical ointment of atorvastatin and cholesterol

An ointment containing 2% cholesterol (Nippon Fine Chemical) and 1% atorvastatin (Aurobindo), approved for use by individuals with porokeratosis by the Division of Clinical Safety of Keio University Hospital, was dispensed during clinical visits and was self-administered between visits. Individuals were instructed to apply a thin layer of ointment twice a day to cover porokeratosis lesions completely. Therapy continued for 14–22 months. Individuals were examined for clinical response and potential adverse effects throughout the treatment period.

Statistical analysis

Statistical analyses were performed with R4.4.1 software (The R Foundation for Statistical Computing). Categorical and continuous data were compared using the two-sided Fisher’s exact test and two-sided Welch’s t test, respectively, unless otherwise specified.

Results

Clinical phenotypes of individuals included in this study

We collected samples from 8 individuals diagnosed with porokeratosis who had no germline pathogenic variants in previously reported genes (MVK, PMVK, MVD, and FDPS) and no family history of porokeratosis (Table 1; Figures 1 and S1). The size and distribution of skin lesions, as well as the age of onset, were heterogeneous across samples, while redness of the lesional skin was common among the individuals. Generalized skin lesions were observed in individuals 1 and 2 (generalized form), whereas a single lesion or mostly unilateral distribution was observed in individuals 3–8 (localized form). The skin lesions were solitary large plaques (individuals 3 and 4), multiple linearly distributed large plaques (individuals 5 and 6), and multiple small circular lesions (<5 mm; individuals 1, 2, 7, and 8). The skin lesions of individuals 3, 5, and 6 were present since infancy or early childhood, whereas those of others developed between their 40s and 80s. In individuals 7 and 8, skin lesions emerged following chemotherapy—administered for acute lymphocytic leukemia and esophageal cancer, respectively. In all individuals, the skin lesions were intensely erythematous with hyperkeratotic rims, which differed from the normal-colored or pigmented lesions seen in porokeratosis caused by variants of previously reported genes (Figure S2A). The clinical diagnoses were disseminated porokeratosis in individuals 1 and 2, porokeratosis of Mibelli in individuals 3 and 4, a linear variant of porokeratosis of Mibelli in individuals 5 and 6,^7,8 and segmentally distributed disseminated porokeratosis in individuals 7 and 8.

Table 1.

Individuals and samples included in the study

Individual	Age at onset	Age at diagnosis	Sex	Age at biopsy	Samples	NGS	FDFT1 (GenBank: NM_004462.5)
							Genetic status		Promoter methylation status
							Variants	Copy-number alterations
1	71	72	female	72	lesion 1	amplicon-seq	c.800T>C (p.Leu267Pro) c.880−2A>T (splice site)	not detected	hypomethylated
					lesion 2	amplicon-seq	c.800T>C (p. Leu267Pro) c.510+2T>G (splice site)	not detected	hypomethylated
					lesion 3	amplicon-seq	c.800T>C (p. Leu267Pro) c.271G>T (p.Glu91∗)	not detected	hypomethylated
					lesion 4	amplicon-seq	c.800T>C (p. Leu267Pro) c.197G>A (p.Arg66His)	not detected	hypomethylated
					lesion 5	amplicon-seq	c.800T>C (p. Leu267Pro) c.607_617delinsT (p.Arg203Leufs∗38)	not detected	hypomethylated
					peripheral blood	amplicon-seq, WES	c.800T>C (p. Leu267Pro)	not detected	N/A
2	66	69	female	69	lesion 1	amplicon-seq	c.395_396dup (p.Arg133Leufs∗11) c.356dup (p.Gln120Profs∗12)	not detected	hypomethylated
					lesion 2	amplicon-seq	c.395_396dup (p.Arg133Leufs∗11) c.476A>G (p.Asp159Gly)	not detected	hypomethylated
					peripheral blood	amplicon-seq, WES	c.395_396dup (p.Arg133Lfs∗11)	not detected	N/A
3	0	3	male	32	lesion 1	WES, WGS	not detected	copy-neutral LOH	hypermethylated
					non-lesion 1	amplicon-seq	not detected	not detected	hypomethylated
					peripheral blood	WES, WGS	not detected	not detected	N/A
4	85	85	male	88	lesion 1	WES, WGS	not detected	copy-neutral LOH	hypermethylated
					peripheral blood	WES, WGS	not detected	not detected	N/A
5	0	25	male	25	lesion 1	amplicon-seq	not detected	not detected	hypermethylated
					peripheral blood	amplicon-seq	not detected	not detected	N/A
6	0	16	male	16	lesion 1	WES, WGS	not detected	copy-neutral LOH	hypermethylated
					non-lesion 1	amplicon-seq	not detected	not detected	hypomethylated
					peripheral blood	WES, WGS	not detected	not detected	N/A
7	42a	45	female	45	lesion 1	WES	not detected	homozygous deletion	hypomethylated
					lesion 2	amplicon-seq	c.577G>T (p.Glu193∗)	not detected	hypermethylated
					lesion 3	amplicon-seq	not detected	heterozygous deletion	hypermethylated
					lesion 4	WES, WGS	not detected	heterozygous deletion	hypermethylated
					lesion 5	amplicon-seq	not detected	heterozygous deletion	hypermethylated
					lesion 6	amplicon-seq	not detected	heterozygous deletion	hypermethylated
					non-lesion 1b	WES, WGS	not detected	not detected	hypomethylated
					non-lesion 2b	WES	not detected	not detected	hypermethylated
8	74a	75	male	75	lesion 1	amplicon-seq	c.358C>T (p.Gln120∗)	copy-neutral LOH	hypomethylated
					lesion 2	amplicon-seq, WES	c.751_752dup (p.Asn251Lysfs∗10)	not detected	hypermethylated
					lesion 3	amplicon-seq, WES	c.622_623delinsAA (p.Gly208Asn)	not detected	hypermethylated
					lesion 4	amplicon-seq, WES	c.1015dup (p.Tyr339Leufs∗10)	not detected	hypermethylated
					non-lesion 1c	amplicon-seq	not detected	not detected	hypomethylated
					non-lesion 2c	WES	not detected	N/A	hypermethylated
					non-lesion 3c	amplicon-seq, WES	not detected	N/A	hypermethylated
					peripheral blood	amplicon-seq, WES	not detected	not detected	N/A

Amplicon-seq, amplicon deep sequencing; NGS, next-generation sequencing; SNP, single-nucleotide polymorphism; WES, whole-exome sequencing; WGS, whole-genome sequencing; LOH, loss of heterozygosity; N/A, not applicable.

^aIndividual 7 developed skin lesions 1 year after bone marrow transplantation, and individual 8 developed skin lesions a few months after chemotherapy.

^bNon-lesion 1 was located distal to the lesions, whereas non-lesion 2 was situated proximal to the lesions.

^cNon-lesions 1, 2, and 3 were situated proximal to lesion 1, 2, and 3, respectively.

Figure 1.

Clinical features of the 8 individuals with porokeratosis in this study

Photographs of porokeratosis lesions (top) and schemas showing the distribution of porokeratosis lesions in the 8 individuals included in this study. The positions of the biopsied lesions are shown in the schema. Multiple lesions were subjected to skin biopsy in individuals 1, 2, 7, and 8. Detailed information on each individual and their skin lesions is provided in Table 1. See also Figure S1.

Bi-allelic genetic alterations of FDFT1 as a cause of the generalized porokeratosis

Our analysis began with individuals 1 and 2, who presented with the generalized form of porokeratosis. To identify germline pathogenic variants associated with porokeratosis, we performed WES on peripheral blood samples from these 2 individuals. WES identified 25,211 and 25,420 coding variants in individuals 1 and 2, respectively. After eliminating putative common SNPs and synonymous variants, 57 and 45 coding variants were detected in 53 and 45 genes, respectively (material and methods). Among these variants, FDFT1 was the only gene mutated in both individuals (Figure S2B). Individual 1 had a heterozygous missense variant at a highly conserved position (c.800T>C [GenBank: NM_004462.5]; p.Leu267Pro) that was predicted to be “disease causing” by MutationTaster (Figures 2A and 2B; Table S4). Individual 2 had a heterozygous frameshift insertion (c.395_396dup [p.Arg133Leufs^∗11]) in FDFT1 (Figure 2A; Table S4). These germline variants were confirmed using amplicon deep sequencing (Table S1).

Figure 2.

Somatic and germline variants of FDFT1 identified in individuals with generalized porokeratosis

(A) Distribution of somatic and germline variants in FDFT1 in individual 1 (top) and individual 2 (bottom).

(B) Multi-alignment analysis of FDFT1 and its orthologous protein sequences from different species: cow (bos taurus), chimpanzee (pan troglodytes), gorilla (gorilla gorilla gorilla), rabbit (oryctolagus cuniculus), mouse (mus musculus), chicken (gallus gallus), frog (xenopus tropicalis), and zebrafish (danio rerio). Color indicates the BLOSUM62 score.

(C) Genes related to cholesterol synthesis. FDFT1 is shown in red, and the other genes associated with porokeratosis are shown in blue.

(D) Allelic configurations (cis versus trans) of germline and somatic variants assessed by Sanger sequencing of the cloned gDNA (for individual 1 lesions 1, 2, and 5) and next-generation sequencing reads (for individual 2 lesion 2). Proportions of mutant and reference alleles are shown. Examined numbers are shown in parentheses. For individual 1 lesion 5, deletion-spanning PCR primers were used to amplify the gDNA of the allele with the somatic deletion. See also Figure S2.

FDFT1 encodes farnesyl-diphosphate farnesyltransferase 1, a membrane-associated enzyme that synthesizes squalene from 2 molecules of farnesyl pyrophosphate. This enzyme is involved in cholesterol biosynthesis and is located in the mevalonate pathway, downstream of the enzymes encoded by MVK, PMVK, MVD, and FDPS, whose pathogenic variants have been reported to cause porokeratosis (Figure 2C).

Next, we performed amplicon deep sequencing on 5- and 2-lesion samples from individuals 1 and 2, respectively. Interestingly, the epidermis of all 7 lesions had unique somatic variants in FDFT1 (Figure 2A; Table S4). The identified somatic variants were 2 splice site variants, 2 missense variants, 1 nonsense variant, 1 frameshift insertion, and 1 frameshift deletion. The somatic missense variants (c.197G>A [p.Arg66His] in individual 1 lesion 4 and c.476A>G [p.Asp159Gly] in individual 2 lesion 2) were located at moderately to highly conserved amino acids and were predicted to be disease causing (Figure 2B). We also performed SNP array analysis, which identified no copy-number alterations (CNAs) (Figure S2C).

Somatic and germline variants can arise in cis (on the same allele) or in trans (on opposite alleles). Therefore, we next investigated the allelic configurations of these variants using next-generation sequencing or Sanger sequencing of the cloned gDNA (material and methods). We evaluated 4 lesions in which the genomic distances between germline and somatic variants were <10 kb, which revealed that the somatic and germline variants were all in trans (Figure 2D). Together, these results indicate that a heterozygous germline pathogenic variant of FDFT1 is associated with generalized porokeratosis and that bi-allelic loss-of-function variants of FDFT1 are required for the development of skin lesions, as in the case of previously reported genes.^16,17

Lesion-specific somatic genetic alterations in FDFT1 in localized porokeratosis

Next, we performed amplicon deep sequencing and/or WES of 14 lesion samples (1–6 samples per individual) and germline control samples from 6 individuals with localized porokeratosis (individuals 3 to 8, Table 1). Intriguingly, we identified FDFT1 heterozygous somatic variants in all 4 lesion samples from individual 8 (Figures S3A and S3B; Table S4). Additionally, we identified a heterozygous somatic variant in FDFT1 in lesion 2 of individual 7 (Figure S3A; Table S4). However, FDFT1 germline variants were not detected in either individual.

Copy-number analysis identified copy-neutral loss of heterozygosity (LOH) of chromosome 8 short arm (8p) spanning FDFT1 in individual 3 lesion 1, individual 4 lesion 1, individual 6 lesion 1, and individual 8 lesion 1, in which 1 allele was deleted and the other was duplicated coextensively (Figures S4A–S4D). We also identified a segmental deletion of chromosome 8p resulting in homozygous and heterozygous deletions of the FDFT1 locus in lesion 1 and lesions 3–6 of individual 7, respectively (Figure S4E). The length of chromosomal deletions varied among lesion samples, confirming that they were genetically independent.

Together, we identified somatic variants, deletions, and/or copy-neutral LOH in 13 of the 14 lesion samples (92.9%) of localized porokeratosis (Figure S5). However, bi-allelic genetic alteration of FDFT1 was observed in only 2 of the 14 lesion samples, suggesting different etiologies of localized porokeratosis from its generalized counterpart.

Hypermethylation at CpG islands of the FDFT1 promoter in skin lesions

Because we observed mono-allelic genetic alterations of FDFT1 in the majority of lesion samples from individuals with localized porokeratosis, rather than bi-allelic alterations, we hypothesized that transcriptional silencing of the other FDFT1 allele contributes to the pathogenesis of localized porokeratosis. FDFT1 promoter typically remains unmethylated across various tissues and cell lines, as demonstrated by public bisulfite sequencing data (Figure S6A; Table S5). We next performed genome-wide methylation analysis of all 21 lesion samples from the 8 individuals using the Infinium MethylationEPIC Kit, which covered >850,000 CpGs in the genome. Additionally, we evaluated 7 non-lesion samples from morphologically normal epidermis adjacent to and distant from lesions (n = 6 and 1, respectively) from 4 individuals and 2 normal epidermal samples from individuals with other skin diseases.

Twelve lesion samples without bi-allelic genetic alterations of FDFT1 were compared with 7 non-lesion samples, which identified prominent hypermethylation of the CpG island at the FDFT1 promoter region (chr8:11,659,676–11,660,795 [GRCh37]) in all 12 lesion samples (Figures 3A, 3B, and S6B–S6F; Table S6). The β value, indicative of methylation levels on a scale from 0 (no methylation) to 1 (full methylation) at a specific CpG site, was approximately 0.5 or higher at the FDFT1 promoter in these 12 samples, in contrast to control samples that presented a β value of approximately 0.1 or less (Figure 3B). Notably, at least 8 of these lesion samples (66.7%) without bi-allelic genetic alterations appeared to have homozygous or hemizygous methylation of the FDFT1 promoter, as the β values of these samples were higher than 0.5. FDFT1 methylation levels were low in the 9 lesion samples with FDFT1 bi-allelic genetic alteration (Figures 3B and S6D–S6F).

Figure 3.

Bi-allelic genetic and/or epigenetic alterations of FDFT1 in all lesion samples of localized porokeratosis

(A) Volcano plot of methylation differences between lesion samples without bi-allelic genetic alterations of FDFT1 (n = 12) and non-lesion samples (n = 7). The x axis shows the magnitude of the effect and the y axis shows the −log₁₀(q value). Each dot represents an autosomal CpG island. The red dot represents the CpG island located in the FDFT1 promoter region (chr8:11,659,676–11,660,795 [GRCh37]). The significance of methylation levels was assessed using a two-sided analysis of variance (ANOVA), which included sample type (lesion and non-lesion), individual ID, and probe ID as explanatory variables. Multiple-testing correction was performed using the Benjamini-Hochberg method.

(B) DNA methylation levels (β values) of a representative probe (cg24123057; chr8:11,660,089) at CpG islands of the FDFT1 promoter in the epidermis of 2 normal skins (N), 7 non-lesions (nL), 2 lesions from localized porokeratosis with bi-allelic genetic alterations (LL1), 12 lesions from localized porokeratosis without bi-allelic genetic alterations (LL2), and 7 lesions from generalized porokeratosis (GL).

(C) Allelic configurations (cis versus trans) of the methylation status of the CpG island at the FDFT1 promoter (chr8:11,660,733) and adjacent heterozygous SNP (chr8:11,660,764) assessed by amplicon deep sequencing of enzymatically converted gDNA of individual 8 lesions 2–4. Proportions of methylated and unmethylated reads are shown.

(D) Allelic configurations of the somatic variant (chr8:11,683,644) and adjacent heterozygous SNP (chr8:11,683,818) assessed by next-generation sequencing reads in individual 8 lesion 3.

(E) Allelic configurations of the somatic variant (chr8:11,687,800) and adjacent heterozygous SNP (chr8:11,687,959) assessed by Sanger sequencing of the cloned gDNA in individual 8 lesion 2.

(F) Allelic configurations of the somatic variant (chr8:11,689,161) and adjacent heterozygous SNP (chr8:11,689,119) assessed by next-generation sequencing reads in individual 8 lesion 4. (D–F) Proportions of mutant and reference reads are shown. (C–F) Examined numbers are shown in parentheses.

(G) Phasing of heterozygous SNPs using long-read WGS reads of the germline control sample of individual 8. Positions of heterozygous SNPs (blue) according to their genomic position (top) and the number of long-read WGS reads between them (bottom).

(H) Schematic view of FDFT1 status in 21 lesion samples from the 8 individuals, revealed by integrated genetic and epigenetic analyses. Individuals 1 and 2 had germline pathogenic variants of FDFT1, and they developed multiple small lesions by acquiring independent somatic variants of FDFT1 on opposite alleles. Individuals 3–6 developed a single large lesion due to abnormal hypermethylation of the FDFT1 promoter, coupled with copy-neutral LOH of the FDFT1 locus in individuals 3, 4, and 6. Individuals 7 and 8 developed multiple lesions due to homozygous deletions of FDFT1 (individual 7 lesion 1), bi-allelic somatic variants of FDFT1 coupled with copy-neutral LOH (individual 8 lesion 1), and abnormal hypermethylation of the FDFT1 promoter coupled with heterozygous deletion (individual 7 lesions 3–6) or somatic pathogenic variant (individual 7 lesion 2 and individual 8 lesions 2–4) of FDFT1. See also Figures S3–S6.

Allelic configuration of mono-allelic methylation and somatic variants

Four lesion samples from 2 individuals (individual 7 lesion 2 and individual 8 lesions 2–4) had an intermediate methylation level. Therefore, we investigated the methylation status (mono-allelic or bi-allelic) in 3 samples from individual 8. The allelic configuration of methylation at the CpG site and adjacent heterozygous SNPs revealed high mono-allelic methylation in these samples (Figure 3C).

Given that these samples from individual 8 also harbor somatic variants, we next evaluated the allelic configuration of mono-allelic methylation and somatic variants. The allelic configurations of somatic variants and adjacent heterozygous SNPs (Figures 3D–3F), along with the direct phasing of heterozygous SNPs at the FDFT1 locus (Figure 3G) and the allelic configurations of methylation and adjacent heterozygous SNPs (Figure 3C), demonstrated that the methylation was in trans with the somatic variants in all 3 samples.

Together, the combined genetic and epigenetic analyses identified bi-allelic genetic and/or epigenetic alterations of FDFT1 in all lesion samples of localized porokeratosis (Figure 3H). FDFT1 promoter was bi-allelically methylated in 4 lesion samples, 3 of which were coupled with copy-neutral LOH. In contrast, FDFT1 promoter was mono-allelically methylated in 8 lesion samples, all of which also harbored heterozygous loss-of-function somatic variants or heterozygous deletions of FDFT1 on the opposite allele.

Decreased FDFT1 expression in lesional epidermis

We next evaluated FDFT1 at the RNA and protein levels. RNA was extracted from the epidermis of 2 skin lesions in 2 individuals (individual 3 lesion 1 and individual 6 lesion 1), both showing bi-allelic FDFT1 promoter hypermethylation (Figure 3H). Notably, both samples exhibited a marked reduction in FDFT1 RNA expression levels (Figures 4A, S7A, and S7B). Immunohistochemical staining demonstrated reduced FDFT1 staining in the epidermis of almost all evaluated lesions, except for those involving missense variants (Figures 4B and S7C–S7K). Importantly, FDFT1 localization was uniformly and specifically diminished within the lesional epidermis (Figure 4B), which confirmed that porokeratosis is a clonal disorder.⁴ At the edge of skin lesions, FDFT1-positive cells and FDFT1-attenuated cells showed a clear boundary, where the FDFT1-positive cells bordering the boundary showed aberrant keratinization and formed cornoid lamella (Figures 4B and S7E–S7G). In summary, our data showed that genetic alterations or promoter hypermethylation resulted in decreased expression of FDFT1 in a lesional epidermis-specific manner.

Figure 4.

Decreased FDFT1 mRNA and protein expression in lesional epidermis

(A) Expression of FDFT1 relative to GAPDH measured by quantitative reverse-transcription PCR (RT-qPCR) for lesion and non-lesion samples of individual 6. (A–D) indicates primer pairs. See Figure S7A and Table S2 for primer positions and sequences. Each circle indicates a technical replicate.

(B) Hematoxylin and eosin (H&E) staining (top left) and FDFT1 immunohistochemistry (bottom left and right) of individual 5 lesion 1. Asterisks indicate cornoid lamella, which is the sharp demarcation of the lesions. Arrows indicate the range of lesional skin with abnormal keratinization determined from H&E staining. The dotted box area in the lower left image is magnified in the right image. Scale bar, 200 μm. See also Figure S7.

No detection of genetic alteration leading to FDFT1 promoter hypermethylation

Alterations of normal methylation can arise in the absence of DNA sequence changes (primary epimutation) or secondary to genetic variants (secondary epimutation). To evaluate potential genetic variants related to aberrant hypermethylation of FDFT1, we performed WGS of 4 lesion samples from 4 individuals exhibiting aberrant hypermethylation of FDFT1. In these lesion samples, we found no somatic genetic alterations within 5 kb of FDFT1, including its adjacent genes NEIL2 (MIM: 608933) and CTSB (MIM: 116810). We also found no somatic genetic alterations in DNA methyltransferases, histone modifiers, or transcription factors. These results suggested that the FDFT1 promoter hypermethylation arose in the absence of DNA alterations.

FDFT1 methylation mosaicism in morphologically normal epidermis

As hypermethylation of the FDFT1 promoter was detected in multiple skin lesions, whereas somatic genetic alterations were specific to each skin lesion, the occurrence of hypermethylation was suspected to precede the somatic alterations. Therefore, we investigated whether there are cells showing hypermethylation of the FDFT1 promoter in the normal epidermis of the individual. Methylation analysis of morphologically normal epidermis adjacent to lesions revealed that the FDFT1 promoter was methylated in 3 non-lesion samples from 2 individuals (individuals 7 and 8) (Figures 3B, 5A, 5B, S6D–S6F, and S8). FDFT1 somatic variants or CNAs detected in the lesions were not detected in these samples, confirming that this hypermethylation was not due to cross-contamination of DNAs from the lesion samples. FDFT1 promoter methylation was not observed in normal epidermis obtained from other skin diseases (n = 2, Figures 3B and S6D–S6F). Notably, in individual 7, non-lesion sample adjacent to lesions (non-lesion 2) showed hypermethylation of the FDFT1 promoter, while that distal from lesions (non-lesion 1) showed low methylation levels of the FDFT1 promoter (Figure S8). Together, these results suggest that asymptomatic FDFT1 methylation mosaicism had existed in the epidermis of individuals 7 and 8 prior to the development of skin lesions.

Figure 5.

Epigenetic mosaicism of FDFT1 in morphologically normal epidermis

(A) Schematic representations of the distribution of porokeratosis lesions in individual 8. The positions of the biopsied lesions and non-lesions are indicated. In individual 8, lesions and non-lesions were separated from the same biopsy samples, as shown in the image and schematic diagram on the right.

(B) DNA methylation levels (β values) of 16 probes at the CpG island in the FDFT1 promoter region in lesions and non-lesions from individual 8.

(C) Number of somatic variants detected by WES in 3 lesions and 2 non-lesions from individual 8. One variant was detected in all evaluated lesion and non-lesion samples.

(D) Schematic representations showing the timing of methylation and acquisition of somatic variants. White and gray circles represent unmethylated and methylated cells, respectively. Somatic variants acquired before methylation become clonal in methylated cells. Hence, only a few somatic variants become clonal, or none at all, if methylation occurs early in development (top), whereas many somatic variants become clonal if methylation occurs late in life (bottom).

(E) Median DNA methylation levels (β values) at the CpG islands in the MVK, PMVK, MVD, FDPS, and FDFT1 promoter regions in 38 epidermal samples from 19 unaffected individuals. See also Figure S8.

Next, we estimated the timing of the hypermethylation in these samples. We performed WES of 3 non-lesion and 5 lesion samples from individuals 7 and 8. WES identified 0–1 (median 1) and 2–57 (median 25) somatic variants in non-lesions and lesions, respectively. While most of the somatic variants were unique to each sample, 1 variant was shared between 3 lesions and 2 non-lesions in individual 8 (Figure 5C). Our results suggest that the hypermethylation occurred very early in the development before the accumulation of most variants (Figure 5D).

Finally, we evaluated the frequency of hypermethylation in FDFT1 in unaffected individuals using publicly available methylation array data.²⁶ Among 38 epidermal samples from 19 individuals, 1 sample from an older individual showed hypermethylation of FDFT1, while another sample from the same individual showed no hypermethylation. Moreover, no hypermethylation was observed in genes previously reported to be associated with porokeratosis, including MVD and MVK. While the number of samples is limited, the results suggest the existence of FDFT1 gene-specific methylation mosaicism even in unaffected individuals (Figure 5E).

Lesion-derived keratinocytes showed transcriptomic changes compatible with the etiology of porokeratosis

Keratinocytes were isolated from 2 skin biopsies (lesion and non-lesion) from individual 5 and cultured in vitro. We confirmed a marked reduction in FDFT1 mRNA and protein expression in cultured lesion-derived keratinocytes using RT-qPCR and immunoblotting, respectively (Figures S9A and S9B). The methylation array revealed that the global methylation status was maintained in cultured keratinocytes (Figure S9C), although the methylation level of the FDFT1 promoter region was slightly higher in lesion-derived keratinocytes than that in the original lesional epidermis (Figure S9D). This is probably because the original lesional epidermis contained cells other than keratinocytes that showed low FDFT1 methylation. Together, our results suggest that cultured keratinocytes from both lesion and non-lesion samples maintain transcriptomic and epigenetic phenotypes ex vivo.

Next, we performed RNA-seq on keratinocytes derived from the lesion and non-lesion of individual 5. The analysis identified 418 DEGs, among which 217 were downregulated and 201 were upregulated in lesion-derived keratinocytes (Figure 6A; Table S7). Consistent with the RT-qPCR results, FDFT1 was one of the most downregulated genes. Functional enrichment analysis of DEGs identified that cholesterol biosynthesis pathways were upregulated in lesion-derived keratinocytes (Figures 6B and S9E; Table S8), suggesting feedback regulation of cholesterol biosynthesis. Interestingly, genes related to skin development, epidermis development, and keratinization were also overrepresented, which was consistent with the abnormal hyperkeratinization observed in porokeratosis. Conversely, genes related to cell-cycle progression were downregulated, suggesting cell-cycle alterations in the lesion-derived keratinocytes.

Figure 6.

Transcriptomes and phenotypes of lesion-derived keratinocytes

(A) Volcano plot of differentially expressed genes (DEGs) in keratinocytes isolated from the lesion and non-lesion of individual 5 (n = 3 for each). The x axis shows the log₂ fold change and the y axis shows the −log₁₀(q value). Red dots represent significantly upregulated genes in lesion-derived keratinocytes (201 genes), whereas blue dots represent significantly downregulated genes (217 genes). Significant DEGs were selected at the thresholds of absolute log₂ fold change >1 and q value <1.0 × 10⁻²⁰.

(B) Pathway enrichment analysis of DEGs between keratinocytes isolated from the lesion and non-lesion of individual 5. The x axis shows the p value obtained using gProfiler2 with multiple-testing corrections. Pink and blue indicate pathways upregulated and downregulated in lesion-derived keratinocytes, respectively.

(C) Proliferation of lesion- and non-lesion-derived keratinocytes cultured in medium supplemented with or without cholesterol for 5 days. Two-sided Welch’s t test.

(D) Light microscopy images of lesion- and non-lesion-derived keratinocytes cultured in medium supplemented with or without cholesterol for 120 h. Scale bar, 100 μm. (C and D) 50,000 cells were plated in a 6-well plate and cultured in the aforementioned medium for 120 h. See also Figure S9.

Cholesterol dependency of lesion-derived keratinocytes for growth

Prompted by transcriptomic and epigenetic changes in lesion-derived keratinocytes, we evaluated the phenotypes of these cells. Lesion- and non-lesion-derived keratinocytes were morphologically identical and showed similar growth rates in a normal culture medium. However, after cholesterol depletion, lesion-derived keratinocytes showed markedly reduced proliferation (p = 0.0035; Figures 6C and 6D). This result suggests that cholesterol biosynthesis is largely interrupted by FDFT1 deficiency in lesion-derived keratinocytes (Figure 2C).

Skin lesion improvement by topical cholesterol and atorvastatin treatment

Recently, a topical ointment containing cholesterol and statin has been reported to be effective against the cutaneous symptoms of porokeratosis and congenital hemidysplasia with ichthyosiform erythroderma and limb defects (CHILD) syndrome (MIM: 308050), both of which are caused by deficiencies of enzymes involved in cholesterol biosynthesis.^31,32,33 Statins, a widely used class of cholesterol-lowering drug, inhibit 3-hydroxy-3-methylglutaryl coenzyme A reductase (HMGCR; MIM: 142910), an enzyme in the upper mevalonate pathway. It is proposed that statins prevent the accumulation of toxic metabolic intermediates in the mevalonate pathway (Figure 2C). Thus, we conducted a therapeutic evaluation of atorvastatin and cholesterol ointment in 3 individuals with FDFT1-deficient porokeratosis (Figure 7). All individuals exhibited reduced skin redness and thickening, pruritus, and scaling within 4–12 weeks of treatment initiation, and no relapse was observed with continued use of the ointment. In addition, they also demonstrated good tolerance to the ointment, with no side effects such as contact dermatitis.

Figure 7.

Skin lesion improvement by topical atorvastatin and cholesterol treatment

(A–C) Representative images of skin lesions before and after topical atorvastatin and cholesterol treatment in individual 1 (A), individual 2 (B), and individual 6 (C).

Discussion

Through comprehensive genetic and epigenetic profiling, we identified both germline and somatic pathogenic variants in FDFT1 associated with porokeratosis. Furthermore, we discovered that postzygotic gene-specific epigenetic silencing of FDFT1 underlies non-hereditary localized forms of porokeratosis. Skin lesions developed as a result of bi-allelic FDFT1 inactivation, with the cause being heterogeneous across individuals and skin lesions (Figure 3H). Individuals with heterozygous germline pathogenic variants of FDFT1 developed the generalized forms by acquiring somatic variants of FDFT1 on opposite alleles. Conversely, individuals without germline pathogenic variants developed the non-hereditary localized forms mainly due to postzygotic gene-specific hypermethylation of the FDFT1 promoter, sometimes coupled with copy-neutral LOH or somatic variants of FDFT1 on opposite alleles. Importantly, we identified hypermethylation of the FDFT1 promoter in several normal tissue samples adjacent to the methylation-related skin lesions but not in those distal from the skin lesions in 2 individuals. This finding suggests the existence of somatic epigenetic mosaicism in the normal epidermis that occurs early in life and serves as an initiating event in the pathogenesis of localized porokeratosis.

One of the major findings in our study is the discovery of a disorder related to epigenetic mosaicism. While the exact timing of acquiring epigenetic mosaicism is unknown, WES analysis and the presence of congenital skin lesions in specific individuals collectively suggest that it occurred during a developmental period. It is speculated that FDFT1 hypermethylation was acquired in a subset of cells of the ectoderm lineage or cells that subsequently differentiate into this lineage during development. In our study, epigenetic mosaicism became apparent by the onset of porokeratosis, and there may be more instances of epigenetic mosaicism that do not involve phenotypic changes. Further studies are warranted to understand the landscape of epigenetic mosaicism not only in the skin but also in other organs.

In our study, among the 6 individuals whose lesions showed FDFT1 hypermethylation, 2 exhibited solitary lesions, 2 displayed linear lesions, and the other 2 displayed linearly distributed multiple lesions. The linear pattern appears to follow the lines of Blaschko, which represent the lines of epidermal cell migration and proliferation during the development.^1,3 Therefore, the well-established correlation between cutaneous mosaicism patterns and mutated cell components in genetic mosaicism could potentially be applied to epigenetic mosaicism.

One remaining question was the cause of the hypermethylation of the FDFT1 promoter. Importantly, in the individuals examined, the global methylation pattern was not altered, and the hypermethylation occurred in a gene-specific manner. WGS identified no genetic alterations near FDFT1, including its neighboring genes, suggesting that the hypermethylation was likely a primary epimutation rather than a secondary epimutation caused by genetic alterations.

While most autosomal genes are expressed from both alleles, recent studies have identified autosomal genes showing random mono-allelic expression (RME),³⁴ which is stably maintained across cell generations and is observed in a cell-type or tissue-specific manner. Although the cause of RME is not fully understood, promoter methylation levels and expression status have shown a correlation in some, but not all, RME genes.³⁵ Therefore, while FDFT1 has not been described as an RME gene, there might be a shared molecular mechanism between FDFT1 hypermethylation and RME.

Monogenic disorders caused by epigenetic alterations are rare, with most reported cases being imprinting disorders such as Beckwith-Wiedemann syndrome (MIM: 130650).³⁶ However, they can also occur in non-imprinted genes, as in the case of mono-allelic hypermethylation of MLH1 (MIM: 120436) in Lynch syndrome (MIM: 609310).^37,38 Both primary epimutations and secondary epimutations have been reported for MLH1 hypermethylation.^37,39 Because hypermethylation of the MLH1 promoter occurs in a gene-specific manner, there might be a shared mechanism with FDFT1 promoter hypermethylation in porokeratosis. Our findings, along with reports of MLH1 hypermethylation in Lynch syndrome, suggest the existence of a category of diseases associated with hypermethylation of non-imprinted genes.

In our findings of localized porokeratosis, in the keratinocytes with mono-allelic FDFT1 promoter hypermethylation, acquisition of secondary somatic alterations on opposite alleles led to the clonal expansion of keratinocytes and subsequent skin lesion formation. Previously, the contribution of methylation to clonal expansion has been described in individuals with Wilms tumor (MIM: 194070), where hypermethylation of the H19 locus is implicated in premalignant clonal expansion of nephrons.⁴⁰ However, unlike FDFT1, the H19 locus is regulated by genomic imprinting and exclusively expressed from the maternal allele. Therefore, our study expanded the concept of methylation-related clone expansion to non-imprinted genes.

Our study has several important findings for the understanding of porokeratosis. First, immunostaining of FDFT1 confirmed the long-standing hypothesis that the porokeratosis is a clonal disorder and that the cornoid lamella is located at the border of clonally expanded mutant cells.⁴ While previous genetic studies have identified clonal genetic alterations, our study histologically confirmed that the mutant cells are clonally distributed and not mixed with FDFT1-expressing cells within the skin lesions. Notably, the cornoid lamella is located at the border and is formed by the FDFT1-expressing cells surrounding the skin lesions. Future studies are needed to elucidate the molecular signals that induce the formation of the cornoid lamella. Second, we noticed that 2 individuals (individuals 7 and 8) developed multiple lesions following the lines of Blaschko after chemotherapy. The multiple developments of skin lesions in a limited skin area suggest that chemotherapy-induced mutagenesis and/or immunosuppression-induced failure to eliminate mutant cells presumably accelerated the development of numerous skin lesions. Third, whereas linear porokeratosis has been reported to be caused by secondary somatic alterations occurring in individuals with heterozygous germline causative variants,^16,17 the linear porokeratosis presented here is associated with postzygotic epigenetic silencing of FDFT1. Therefore, the former is an autosomal-dominant trait and the latter is non-hereditary, which is important information for genetic counseling. Finally, we identified a therapeutic option for FDFT1-deficient porokeratosis. While sporadic studies have employed the topical application of statin and cholesterol for porokeratosis, the results have been controversial.^{32,33,41,42,43,44} Given that statins presumably prevent the accumulation of toxic metabolic intermediates upstream of the deficient enzyme,³¹ the effect of this treatment needs to be evaluated for each of the deficient enzymes encoded by genes associated with porokeratosis. Although the number of individuals evaluated in this study was limited, our results suggest that topical application of statin is a promising therapeutic option for FDFT1-deficient porokeratosis.

In summary, our study identified bi-allelic genetic and/or epigenetic alterations in FDFT1 as causes of porokeratosis, supported by transcriptomic and functional studies as well as the discovery of effective therapies. Our findings highlight the presence of gene-specific asymptomatic epigenetic mosaicism in non-imprinted genes that predisposes cells to clonal expansion. This study not only advances our understanding of porokeratosis pathogenesis but also contributes to our knowledge of epigenetic mosaicism and clonal expansion of non-cancerous cells in general.

Data and code availability

Sequencing data and array data have been deposited in the Japanese Genotype-Phenotype Archive (JGA) under accession number JGAS000684.

Published: April 22, 2024