The recruitment of the families was done at the Genodermatoses Unit at the Department of Dermatology at the American University of Beirut Medical Center. Clinical phenotypes were provided by the referring physician. The project was reviewed and approved by the Institutional Review Board (IRB) at the American University of Beirut Medical Center (Protocol Number: DER.MK.01), and written informed consent was obtained from the participants, or from their parents if minors, to collect blood samples and pictures. The patients included in this study were unrelated, representing distinct familial backgrounds. Patient 1 was affected by androgenetic alopecia (AGA), while Patient 2 was affected by vitiligo and a family history of Hermansky-Pudlak Syndrome (HPA).

Peripheral blood samples were collected from patients and stored at 4° C. DNA was extracted from the specimens within one hour of collection. Whole exome sequencing (WES) was conducted for the probands and family members by Macrogen Laboratory, Seoul, Korea. The FASTQ files were mapped to Human GRCh37/hg19 reference assembly using CLC Genomics Workbench (version 20.0.4). Variant calling and annotation were performed using Illumina VariantStudio software version 3.0. To identify the mutations/variants that might lead to the diseases, stringent filter was applied ( Figure 1 ). Detailed methods for this analysis can be found elsewhere [ 13 ] .

To explore the biological processes and pathways that are regulated in the diseases, we analyzed and the Gene Expression Omnibus (GEO) DataSets using Ingenuity Pathway Analysis (IPA) software [ 14 ] . We curated datasets based on: (1) studies that have skin biopsies/blood samples to conduct expression profiling by microarray or RNA sequencing; (2) studies with information about the technology used; (3) studies including normal and control groups. From that, we curated these four datasets GSE90594, GSE36169, GSE90880, and GSE75819 ( Table 1 ).

The proband (III.1, Figure 2A ) is a 3-year-old female who visited the clinic due to features like adult male androgenetic alopecia, exhibiting hair loss and widening at the vertex and frontotemporal areas and had sparse eyebrows and eyelashes ( Figure 2A and D ). The results of the exome analysis yielded 75105 variants per sample before stringent filtering and 613 variants per sample after stringent filtering ( Figure 1 ). After the first round of filtering, as we did not find any homozygous mutation in the patient that fit the stringent criteria, we looked for compound heterozygous mutations. Two potential variants were detected; the first variant is FOXC1 c.1450C>T, p.H484Y, which leads to a heterozygous missense mutation with a predictive CADD score of 26.8 ( Table 1 and Figure 2B and E ). The variant is predicted to be likely pathogenic according to the ACMG [ 15 ] classifications and has a deleterious effect with a tree vote of 58|42 (del | benign) [ 16 ] . The variant was absent from the genomes/exomes databases in gnomad [ 17 ] , ExAC [ 18 ] , 1000G [ 19 ] , and 300 Lebanese in-house exomes ( Table 1 ). The other deleterious variant, SMARCD1 c.1051C>T, p.R351C (g.5537C>T), results in a heterozygous missense mutation with a CADD score of 33 ( Table 1 and Figure 2C and F ).

The analysis of the GEO datasets GSE90594 and GSE36169 of individuals affected with androgenic alopecia compared to controls showed that FOXC1 , an upstream regulator that leads to activation, had a z-score of 0.515, p-value 3.07E ⁻⁰³ and z-score of 4.233, p-value 3.64E−22, respectively. In the GSE90594 dataset, under the top diseases and biofunction, hair and skin development ranked as the second top function, with a p-value of 1.78-07-2.65E ⁻⁶⁵ , and dermatological diseases ranked as the third top diseases, with a p-value of 2.22E ⁻⁰⁵ –1.80E ⁻⁶⁸ ( Table 2 ). For the network analysis, we found that there was a protein-protein interaction (PPI) between FOXC1 and SMARCD1 through SOX2 and YAP1 genes. We found that FOXC1 played a role in skin formation and keratinocyte differentiation, and that both FOXC 1 and SMARCD1 increased the risk of skin cancer ( Figure 3 ).

The patient (III.1, Figure 4A ) was a 12-year-old female who presented to the clinic with vitiliginous lesions accompanied by immune deficiency. Her paternal uncle, patient II.3 ( Figure 4A ), is a 48-year-old male diagnosed with albinism, with no immune deficiency. The exome analysis and filtering criteria is shown in Figure 2 . The patient and her uncle harbored pathogenic nonsense variant HPS1 c.1697C>A, p. Ser566Ter, that can lead to nonsense-mediated mRNA decay (NMD) ( Figure 4B/D ) [ 16 ] with a CADD score of 37. This variant was not present in the public databases. To explain the immune deficiency phenotype in the proband, we searched for immunoregulatory genes and found a deleterious homozygous missense variant ITK p. Pro521Leu with a CADD score of 29.6. ( Figure 4C and E ). The variant was present with a rare MAF in gnomad and 1000 Genome, and ExAC databases 0.00000796, 0.000007957, and 0.000008237, respectively ( Table 1 ).

IPA analysis of the GEO dataset GSE90880 revealed that immune cell trafficking was ranked the top function in the top diseases and biofunction, with a p-value range of 2.15E ⁻⁵⁴ – 5.91E ⁻¹⁰⁹ ( Table 2 ), confirming the relevance to the underlying trait. No significant association was detected for GSE75819. For the upstream analysis, we found that ITK was an upstream regulator that leads to inhibition of the trait, with a z-score: −2.08, p-value 1.19E ⁻⁰⁶ and z-score: −1.813, p-value 9.91E ⁻⁰² , respectively. Finally, the network analysis showed that ITK and HPS1 were upstream of the TNF gene ( Figure 5 ).

AGA is well-known in adults but under-reported in pediatric and prepubertal children, making prevalence estimates challenging. A retrospective report identified 57 of 483 pediatric patients (13%, ages 8–19) with AGA and a single case of a 6-year-old female with AGA, all diagnosed based on clinical manifestations, strong family history, genetic predisposition, or endocrinological abnormalities [ 21 , 22 ] . Our patient represents the youngest case reported in the literature affected with AGA (consistent with male androgenic alopecia hair loss) with no prepubertal status or a family history of AGA, despite consanguinity originating from the granparents’ generation, making it an intriguing case to further investigate.

AGA is characterized by progressive hair follicular miniaturization due to shortened anagen phase during hair cycle leading to disrupted hair growth cycles, dermal papilla size, and keratinocyte activities, which is mainly controlled by androgens through intracellular signaling of the hair follicle target cells. Androgen receptors (AR) are localized in dermal papilla cells (DPC) in a hair follicle, indicating that DPC are the main target cells for androgen, which is shown in patient with AGA compared to non-AGA patients [ 23 – 27 ] . Androgens such as progesterone, androstenedione, and testosterone are converted into a more potent androgen, dihydrotestosterone (DHT), by the cytoplasmic 5α-reductase enzyme. DHT binds to AR in the cytoplasm, making AR-DHT complex, which translocates to the nucleus after dimerization. AR coactivators are recruited to this complex, which binds to the androgen-response element consistent with the DNA sequence. The coactivators connect AR and transcription factors and RNA polymerase, which results in gene transcription followed by protein translation, which exerts the biological activity. This multi-step molecular pathway can be implicated in AGA pathogenesis. [ 25 , 28 ] . As such, the genes previously reported were linked to androgen signaling pathways such as (AR )/ EDAR2 , SRD5A1, SRD5A2 , HSD17B2, HSD17B3, SHBG, AKR1C1, AKR1C2, AKR1C3 , and FOXA2 [ 26 , 29 – 31 ] . Pathogenic variants in these genes were absent in our sequencing results, leading us to investigate other genes involved in AGA pathogenesis. Our analysis revealed a potential combinatorial role for FOXC1 and SMARCD1 in the phenotype ( Figures 3 and 6 ). SMARCD1 (SWI/SNF-related matrix-associated, actin-dependent regulator of chromatin, subfamily D, member 1) gene encodes BAF60A, a key protein in nucleosome-DNA interactions and found to have the highest induction of a strong interaction with AR. SMARCD1 directly interacts with the coactivator groove in the AR cofactor via its FxxFF motif directly in a hormone-dependent manner and acts as a promoter for the expression of TMPRSS2 gene [ 32 ] . The SMARCD1 mutation lies within the SWIB domain of the protein. A previous study identified several SMARCD1 variants linked to a neurodevelopmental disorder with sparse or temporal hair deficiency as a phenotype ( Figure 2B ) [ 33 ] .

The other player in this study is FOXC1 (Forkhead box C1), a transcriptional factor involved in cell differentiation and embryogenesis that plays a key role in human keratinocytes terminal differentiation [ 34 ] . It has been demonstrated that FOXC1 maintains and reinforces quiescence in self-renewing hair follicle stem cells in mice and humans leading to premature aging of these stem cells [ 35 – 37 ] . More importantly, a recent study showed that FOXC1 is highly expressed in human dermal papilla cells (DPCs) at the mRNA and protein level of patients affected with AGA [ 38 ] , but no previous reports identified FOXC1 mutations with AGA. Our group previously showed a familial case of Axenfeld-Rieger Syndrome (ARS) where degenerated hair follicle cells were observed, which was explained through FOXC1/NFATC1 genetic axis [ 37 ] . Another report showed similar digenic inheritance of a mutation in FOXC1 , but with PITX2 gene in ARS patient where the phenotypic severity increased due to this inheritance [ 39 ] . Our mutation is in the activation domain ( Figure 2B ), and GEO DataSets show that FOXC1 is highly activated in AGA patients ( Table 2 ). This led us to hypothesize that the mutation represented a partial gain of function. The involvement of both genes in AGA pathogenesis and in silico predictions support our hypothesis. Interestingly, a recent report identified SMARCD1 as one of the top 360 FOXC1 interactors, [ 40 ] and our in silico analysis supports this interaction ( Table 2 and Figure 3 ). Whether this interaction is crucial for hair follicle development remains to be explored.

A permissive environment is always a cordon for developing differential phenotypes within a familial case. This is the case of our second family presented with two kinds of pigmentation disorders: albinism and vitiligo, where epidermal/hair follicle melanocyte disruption is a common pathway between both disorders [ 41 , 42 ] .

Since the hallmark of both diseases is melanocyte disruption, we looked first for the genes that encode melanocyte proteins associated with vitiligo such as PTPN22, TYR , and MC1R [ 43 ] . We then looked for specific genes associated with OCA. We did not find any pathogenic variant neither in any of the genes that encode components of the four protein complexes: adapter protein 3 (AP-3) and Biogenesis of Lysosome-related Organelles Complex 1, 2, and 3 (BLOC-1, BLOC-2, and BLOC-3) that are associate with the monogenic type, nor in the genes involved in the digenic mode of inheritance like TYR and OCA2 or SLC24A5 and OCA2 [ 44 – 47 ] . The only pathogenic mutation we uncovered is in the HPS1 gene which lead to a premature stop codon and a truncated protein HPS1 mutations which have been associated with OCA in several reports. HPS1 (c.9C > A) mutation counted for 5% in nonconsanguineous Chinese patients, two frameshift variants in HPS1 (c.9delC and c.1477delA) in children with OCA, and two other gene mutations in Puerto Ricans, a 16-base pair (bp) duplication in HPS1 which accounted for 42.8% of OCA cases [ 48 , 49 ] . The homozygous HPS1 mutation accounts for the OCA phenotype in our family, but the heterozygous form alone does not explain the vitiligo phenotype. We hypothesized that HPS1 could be a novel candidate gene in vitiligo since it is implicated in melanocyte activation, proliferation, and differentiation. HPS1 is a BLOC-3 subunit which support the intracellular biogenesis and trafficking of lysosome and lysosome-related organelles (LROS) such as melanosomes, platelet dense bodies (also called delta granules), lamellar bodies of type II pneumocytes, and granule proteins of cytotoxic and suppressor T cells and natural killer (NK) cells. As such, mutations in HPS1 leads to defective LROS trafficking and assembly [ 47 , 50 ] . By that, the proteins are mistargeted i.e., the proteins that are trafficked to plasma membrane are mistakenly targeted to intracellular organelles like lysosome. This disrupts normal cell-cell interactions and chemotactic detection of migrating melanocytes, affecting cell survival and proliferation. Consequently, it results in decreased melanocyte numbers in the epidermis and dermis and delayed melanocyte protein function. [ 51 – 53 ] . The disruption of the melanocytes occur due to factors like oxidative stress from Reactive Oxygen Species (ROS), innate immunity, or adaptive immunity and gene responsible for that are XRP1, IFIH1, NLRP1, PTPRC , RERE , CTLA4, FOXP1, LPP, TSLP, IL12RA, GZMB , [ 54 – 56 ] however none of these genes were altered in our case. Interestingly, we found a novel missense variant in ITK p. Pro521Leu in our indexed patient (homozygous) which was absent in the uncle ( Figure 4E ).

ITK (IL2 inducible T-cell kinase) belongs to TEC family kinase and is highly expressed in T cells and involved in T-cell receptor (TCR) signaling, cytokine release and differentiation regulation [ 57 , 58 ] . Mutations in ITK have been associated with benign inflammatory dermatoses that mimics cancer, eczema, lymphoproliferative disorder [ 59 – 62 ] , and most importantly, a recent report have identified the same variant p.Pro521Leu in a patient with lymphoproliferative disorder where vitiligo was part of the phenotype [ 63 ] . Our IPA analysis revealed that ITK is inhibited in vitiligo patients ( Table 2 ), suggesting that our mutation acts as a partial loss of function. From melanocyte disruption by immunity involvement, we propose a di-genic inheritance pattern between HPS1 and ITK genes in vitiligo and found that ITK and HPS1 interacts with each other through TNF ( Figures 5 and 7 ).

We thus propose that a partial disruption of the TNF inhibition pathway imposed by the missense mutation in ITK combined with a partial gain of function through the truncated HPS1 protein will lead to an exaggerated increase in TNF and other cytokines such as IL-2, IL-6, IL-17 which are frequently observed in patients with vitiligo [ 64 ] . We cannot exclude a direct interaction between the proteins that would potentially be potentiated by the loss of the C-terminal domain of HPS1 . Despite the small sample size, this study validated the technologies and methodologies, identified challenges with software and sequencing, and established a reference center to serve the underserved MENA region. Moreover, the samples were well preserved despite the region’s fluctuating temperatures, demonstrating the robustness of the protocols and paving the way for future large-scale studies.