Frequency of Truncating FLCN Variants and Birt-Hogg-Dubé-Associated Phenotypes in a Healthcare System Population

Abstract

Purpose:

Penetrance estimates of Birt-Hogg-Dubé (BHD)-associated cutaneous, pulmonary, and kidney manifestations are based on clinically ascertained families. In a healthcare system population, we used a genetics-first approach to estimate the prevalence of pathogenic/likely pathogenic (P/LP) truncating variants in FLCN, which cause BHD, and the penetrance of BHD-related phenotypes.

Methods:

Exomes from 135,990 patient-participants in Geisinger’s MyCode® cohort were assessed for P/LP truncating FLCN variants. BHD-related phenotypes were evaluated from electronic health records. Association between P/LP FLCN variants and BHD-related phenotypes was assessed by Firth’s logistic regression.

Results:

P/LP truncating FLCN variants were identified in 35 individuals (1 in 3,234 unrelated individuals), 68.6% of whom had BHD-related phenotype(s) including: cystic lung disease (65.7%), pneumothoraces (17.1%), cutaneous manifestations (8.6%), and kidney cancer (2.9%). Four (11.4%) had prior clinical BHD diagnoses.

Conclusion:

In this healthcare population, the frequency of P/LP truncating FLCN variants is 60 times higher than the previously reported prevalence. Although most variant-positive individuals had BHD-related phenotypes, a minority were previously clinically diagnosed, likely because cutaneous manifestations, pneumothoraces, and kidney cancer were observed at lower frequencies than in clinical cohorts. Improved clinical recognition of cystic lung disease and education concerning its association with FLCN variants could prompt evaluation for BHD.

Graphical Abstract

Introduction

Birt-Hogg-Dubé syndrome (BHD; OMIM 135150) is an autosomal dominant condition associated with benign cutaneous manifestations, pulmonary cysts/blebs/bullae (hereafter referred to as “cystic lung disease”), spontaneous pneumothoraces, and kidney neoplasms.^1,2 BHD is caused by germline variants in the folliculin (FLCN) gene, a tumor suppressor on chromosome 17p11.2 (OMIM 607273), that encodes the folliculin protein that is involved in numerous signaling pathways.^3–6 Variants resulting in truncation of the protein (e.g., nonsense, frameshift, canonical splice site variants) are most common in individuals with BHD.²

BHD has an estimated prevalence of 1:200,000⁷, and, as of 2017, over 600 families with BHD had been described.⁸ The most common phenotypes in individuals with BHD are benign cutaneous manifestations, including fibrofolliculomas and trichodiscomas. Such findings have been identified in 58–90% of clinically ascertained patients and typically appear after age 25.⁹ Macroscopically, trichodiscomas and fibrofolliculomas are indistinguishable from one another and are believed to be on the same morphological spectrum as perifollicular fibromas.¹⁰

Patients with BHD also have increased risk of developing cystic lung disease and spontaneous pneumothoraces. Pulmonary cysts occur in 67–90% of patients with BHD, vary in number, and are typically bilateral and located in the lower basal zones^2,9,11 Between 24–38% of patients with BHD experience a spontaneous pneumothorax at a median age of 38 years and earliest age of seven.^2,11

Additionally, patients with P/LP FLCN variants are at an increased risk for benign and malignant kidney findings. Bilateral, multifocal renal masses are typical and are histologically diverse with most being hybrid oncocytic masses, chromophobe renal cell carcinoma, and clear cell carcinoma.² It is estimated that 12 to 34% of individuals with BHD develop kidney cancers^2,9,11 with a mean age of diagnosis of 50.7 years and earliest diagnosis at age 20.⁹ Additionally, individuals with BHD are at risk of developing oncocytosis, microscopic foci of oncocytic cells in the renal parenchyma, that is a precursor to kidney cancer.² Other phenotypes have been reported in patients with BHD including lipomas, parathyroid adenomas, thyroid nodules, thyroid cancer, and parotid oncocytomas, but additional data are needed to clarify their association with BHD.²

To date, penetrance estimates of BHD-related features in individuals with P/LP FLCN variants have been based on families clinically ascertained due to a personal and/or family history. As genetic testing becomes widely integrated into clinical care, understanding variant prevalence and phenotypic presentation in broader populations will be necessary to enable accurate counseling and appropriate management. Population-based testing for pathogenic variants in other disease-associated genes has suggested that clinical penetrance estimates might be higher than penetrance in variant-positive individuals from broader cohorts due to ascertainment bias in clinical populations.^12,13 Here, we explore the prevalence of P/LP truncating FLCN variants in a healthcare system-based population and utilize electronic health record (EHR) data to investigate the frequency of BHD-related phenotypes and prior clinical identification of the FLCN variant.

Materials and Methods

MyCode Participants

The Geisinger MyCode® Community Health Initiative (MyCode) serves as a biobank of blood and other samples from over 300,000 patient-participants who consent to health-related research.^14–16 Participants are recruited throughout Geisinger regardless of disease or phenotype.^14–17 The population for this study includes a subset of MyCode patient-participants, referred to as the DiscovEHR cohort¹⁵, with available exome sequencing and linked EHR data. Most patient-participants also have available SNP genotyping data. MyCode participants have higher rates of self-reported White race and non-Hispanic ethnicity, older median age, and higher comorbidity index compared to the overall Geisinger population.¹⁷ MyCode and the research outlined are governed by the Geisinger Institutional Review Board; all participants provided written informed consent.

Exome Sequencing and Intragenic CNV Calling

Exome sequencing for 135,990 adult MyCode patient-participants was performed in collaboration with the Regeneron Genetics Center, as described previously.^14,15 Exome sequence data were aligned to human reference GRCh38 with BWA-MEM, and the resultant BAM files were processed using the Picard MarkDuplicates tool (http://broadinstitute.github.io/picard) to flag duplicate reads. After exome data were aligned, variants were called using weCall.¹⁸ Joint genotyping was performed across the cohort using GLnexus (Lin et al., 2018, unpublished manuscript, https://www.biorxiv.org/content/10.1101/343970v1.full.pdf) . CNV calling from exome data was carried out using Copy number estimation using the Lattice-Aligned Mixture ModelS (CLAMMS) algorithm¹⁹ that demonstrates high sensitivity and specificity down to single-exon level CNVs (Maxwell et al., 2017, unpublished manuscript, https://www.biorxiv.org/content/biorxiv/early/2017/03/22/119461.full.pdf). CNV calls were validated with SNP-array CNV calls using Penn CNV.²⁰

Identification of P/LP Truncating FLCN Variants

Exome sequencing data (n=135,990) were reviewed for truncating FLCN variants. Variants in VCF files were annotated with Ensembl Variant Effect Predictor (VEP) version 96 using RefSeq transcripts. High impact variants (nonsense, frameshift, +/− 1,2 splice site) based on the FLCN transcript most expressed in adult tissues (transcript variant 1, GenBank: NM_144997.7) in genome build GRCh38.⁵ Sequence variants with an allele balance greater or equal to 0.20, genotype quality greater than or equal to 80, coverage greater than or equal to 20, and maximum gnomAD subpopulation allele frequency of less than 0.001 were identified. Two FLCN variants that nearly met our quality thresholds were included in our analysis upon confirmation of the FLCN variant from a clinical genetic testing report documented in the EHR (n=1) and validation of a variant by evaluation of IGV plots (n=1). Truncating variants were then reviewed (JMS and HS) using the American College of Medical and Genomics and Association for Molecular Pathology variant interpretation guidelines and classifications were confirmed by a third reviewer (NTS), who is board certified in laboratory genetics and genomics (Supplemental Table 2).^21–23 Truncating variants with P/LP classifications were included in the analysis.

While this study focuses the most common BHD-causative variants (truncating FLCN variants), a secondary analysis of non-truncating variants reported in the National Center for Biotechnology Information ClinVar²⁴ database as P/LP was conducted (Data not incorporated into further analyses, Supplementary Materials and Methods, Supplemental Tables 1, 2, 6).

Determination of Relatedness

To report on the number of unique families with P/LP truncating FLCN variants, the frequency of phenotypes within families, and P/LP truncating variant prevalence in an unrelated subset of the MyCode cohort, first- and second-degree familial relationships were identified from genotype data using Pedigree Reconstruction and Identification of the Maximally Unrelated Set (PRIMUS).²⁵

Phenotype Evaluation – EHR and Imaging Review

BHD-related pulmonary, cutaneous, and kidney phenotypes were characterized in variant-positive individuals via manual review of patients’ EHRs. A chart review guide was developed to collect variables of interest and independent, double reviews were completed by a genetic counselor and board-certified neurodevelopmental pediatrician (JMS, SMM) between March and November 2020 (Supplemental Materials and Methods). Discrepancies were resolved through joint review and consensus.

Cystic lung disease and renal masses might be incidentally identified on imaging studies of adjacent body parts (e.g., Computed Tomography (CT) of the abdomen includes the lung bases and lumbar spine Magnetic Resonance Imaging (MRI) includes part of the kidneys). Additionally, cystic lung disease may be identified, but sometimes not mentioned, in radiology reports or elsewhere in the EHR. Because of this, dedicated review of imaging studies including all or part of the kidneys and/or lungs was completed by a board-certified radiologist (AMDL) for all patients with available imaging (Supplemental Materials and Methods).

Phenotype Evaluation – ICD-9/10 Diagnostic Codes

While chart review provides more accurate phenotyping, it is low throughput and not feasible for large sample sizes. As such, clinical manifestations of BHD-related phenotypes were also evaluated in variant-positive and negative patient-participants by extracting relevant International Statistical Classification of Diseases and Related Health Problems 9 and 10 codes (ICD-9/10) from the EHR in February 2021. Selected ICD-9/10 diagnostic codes were defined prior to manual chart review to capture BHD-related pulmonary, cutaneous, and kidney phenotypes (Supplemental Table 3). While ICD-9/10 codes can provide an estimate of BHD-related phenotypes, some codes that correspond to BHD-features are non-specific and could be used for other, unrelated clinical findings (e.g., ICD-10 codes for trichodiscoma and fibrofolliculoma are broadly used for benign neoplasms of the skin), and others might not be applied even when such a feature is identified due to its indolent or benign nature (e.g., ICD-9/10 codes for pulmonary cyst). Since use of ICD-9/10 codes alone might not correctly reflect patient phenotypes,²⁶ we evaluated the ability of ICD-9/10 codes in the EHR to accurately identify BHD-related phenotypes by comparing diagnoses captured using diagnostic codes to those captured with manual review of EHR data in variant-positive individuals.

Statistical Analysis

Continuous variables are summarized as median and interquartile range and categorical variables are described as frequency and percentage. Two-tailed Fisher’s exact and Wilcoxon rank-sum tests were performed to test significant differences in demographics between variant-positive and variant-negative individuals. Associations between P/LP truncating FLCN variants and BHD phenotype were assessed by comparing the frequency of BHD-related phenotypes in variant-positive individuals with the variant-negative group using relevant ICD-9/10 diagnostic codes. The statistical significance of associations between P/LP truncating FLCN variants and BHD-related phenotypes was calculated using Firth logistic regression model adjusting for age, sex and the first four principal components of ancestry. To reduce the impact of population stratification on the results, the European subset, as determined by genetic ancestry, (121,876) was used for the association analysis. The European subset used in the association analysis was identified by principal component analysis and clustering with the 1000 Genomes reference population using the same protocol applied to the UK Biobank.²⁷ Bonferroni correction was used to account for multiple testing (α = 0.05). All analyses were conducted using R version 4.0.1.

Results

Prevalence of P/LP Truncating FLCN Variants

Of the 135,990 patient-participants, 35 (0.03%, 1 in 3,885) had a P/LP truncating variant in FLCN (1 patient-participant not included in analyses had a non-truncating P/LP variant - Supplemental Tables 1, 6). Based on genetic relatedness data,²⁵ we determined that these 35 individuals are from 28 families, representing 0.03%(1 in 3,234) of the 90,563 first and second-degree unrelated subset of the cohort. Thirteen unique truncating FLCN variants were identified, including five frameshift variants in 17 individuals, five nonsense variants in six patient-participants, two canonical splice site variants in 12 individuals, and a single intragenic deletion of exons 4–11 in one participant (Figure 1). The individual with the intragenic deletion also had a frameshift variant in cis. Nine of the thirteen variants identified in 32 of the 35 participants were previously reported in the literature and/or submitted to the ClinVar²⁴ database. Table 1 summarizes demographics of variant-positive and variant-negative individuals in the MyCode cohort. Variant-positive participants were 62.9% female, 97.1% self-reported their race as White, 100% self-reported non-Hispanic ethnicity, and had a median age of 62 years. Collection of patient-reported categorical race and ethnicity data occurs as part of participants’ interaction with the healthcare system. Eighty-nine percent of variant-positive patient-participants were alive, and the median length of EHR data was 17.43 years at the time of manual chart review. No statistically significant demographic differences were identified between the variant-positive and variant-negative individuals.

Figure 1. Pathogenic/Likely Pathogenic truncating variants in FLCN identified in the MyCode cohort.

Based on build GRCh38 (NC_000017.11) and FLCN transcript most expressed in adult tissues (transcript variant 1, GenBank: NM_144997.7, 14 exons total with 11 coding exons, 579 amino acid residues) P=number of patient-participants with that variant and F=number of Families with that variant. ∓denotes variants that were previously reported in the literature and/or submitted to ClinVar. ^a denotes two variants identified in the same participant in cis.

Table 1.

Demographic Details in Truncating FLCN Variant-Positive Patient-Participants and the Overall MyCode Cohort

Demographics	P/LP Truncating FLCN Variant-Positive Patient-participants  (n=35)	Remaining MyCode Cohort  (n=135,955)	P-value
Median Age (IQR)	62 years (54–76)	61 years (45–72)	0.20
Sex  Female	22 (62.9%)	82,958 (61.0%)	0.86
Racea  American Indian or Alaska Native  Asian  Black or African American  Native Hawaiian or Pacific Islander  Other  Unknown  White	1 (2.9%) -- -- -- -- -- 34 (97.1%)	161 (0.1%) 487 (0.4%) 2,689 (2.0%) 231 (0.1%) 3 (0.0%) 325 (0.2%) 132,059 (97.1%)	-- -- -- -- -- -- 0.99
Ethnicity  Hispanic or Latino  Not Hispanic or Latino  Unknown	-- 35 (100%) --	2,978 (2.2%) 131,090 (96.4%) 1,887 (1.4%)	-- 0.99 --
Alive During Study Period	31 (88.6%)	125,455 (92.3%)	0.34
Smoking Status  Current Smoker	4 (11.4%)	23,094 (17.0%)	0.50
Median Length of EHR (Range)  Number with <2 years	17.43 years (0.05–22.61) 2 (5.7%)	14.68 years (0.01–22.58) 5,301 (3.9%)	-- 0.40

^a,Collection of race and ethnicity data occurs as part of participants’ interaction with the healthcare system. Race and ethnicity are participant-provided and categories were defined based on what is included in the EHR. “Other” is a category available within the EHR.

BHD-related Phenotypes – Chart Review of Patient-participants with Truncating FLCN Variants

The frequencies of major BHD-related phenotypes were evaluated by manual chart review of each variant-positive patient-participant. We found that 68.6% (n=24/35) of variant-positive individuals had EHR-documentation of a BHD-related phenotype (Figure 2 and Table 2).

Figure 2. Frequency of BHD-related phenotypes in the literature compared to that in MyCode patient-participants with truncating P/LP FLCN variants and in unique MyCode families.

^a Frequency of BHD-associated cutaneous manifestations is presented in all individuals with truncating P/LP FLCN variants. When restricting to those who were examined by a dermatologist and/or had a skin biopsy, 14.3% (n=3/21) had a BHD-related cutaneous finding. ^b An additional three participants had renal masses identified through manual EHR review. If we assume the three participants with a renal mass noted on imaging have an undiagnosed kidney cancer and only consider those with bilateral kidney imaging as being assessed for kidney cancer, an upper limit of potential kidney cancer in variant-positive individuals is 18.2% (n=4/22). Exome data was used to identify first-second degree

Table 2.

EHR Review for BHD-related Phenotypic Features in Truncating Variant-Positive Patient-Participants

Phenotypic Feature	Frequency in Participants with P/LP Truncating FLCN Variant (n=35)
BHD Diagnosis and/or Clinical Identification of FLCN Variant	4/35 (11.4%)
Any Personal History of BHD Cutaneous, Pulmonary, and/or Renal Phenotypes	24/35 (68.6%)
Any Cutaneous Phenotype Dermatologic Assessment (Dermatologist Exam and/or Skin Biopsy)  Fibrofolliculoma, Perifollicular Fibroma, Trichodiscoma	3/35 (8.6%) 21/35 (60.0%)  3/21(14.3%)
Any Pulmonary Phenotype  Basilar Cystic Lung Disease  Single Spontaneous Pneumothorax  Multiple Spontaneous Pneumothoraces Lung Imaging (Adequate Visualization and Any CT/MRI Imaging)  Basilar Cystic Lung Disease	23/35 (65.7%)  23/35 (65.7%)  2/35 (5.7%)  4/35 (11.4%) 13/35 (37.1%) and 29/35 (82.9%)  13/13 (100%) and 23/29 (79.3%)
Any Kidney Phenotype  Kidney Cancer  Renal Massa Renal Imaging (Adequate Visualization and Any Imaging)a  Renal Massa	4/35 (11.4%)  1/35 (2.9%)  3/34 (8.8%) 23/34 (67.6%) and 27/34 (79.4%)  3/23 (13.0%) and 3/27 (11.1%)
Documented Family History of BHD-related Phenotypic Feature  Family History of Renal Cancer  Family History of Pneumothorax  Family History of Fibrofolliculomas/Papules Documented Family History of Clinical BHD or FLCN Variant	7/35 (20.0%)  2/35 (5.7%)  3/35 (8.6%)  2/35 (5.7%) 2/35 (5.7%)

^aexcludes individual with renal cancer diagnosed

BHD-related cutaneous findings including fibrofolliculomas, perifollicular fibromas, and trichodiscomas were identified in three patient-participants (8.6%). Among all variant-positive participants, 60% (n=21/35) had documentation of an evaluation with a dermatologist and/or at least one skin biopsy. When restricting to those who were examined by a dermatologist and/or had a skin biopsy 14.3% (n=3/21) had a BHD-related cutaneous finding. BHD-related cutaneous findings were identified in 10.7% (n=3/28) of unique families with FLCN variants.

Pulmonary phenotypes, including basilar cystic lung disease and/or pneumothorax, were identified in 65.7% (n=23/35) of patient-participants. Of the 23 individuals with pulmonary phenotypes, six had at least one spontaneous pneumothorax; two had exactly one and four had more than one. Participants experienced their first pneumothorax between the ages of 21 and 79 years (median 27.5 years). The individual that experienced their first pneumothorax at age 79 had a diagnosis of chronic obstructive pulmonary disease. All 23 individuals with a pulmonary phenotype had basilar cystic lung disease. Eighty-three percent (n=29/35) of variant-positive patient-participants had CT or MRI imaging that included at least a portion of the lungs (e.g., abdomen CT includes lung bases) and 37.1% (n=13/35) had adequate visualization of the entire lungs. Among those with any lung imaging and those with adequate imaging reviewed, 79.3% (n=23/29) and 100% (n=13/13) had basilar cystic lung disease, respectively. Among unique families with FLCN variants, 21.4% (n=6/28) had pneumothorax and 67.9% (n=19/28) had cystic lung disease in at least one family member (Figure 2).

Kidney cancer was identified in one patient-participant (2.9%) with a chromophobe renal cancer diagnosed at age 71. Among the remaining 34 patient-participants without kidney cancer, 79.4% (n=27/34) had imaging that included the kidneys and 67.6% (n=23/34) had adequate visualization of both kidneys in their entirety. Renal masses were identified in 11.1 % (n=3/27) and 13.0% (n=3/23) of those who had any kidney imaging or bilateral imaging of the entire kidneys, respectively. These renal masses had not been pathologically confirmed to be kidney cancer or benign BHD-associated renal masses. In one case, follow-up imaging suggested the mass was likely a cyst while, in the other two cases, unrelated, benign renal cysts had not been ruled out.

In addition to core BHD-related phenotypes, the EHR review included evaluation of other phenotypes that have been reported in patients with BHD including thyroid cancer and parotid oncocytomas. Bilateral parotid oncocytosis was noted in a single participant with bilateral cystic lung disease. Thyroid cancer was not noted in any individuals.

Four individuals (11.4%) from three families had a diagnosis of BHD or documentation of a germline FLCN variant in their EHR. Two of those individuals had EHR documentation of multiple spontaneous pneumothoraces and bilateral blebs; one had a fibrofolliculoma, bilateral blebs, and renal mass; and one had a perifollicular fibroma.

BHD-related Phenotypes – Diagnostic Code Comparison

Comparison of BHD-related phenotypes captured by ICD codes with manual chart review (Supplemental Table 4) revealed that ICD-9/10 codes for BHD-related cutaneous features lack specificity and, as such, estimate the rates of such findings in variant-positive individuals at a higher rate than observed via manual review. We found that 42.9% (n=15/35) versus 8.6% (n=3/35) had BHD-related cutaneous features based on the ICD code approach and chart review respectively. ICD codes underestimated the frequency of cystic lung disease [20.0% (n=7/35) based on ICD codes versus 65.7% (n=23/35) chart and radiology review], likely due to the benign nature and incidental identification on imaging. Diagnostic codes for pneumothorax were found to accurately identify this phenotype in variant-positive individuals [17.1% (n=6/35) ICD codes versus 17.1% (n=6/35) chart and radiology review]. All six individuals with pneumothoraces identified on manual chart review had an appropriate ICD-9/10 code in their EHR. ICD diagnostic codes for kidney cancer appropriately identified the single individual with kidney cancer. Diagnostic codes associated with a renal mass underestimated this phenotype [2.9% (n=1/34) ICD codes versus 8.8% (n=3/34) chart review].

The associations between FLCN P/LP variants and ICD-9/10 based BHD-related phenotypes were assessed by Firth logistic regression in the unrelated, European subset of the cohort (Table 3). P/LP variants in FLCN were associated with a higher prevalence of pulmonary phenotypes (OR=4.16; 95% CI:11.65–9.43; p=0.026), and, more specifically, spontaneous pneumothorax (OR=15.63; 95% CI: 5.96–35.69; p=1.92×10⁻⁵) in variant-positive patient-participants. Association of P/LP truncating FLCN variants with other pulmonary, cutaneous, and kidney phenotypes were not statistically significant when correcting for multiple testing. We performed the association analysis in the full unrelated subset of the cohort including European and non-European minority populations. The results of the full unrelated cohort were consistent with those of the European subset (Supplemental Table 5).

Table 3.

Associations Between P/LP Truncating FLCN Variants and Diagnosis of BHD-related Phenotypes^a in the First- and Second-Degree Unrelated European Subset of the MyCode Cohort

Phenotype	Patient-Participants with P/LP Truncating FLCN Variant  N=27	Remaining MyCode Cohort  N=79496	Odds Ratio (95% CI)	p-valueb
Any Pulmonary Phenotype	7 (25.9%)	6,422 (8.1%)	4.16 (1.65–9.43)	0.026*
Spontaneous Pneumothorax	6 (22.2%)	1,523 (1.9%)	15.63 (5.96–35.69)	1.92×10−5*
Pulmonary Cyst	5 (18.5%)	5,245 (18.5%)	3.40 (1.17–8.32)	0.185
Cutaneous Phenotype	10 (37.0%)	16,243 (20.4%)	2.19 (0.97–4.70)	0.405
Any Kidney Phenotype	1 (3.7%)	2,200 (2.8%)	1.93 (0.21–7.43)	>0.99
Renal Mass	0 (0%)	1,693 (2.1%)	0.81 (0.01–5.81)	>0.99
Kidney Cancer	1 (3.7%)	852 (1.1%)	4.94 (0.55–19.37)	0.891

^aAbility of ICD-9/10 codes in the EHR to accurately identify BHD-related phenotypes varied between phenotypes.

^bBonferroni corrected for multiple testing (7-tests).

Discussion

Frequency of P/LP Truncating FLCN Variants

BHD is considered a rare disorder with an estimated prevalence of 1 in 200,000.⁷ BHD prevalence and penetrance estimates have been based on clinically ascertained families to date. This study examined exome sequencing data from a large healthcare population to estimate the prevalence of P/LP truncating FLCN variants and frequency of BHD-related phenotypes. Truncating FLCN variants were identified in 28 out of 90,563 first- and second-degree unrelated individuals (1 in 3,234 in unrelated cohort) suggesting that such variants are significantly more common than previously reported. This finding is consistent with the increased frequency of variants associated with other genetic syndromes identified in broader, population cohorts via genome-first approaches.^28–30

Prevalence of BHD-related Phenotypes

In participants with P/LP truncating FLCN variants from this healthcare cohort, BHD-related lung findings were the most common phenotype (65.7%, n=23/35). The frequency of basilar pulmonary cystic lung disease in those with CT and MRI lung imaging (79.3%, n=23/29) overlapped with the reported frequency in clinically ascertained individuals (67–90%)^2,9,11 and was even higher than the reported frequency in those with adequate visualization of the lungs (100%). The frequency of pneumothorax (17.1%, n=6/35) was lower than that reported in the literature (24–38%). Pneumothorax-associated ICD-9/10 codes were present in the EHR of all patient-participants with pneumothoraces identified by chart review, suggesting that these codes capture the phenotype with high sensitivity. When we assessed the association between P/LP truncating FLCN variants and pneumothorax, we identified a higher prevalence of spontaneous pneumothorax diagnosis (OR=15.63 95%CI (5.96–35.69), p=1.92×10⁻⁵) in variant-positive individuals. Together, these data suggest that while patient-participants with truncating FLCN variants in this healthcare-based cohort may have lower rates of pneumothorax compared to the clinically ascertained BHD population and higher rates compared to the variant-negative population. The prevalence of diagnosis codes associated with cystic lung disease was not statistically different between the variant-positive and variant-negative patient-participants. However, this could be due to inability of the selected ICD-9/10 codes to identify all individuals with the phenotype, as a minority of variant-positive participants with this phenotype based on manual chart review had the relevant ICD codes in the EHR. Therefore, this pulmonary finding, which can be a precursor to spontaneous pneumothorax, may also occur at higher rates in individuals with P/LP truncating FLCN variants. Additional research is needed to determine whether P/LP FLCN variants are associated with increased prevalence of cystic lung disease.

Other BHD-related phenotypes, including cutaneous and kidney manifestations appeared to be less common in the variant-positive individuals in this cohort compared with BHD probands in the reported literature. BHD-related dermatologic findings were identified in only 8.6% (n=3/35) of FLCN variant-positive individuals compared to 58–90% of patients with BHD in the literature.⁹ When restricting the analysis to patient-participants who were examined by a dermatologist and/or had a skin biopsy that would have likely identified these benign phenotypes, only 14.3% (n=3/21) had BHD-related skin findings. Since lack of clinical assessment does not alone account for differences in frequency, this finding is consistent with a lower penetrance of P/LP truncating FLCN variants than previously reported. P/LP truncating FLCN variants were not associated with an increased rate of cutaneous diagnosis, based on ICD-9/10 codes. Given the non-specific nature of cutaneous ICD-9/10 codes and their overestimation of BHD-related skin-findings in variant-positive individuals, however, additional assessment comparing the frequency of BHD-related cutaneous findings between variant-positive and variant-negative populations is needed.

Only one participant with a FLCN variant was identified to have kidney cancer (2.9%, n=1/35). This is a lower frequency than reported in the literature (12–34%).^2,9,11 Although some participants are younger (e.g., five variant-positive participants <40 years of age) and could go on to develop a kidney cancer in their lifetime, the majority of FLCN variant-positive individuals in the cohort are older than 50.7 years (74.3%, n=26/35) which is the median age of kidney cancer diagnosis reported in individuals with BHD. Due to the indolent nature of BHD-related kidney cancers,² some variant-positive patient-participants might have yet to come to clinical attention. If we assume the three participants with a renal mass noted on imaging have an undiagnosed kidney cancer and only consider those with bilateral renal imaging as being assessed for a kidney cancer, an upper limit of potential kidney cancer in variant-positive individuals is 18.2% (n=4/22), which overlaps with the frequency of kidney cancer reported in the literature. P/LP truncating FLCN variants were not associated with increased prevalence of kidney cancer in the MyCode cohort based on EHR data. Although additional studies are needed due to small numbers, this suggests the rates of kidney cancer among individuals with truncating FLCN variants may not be as high as previously estimated.

These data suggest that, among individuals in a healthcare-system population who volunteered to participate in genetic research but were not selected for any disease state, the majority of individuals with P/LP truncating FLCN variants have a phenotype consistent with BHD. Of all MyCode participants without considering relatedness, 1 in 5,663 had a P/LP truncating FLCN variant and an associated phenotype. Individual BHD-related phenotypes including pneumothorax, cutaneous findings, and kidney cancer may be less common among individuals with P/LP truncating FLCN variants than previously reported, a finding that is consistent with other studies examining frequency of disease-associated phenotypes in population-based cohorts.^12,13 Although we are not aware of any prior population-based studies of FLCN variants, the FLCN gene is located within the recurrent copy number variant region on 17p11.2 that is deleted in 90% of individuals with Smith-Magenis syndrome (SMS), hence these individuals are haploinsufficient for FLCN; however, there are few published reports of BHD-related phenotypes in individuals with SMS.^31-34Although SMS cohorts may be younger in age and not all individuals with SMS have undergone clinical evaluation for BHD, these data also suggest that published penetrance of BHD-related phenotypes may be overestimated.

Clinical Identification of Truncating FLCN Variants

Even though most participants with P/LP truncating FLCN variants had features consistent with BHD documented in the EHR, the majority of variant-positive patient-participants (88.6%, n=31/35) did not have a clinical diagnosis of BHD. The majority of those with BHD-related features had cystic lung disease (95.8%, n=23/24 with phenotype). Although cystic lung disease is included in the diagnostic criteria for BHD,^2,35 these findings did not prompt a referral to genetics or clinical genetic testing in any variant-positive patient-participants. The four individuals with clinically identified FLCN variants or BHD diagnoses all had cystic lung disease but were evaluated due to other phenotypes (fibrofolliculoma, perifollicular fibroma, pneumothoraces, family history of BHD). Although cystic lung disease was identified in 23 individuals in total, nine patient-participants were only identified on radiologist review of imaging data, meaning these findings were not documented in the EHR and were not commented on in the initial radiology report. Improved clinical recognition of these benign findings and education concerning their potential association with truncating FLCN variants could prompt additional evaluation for BHD. Identifying patients at risk for BHD-phenotypes could inform appropriate management. For example, in patients with BHD, a first pneumothorax might be treated more aggressively with pleurectomy/pleurodesis given the risk for recurrent pneumothoraces^36,37 and identification of individuals with FLCN variants could enable earlier screening for kidney cancer, potentially allowing detection of cancers at earlier stages.³⁸

Limitations

This study has several limitations. Because the MyCode population is primarily of European ancestry, from a single healthcare system participating in genetics research, it is unclear if these findings are generalizable to more diverse patient populations. Future work that combines data from multiple, more diverse cohorts are needed. Our estimated phenotype frequency relies on data within the EHR; as such, not all patient-participants have had assessments that would identify BHD-related features and imaging studies reviewed differ in their ability to detect lung and kidney phenotypes. Finally, this study focused on truncating variants since they are most common in individuals with BHD; additional variant types could be considered in future work.

Conclusions

P/LP truncating FLCN variants were identified in 1 in 3,234 individuals unrelated cohort in a healthcare population. Of those with such variants, 65.6% were identified to have BHD-related phenotypes (1 in 5,663 individuals with a truncating FLCN variant and an associated phenotype); however, only 11.4% had been clinically diagnosed with BHD. This limited ascertainment is likely because classic BHD-related phenotypes, such as cutaneous manifestations, pneumothorax, and kidney cancer, were observed at lower frequencies. Cystic lung disease was the most common phenotype in our cohort and could be used to prompt evaluation for BHD. This finding could inform future screening strategies for BHD in broad patient populations.

Data Availability

The data that support the findings of this study are available within the article and/or are available on request from the corresponding author, JMS.