X-chromosome inactivation patterns in females with Fabry disease examined by both ultra-deep RNA sequencing and methylation-dependent assay

Fabry disease is an X-linked inherited lysosomal storage disorder caused by mutations in the gene encoding α-galactosidase A. Males are usually severely affected, while females have a wide range of disease severity. This variability has been assumed to be derived from organ-dependent skewed X-chromosome inactivation (XCI) patterns in each female patient. Previous studies examined this correlation using the classical methylation-dependent method; however, conflicting results were obtained. This study was established to ascertain the existence of skewed XCI in nine females with heterozygous pathogenic variants in the GLA gene and its relationship to the phenotypes. We present five female patients from one family and four individual female patients with Fabry disease. In all cases, heterozygous pathogenic variants in the GLA gene were detected. The X-chromosome inactivation patterns in peripheral blood leukocytes and cells of urine sediment were determined by both classical methylation-dependent HUMARA assay and ultra-deep RNA sequencing. Fabry Stabilization Index was used to determine the clinical severity. Skewed XCI resulting in predominant inactivation of the normal allele was observed only in one individual case with low ⍺-galactosidase A activity. In the remaining cases, no skewing was observed, even in the case with the highest total severity score (99.2%). We conclude that skewed XCI could not explain the severity of female Fabry disease and is not the main factor in the onset of various clinical symptoms in females with Fabry disease.


Introduction
Fabry disease (OMIM #301500), also known as Anderson-Fabry disease or angiokeratoma corporis diffusum, is a rare lysosomal disorder with multi-system involvement due to deficiency of ⍺-galactosidase A, which is required for the degradation of globotriosyl-ceramide (Gb3) [1,2]. This enzyme is encoded by the GLA gene, located on the X-chromosome. More than 400 private mutations have been defined in patients globally [3]. The incidence of Fabry disease was reported to be approximately 1:117,000 based on live male births, but when considering heterozygotes, the incidence may be as high as 1:58,000 [4]. The most recent studies have reported higher frequencies of 1:7000 and 1:3000 in neonatal mass screening [5][6][7].

3
The accumulation of undegraded substrates, Gb3 and deacylated globotriaosylsphingosine (lyso-Gb3), results in the earliest symptoms observed in patients with the classic form of this disorder, in childhood and adolescence (earlier in males than in females). These include neuropathic pain, vascular skin lesions, sweating abnormalities, characteristic ocular changes, gastrointestinal problems, temperature intolerance, or proteinuria [1]. These common signs and symptoms are noted in both male patients and female carriers; however, their progression and severity may be attenuated in females, so they may be presented a decade or more later in females than in males. The average life span was also reported to be reduced by 22% in males and 5.75% in females, with cardiovascular disease being the most common cause of death in both sexes [8,9].
In the past, Fabry disease was considered to be transmitted as an X-linked recessive trait. When it was found that heterozygous females may be affected in the same manner as hemizygous males, it was postulated that the disease could be classified as X-linked dominant [10] or X-linked semi-dominant [11]; however, at present, it is usually simply described as X-linked [12].
The majority of X-linked diseases produce symptomatic disease only in males [13]. However, by a molecular mechanism known as skewed X-chromosome inactivation (XCI), wherein random transcriptional silencing of one X-chromosome occurs in each female cell, females exhibit mosaic expression of X-linked genes. Although there are numerous biological events other than skewed XCI that may influence the penetrance and expression in heterozygous females, skewed X-inactivation cannot be disregarded in terms of it possibly playing a major role in phenotypic expression [11].
In cases of Fabry disease, conflicting results have been reported regarding the correlation between X-inactivation profiles and disease severity in female heterozygotes. One study found that X-inactivation was a major factor in determining the clinical severity of female heterozygotes [14,15]; whereas, another study showed no correlation between these variables [16]. These studies examined X-inactivation by using human androgen receptor (HUMARA ) assay, a polymerase chain reactionbased XCI assay using a methylation-sensitive restriction enzyme [17]. Recently, we developed a novel method for determining X-inactivation patterns by ultra-deep RNA sequencing [18]. The purpose of this study was to ascertain whether skewed X-inactivation favoring the mutant α-galactosidase A allele occurs in our cohort of female heterozygotes with Fabry disease. To clarify this, we applied both the conventional HUMARA assay and ultradeep RNA sequencing using peripheral blood leukocytes and urine sediments.

Genomic DNA and total RNA isolation from leukocytes and urine sediment
Genomic DNA was extracted from peripheral blood leukocytes and urine sediment of patients using the Quick Gene Mini 80 system (Wako Pure Chemical Industries, Ltd., Tokyo, Japan) and Quick-DNA Urine kit (Zymo Research Corporation, Irvine, CA, USA). A Ribopure Blood Kit (Invitrogen, Carlsbad, CA, USA) and RNA stabilization agent (RNAlater; Invitrogen) were used for total RNA extraction from blood leukocytes, and the ZR Urine RNA Isolation Kit (Zymo Research Corporation) was used for urine sediment. The obtained genomic DNA was used for targeted sequencing, Sanger sequencing, and HUMARA assay. Ultra-deep targeted RNA sequencing was performed using the total RNA.

Calculation of the clinical severity score by The FAbry STabilization indEX (FASTEX)
Physical and laboratory examinations were used to assess patients' clinical severity by FASTEX, according to a four point severity scale in three domains: nervous system, renal, and cardiac domains. In short, nervous events included pain and other neural events; renal events included the albumin-creatinine ratio (ACR)/protein-creatinine ratio (PCR) and eGFR; and cardiac events included left ventricular hypertrophy (LVH), echocardiography parameters/arrhythmia, and cardiac functioning categorized by the New York Heart Association (NYHA). Overall severity of the disease was calculated as the sum of each severity domain corrected by their interaction (combination of raw and weighted scores), reported as a percentage, with maximum severity of 100% [19].

Detection of pathogenic variants of GLA
Comprehensive analysis was conducted by next-generation sequencing (NGS) using MiSeq (Illumina, San Diego, CA, USA) for all patients. The sample library for NGS analysis was prepared using HaloPlex Target Enrichment Kit 500 kb for Illumina (Agilent Technologies, Santa Clara, CA, USA), in accordance with the manufacturer's workflow. Briefly, 225 ng of genomic DNA was used for a restricted reaction and hybridized at 54 °C for 16 h with NGS probes. All indexed DNA samples were amplified by polymerase chain reaction and sequenced using the MiSeq platform. The results were analyzed using SureCall 3.0 (Agilent Technologies).

Sanger sequencing
The results of mutational analyses of GLA obtained by NGS were confirmed using Sanger sequencing. Mutational sites of GLA in each patient were amplified by PCR using the primer set for GLA according to the mutation site. After 40 cycles of amplification, PCR products were separated on 1.5% agarose gel and subjected to direct sequencing using dye terminator cycle sequencing kit (Amersham Biosciences, Piscataway, NJ, USA) and an automatic DNA sequencer (ABI Prism 3130; PerkinElmer, Applied Biosystems, Foster City, CA, USA). For variant description, NM_000169 was used as a reference sequence.

HUMARA assay
The HUMARA assay was performed as described by Allen et al. [20]. Briefly, 200 ng of genomic DNA from blood leukocytes and urine sediment was digested by a methylationsensitive enzyme (Hpa2) and the same volume of undigested DNA was amplified with FAM-labeled forward primer and reverse primer specific for regions either side of polymorphic CAG repeats (forward primer: 5'-TCC AGA ATC TGT TCC AGA GCG TGC -3'; reverse primer: 5'-GCT GTG AAG GTT GCT GTT CCT CAT -3'). The PCR product was mixed with internal size standard (GeneScan 500 LIZ Dye Size Standard; PerkinElmer, Applied Biosystems). Quantification of data and their visualization in graphs were performed using GeneScan software. The XCI pattern was defined as random (anything less than 50:50-80:20), skewed (from 80:20 to 90:10), or extremely skewed (more than 90:10).

Ultra-deep targeted RNA sequencing using NGS
We recently developed a novel assay of ultra-deep RNA sequencing for examining XCI patterns [18] that uses the quantification of transcript expression for wild-type and variant alleles. Reverse transcription of total RNA to cDNA was performed using Ecodry Premix (Double Primed; Clontech Laboratories Inc., Mountain View, CA, USA). cDNA was amplified by PCR using the following pair of glyceraldehyde 3-phosphate dehydrogenase (GAPDH)-specific primers, which were designed to amplify both genomic DNA and cDNA to confirm genomic DNA contamination (forward primer: 5'-CCC TTC ATT GAC CCT CAA C-3', reverse primer: 5'-TTC ACA CCC ATG ACG AAC -3'). Nested RT-PCR was then performed using forward and reverse primers designed according to the mutation site (Supplementary Table 1). To create an ultra-deep read in the sequence, we amplified a short fragment in 150 bp of PCR product from the cDNA. The product was then purified on 1.5% agarose gel; the ends of the cDNA fragments were repaired and adenyl nucleotides were added, followed by ligation of the adapters using TruSeq Nano DNA Library Prep for Illumina (Santa Clara, CA, USA) for Cases 1-8, and SureSelect NGS Target Enrichment Workflow (Agilent Technologies) for Cases 9 and 10. The results were analyzed using SureCall 3.0 (Agilent Technologies). We analyzed the ratio of variant to wild-type alleles to estimate XCI.

Patient history
We present five female patients from one family (Cases 2-6, Supplementary Fig. 1) and four individual female patients (Cases 1 and 7-9) with Fabry disease ( Table 1). The average age at onset was 37.5 years old. All cases have a family history of Fabry disease. The manifestations of Fabry disease, including cardiac, renal, and cerebrovascular symptoms, varied among the patients, even within the same family. A very low level of α-galactosidase A was found in one individual (Case 1).

Calculation of the clinical severity score by FASTEX
On the basis of the clinical characteristics, the FASTEX combination score was summed from each domain. This score ranged from 35.4% (Case 7) to 99.2% (Case 3) ( Table 1, Supplementary Table 2).

Detection of pathogenic variants of GLA
Targeted sequencing by NGS and Sanger sequencing revealed that all cases were heterozygous for likely pathogenic variants of GLA (Case 1-Case 8) and pathogenic variant for Case 9. (Table 1).

HUMARA assay analysis
HUMARA assay was performed using genomic DNA derived from blood leukocytes and urine sediment ( Table 2). Cases 1 and 8 had homozygous CAG repeats, for which examination of XCI patterns could not be performed. For the HUMARA assay, we determined the XCI pattern in each allele divided by molecular size after running capillary electrophoresis. Therefore, when the CAG repeat number was the same in each allele, we could not examine the pattern. For Case 7, we failed to extract genomic DNA from the urine. For Cases 8 and 9, we could not obtain fresh urine samples because these individuals lived far from our hospital. No skewed pattern of Corneal opacity White matter lesion XCI was revealed in this assay, except for in Case 1, in which low α-galactosidase A activity was also exhibited.

Ultra-deep targeted RNA sequencing
We conducted targeted RNA sequencing for all cases. The depth of sequencing in the patients ranged from 102,834 to 694,692 and from 2032 to 3323 for Haloplex and SureSelect NGS Target Enrichment Workflow in the NGS analysis, respectively, representing ultra-deep sequencing. A skewed pattern of XCI involving predominant inactivation of the wild-type allele was revealed in only one individual case (Case 1), in which a very low α-galactosidase A level was exhibited ( Table 2).

Discussion
Recently, we have developed a novel method of analyzing XCI using ultra-deep RNA sequencing [18]. Using this assay, we directly measured the ratio of mutated to wild-type alleles in the GLA gene at the transcript level, which functionally and significantly reflects the influence of XCI on clinical manifestations [18]. We also carried out HUMARA assay to evaluate XCI in the HUMARA gene. The two sets of results were comparable. Conventionally, females heterozygous for a variant causative of an X-linked disease are expected to be virtually asymptomatic because of the input of the non-mutated allele. In Fabry disease, we know that the vast majority of female heterozygotes will suffer from the classical signs and symptoms of Fabry disease, albeit with delayed onset compared with that in male hemizygous patients [11]. Non-random XCI was believed to potentially underlie this phenomenon [11]. Female patients with Fabry disease may display marked variability in disease onset, severity, and progression [21], which was also found in this study. Our patients suffered from a wide range of Fabry-related manifestations.
Despite the small number of patients in our study, the ranges of ages upon onset of the first Fabry symptoms and diagnosis were nearly consistent with those in a larger study encompassing patients enrolled in the Fabry Registry (32 vs. 40 years) [22,23]. Most of the variants were categorized as type 1 classic according to the International Fabry Disease Genotype-Phenotype Database (dbFGP) (www. dbfgp. org).
We determined the clinical severities of patients using a smaller number of variables, based on a faster, objective, and more practical method by calculating the disease severity with FASTEX. To date, two severity scoring systems, the Mainz Severity Score Index (MSSI) and Fabry Disease Severity Scoring System (DS3), have been validated to measure the disease burden of Fabry disease. However, both of these evaluations require the assessment of several  1 3 domains, which include a large number of items (5 domains and 12 items for DS3 and 4 domains and 24 items for MSSI), resulting in difficulties in applying the indices and making the process time-consuming. Additionally, some of the items are largely based on patients' self-reported evaluation of symptoms. One study has proven that, even when using only a small number of variables, the score in FASTEX was significantly correlated with MSSI and DS3, showing that this approach can achieve the same accuracy as these other tools [19].
The development of severe clinical symptoms in heterozygous Fabry disease and some other X-linked diseases has been proposed to occur via a mechanism involving XCI. By studying the methylation status of the polymorphic (CAG)n repeat region located within exon 1 of the HUMARA gene, skewed XCI patterns were determined in 29% of cases [24]. In the present study, all patients except Case 1 showed random X-chromosome inactivation in the range of 45:55-78:22, irrespective of the disease severity. Our findings are comparable to those of the study by Maier et al. [16], who found that only 18% of female cases showed highly skewed XCI ratios, and to those of other studies with slightly higher ratios [15].
From these results, in the presented analysis, the clinical involvement of Fabry disease in females did not correlate with the XCI profiles. Others previously showed that, in most cases (27/36, 75%), the methylation of the non-mutated allele directly correlated with the FASTEX severity score [14]. In our cohort, a patient with skewed XCI (Case 1) had moderate symptoms. Case 3, showing the highest severity score by FASTEX (99.2%), had random XCI. This is consistent with the results presented by Maier et al. [16] and Elstein et al. [11]. However, most recent research has shown that, in females with skewed XCI, Fabry disease progression correlated with the direction of skewing [15,24]. It has been reported that the measurement of α-galactosidase A enzyme activity cannot be used to predict the clinical phenotype of heterozygous females [14]. However, it was previously reported that there was a clear correlation between XCI and the accumulation of lyso-Gb3. There have been some successful attempts to use such accumulation as a prognostic marker to elucidate disease severity and progression in males [25]. Such accumulation is considered a potential biomarker that may improve the initial diagnosis of clinically relevant Fabry disease, particularly in females. Nonetheless, lyso-Gb3 levels tend not to consistently correlate with disease burden and progression in female patients [21]; especially in heterozygotes with later symptom onset, the lyso-Gb3 levels may be very low or normal [25].
As Fabry disease is a multiorgan disorder with severe renal, cardiac, and cerebrovascular involvement, studying tissue-specific XCI patterns in affected organs would give the best perspective on the contribution of XCI to the clinical variability in female patients. However, such analysis would require invasive biopsies [26]. As all patients have renal symptoms (proteinuria and low eGFR), we carried out assays with genomic DNA extracted from urinary cells as it is derived from cellular (podocytes, proximal tubule epithelial cells, undifferentiated kidney cells) [27] and/or cell-free DNA originating from cells sloughing into urine from the genitourinary tract [28,29]. Because tissue affected by Fabry disease may be difficult to access (heart and kidney) or inaccessible (brain), we also selected another easily accessible sample (peripheral blood leukocytes) to reflect the XCI pattern. Using urinary cells and peripheral blood leukocyte, we can detect intertissue variation. In one study, XCI patterns were similar among the tissues [30]. In contrast, previous studies showed that XCI patterns vary widely across different tissues [31][32][33], making it difficult to extrapolate XCI status solely from the blood. Thus far, only one previous study examined the intraindividual differences in XCI patterns in females heterozygous for a variant causative of Fabry disease. The results did not show any significant variations between tissues [15]. Concordance with the data obtained in this study suggests that the patterns of XCI in blood leukocytes and urinary cells were similar. It is recommended to study XCI in at least one other tissue type in addition to leukocytes, preferentially one obtained noninvasively [24].
Given our limited number of patients and the finding of only one patient with skewed XCI, our results do not enable the prediction of female Fabry cases. Our female Fabry cases presented a wide range of clinical severity. As our study limitation, we did not carry out functional study for each variant. The degree of protein dysfunction itself can affect the clinical severity in female cases [24]. We did not find another benign heterozygous variant in the GLA (Electronic supplementary materials 1-7). Further investigations are needed to elucidate the reason for the variable clinical expression of Fabry disease in females.
We conclude that skewed XCI could not explain the severity of female Fabry disease and is not the main factor determining the onset of various clinical symptoms in females with Fabry disease.
Ethical approval All procedures performed in this study were reviewed and approved by the Institutional Review Board of Kobe University Graduate School of Medicine (IRB approval number 019-301) and performed in accordance with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. Written informed consent for conducting this study was obtained from all participants.