Visual Overview
Abstract
Maintenance of the composition of inner ear fluid and regulation of electrolytes and acid-base homeostasis in the collecting duct system of the kidney require an overlapping set of membrane transport proteins regulated by the forkhead transcription factor FOXI1. In two unrelated consanguineous families, we identified three patients with novel homozygous missense mutations in FOXI1 (p.L146F and p.R213P) predicted to affect the highly conserved DNA binding domain. Patients presented with early-onset sensorineural deafness and distal renal tubular acidosis. In cultured cells, the mutations reduced the DNA binding affinity of FOXI1, which hence, failed to adequately activate genes crucial for normal inner ear function and acid-base regulation in the kidney. A substantial proportion of patients with a clinical diagnosis of inherited distal renal tubular acidosis has no identified causative mutations in currently known disease genes. Our data suggest that recessive mutations in FOXI1 can explain the disease in a subset of these patients.
An important homeostatic function of the kidney is the regulation of acid-base balance. This occurs in the proximal tubule via bicarbonate reabsorption and in the collecting duct system via proton secretion. The intercalated cells of the collecting duct use specific transporters to regulate acid-base balance that can secrete protons (type A intercalated cells) and bicarbonate (type B intercalated cells).1 Impaired type A intercalated cell function may lead to an inability to acidify the urine and secrete acids, with consequent development of hyperchloremic non-anion gap metabolic acidosis. Mutations known to cause inherited distal renal tubular acidosis (dRTA) have been identified in transporters present in the type A intercalated cells and include the B1 and a4 subunits of the vacuolar H+-ATPase (V-ATPase), anion exchanger 1 (AE1), and the cytosolic carbonic anhydrase.2–4 Interestingly, mutations in the V-ATPase B1 subunit are typically associated with sensorineural deafness as can be those in a4,2,5 emphasizing functional similarities between the epithelium of the inner ear and that of the collecting ducts of the kidney.6 This is due to the expression of a common set of membrane transport proteins in distinct cell types of the epithelium of the inner ear and the renal collecting duct. Specifically, Forkhead-related cells of endolymphatic duct and sac epithelium of inner ear7 and intercalated cells express B1 and a4 subunits of the V-ATPase proton pump as well as the anion exchange protein AE4 and pendrin.7,8
A genetic impairment of urine acidification can be associated with an altered pH/ionic composition in the inner ear fluid and hence, prevent proper sound perception. Another common feature of these cell types is that they each express FOXI1—a transcription factor that has been shown to be crucial for regulation of AE1, AE4, and the V-ATPase subunits B1, a4, A, and E2.8 Consistent with this, mice with a germline deletion of Foxi1 display sensorineural deafness and dRTA9,10 together with reduced expression of these target genes. Although a family has been identified in which Pendred syndrome (PS) segregates with a composite heterozygous genotype of single-allele mutations in SLC26A4 and FOXI1,11 biallelic recessive loss-of-function mutations in FOXI1 have not been described in human patients.
In this report, we describe two novel missense mutations that do not directly affect membrane transport proteins but rather, disrupt cell function at the transcriptional regulatory level—in that they prevent the transcription factor FOXI1 from binding to regulatory DNA cis-elements of target gene promoters. This leads to a severely reduced activation of those target genes, including membrane transport proteins, which require FOXI1 interactions for proper expression. In this way, these two novel loss-of-function mutations induce a severe syndrome of deafness and acidosis with a phenotype very similar to that of mice that altogether lack Foxi1 expression.
Results
Patient Reports
Two consanguineous kindred with an autosomal recessive pattern of dRTA and sensorineural deafness were evaluated (Figure 1). In family 1, the proband (patient 1.1), a boy, was the eldest of three children, with one affected sister (patient 1.2) (Figure 1A). In family 2, the proband (patient 2.1), a girl, was the only affected family member (Figure 1B). Key clinical features in all affected patients included early-onset sensorineural hearing loss requiring cochlear implants; a hypokalemic, hyperchloremic metabolic acidosis at presentation with inappropriately high urine pH; failure to thrive; and bilateral nephrocalcinosis (Figure 1E), consistent with dRTA. The deafness was profound and associated with dilation of the vestibular aqueduct (Figure 1F, Supplemental Figures 1–4). The dRTA was associated with hypercalciuria and nephrocalcinosis noted at presentation but responded to treatment with both alkali and potassium supplements, with subsequent catchup in growth of the children. In addition, patients 1.1 and 2.1 also developed medullary cystic changes within the kidney (Figure 1, E and F). Clinical data are summarized in Table 1.
Clinical, molecular and modeling data of FOXI1 families and their mutations. A and B show the family pedigrees and status with respect to the FOXI1 mutation genotype. Squares denote family members who are men/boys, and circles denote family members who are women/girls. Shaded symbols denote affected members with homozygous mutation, and semishaded symbols denote heterozygous carriers. Arrows denote the probands (patient 1.1 in family 1 and patient 2.1 in family 2). C and D show sequence chromatograms of the affected and parental genotype. Mutations are marked with arrowheads. E and F show renal ultrasound scans showing medullary nephrocalcinosis in all affected and cystic change within the kidneys in patients 1.1 and 2.1. Computed tomography (CT) volumetric rendering of the right labyrinth in patient 2.1 revealed (F, white arrow) dilation of the endolymphatic sac compared with (F, lower right CT scan) healthy control. Note the normal appearance of the cochlea (C) and lateral semicircular canals (SCs) in patient 2.1. G depicts the secondary structural elements of the Foxa3 DNA binding domain (α-helices, cylinders; β-sheets, arrows; and wing domains wing 1 [W1] and W2) above sequence alignments with FOXI1 proteins from various species. FOXI1 L146 in helix 2 (α-helix 2 [α2]) is indicated by the red arrow, and FOXI1 R213 in W2 is indicated by the black arrow. H shows superposition of the human FOXI1 homology model (green) with the crystal structure of Foxa3 (orange) bound to DNA. (I) The recognition helix (α3) of Foxa3 is positioned in the major groove of the cognate DNA and stabilized in part by a 3.4-Å interaction between Foxa3 L142 and the sugar-phosphate backbone. (J) Mutation of the homologous leucine residue in FOXI1 to phenylalanine (L146F; colored magenta) is predicted to disrupt DNA binding. K shows a close-up view of the W2 domain of Foxa3 and FOXI1. L209 in Foxa3 forms a 4.3-Å interaction with the DNA backbone. (L) The structurally homologous FOXI1 R213 is predicted to facilitate DNA binding, similar to the role of R210, R211, and R214 in W2 of Foxa3. Mutation of FOXI1 R213 to proline (R213P; colored magenta) will destroy the electrostatic interaction with DNA.
Clinical and genetic characteristics of affected patients
Structural Analyses of Mutated FOXI1 Proteins
The novel FOXI1 homozygous missense mutations, p.L146F and p.R213P, were identified after whole-exome sequencing in the affected patients with segregation of the alleles from the parents (Figure 1, C and D). The mutations were not found in the Exome Aggregation Consortium database. Both missense changes were located within evolutionary highly conserved residues of the FOXI1 protein (Figure 1G). Consistent with FOX protein sequence alignments (Figure 1G, Supplemental Figure 1), in silico modeling of FOXI1, using the crystal structure of rat Foxa3,12 predicted that FOXI1 L146 occupies a homologous position to L142 in α-helix 2 of the Foxa3 DNA binding domain (Figure 1, H and I). A stabilizing interaction between Foxa3 and its cognate DNA is formed between the side-chain δ1 carbon atom of L142 and the ribose DNA backbone.12 Identical results were obtained when modeling FOXI1 on the structures of human FOXM1,13 human FOXO4,14 human FOXK2,15 and rat Foxd316 (Supplemental Figure 2), with similar protein-DNA interactions identified for the homologous leucine residues in these structures. The FOXI1 L146F mutation was predicted to disrupt this stabilizing protein-DNA interaction, with the larger (approximately 30 Å3),17 rigid aromatic ring of phenylalanine predicted to impinge on the ability of the recognition helix (α-helix 3) to fully engage the major groove of the target DNA (Figure 1J). The functional importance of a small hydrophobic residue at this position in the DNA binding domain is highlighted by the almost complete conservation of a leucine residue in the FOX protein family (Supplemental Figure 1). FOXI1 R213 was predicted to occupy a homologous position to L209 in the wing 2 domain of Foxa3 (Figure 1K). Similar to Foxa3 L142, the side-chain β-carbon atom of Foxa3 L209 interacts with the ribose DNA backbone.12 Therefore, we hypothesized that FOXI1 R213 also served to stabilize the interaction with its target DNA, similar to the role of R210, R211, and R214 in Foxa3 (Figure 1K). Substitution of the positively charged side chain of arginine with proline (p.R213P) will destroy the predicted electrostatic interaction with the DNA, thereby disrupting DNA binding (Figure 1L). (Additional modeling information is detailed in Supplemental Appendix.)
Functional Analyses of Mutated FOXI1 Proteins
We further investigated the functional effect of the mutations in vitro. First, we assessed protein expression levels of the transfected wild type and the mutations p.L146F and p.R213P of FOXI1 in the human kidney cell line HEK 293T, which did not differ (Figure 2A). Second, we compared trans-activation of wild-type and mutant FOXI1 using previously validated promoter constructs8,10 for the human a4 subunit of the H+-ATPase (ATP6V0A4), the human anion-exchange protein pendrin (SLC26A4), and the human anion-exchange protein 1 (AE1/SLC4A1). Consistent with our previous data,8,10 wild-type FOXI1 increased reporter gene activity several fold above empty vector–transfected cells, whereas the p.L146F or p.R213P mutant constructs show significantly reduced capability to trans-activate these promoters (Figure 2B). In an electrophoretic mobility shift assay, we assessed binding of wild-type and mutant FOXI1 proteins to DNA in the form of a canonical FOXI1 binding cis-element. Wild-type FOXI1 bound to the oligonucleotide (Figure 2C, lane 1), whereas no such interaction was detected for the p.L146F or p.R213P mutants (Figure 2C, lanes 2 and 3, respectively). Third, we used GFP fusion constructs to determine the intracellular distribution of the wild-type and mutant FOXI1 proteins. Distinct nuclear localization was observed for wild-type and mutant FOXI1 proteins (Figure 2D), precluding defective nuclear import as a cause for the reduced trans-activational capacity of the p.L146F and p.R213P mutants. A morphometric analysis of data from Figure 2D is shown in Supplemental Figure 5, and an immunochemistry-based approach to intracellular localization of FOXI1 and the p.L146F and p.R213P mutants are shown in Supplemental Figure 6. Taken together, these experiments show that these mutants have a similar intracellular distribution as wild-type FOXI1. Our patients with loss-of-function mutations in FOXI1 recapitulate the phenotype described in a mouse model that lacks Foxi1 expression with early-onset sensorineural deafness and dRTA.7,10
Effect of FOXI1 mutations on protein stability, promoter transactivation, DNA binding, and cellular localization. (A) Western blot (WB) analysis of FOXI1 protein expression in whole-cell lysates from HEK 293T cells transfected with FOXI1 wild type (WT), L146F mutant, R213P mutant, or empty vector (EV). To determine equal protein loading, the membrane was stripped and reprobed for β-actin. *Incomplete stripping of signal from the FOXI1 labeling. B shows reporter gene activity, expressed in relative luciferase units (RLUs), on cotransfection of increasing amounts (15, 30, and 75 ng) of FOXI1 expression plasmid and 25 ng of ATP6V0A4, pendrin, or AE1 promoter reporter constructs in HEK 293T cells. Level of significance, calculated by two-way ANOVA using GraphPad Prism 5, is indicated. **P<0.01; ***P<0.001; ****P<0.001. C shows electrophoretic mobility shift assay using in vitro–transcribed/translated FOXI1 WT, L146F, or R213P with EV as control and radiolabeled oligonucleotides harboring an FOXI1 binding site. Bound and unbound oligonucleotides (oligos) are marked. Lower panel shows a WB of the transcribed/translated products. D illustrates the intracellular localization of FOXI1-GFP fusion proteins after transfection of HEK 293T cells with FOXI1 WT, L146F, or R213P mutants or EV. Green channel, GFP signal (insets show zoomed in views of a single cell); red channel, nuclear staining (To-PRO-3); merge channel, all channels together. Scale bar, 10 μm.
Discussion
The human phenotype that we report mimics the renal findings of patients with mutations in the B1 or a4 subunits of the V-ATPase, and thus, FOXI1 mutations are important in the differential diagnosis of dRTA. The phenotypic mimicry is not surprising, considering the previously reported important role of Foxi1 in the regulation of genes associated with dRTA.3,5,10,18 The discrepancy with respect to the absence or later onset of sensorineural deafness in patients with a4 or more typically, B1 mutations is likely due to FOXI1 mutations affecting the regulation of several important membrane transporters, including those critical for inner ear function, whereas isolated impairment of the a4 subunit and B1 accordingly may generate a less severe phenotype that, to some extent, can be tolerated in the ear. This emphasizes the difference in phenotypic scope between a mutation in a distinct membrane transport protein and that of a transcription factor that regulates several target genes.
In murine studies, in the intercalated cells of the renal tubules and the Forkhead-related cells of the endolymphatic duct and sac of the inner ear, Foxi1 regulates expression of B1 and a4 subunits of the V-ATPase proton pump and also, the anion-exchange proteins AE1, AE4, and pendrin. Without expression of a functional Foxi1 protein, there is a failure to express these target genes, and as a consequence, this results in severely impaired cellular function, with altered ionic/pH composition of inner ear fluid, and a significantly reduced capacity of the kidneys to secrete an acid load.9,10 The inner ear phenotype seen in mice lacking Foxi1 is remarkably similar to what we show here for one of the probands (Figure 1F), with dRTA and an enlargement of the endolymphatic sac and duct resulting in an enlarged vestibular aqueduct (EVA) syndrome (Supplemental Figures 3 and 4).7,9
The biochemical analysis of mutated FOXI1 protein (Figure 2) showed a severe reduction in the ability to interact with bona fide FOXI1 cis-elements with consequent reduced expression of target genes. This is consistent with the in silico analysis showing impaired interactions of mutant FOXI1 with the major groove of target DNA (p.L146F) and disturbance of electrostatic interactions with DNA (p.R213P). Together, these data show that, in patients homozygous for these missense FOXI1 mutations, there is very limited FOXI1 DNA binding activity, likely resulting in a failure to trans-activate a set of membrane transport proteins needed for normal inner ear and kidney function.
The patients in this study did not show any signs of goiter, and available data on thyroid function were normal. Furthermore, we show that patients with recessive loss-of-function mutation in FOXI1 display dRTA and deafness with EVA, suggesting that complete loss of FOXI1 activity induces a unique and distinct phenotype. In a previous report, PS associated with a digenic pattern of inheritance, in which a composite heterozygous genotype with mutations in FOXI1 and SLC26A4 segregates with deafness and an EVA.11 In this patient, goiter was not noted, and the contributing missense mutation in FOXI1 was located approximately 50 amino acids C terminally of the DNA binding domain. In addition, the authors identify five patients with PS or nonsyndromic EVA with a single mutation in one of the FOXI1 alleles, similar to another study.19 Although the finding of rare FOXI1 variants in this patient group is intriguing, the lack of a family history in these patients argues against the pathogenicity of single heterozygous mutations, consistent with the apparent lack of a clinical phenotype in the parents of our patients. Additional evidence against the pathogenicity of isolated heterozygous loss of function of FOXI1 is provided by observations in mice with a heterozygous germline deletion of Foxi1 that also have no phenotype.7,9 On the basis of this, we find it very unlikely that a single mutation in one FOXI1 allele will cause disease. However, in combination with another single-allele mutation (for instance, in a gene that regulates SLC26A4), a single-allele mutation in FOXI1 could be part of the pathogenetic mechanism underlying PS or nonsyndromic EVA.
Of interest, in mice lacking Foxi1, males have been shown to be infertile.20 This is due to a failure of clear cells in the epididymis to acidify the extracellular luminal fluid. This leads to a pathologic post-testicular sperm maturation, and as a consequence, spermatozoa from Foxi1 null males fail to reach the female genital tract in sufficient number to allow fertilization.20 Whether mutations in FOXI1 will be associated with reduced fertility in men also remains an open question, because for the one patient in our study who was a boy, analysis of spermatozoa function was not available. FOXI1 is highly expressed in human renal kidney cortex (RNA-Seq expression data from GTEx; https://www.gtexportal.org). Other human tissues that express FOXI1 at more modest levels include breast tissue, salivary gland, prostate, and skin (https://www.gtexportal.org/home/gene/ENSG00000168269.7). The human protein atlas reveals high levels of FOXI1 protein expression in kidney and low expression levels in nasopharynx, bronchus, salivary gland, and breast tissue (https://www.proteinatlas.org/ENSG00000168269-FOXI1/tissue). Thus, it is possible that other tissues/cell types than those reported here display pathologic phenotypes in response to loss-of-function mutations in FOXI1.
All three patients in this study were noted to have medullary renal cysts, raising the question of whether FOXI1 could also be involved in cystogenesis. However, an association of dRTA with cysts is well recognized: in our own cohort of children with dRTA, over one third developed cysts during childhood.21 Typically, this is ascribed to hypokalemia,22 although other factors, such as hypercalciuria/nephrocalcinosis, have also been implicated.23 Observations in more patients with dRTA are needed to better assess the frequency of cysts in FOXI1-based dRTA compared with other genetic forms.
Approximately one third of patients with dRTA have no identified causative mutation in recognized disease genes21,24,25 and thus, should be tested for FOXI1 mutations to help establish a precise diagnosis with consequent genetic counseling—especially in patients with cases involving hearing and possibly, fertility problems in men.20
Concise Methods
This study was approved by the following institutional review boards: Institute of Child Health/Great Ormond Street Hospital Research Ethics Committee 05/Q0508/6 and Newcastle and North Tyneside Research Ethics Committee 2003/163.
Parents provided written informed consent. Three subjects with dRTA and sensorineural deafness without identified mutations in known disease genes were investigated. We performed whole-exome sequencing using peripheral blood DNA from affected family members. Sanger sequencing was used to validate the variants identified by whole-exome sequencing in both affected children. In silico modeling and biochemical analysis were performed with human FOXI1 protein as a control compared with the two mutated FOXI1 proteins.
Exome Capture and Next Generation Sequencing
Whole-exome sequencing of affected patients 1.1 and 1.2 was performed by GATC Biotech using the Illumina HiSeq-2000 (Illumina Inc., San Diego, CA). The genomic DNA was captured on the SureSelectXT Human All Exon V5 Kit (Agilent, CA) to capture the target sequence of exonic regions and noncoding RNAs in the human genome. The paired end Illumina libraries were captured in solution according to the Agilent SureSelect protocol with 125-bp read length. Perkin Elmer provided commercially available exome sequencing for patient 2.1 (exome capture: Agilent SureSelect V4; read length/library construction: 75–100 bp per paired end library and reads; sequence coverage: approximately 35 times mean coverage; detection of sequencing fragments via the Illumina Hiseq2000).
Single-Nucleotide Polymorphism Detection
The base quality of each sequence read was inspected for low-quality calls and subsequently removed before proceeding with further processing. Using a sliding window approach, bases with low quality were removed from the 3′ and 5′ ends. Bases were removed if the average quality was below the threshold of 15. Finally, only reads with both forward and reverse read were used for the next analysis step. Mapping to the reference database (hg19, NCBI37) was performed using the Burrows–Wheeler Alignment method with default parameters. The single-nucleotide polymorphism and insertion and deletion selection calling was performed using GATK UnifiedGenotyper using a Bayesian genotype likelihood model to estimate simultaneously the most likely genotypes and allele frequency in a population of N samples. Variants detected were annotated on the basis of their gene context using snpEff. The total reads were approximately 145 million, with the percentage of mapped reads of 99%. Data were pipelined to allow insertion and deletion selection and single-nucleotide polymorphism vcf files to be generated, which were viewed using Ingenuity Variant Analysis software (Qiagen) (Tables 2 and 3). Additional information is in Supplemental Appendix.
Filtering criteria used for single-nucleotide variants and insertion and deletion selection
PCR primers used for Sanger sequencing
Acknowledgments
We thank the patients and their family members for their help in this study. J.A.S. is funded by the Medical Research Council (MR/M012212/1), Kidney Research UK, Kids Kidney Research, Northern Counties Kidney Research Fund and the Newcastle upon Tyne Hospitals NHS Charity. Funding for this study to R.K. and D.B. was kindly provided by the European Union, FP7 (grant agreement 2012-305608 “European Consortium for High-Throughput Research in Rare Kidney Diseases (EURenOmics)“) and Kids Kidney research, Kidney Research UK and The John Moorhead Trust. S.E. is supported by the Swedish Research Council (2014–2516), The Knut and Alice Wallenberg Foundation, Sahlgrenska's University Hospital (LUA-ALF), The Inga Britt and Arne Lundgren Foundation, The Torsten Söderberg Foundation, Novo Nordisk Foundation, and The King Gustaf V and Queen Victoria Freemason Foundation. CAW is supported by the Swiss National Science Foundation (155959 and NCCR Kidney.CH) and the European Union FP7 project EURenOmics.
Disclosures
None.
Footnotes
C.A.W., R.K., D.B., and J.A.S. have shared senior authorship.
Published online ahead of print. Publication date available at www.jasn.org.
This article contains supplemental material online at http://jasn.asnjournals.org/lookup/suppl/doi:10.1681/ASN.2017080840/-/DCSupplemental.
- Copyright © 2018 by the American Society of Nephrology