3′ Untranslated Region Structural Elements in CYP24A1 Are Associated With Infantile Hypercalcemia Type 1

ABSTRACT

Loss-of-function mutations in the CYP24A1 protein-coding region causing reduced 25 hydroxyvitamin D (25OHD) and 1,25 dihydroxyvitamin D (1,25(OH)₂D) catabolism have been observed in some cases of infantile hypercalcemia type 1 (HCINF1), which can manifest as nephrocalcinosis, hypercalcemia and adult-onset hypercalciuria, and renal stone formation. Some cases present with apparent CYP24A1 phenotypes but do not exhibit pathogenic mutations. Here, we assessed the molecular mechanisms driving apparent HCINF1 where there was a lack of CYP24A1 mutation. We obtained blood samples from 47 patients with either a single abnormality of no obvious cause or a combination of hypercalcemia, hypercalciuria, and nephrolithiasis as part of our metabolic and stone clinics. We used liquid chromatography tandem mass spectrometry (LC-MS/MS) to determine serum vitamin D metabolites and direct sequencing to confirm CYP24A1 genotype. Six patients presented with profiles characteristic of altered CYP24A1 function but lacked protein-coding mutations in CYP24A1. Analysis upstream and downstream of the coding sequence showed single nucleotide variants (SNVs) in the CYP24A1 3′ untranslated region (UTR). Bioinformatics approaches revealed that these 3′ UTR abnormalities did not result in microRNA silencing but altered the CYP24A1 messenger RNA (mRNA) secondary structure, which negatively impacted translation. Our experiments showed that mRNA misfolding driven by these 3′ UTR sequence-dependent structural elements was associated with normal 25OHD but abnormal 1,25(OH)₂D catabolism. Using CRISPR-Cas9 gene editing, we developed an in vitro mutant model for future CYP24A1 studies. Our results form a basis for future studies investigating structure–function relationships and novel CYP24A1 mutations producing a semifunctional protein. © 2023 The Authors. Journal of Bone and Mineral Research published by Wiley Periodicals LLC on behalf of American Society for Bone and Mineral Research (ASBMR).

Introduction

Vitamin D plays a key role in classical calciotropic processes including calcium and bone metabolism and is postulated to contribute to nonclassical disorders including cancer, diabetes, multiple sclerosis, and coronavirus disease 2019 (COVID-19).^(1-3) Vitamin D is obtained from either the diet (ergocalciferol) or in the skin (cholecalciferol) after photochemical conversion of 7-dehydrocholesterol. Vitamin D is transported to the liver where it is hydroxylated by cytochrome P450 family 2 subfamily R member 1 (CYP2R1) to form 25 hydroxyvitamin D (25OHD). A further hydroxylation in the kidney by cytochrome P450 family 27 subfamily B member 1 (CYP27B1) generates the active systemic metabolite 1,25 dihydroxyvitamin D (1,25(OH)₂D) essential for calcium homeostasis. Vitamin D metabolism is regulated by the activity of CYP27B1 and cytochrome P450 family 24 subfamily A member 1 (CYP24A1). CYP24A1 converts the precursor 25OHD into 24,25 dihydroxyvitamin D (24,25(OH)₂D) and 1,25(OH)₂D into 1,24,25 trihydroxyvitamin D (1,24,25(OH)₃D). Both 24,25(OH)₂D and 1,24,25(OH)₃D are subject to further hydroxylation followed by bile and urinary excretion to prevent vitamin D toxicity. CYP27B1 and CYP24A1 activity is controlled by 1,25(OH)₂D, calcium, parathyroid hormone (PTH) and fibroblast growth factor 23 (FGF23). Vitamin D metabolite relative ratios (VMRs) such as 25OHD:24,25(OH)₂D and 1,25(OH)₂D:24,25(OH)₂D are critical in the differential diagnosis of the vitamin D hydroxylation pathways.⁽⁴⁾

Vitamin D toxicity and/or sensitivity manifesting as hypercalcemia, hypercalciuria, and/or nephrolithiasis caused by CYP24A1 loss-of-function mutations resulting in elevated serum 1,25(OH)₂D⁽⁵⁾ is a disorder known as infantile hypercalcemia type 1 (HCINF1, OMIM #143880). Type 2 (HCINF2, OMIM #616963) differs in that there is a loss-of-function in solute carrier family 34 member 1 (SLC34A1) and the phenotype includes hypophosphatemia. HCINF1 infant presentation includes vomiting, failure to thrive, colic, and in rare cases death. HCINF1 adult presentation can include flu-like symptoms, hypercalciuria, and renal stone formation. In some female patients these symptoms are triggered in pregnancy, likely uncovered by vitamin D supplementation rather than the pregnancy itself.⁽⁶⁾ A future discussion required in the field is the removal of the word “infantile” from HCINF1 given the increasing adult presentation frequency.

In the current study we used the VMR to select patients with suspected HCINF1 for genetic analysis to confirm the diagnosis as part of their care in metabolic and stone former clinics. We also investigated two pediatric patients referred for investigation of nephrocalcinosis and calcium metabolism abnormalities. Through CYP24A1 direct sequencing, we identified in the two children and four out of 47 adults (8.5%) single nucleotide variants (SNVs) in the 3′ untranslated region (UTR) with unknown clinical significance, which prompted nonclinical studies described here. The 3′ UTR is of significant messenger RNA (mRNA) regulatory importance plus single-stranded RNAs fold into complex three-dimensional structures that are critical for their function/regulation including posttranscriptional modification, nuclear export, cellular localization, translation, and degradation.^(7-12) We hypothesized that SNVs in the 3′ UTR affected the CYP24A1 mRNA secondary structure with mRNA misfolding leading to the heterogeneous phenotypes observed in some HCINF1 cases. After performing bioinformatics and computational modeling to demonstrate mRNA structural abnormalities in these patients, we generated a CRISPR-Cas9 mutant HEK293T cell line to mimic patients with a CYP24A1 3′ UTR variant. This mutant model provides a tool for in vitro investigation into noncanonical and pathogenic CYP24A1 phenotypes.

Materials and Methods

Clinical samples

Forty-seven patient blood samples were collected as part of routine requests for 25OHD liquid chromatography tandem mass spectrometry (LC-MS/MS) analysis from the Department of Laboratory Medicine at the Norfolk and Norwich University Hospital between 2016 and 2017. Patients were referred from the metabolic or stone-former clinics. There were 21 patients with a diagnosis of primary hyperparathyroidism including five who formed renal stones. There were eight patients with osteoporosis including two who had high urinary calcium excretion. There were five patients who had hypercalcemia of nonparathyroid and nonmalignant origin. Eight patients were being investigated for renal stones, three were hypoparathyroid, and two were under investigation for secondary hyperparathyroidism. Blood samples were collected into serum gel separator tubes (BD Vacutainer; BD Biosciences, San Jose, CA, USA) and immediately centrifuged. The serum layer was aliquotted and stored at −20°C. Two samples from children were referred to the regional metabolic biochemistry service for investigation of causes of nephrocalcinosis and abnormal calcium biochemistry.

Biochemical analysis

Serum vitamin D metabolites 25OHD and 24,25(OH)₂D were measured simultaneously by LC-MS/MS using the Micromass Quattro Ultima Pt electrospray ionization (ESI) tandem mass spectrometer (Waters Corp., Milford, MA, USA) as described.⁽⁴⁾ MassLynx v4.1 and QuanLynx (Waters Corp.) were used for assay performance, system control, data acquisition, baseline integration, and peak quantification. The LIAISON® XL 1,25(OH)₂D chemiluminescent immunoassay (DiaSorin, Saluggia, Italy) method was used to measure 1,25(OH)₂D in serum samples. This sandwich assay utilizes a recombinant fusion protein for 1,25(OH)₂D capture and a murine monoclonal antibody detection system. The assay measures total 1,25(OH)₂D between 12 and 480 pmol/L. The interassay/intraassay coefficient of variation (CV) was ≤9.2% and the mean assay recovery was 94% ± 2% across the analytical range. Cell line 1,25(OH)₂D catabolism was measured by LC-MS/MS using the Xevo TQ-XS (Waters Corp.) as described.⁽¹³⁾ Intact PTH and albumin-adjusted calcium (ACa) were analyzed on the COBAS® 6000 (Roche Diagnostics, Burgess Hill, UK) platform. Ethylenediamine tetraacetic acid (EDTA)-plasma PTH was measured using electrochemiluminescence immunoassay (ECLIA). The interassay CV was ≤3.8% across the analytical range of 1.2 to 5000 pg/mL. Total calcium (Ca) and albumin were measured using spectrophotometric methods. The interassay CV for Ca was ≤1.6%, albumin was ≤1.1% across the working ranges of the assays.

Study participants

The University of East Anglia Faculty of Medicine and Health Sciences Research Ethics Committee approved the collection and study of DNA samples for nonclinical procedures (ref: ETH2122-1884); ie, the four adult and two pediatric patients. We obtained whole blood from the metabolic and stone clinics for genetic analysis from patients presenting with elevated 1,25(OH)₂D and high/low VMR plus HCINF1 clinical presentation and who were negative for protein-coding mutations (patients 1–4) (Table 1). We obtained genomic DNA from an infant presenting with Williams syndrome and nephrocalcinosis (patient 5, Croydon Hospital) and an infant presenting with nephrocalcinosis and polyuria (patient 6, Royal Hospital for Children) (Table 1). All adults or infant parents/guardians provided written informed consent to donate samples for this study. Anonymized negative control blood samples were collected at the Norfolk and Norwich University Hospital blood typing service (n = 10). Exclusion criteria for the negative control samples were those with a vitamin D, calcium, or metabolic clinical history.

Table 1. Serum biochemistry and mutational analysis

Patient	Sex	Age (years)	Total 25OHD (50–120 nmol/L)	1,25(OH)2D (55–139 pmol/L)	Total 24,25(OH)2D (1.1–13.5 nmol/L)	Total 25OHD:24,25(OH)2D relative ratio (7–23)	1,25(OH)2D:24,25(OH)2D relative ratio (11–62)	Adjusted calcium (2.1–2.6 mmol/L)	Phosphate (0.8–1.5 mmol/L) (adult); 1.45–2.1 (child)	SNV	Clinical notes
1	F	33	104	83	3.3	32	25	3.27	1.04	c.*131T>C	Hypercalciuria; recurrent nephrolithiasis
2	F	28	91	262	10.6	9	25	2.33	1.29	c.*41C>T; c.*706C>G	Hypercalciuria; recurrent nephrolithiasis
3	F	33	97	177	7	14	25	2.32	0.93	c.*131T>C; c.*560T>A	Hypercalciuria; recurrent nephrolithiasis; osteopenia
4	M	55	73	171	9.4	8	18	2.44	0.93	c.*706C>G	Recurrent nephrolithiasis; osteopenia
5	M	<1	122	616	3.5	35	176	3.41	1.98	c.*739G>A	Williams syndrome; IIH; nephrocalcinosis
6	M	<1	108	175	5.8	19	30	3.2	Not tested	c.368insC; c.1144insT; c.*131T>C	Nephrocalcinosis; high creatinine

• Note: A low 25OHD:24,25(OH)₂D VMR was associated with hypercalciuria and nephrolithiasis. A high 25OHD:24,25(OH)₂D VMR was associated with hypercalcemia and nephrocalcinosis. All but one patient had significantly reduced 1,25(OH)₂D catabolism and some patients presented with abnormal calcium handling. Reference ranges are given in brackets. Bold font shows abnormal measurement.
• Abbreviation: SNV = single nucleotide variant.

DNA and RNA extraction

Genomic DNA was isolated from nucleated blood cells using the Purelink genomic DNA kit (Invitrogen, Carlsbad, MA, USA). Total RNA was extracted using the miRNeasy mini kit (Qiagen, Manchester, UK). DNA and RNA concentration and integrity was measured on the NanoDrop 8000 (Thermo Fisher Scientific, Loughborough, UK). DNA was stored at −20°C. RNA was stored at −80°C.

Sequencing and variant calling

CYP24A1 direct sequencing was performed using primers as described.⁽¹⁴⁾ For direct sequencing analysis we aligned FASTA reads with CYP24A1 transcript variant 1 (NM_000782.5) using BioEdit (https://bioedit.software.informer.com/), a biological sequence alignment editor. MutationTaster (https://www.mutationtaster.org/), ClinVar (https://www.ncbi.nlm.nih.gov/clinvar/), and dbSNP (https://www.ncbi.nlm.nih.gov/snp/) were used to assess SNV disease-causing potential.

Bioinformatics

RNAfold⁽¹⁵⁾ was used to determine the predicted CYP24A1 mRNA structures. For mRNA secondary structure prediction, RNAfold provides three dynamic programming algorithms including (i) minimum free energy (MFE), which generates a single optimal structure using thermodynamic predictions based on the MFE generated by the nucleotide composition of the input sequence; (ii) partition function, which calculates base pair probabilities in the thermodynamic ensemble; and (iii) suboptimal folding, which generates all suboptimal structures within a given energy range of the optimal energy. For quantifying mRNA secondary structure comparison (to the control/wild-type), the package contains several measures of distance (ie, dissimilarities) using either string alignment or tree-editing. RNAfold performance was extensively tested and validated by comparing MFE predictions between RNAfold 1.8.5, RNAfold 2.1.8, UNAfold 3.8, and RNAstructure 5.7 including accuracy; sensitivity, positive predictive value, Phi coefficient, and F-measure. The test set was based on 1919 nonmultimer sequence/structure pairs obtained from the RNAstrand database (all without pseudoknots in the reference structure). Both versions of RNAfold were run with -d2 option whereas UNAfold and RNAstructure were run with default options.⁽¹⁶⁾ RBPmap⁽¹⁷⁾ was used to determine whether the 5′ UTR and 3′ UTR mutations impaired protein-RNA interaction. RBPmap employs an algorithm for mapping protein binding motifs on RNA transcripts while considering the motif clustering propensity and the overall tendency of the regulatory region to be conserved. miRDB^{(18, 19)} was used to elucidate whether the 3′ UTR variants altered or introduced microRNA (miRNA) recognition elements (ie, target sites). miRDB is a database for miRNA target prediction and functional annotation. All mRNA targets in miRDB (https://mirdb.org/) were predicted by miRTarget, which was developed by analyzing thousands of miRNA-mRNA functional interactions from next generation sequencing studies.

Digital polymerase chain reaction

Total RNA was quantified by density measurement after separation by agarose gel electrophoresis with ethidium bromide staining. Equal RNA amounts were reverse transcribed using the high-capacity RNA to complementary DNA (cDNA) kit (Thermo Fisher Scientific). CYP24A1 transcript expression was quantified in triplicate using a TaqMan gene expression assay (Thermo Fisher Scientific). Digital polymerase chain reaction (dPCR) was performed on the QuantStudio 3D Digital PCR System using the GeneAmp PCR System 9700 (Thermo Fisher Scientific). After PCR, chips were imaged on the QuantStudio 3D instrument to convert raw data into the concentration of the cDNA sequence targeted by fluorescein amidites (FAM)-labeled and 2′-chloro-7′phenyl-1,4-dichloro-6-carboxy-fluorescein (VIC)-labeled probes according to Poisson distribution.⁽²⁰⁾ The QuantStudio 3D AnalysisSuite was used to convert the data into copies per microliter (μL). Experiments were performed in triplicate in three independent experiments.

CYP24A1 protein expression

Peripheral blood mononuclear cells (PBMCs) were separated from whole EDTA blood samples by Ficoll-Paque PLUS (GE Healthcare Life Sciences, Marlborough, MA, USA). PBMCs were lysed using M-PER lysis buffer (Thermo Fisher Scientific) with complete protease inhibitor cocktail (Roche Diagnostics) and were clarified by centrifugation. Protein concentrations were determined from the supernatants using the bicinchoninic acid (BCA) protein assay system (Thermo Fisher Scientific). Proteins were separated on a precast 4%–12% gradient sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) gel (Thermo Fisher Scientific) and transferred onto Immobilon polyvinylidene fluoride (PVDF) (Millipore, Watford, UK) to blot. CYP24A1 monoclonal antibody #WH0001591M7 (Sigma-Aldrich, Gillingham, UK) was used to probe membranes for 48 hours at 4°C. Actin and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) were included as loading controls. IRDye (LI-COR, Cambridge, UK) labeled secondary antibodies were used to detect primary antibodies. Proteins were visualized using the Odyssey infrared system (LI-COR, Cambridge, UK). CYP24A1 was also quantified by enzyme-linked immunosorbent assay (ELISA) (Cusabio, Houston, TX, USA). The detection range of this sandwich assay was 7.8–500 pg/mL and the sensitivity was 1.95 pg/mL. The interassay/intraassay CV was ≤10% and the mean assay recovery was 86% ± 6% across the analytical range. Sample absorbance was detected at 450 nm using a plate reader. Concentrations were calculated using a four-parameter logistic (4PL) curve ranging 0–500 pg/mL. ELISAs were performed in duplicate according to manufacturer's instructions.

CRISPR-Cas9

Plasmid guide RNA (gRNA) Cas9 constructs based on the vector GE100002 and donor 100-basepair (bp) single-stranded DNA (ssDNA) oligonucleotides (oligos) (Origene, Herford, Germany) were designed to introduce CYP24A1 3′ UTR alterations (File S1). To anneal forward/reverse ssDNA oligos, 2 μg of each oligo were combined with 50 μL annealing buffer (10nM Tris, 1mM EDTA, 50mM NaCl), heated at 95°C for 2 minutes and then allowed to cool to 25°C, over 1 hour. Human embryonic kidney (HEK293T) cells were cultured as a monolayer in Dulbecco's modified Eagle medium (DMEM) GlutaMax medium (Thermo Fisher Scientific) supplemented with 1% (vol/vol) penicillin–streptomycin (Thermo Fisher Scientific) and 10% (vol/vol) fetal bovine serum (Sigma-Aldrich). HEK293T cells were seeded at 3.5 × 10⁵ cells/well into six-well plates and were transfected with 7.5 μL Lipofectamine 3000 (Thermo Fisher Scientific), 1 μg gRNA/Cas9 and 1 μg annealed DNA oligo in 250 μL Opti-MEM (Thermo Fisher Scientific). Cells were incubated for 48 hours at 37°C in 5% CO₂. Control cells were treated with Lipofectamine 3000 minus the gRNA/Cas9 vector or oligos. After 48 hours, cells were split 1:10 with fresh DMEM GlutaMax medium (Thermo Fisher Scientific). Array dilution was performed to isolate single-cell colonies. To confirm CRISPR-Cas9 transfection in each single-cell colony, CYP24A1 direct sequencing was performed (c.*74_80del, c.*83_85del, and c.*88del).

Mutant 25OHD and 1,25(OH)₂D catabolism

Both wild-type HEK293T (ie, controls) and CRISPR-Cas9 mutant cell lines were treated with 200nM 25OHD or 10nM 1,25(OH)₂D over a 48-hour period. Culture medium was collected at intervals (1, 4, 12, 24, and 48 hours) to assess vitamin D metabolite metabolism as described under biochemical analysis.

Single-molecule fluorescence in situ hybridization

Single-molecule fluorescence in situ hybridization (smFISH) for CYP24A1 mRNA transcripts was performed in HEK293T cells, CYP24A1 mutant cells and human PBMCs using a pool of 48 Stellaris RNA FISH probes (LGC Biosearch Technologies, Petaluma, CA, USA). Probes recognizing the housekeeping gene RNA polymerase II subunit A (POLR2A) labeled with Quasar™ 570 dye (LGC Biosearch Technologies, Petaluma, CA, USA; #SMF-2003-1) were used as a positive control following the manufacturer's instructions available online at https://www.biosearchtech.com/support/resources/stellaris-protocols (access date 2021). Parameter design for POLR2A probes were against the coding sequence of NM_000937.4 (NCBI gene ID: 5430, nucleotides 387–6299). Custom probes were designed against CYP24A1 (NM_000782.5; NCBI gene ID: 1591; nucleotides 1–643 and 1842–1962) (File S2) using the Stellaris RNA FISH probe designer available online at www.biosearchtech.com/stellarisdesigner and labeled with Quasar™ 670 dye. Cells were grown to ~80% confluency on 18-mm-round glass coverslips before fixing with 4% paraformaldehyde for 30 minutes. Fixed cells were permeabilized with 90% (vol/vol) ethanol for 1 hour. Hybridization with 125nM each probe was carried out for 18–24 hours in a blackout humidified chamber at 37°C. Coverslips were placed into fresh 12-well plates containing wash buffer A and incubated at 37°C for 30 minutes. Samples were counterstained with 4′, 6-diamidino-2-phenylindole (DAPI) for nuclear detection. Coverslips were then washed with wash buffer B and incubated for 5 minutes. Coverslips were mounted with 15 μL Vectashield mounting medium (Vector Laboratories, Burlingame, CA, USA) onto a microscope slide cell-side down. Images were acquired with a Zeiss Elyra PS1 inverted microscope (Carl Zeiss Microscopy, Inc., Dublin, CA, USA) using 100× oil-immersion objective (1.46 NA) and cooled electron multiplying–charge-coupled device (CCD) Andor iXon 897 camera (512 × 512 QE>90%). Images were acquired using standard widefield rather than super-resolution mode. For fluorescence detection, Quasar™ 570 probes were detected using an excitation line of 561 nm with the signal detected at 570–640 nm. Quasar™ 670 probes were detected using an excitation line of 642 nm and signal detection at 655–710 nm. A 405 nm excitation line was used to detect the nuclear stain DAPI with emission detected between 420 and 480 nm. For all experiments, z-steps of 0.2 μm series were collected. mRNA counting was performed using ImageJ software (NIH, Bethesda, MD, USA; https://imagej.nih.gov/ij/).⁽²¹⁾

Statistics

Unless otherwise stated in the methods, statistical analyses and graphical representations were performed using SPSS Statistics v25.0.0.1 (IBM, Armonk, NY, USA) and Prism v9.0 (GraphPad Software, Inc., La Jolla, CA, USA). Statistical significance was considered as a two-tailed p value <0.05.

Results

High/low VMRs were associated with CYP24A1 genetic abnormalities

Forty-seven patient biochemical profiles were analyzed using LC-MS/MS as part of their standard of care from our metabolic and stone former clinics. Patients with elevated 1,25(OH)₂D and/or high/low VMR concentrations were indicative for HCINF1. Cases were sent for genetic screening to support the CYP24A1 diagnosis. Four of 47 (8.5%) displayed a lack of CYP24A1 protein-coding region mutations despite the clinical presentation and biochemistry, which prompted nonclinical studies. The other 43 patients were positive for coding sequence mutations in CYP24A1 consistent with previous reports and/or other genes including PHEX associated with their condition. The four adult cases (patients 1–4) were examined alongside two infants with suspected CYP24A1 pathology (patients 5 and 6) (Table 1); patient 5 also had Williams syndrome. Except for one adult (patient 1) who was within the reference range, 1,25(OH)₂D was elevated in all patients providing evidence for reduced 1,25(OH)₂D catabolism (Table 1). Patient 1, who was within the 1,25(OH)₂D reference range, had markedly elevated adjusted calcium (Table 1). The two infants (patients 5 and 6) also displayed abnormal calcium handling (Table 1). Except for patient 1, adults presented a 25OHD:24,25(OH)₂D VMR in the lower 25th percentile, but within the reference range demonstrating some CYP24A1 protein function (Table 1). In both infants (patients 5 and 6) and patient 1, the 25OHD:24,25(OH)₂D VMR was in the upper 75th percentile or above the upper limit, but within the reference range (Table 1). In summary, all six patients presented with either an abnormality of 1,25(OH)₂D or calcium handling with elevated 1,25(OH)₂D plus low normal 25OHD:24,25(OH)₂D VMR associated with hypercalciuria and nephrolithiasis. Markedly elevated 1,25(OH)₂D plus high normal and elevated 25OHD:24,25(OH)₂D VMR was associated with hypercalcemia (in three patients) and nephrocalcinosis/renal stones (all patients). Five patients displayed normal phosphate levels (Table 1). Patient 6 was not tested for phosphate. All six patients harbored SNVs in the CYP24A1 3′ UTR (patient 1 c.*131T>C; patient 2 c.*41C>T and c.*706C>G; patient 3 c.*131T>C and c.*560T>A; patient 4 c.*706C>G; patient 5 c.*739G>A; patient 6 c.*131T>C) with additional mutations in the protein-coding region in patient 6 (c.368insC and c.1144insT) (Table 1).

CYP24A1 3′ UTR mutations do not introduce de novo miRNA target sites

We first speculated that the 3′ UTR SNVs might have introduced de novo and/or mutated endogenous miRNA recognition elements (MREs) causing hypersilencing by miRNAs known to target CYP24A1, such as miR-30b and miR-125b; however, our results did not support the hypothesis that there was MRE interruption with the 3′ UTR mutations.

CYP24A1 3′ UTR mutations are associated with mRNA misfolding

Given the regulatory importance of the 3′ ends of mRNA transcripts, we assessed all six patients using RNAfold, which produced graphical (Fig. 1A–G) and quantitative (Fig. 2A–G) outputs visualizing the predicted mRNA structures. Each patient's predicted mRNA structure was visibly misfolded when compared to the reference human genome (Fig. 1A–G) and measurably misfolded after producing mountain plots when compared to the reference human genome (Fig. 2A–G). We investigated both the thermodynamic mRNA structures and the centroid structures for each genotype. Both parameters consistently showed that the patients' genotypes altered the mRNA structure.

Fig. 1

3′ UTR mutations and mRNA structures. (A) Reference human genome CYP24A1 transcript. (B) Patient 1 (c.*131T>C). (C) Patient 2 (c.*41C>T; c.*706C>G). (D) Patient 3 (c.*131T>C; c.*560T>A). (E) Patient 4 (c.*706C>G). (F) Patient 5 (c.*739G>A). (G) Patient 6 (c.368insC; c.1144insT; c.*131T>C). Red circles highlight the misfolded region when compared to the reference.

Fig. 2

mRNA misfolding quantification using mountain plots. mRNA structure thermodynamic ensemble and the centroid structures. Mountain plots represent structure in a plot of height versus position where the height m(K) is given by the base pair number enclosing the base at position k; ie, loops correspond to plateaus, hairpin loops are peaks, helices to slopes. We also show the entropy for each position. Patient genotypes are visibly different when compared to the reference structure. (A) Reference human genome CYP24A1. (B) Patient 1. (C) Patient 2. (D) Patient 3. (E) Patient 4. (F) Patient 5. (G) Patient 6.

CYP24A1 mRNA misfolding does not alter transcript expression but is associated with higher CYP24A1 protein abundance

Information movement from genotype to phenotype includes abundant regulatory stages so we tested whether mRNA misfolding caused a reduction of CYP24A1 mRNA through mRNA degradation and/or decay leading to loss-of-function and reduced 1,25(OH)₂D catabolism. We extracted total RNA from whole blood from controls (n = 5) plus the six patients and performed CYP24A1 dPCR analysis. This experiment showed a nonsignificant difference in mRNA levels between the controls and patients (p = 0.07) indicating that the mRNA structural changes were not causing reduced mRNA expression or stability (Fig. 3A). We next sought to determine whether mRNA misfolding was associated with abnormal translation. Western blot analysis performed on controls and three of the adult patients available for resampling showed a significant increase and/or accumulation of CYP24A1 (p = 0.023) (Fig. 3B,C). To support this result, we also performed ELISAs for CYP24A1, which confirmed a significant increase in CYP24A1 protein associated with CYP24A1 mRNA misfolding (p = 0.008) (Fig. 3D).

Fig. 3

mRNA misfolding mechanisms. Three patient samples were available for ex vivo studies. (A) Total RNA extracted from whole blood samples used for dPCR analysis for CYP24A1 transcript expression shows no significant difference (p = 0.07) between five controls and three patients. Data is reported as copies/μL as calculated by Poisson distribution. Data presentation includes the 75th and 25th percentiles with the median values (50th percentile) indicated by the central line. The x represents the mean value. Due to the accuracy and sensitivity of dPCR, CYP24A1 expression across the samples is invariable. Statistical significance was calculated using an unpaired t test. (B) PBMC Western blot using a mouse monoclonal anti-CYP24A1 antibody shows an increase, retention, and/or accumulation of CYP24A1 protein in patients when compared to controls. Actin was used as a loading control. (C) Graphical representation of the Western blot data to demonstrate CYP24A1 increase when compared to controls. Error bars represent the standard deviation. Statistical significance was calculated using an unpaired t test (p = 0.023). (D). Plasma samples' ELISA confirmed CYP24A1 increase and/or accumulation in patient samples when compared to controls. Error bars represent the standard deviation. Statistical significance was calculated using an unpaired t test (p = 0.008). For (C) and (D), the top and bottom whisker lines denote the maximum and minimum value in the dataset, respectively, the top and bottom of the box denote the 75th and 25th percentiles of the dataset, the median value (50th percentile) is indicated by the central line of the box plot, the x represents the mean value and the white circles indicate data points that are within the interquartile range.

Development of an in vitro model system for HCINF1 and CYP24A1 studies

For studies to explore how CYP24A1 mRNA misfolding was associated with CYP24A1 abundance, which given the patients' phenotype appeared to be semifunctional (ie, normal 25OHD and 24,25(OH)₂D but abnormal 1,25(OH)₂D catabolism), we used CRISPR-Cas9 to modify the CYP24A1 3′ UTR in a HEK293T cell line. The Human Protein Atlas (https://www.proteinatlas.org/) confirmed that CYP24A1 is enriched in kidney cells, which is why we selected an immortalized human kidney cell line. We designed the mutant cell model to contain three deletions (c.*74_80del, c.*83_85el, and c.*88del), which was a tradeoff between ensuring that the 3′ UTR structure was altered and designing CRISPR-Cas9 oligos that would be successful, which was confirmed and validated by direct sequencing and RNAfold (Fig. 4A,B). The mountain plot and entropy for each nucleotide of the mutant sequence with comparison to the wild type was measured (Fig. 4C,D). Western blotting showed that CYP24A1 was still expressed in the mutant when compared to wild-type HEK293T cells (Fig. 4E). We then compared mutant cell responses to 25OHD and 1,25(OH)₂D stimulation to validate the model. We observed a nonsignificant (p = 0.18) production of 24,25(OH)₂D in 25OHD-treated cells over a 48-hour period (Fig. 4F) (Table 1). 1,25(OH)₂D catabolism was significantly impacted after 24 hours (p = 0.04) in 1,25(OH)₂D-treated mutant cells (Fig. 4G).

Fig. 4

CRISPR-Cas9 mutant cell line. MFE-based mRNA structures of (A) reference human genome CYP24A1 transcript and (B) CRISPR-Cas9 mutant CYP24A1 (c.*74_80del, c.*83_85el, and c.*88del) with the red circle indicating structure alteration. Mountain plot measurements of the MFE-based mRNA structures of (C) Wild-type CYP24A1 and (D) CRISPR-Cas9-modified CYP24A1 (c.*74_80del, c.*83_85el, and c.*88del). (E) Western blot shows that the CYP24A1 protein is still expressed in the mutant consistent with the patient phenotype. GAPDH was used as a loading control. (F) LC-MS/MS analysis of HEK293T and mutant 24,25(OH)₂D in the culture medium over 48 hours when stimulated with 200nM 25OHD. Results are displayed as average 24,25(OH)₂D concentration at each time point measured (n = 3). (G) LC-MS/MS analysis of HEK293T and mutant 1,25(OH)₂D concentration in culture medium over 48 hours after stimulating with 10nM 1,25(OH)₂D (n = 3).

CYP24A1 mRNA transcript subcellular location in vitro

We developed a single molecule fluorescent in situ hybridization (smFISH) protocol to detect single CYP24A1 mRNA transcripts. This protocol was developed to visualize whether mRNA misfolding interfered with mRNA nuclear export, trafficking, and/or localization leading to abnormal protein translation. We stimulated HEK293T cells with 10nM 1,25(OH)₂D for 12 hours before fixing and probing, which showed single CYP24A1 mRNA transcripts (Fig. 5A). These mRNAs were significantly upregulated in response to 1,25(OH)₂D3 (p = 0.008) (Fig. 5B). As well as observing a significant increase in total CYP24A1 mRNA transcripts when compared to the housekeeping gene POLR2A (p = 0.005), we observed markedly high cell-to-cell variation and that most CYP24A1 mRNA transcripts resided in the nucleus (Fig. 5C). POLR2A detection was consistent across all experiments, supporting this molecule as an appropriate control for smFISH studies (Fig. 5C). CYP24A1 localization variation included cytoplasmic accumulation in cells with minimal nuclear expression (Fig. 5D), low abundance CYP24A1 absent from both the nucleus and the cytoplasm (Fig. 5D) plus several CYP24A1 transcription sites in the nucleus with minimal detection in the cytoplasm (Fig. 5D). We also performed smFISH in the mutant and observed CYP24A1 localization in both the nucleus and cytoplasm (Fig. 5D). Strong CYP24A1 probe signal was identified in mutant cells similar to transcription sites observed in the wild type (Fig. 5D). We observed a higher CYP24A1 mRNA cytoplasmic retention in the mutant, but this was nonsignificant (cytoplasm p = 0.54; nuclear p = 0.43) when compared to wild-type HEK293T cells (Fig. 5E).

Fig. 5

CYP24A1 mRNA quantification and subcellular localization. (A) Maximum Z-projected smFISH image showing a HEK293T cell stimulated with 10nM 1,25(OH)₂D with markedly more CYP24A1 mRNAs (green) than control POLR2A transcripts (magenta). (A) High intensity CYP24A1 spots indicate ongoing CYP24A1 transcription (white arrow). Nuclear stain DAPI is shown in blue. (B) Comparison of the average CYP24A1 mRNA transcripts per cell observed before and after stimulation of wild-type HEK293T cells with 10nM 1,25(OH)₂D (p = 0.00842). (C) A significant difference was observed between the frequency of CYP24A1 and POLR2A in the cytoplasm (p = 0.03) and the nucleus (p = 0.04) in 1,25(OH)₂D-stimulated HEK293T cells. There was a significant difference in total CYP24A1 and POLR2A probe frequency between cells (p = 0.005). (D) smFISH images of a group of three HEK293T cells and two mutant cells that demonstrate the observed variation in CYP24A1 mRNA localization. Individual control POLR2A (magenta) and CYP24A1 mRNA transcripts (green) are shown along with the nuclear stain DAPI (blue) in the merged image. In HEK293T cell 1, CYP24A1 mRNA is absent from both the nucleus and the cytoplasm. HEK293T cell 2 has a high abundance of CYP24A1 mRNA in both the nucleus and cytoplasm with a high-intensity spot indicating high levels of ongoing transcription (white arrows). HEK293T cell 3 also has a discrete CYP24A1 mRNA spot in the nucleus indicating ongoing transcription (white arrow), but relatively fewer CYP24A1 mRNAs in the cytoplasm. In the mutant cells 1 and 2 CYP24A1 is also expressed in both the nucleus and the cytoplasm with high-intensity spots indicating high levels of ongoing transcription (white arrows) (E) Data for CYP24A1 in the wild-type HEK293T and mutant cell lines show no significant difference in mRNA localization. For (B), (C), and (E), the top and bottom whisker lines denote the maximum and minimum value in the dataset, respectively, the top and bottom of the box denote the 75th and 25th percentiles of the dataset, the median value (50th percentile) is indicated by the central line of the box plot, the x represents the mean value and the white circles indicate data points that are within the interquartile range. Scale bars = 10 μm.

Discussion

LC-MS/MS and immunoassay-based vitamin D analysis and VMR determination of 47 patients who presented at our metabolic and renal stone clinics and two children being investigated for nephrocalcinosis and calcium abnormalities suggested possible CYP24A1 loss-of-function, consistent with a HCINF1 diagnosis. A small group of patients with suspected CYP24A1 pathology due to abnormal biochemical profiles were absent of protein-coding mutations, which prompted further investigation. A lack of protein-coding mutations in CYP24A1 despite phenotypes was consistent with some other previous reports.^{(22, 23)} We extended our mutational analysis to include the 5′ UTR and 3′ UTR regions in the six patients because these loci are highly important in gene regulation independent of the coding sequence. Direct sequencing identified SNVs in the 3′ UTR in all six patients. The differing 3′ UTR SNVs in our cohort might explain some of the heterogeneous phenotypes reported in the literature in patients with HCINF1.⁽⁵⁾

We observed mRNA structural alterations in each patient with 3′ UTR variants. mRNA structural changes triggered by the presence of 3′ UTR sequence variants (or structural elements) have been shown to induce translational heterogeneity or impair translation completely by mRNA destabilization,⁽²⁴⁾ abnormal mRNA trafficking,^{(24, 25)} and/or reduced ribosome scanning efficacy.⁽²⁶⁾ Our in silico mRNA structural alterations driven by the described 3′ UTR variants may impair proper production of functional CYP24A1 protein resulting in inappropriate 1,25(OH)₂D concentration with a low/high VMR.

We did not observe MRE abnormalities in the 3′ UTR mutations in our cohort, which was the most obvious starting point for nonclinical studies. We observed no significant effect on CYP24A1 mRNA transcription in patients with 3′ UTR structural elements. The lack of evidence supporting transcriptional or posttranscriptional alteration observed in our cohort directed our new hypotheses toward a translational impairment, compromising CYP24A1 enzyme function giving rise to the elevated 1,25(OH)₂D and high/low VMR.^{(27, 28)} We observed a significant increase and/or retention of CYP24A1 in patients with misfolded mRNAs. Our data suggested an accumulation of a partially functional CYP24A1 protein, demonstrated by a somewhat normal catabolism of 25OHD into 24,25(OH)₂D but abnormal 1,25(OH)₂D catabolism.

Molecular investigation into the consequences of 3′ UTR mRNA structural alterations led to two new hypotheses that needed to be addressed: (i) CYP24A1 upregulation is the expected homeostatic response to increased serum calcium therefore increased protein on a western blot, whereas mRNA structural elements signal for an unknown but pathogenic posttranslational modification hindering protein function; and (ii) mRNAs are trafficked from the nucleus to subcellular regions where the subsequent protein will be required more rapidly; mRNA localization to specific regions within a cell provide regulation of protein expression but mRNA misfolding interferes with this trafficking process.⁽²⁹⁾ Given that CYP24A1 is functional in the inner mitochondrial membrane in 1,25(OH)₂D target cells and that some RNA species, particularly long noncoding RNAs, are known to act as structural components to the mitochondrial membrane,⁽³⁰⁾ it is possible that mRNA structural abnormalities physically anchor translational machinery and prevent the protein from proper localization. Improper localization affecting translational machinery could go some way to describe increased CYP24A1 expression with little effect on mRNA transcription as was observed. We investigated some but not all components of these hypotheses here using our newly developed CYP24A1 mutant model.

Although the CYP24A1 knockout mouse models can be useful in further understanding patients with complete loss-of-function CYP24A1 mutations,⁽³¹⁾ our patient cohort presented with hypomorphic mutations associated with compromised CYP24A1 function. Currently no models are commercially available that comprise CYP24A1 3′ UTR variants leading to mRNA structural alterations and partially functional CYP24A1. RNAfold demonstrated that the 3′ UTR deletions transfected into our newly generated CRISPR-Cas9 mutant cell line altered the CYP24A1 mRNA secondary structure. There was a nonsignificant decrease in catabolism of 25OHD to 24,25(OH)₂D but significantly decreased 1,25(OH)₂D clearance over 48 hours. Decreased catabolism rates were consistent with the patient phenotype of detectable 24,25(OH)₂D concentration and elevated 1,25(OH)₂D concentration due to partially functional CYP24A1. Our CRISPR-Cas9 mutant allows for future in vitro investigation into noncanonical CYP24A1 disease pathogenesis.

Using the mutant model we investigated the effect that 3′ UTR structural elements had on CYP24A1 cellular localization by novel single-cell visualization of individual CYP24A1 mRNA transcripts in vitro using smFISH. We observed significant CYP24A1 cell-to-cell expression and localization variability in HEK293T and mutants. The variability in CYP24A1 localization would have been undetected using routine mRNA quantification methods such as quantitative PCR (qPCR) that analyze total mRNA abundance. The cell-to-cell variability could be explained by the progression of the cell cycle, which has been shown to affect mRNA transcription rates⁽³²⁾ or fixing at different stages of the cell cycle. smFISH analysis indicated that alterations to the 3′ UTR likely had little to no effect on CYP24A1 localization in vitro because no significant difference was observed in comparison to wild-type HEK293T cells. This finding supports the lack of CYP24A1 mRNA transcription variability observed in our patient samples analyzed by dPCR. Although no significant localization or abundance differentiation was observed between the mutant and HEK293T cell lines, this work provides insight into the localization of CYP24A1 in cells, which has not been previously reported. CYP24A1 visualization can be applied on a patient-by-patient basis to assess the effect that different CYP24A1 mutations (both protein-coding and noncoding) have on mRNA stability, expression and cellular localization. Future human ex vivo plus in vitro studies using smFISH will allow for the accurate visualization of mRNA abundance and localization relating to structure–function relationships, which is a major advancement on current qPCR techniques in RNA biology. smFISH is therefore a compelling tool in investigating the mRNA lifecycle of CYP24A1.⁽³³⁾

This study investigated the role of mRNA misfolding in human disease. We identified a cohort of patients with CYP24A1-mediated abnormal calcium handling and/or HCINF1 phenotypes harboring noncoding genetic abnormalities. We expanded CYP24A1 analysis to include the UTRs, which revealed a possible mechanism for six patients who lacked coding sequence pathology. Further studies on this group revealed that whilst the amino acid sequence was unaffected (except for patient 6), the mRNA molecule itself was affected, importantly the key regulatory region 3′ UTR, which resulted in predicted three-dimensional structural changes that interfered with proper protein translation. Our findings provide insight into the potential effect mRNA structural abnormalities caused by sequence-dependent structural elements have on biological function. Though we could not fully describe the underlying mechanism linking mRNA structure with partially functional CYP24A1 protein activity, our work presents a pathway for employing ribosome frameshift profiling and/or protein sequencing to determine the core C-term region required for catabolizing 1,25(OH)₂D, which only differs from 25OHD by the additional OH group. Some mRNAs carry specific structural elements in their 3′ ends that cause ribosomes to slip and then readjust the reading frame.⁽³⁴⁾ The frameshift results from a change in the reading frame by one or more bases in either the 5′ (−1) or 3′ (+1) directions during translation.⁽³⁴⁾ The newly developed CRISPR-Cas9 mutated HEK293T cell line could provide the foundation for large scale in vitro studies and will continue to support our understanding of vitamin D metabolism in patients with novel 3′ UTR mutations. The findings of this research provide a framework that can be used to better understand the molecular basis of pathogenesis in patients lacking protein-coding region abnormalities.

Acknowledgments

The Michael Davie Research Foundation and Norwich Medical School PhD Programme funded this study. EW is funded by the Bone Cancer Research Trust. WDF and DG are supported by The Difference Campaign. We thank the John Innes Centre bioimaging facility and staff for their contribution to this work and we thank Matthew Jefferson, Yingxue (Sophia) Wang and Gabriella Oliver-Wilkins for technical support. We are indebted to the patients for their participation in this study.

Authors' roles: Study conception and design: WDF and DG. Experiments: NB, SD, YZ, RP, IP, EW, and JCYT. Data analysis and interpretation: NB, RP, IS, YD, WDF, and DG. Samples, medical classification and pathology: BL, AC, AK, LP, HM, and WDF. Wrote the manuscript: NB and DG. Revised and approved the final manuscript: all authors.

Conflicts Of Interest

The authors declare no competing interests.

Open Research

Peer Review

The peer review history for this article is available at https://publons.com/publon/10.1002/jbmr.4769.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.