Functional Analysis of Genetic Variation in the SECIS Element of Thyroid Hormone Activating Type 2 Deiodinase

Abstract

Context

Thyroid hormone is important for normal brain development. The type 2 deiodinase (D2) controls thyroid hormone action in the brain by activating T4 to T3. The enzymatic activity of D2 depends on the incorporation of selenocysteine for which the selenocysteine-insertion sequence (SECIS) element located in the 3′ untranslated region is indispensable. We hypothesized that mutations in the SECIS element could affect D2 function, resulting in a neurocognitive phenotype.

Objective

To identify mutations in the SECIS element of DIO2 in patients with intellectual disability and to test their functional consequences.

Design, Setting, and Patients

The SECIS element of DIO2 was sequenced in 387 patients with unexplained intellectual disability using a predefined pattern of thyroid function tests. SECIS element read-through in wild-type or mutant D2 was quantified by a luciferase reporter system in transfected cells. Functional consequences were assessed by quantifying D2 activity in cell lysate or intact cell metabolism studies.

Results

Sequence analysis revealed 2 heterozygous mutations: c.5703C>T and c.5730A>T, which were also present in the unaffected family members. The functional evaluation showed that both mutations did not affect D2 enzyme activity in cell lysates or intact cells, although the 5730A>T mutation decreased SECIS element read-through by 75%. In the patient harboring the c.5730A>T variant, whole genome sequencing revealed a pathogenic deletion of the STXBP1 gene.

Conclusions

We report on two families with mutations in the SECIS element of D2. Although functional analysis showed that nucleotide 5730 is important for normal SECIS element read-through, the two variants did not segregate with a distinct phenotype.

Thyroid hormone (TH) is crucial for normal brain development, which has been illustrated by the severe consequences of untreated congenital hypothyroidism (1). TH regulation and action is controlled by TH transporters, deiodinases, and nuclear T3 receptors. Disrupted TH signaling in the brain is associated with neurocognitive impairment, which is exemplified by disorders caused by mutations in the TH transporter MCT8 or in the nuclear T3 receptor-α (2–4).

In the brain, bioactive T3 is locally generated by deiodination of its precursor T4. This process is catalyzed by type 2 deiodinase (D2) (5, 6). Expression of T3-sensitive genes in neurons is dependent on T4 to T3 conversion by D2 in the glial compartment (7). The commonly occurring Thr92Ala single nucleotide polymorphism (SNP) in DIO2, the gene encoding D2, has been linked with changed TH signaling in the brain, although the exact role of this SNP remains to be determined (8–12).

Given its physiological role in TH signaling in the brain, we hypothesized that mutations affecting D2 function would result in a neurocognitive phenotype (13). Previously, Zevenbergen et al. (14) identified two novel heterozygous nonsynonymous mutations in the coding sequence of D2 in a cohort of patients with unexplained intellectual disability (ID). However, functional studies have suggested that these mutations do not affect D2 enzyme activity and are, in fact, rare, but harmless, variants.

Deiodinases are selenoproteins, because they contain a selenocysteine (Sec) residue in the center of their catalytic domain (15, 16). Thus, the presence of a Sec is important for normal enzymatic activity of D2 (17). The Sec is encoded by a UGA triplet, which is an opal stop codon that usually terminates translation. However, if the 3′ untranslated region (UTR) of the mRNA contains a selenocysteine-insertion sequence (SECIS) element, this particular stem loop structure is critically involved in specifying the UGA codon for insertion of Sec (18–21). The SECIS element requires binding with SECIS binding protein 2 (SBP2), which in turn recruits a specific elongation factor, leading to the incorporation of Sec (19). The crucial role of Sec in deiodinases was shown in patients with SPB2 deficiency, who have a multisystem disorder that includes a developmental delay and impaired deiodinase function, resulting in abnormal thyroid function test results (elevated serum T4 and decreased serum T3 concentrations) (22, 23). In the present study, we explored whether mutations in the SECIS element of D2 will result in impaired neurologic development.

We hypothesized that mutations in the SECIS element of D2 could affect the efficiency of Sec incorporation into D2, thereby reducing D2 activity and, ultimately, leading to impaired brain development. Therefore, we screened the D2 SECIS elements in a cohort of patients with unexplained ID (14). We identified two heterozygous mutations in the SECIS element of D2 and tested their functional consequences for D2 activity.

Patients and Methods

Patients

Patients were selected from the TOP-R (Thyroid Origin of Psychomotor Retardation) study, a cohort consisting of 946 subjects with unexplained ID (IQ <50) in whom extensive profiling of TH parameters was performed (24).

Serum analysis

Serum T4, T3, free T4 (FT4), and TSH were measured using chemiluminescence assays (Vitros ECI; Ortho-Clinical Diagnostics Inc., Rochester, NY). rT3 was measured using a commercial RIA (Immunodiagnostic Systems, Scottsdale, AZ).

Screening and genetic analysis

Because the biochemical phenotype in humans with DIO2 mutations is unknown, we used nonstringent criteria to select subjects for sequence analysis to maximize the chance of identifying such mutations. The selection criteria have been previously described (14). We based the selection criteria on the role of D2 in converting T4 to T3 and rT3 to T2, the serum TH abnormalities in patients with SBP2 mutations, and those in D2 knockout mice (13, 22, 25). For inactivating mutations in DIO2, the following criteria were used: (i) a T3/rT3 ratio <20th percentile and a FT4xlnTSH product >80th percentile; or (ii) a T3/T4 ratio <20th percentile; and, for possible activating mutations, (iii) a T3/T4 ratio >90th percentile and a TSH level <10th percentile in the cohort. Applying these arbitrary, nonstringent criteria, 387 patients from our cohort were selected for DIO2 SECIS sequencing.

The SECIS element of DIO2 (NM_013989.3) was analyzed using the following primers: CCAGTTTTGTTTAGTTTTGCATCA (forward, 5′-3′) and CACATAGCACTCAGCACCAA (reverse, 5′-3′). Oligonucleotides were synthesized by Invitrogen (Bleiswijk, Netherlands). PCR and sequence reactions were performed using BaseClear (Leiden, Netherlands). For the PCR reactions of family members, we used the Taq polymerase, deoxyribose adenine triphosphate and buffers from Qiagen (Venlo, Netherlands). The SeciSearch and Mfold programs were used to analyze the effects of mutations on the structure of the SECIS element (26, 27).

DNA constructs and mutagenesis

The cloning of the luciferase SECIS reporter construct in pcDNA3.1 has been previously described (28). This construct contains a wild-type (WT) or mutant SECIS element that is placed downstream of a luciferase coding sequence, where a TGC, coding for Cys, has been replaced by TGA. An active luciferase is created only when the UGA has been recoded as a Sec. This construct is used to quantify the recoding efficiency of the D2 SECIS element. The identified mutations (5703C>T and 5730 A>T) were introduced in this construct using the QuickChange Site-Directed Mutagenesis protocol, according to the manufacturer (Stratagene, Amsterdam, The Netherlands). The artificial 5730A>C and 5679 G>A mutations were introduced using the same protocol. The 5679 G>A is known to inactivate the SECIS element and was used in our experiments as a positive control (29).

To determine the effect of SECIS mutations on D2 activity, we introduced the 5703C>T and 5730 A>T patient variants and the artificial mutations 5679G>A and 5730A>C into the previously described pUHD10-3-D2 expression construct using site-directed mutagenesis. This construct contains the human D2 coding sequence and a part of its 3′UTR (∼2.5 kb), including the SECIS element.

Because the efficiency of Sec incorporation is inversely related to the distance between the SECIS element and the UGA codon (28), we also generated a construct in which the SECIS element was placed directly behind the D2 coding region. Therefore, we used the two Afe1 restriction sites to delete ∼2 kb of the 3′UTR (∼2 kb) from the D2–pUHD10-3 construct. This product was subcloned in pcDNA3 using the HindIII and Xho1 restriction sites. Mutagenesis was performed to introduce the 5730A>T mutation.

SECIS element recoding efficiency assay

COS1 cells were cultured in 96-well plates using 100 µL DMEM/F12 supplemented with 9% heat-inactivated fetal bovine serum, 1% penicillin-streptomycin, and 100 nM Na₂SeO₃. At 70% confluence, the cells were transiently transfected with 25 ng of the luciferase D2 SECIS element reporter construct (WT or mutant) and 25 ng of a control Renilla luciferase reporter. After 48 hours, the cells were washed once with Dulbecco’s PBS and lysed. Luciferase and Renilla signals were subsequently measured in 25 µL of the cell lysate using the Dual-Glo Luciferase Assay System (Promega, Leiden, Netherlands) and a Victor 2 multilabel plate reader (Perkin Elmer, Rotterdam, Netherlands) according the manufacturers’ protocols. Firefly luciferase activity was normalized to Renilla luciferase activity to adjust for transfection efficiency.

D2 activity in cell lysates

COS1 cells were cultured at 37°C and 5% carbon dioxide in 75-cm² flasks with DMEM/F12 (Life Technologies, Bleiswijk, Netherlands) supplemented with 9% fetal calf serum (Sigma-Aldrich, Zwijndrecht, Netherlands), 2% penicillin-streptomycin, and with or without 100 nM Na₂SeO₃. At confluence, the cultured cells were split and seeded in six-well plates. At 70% confluence, the cells were transiently transfected with 500-ng empty vector, WT, or mutant D2 cDNA using X-treme Gene 9 Transfection Reagent (Roche Diagnostics, Almere, Netherlands).

Two days after transfection, the cells were harvested, and D2 activity was measured in lysates using HPLC, as described previously (2). The protein levels were determined using the method of Bradford (30). Transfection efficiency was similar in all conditions as assessed in preliminary experiments performed by cotransfection with human GH cDNA and measurement of GH concentrations in the medium (14). Unlabeled iodothyronines were obtained from Sigma-Aldrich. [¹²⁵I]T4 and [¹²⁵I]rT3 were prepared as previously described (31).

D2 metabolism in intact cells

COS1 cells were cultured in 24-well plates using 500 µL of DMEM/F12 supplemented with 9% fetal calf serum and 100 nM Na₂SeO₃. At 70% confluence, the cells were transiently transfected with 200 ng of WT or mutant D2 cDNA plus 80 ng of the trans-activator pUHD15 (32). Two days after transfection, the cells were incubated for 24 hours with 1 nM [¹²⁵I]T4, and T4 metabolism was analyzed using HPLC, as described previously (2).

Whole genome sequencing and mutation confirmation

Whole genome sequencing (software version 2.5.0.37) was performed as described by Drmanac et al. (33). In brief, the human genome sequencing procedures included DNA library construction, DNA nano-ball (DNB) generation, DNB array self-assembling, combinatorial probe-anchor ligation-based sequencing, and imaging. Imaging data analyses included base calling, DNB mapping, and sequence assembly. Reads were mapped to the National Center for Biotechnology Information reference genome, build 37. Variants were annotated using National Center for Biotechnology Information build 37 and dbSNP build 137. Data were provided as lists of sequence variants (SNPs and short indels) relative to the reference genome. Analysis of the parallel sequencing data was performed using the Complete Genomics analysis tools (cga tools, version 1.8.0, build 1; available at: http://www.completegenomics.com/sequence-data/cgatools/) and TIBCO/Spotfire, version 7.0.1 (available at: http://spotfire.tibco.com/). De novo single nucleotide variants were identified using the Complete Genomics cgatools “calldiff” program as described by Gilissen et al. (34). Homozygous variants and compound heterozygous variant pairs were identified using a custom-made Python script called “multiple genome analysis” using the Complete Genomics cgatools script “listvariants” and “testvariants.”

Structural variant analysis was performed using a custom-made Python script called “Multiple_Genome_Large_Variant_Analysis.” Copy number variations and structural variants were reported by Complete Genomics on the basis of read-depth deviations and discordant read pairs, respectively. Variants were filtered based on the inheritance pattern and a Complete Genomics baseline frequency of 0. The called large variants were ranked on mate pair count. Control genomes in the Structural Variation Baseline Genome Set comprised 52 baseline genomes used by Complete Genomics and 588 Wellderly samples (Scripps Wellderly Genome Resource, The Scripps Wellderly Study, La Jolla, CA; December 2015; funded by Scripps Health and National Institutes of Health/NCATS UL1 TR00114).

The identified heterozygous deletion, which includes a large part of the STXBP1 gene (NM_001032221.3) was confirmed by Sanger sequencing of the PCR product obtained using specific primers targeting the flanking regions of the deletion [ATACAAATGTCGCAGCAGCAT (forward, 5’-3’) and CACCTGGCCGGAAATACTTA (reverse, 5’-3’)]. Control PCR reactions were designed to amplify the region in front [control 1; CAAATCACGTGAGGTCAGGA (forward, 5’-3’) and AGGCCACATTTTGCTCATCT (reverse, 5’-3’)] and behind [control 2; CGGAGGCGCGTGAATCAACATGGC (forward, 5’-3’) and ATTACAGGCGTGAGCCATCGT (reverse, 5’-3’)] of the expected deletion region.

Ethical considerations

The ethical committee of the Erasmus Medical Center, Rotterdam, approved all clinical and genetic studies (approval no. MEC-2006-177). The parents and unaffected brother of the patient provided written informed consent.

Statistical analysis

All results are presented as the mean ± SEM of two to four independent experiments performed in duplicate. Statistical analysis was performed in GraphPad Prism, version 5.01 for Windows (GraphPad Software, San Diego, CA). The statistical significance of WT D2 compared with the various mutants in the luciferase, cell lysate, and intact cell assays was determined using one-way ANOVA, followed by a Bonferroni post hoc test.

Results

Sequence analysis of the SECIS element of DIO2 revealed heterozygous mutations in patient 760 (c.5703C>T) and patient 568 (c.5730A>T; Fig. 1A and 1B). Subsequent sequence analysis also revealed the presence of these mutations in unaffected relatives of both patients (Fig. 1A and 1B). No cosegregation of thyroid function test results was observed in the subjects carrying a mutation (Fig. 1C and 1D). The SECIS element mutations have not been described in the 1000 Genomes Catalog (35) or in the National Heart, Lung, and Blood Institute Exome Variant Server. Although the mutations were not localized in the three clusters of highly conserved nucleotides that are elementary for SECIS element function (Fig. 1E), a prediction tool (SeciSearch) indicated that c.5703C>T and c.5730A>T putatively change the stem-loop structure of the SECIS element (data not shown) (26). The 5679G>A is a known artificial mutation localized within the AUGA motif of the SECIS element that completely abolishes SECIS element activity (29).

SECIS element recoding efficiency assay

To quantify the SECIS element read-through in WT and SECIS mutants, we used an expression vector containing a WT or mutant D2 SECIS element placed after a luciferase gene. In this construct, a TGC, coding for Cys, has been replaced by TGA. An active luciferase is created only when the UGA is recoded as a Sec. Decreased SECIS function will result in less efficient incorporation of Sec and, consequently, in decreased luciferase activity. The SECIS element read-through in the 5703C>T mutant was similar to the WT (Fig. 2A). However, SECIS element read-through decreased by ∼75% in the 5730A>T mutant (Fig. 2A). The 5679G>A artificial mutant used as a positive control lacked any read-through (Fig. 2A). Next, we mutated the 5730A>T mutation back to WT, resulting in 100% activity, confirming the specificity of this mutation (Fig. 2A). Finally, to evaluate the relevance of nucleotide 5730, we created the 5730A>C mutant. This mutant showed an even greater decrease in SECIS read-through compared with the patient mutation (Fig. 2A). Together, these data suggest that the nucleotide at position 5730 in the D2 SECIS element appears relevant for SECIS element read-through in an artificial SECIS element recoding efficiency assay.

D2 activity in cell lysates and intact cells

Next, we studied the effects of the SECIS mutants on D2 activity. In cell lysates, we observed no substantial difference in D2 activity between WT and SECIS mutants (Fig. 2B). As anticipated, the 5679G>A mutant showed almost no activity in all experiments. Because D2 activity depends on Se availability, we also tested the function of the SECIS variants under Se-deficient culture conditions. In Se-deficient cultures, the total D2 activity in cell lysates decreased by ~50%. However, no substantial difference was found between WT and mutants (Fig. 2C).

Next, we used the same approach in intact cell metabolism studies. In normal and Se-deficient culture conditions, no difference in function between WT and SECIS mutants was observed (Fig. 2D and 2E). Also, kinetic studies revealed no differences in apparent Michaelis constant and maximum velocity values between WT D2 and SECIS mutants (data not shown). As luciferase activity was more affected by the artificial 5730A>C mutant than by the 5730A>T mutant, we also assessed 5730A>C mutant D2 activity. In cell lysates, a nonsignificant trend was observed toward a decreased D2 activity compared with the WT (Fig. 2F). However, this could not be confirmed in the intact cell metabolism assay (Fig. 2G).

Interassay variability by different constructs

Because we observed discordant results for the 5730A>T mutant in the SECIS element read-through assay vs D2 activity, we aimed to rule out trivial differences between these constructs. In the luciferase construct, the SECIS element is placed directly after the coding sequence. In contrast, the D2-pUHD10-3 vector contains ∼2.5 kb of the DIO2 3′UTR proximal of the SECIS element. Previously, it was shown that the distance between UGA and SECIS is inversely related to deiodinase activity (36). Therefore, we removed most of the 3′UTR (∼2 kb) in the D2–pUHD10-3 constructs.

In accordance with previous studies, we observed greater total D2 activity in lysates from cells expressing the short vs the long 3′UTR. However, no differences were seen between WT and 5730A>T mutant D2 activity (Fig. 3A). Similar findings were obtained from the intact cell metabolism studies (Fig. 3B). To rule out vector differences, we subcloned D2 in pcDNA3 (matching the luciferase pcDNA3 vector) and repeated the experiments. However, no differences between WT and SECIS mutants in both cell lysates and intact cell metabolism assays were detected (Fig. 3C and 3D).

Whole genome sequencing

Given the functional effect of the 5730A>T in the SECIS element read-through assay, we hypothesized that the clinical features of patient 568 could be explained either by another genetic mutation or by an additional complementary mutation in the D2 signaling pathway. Therefore, we performed whole genome sequencing in that family (unaffected father, unaffected mother, unaffected brother, and index patient). Bioinformatic analysis revealed a deletion of ~114 kb in the patient (chr9:130384058-130498502; 13 mate pair counts) containing part of the STXBP1 gene but not in the other family members nor in a control data set (Fig. 4A). In the patient, the coverage in the deletion region was 50% compared with that in the surrounding regions.

To confirm the presence of this deletion, we performed PCR analysis using primers directed against the flanking regions of the deletion. This resulted in an amplicon of 500 bp in the patient but not in the family members (Fig. 4B). Sequencing of this amplicon refined the borders of the deletion, demonstrating a deletion of 114,372 bp (chr9:130384138-130498510) affecting seven transcripts, most importantly STXBP1. Revisiting the medical history and clinical features of patient 568 revealed a strong similarity to other described cases with STXBP1 mutations. At the age of 7 years, epileptic seizures had started and had remained refractory despite three antiepileptic drugs. Most seizures were partial frontal seizures. She had a stationary developmental age of 11 months, with no eye contact or language skills. However, she was able to walk assisted for a few meters and to hold a toy. Physical examination revealed extrapyramidal symptoms (chorea-athetosis, tremors, and rigidity), spasticity, and an overtly ataxic gait. Brain MRI performed at the age of 20 years showed a structurally normal cerebrum and cerebellum, with a slightly thin corpus callosum and slight hyperostosis of the skull. Electroencephalography at age 7 years had revealed left temporal slowness and sharp activity, as well as frontal slow activity, but no electrographic seizures during the registration.

Discussion

Because D2 is important for the activation of T4 to T3 in the brain (1), and T3 is necessary for normal neurologic development (37), we hypothesized that diminished D2 activity might result in neurocognitive impairment. Previously, we reported on the existence of rare, but harmless, variants in the coding sequence of D2 in a cohort of patients with ID (14). Because the SECIS element of D2 is crucial to recode the UGAs contained in the DIO2 mRNA sequence into a Sec, we hypothesized that mutations in the D2 SECIS element could affect D2 function by changing the catalytic activity. Thus, such mutations could affect T3 availability in the brain and, thereby, result in neurodevelopmental delay. Therefore, we performed a genetic analysis of the D2 SECIS element in a large cohort of patients with ID. This revealed heterozygous variants in the D2 SECIS element in 2 patients.

However, both D2 SECIS genetic variants were also present in the unaffected family members and did not segregate with the clinical or biochemical phenotype. Functional analysis confirmed that the 5703C>T mutation did not interfere with normal SECIS element function and D2 activity. In contrast, we noted a 75% decrease in SECIS element read-through in the 5730A>T mutant. However, D2 activity was not affected as assessed by in vitro D2 activity assays in cell lysates or intact cells under both Se-replete and Se-deplete conditions. We did not investigate whether the interaction between SBP2 and D2 SECIS element was disrupted, because the D2 activity was not altered. The artificial 5730A>C mutant showed an even more pronounced decrease in SECIS element read-through. Furthermore, in cell lysates, in which the synthetic cofactor DTT was used, we observed a decrease, although statistically insignificant, in the D2 activity of the 5730A>C mutant.

As it has been shown that spacing between the UGA codon and the SECIS element interferes with the efficiency of Sec incorporation (36), it is possible that the SECIS element is more vulnerable to mutations when no 3′UTR is present in the constructs. We tested this potential explanation by changing the various constructs in containing a 3′UTR of similar size. However, this did not result in a different D2 activity between the WT and 5730A>T mutant in the cell lysates or intact cells. Although the D2 activity assessed in intact cell metabolism assays likely represents the physiology more adequately than that in the cell lysates, the supply of substrate (T4) to the intracellular compartment is dependent on transporters. Therefore, if transport is the rate-limiting step in this system, small differences in D2 activity might go unnoticed in this system.

Our data suggest that position 5730 is relevant for the SECIS element under certain conditions. However, although mutations at this position affected read-through in an artificial SECIS element recoding efficiency assay, they did not seem to be related to a clinically relevant impaired D2 function. Also, the 5703C>T mutation did not show any functional consequences, and this mutation can, therefore, be regarded as a harmless, rare variant.

Given the discrepancy between testing systems, we used whole genome sequencing to rule out a secondary independent mutation in the D2 signaling pathway, which, together with the 5730A>T D2 SECIS mutation, could explain the phenotype of patient 568. The whole genome sequencing analysis revealed a deletion involving the STXBP1 gene. Reevaluation of the clinical phenotype, brain imaging studies, and neurophysiological tests, the features were consistent with the syndrome caused by mutations (including deletions) of the STXBP1 gene (38, 39).

In conclusion, our functional studies of the two heterozygous genetic variants in the SECIS element of D2 indicate that nucleotide 5730 is relevant for normal SECIS element read-through but that mutations at this position are unlikely to result in a clinically relevant phenotype. Our studies reinforce the notion that functional testing is crucial to linking rare genetic variants to the presence of disease. Future studies might reveal whether mutations that affect D2 function are associated with a distinct phenotype.

Abbreviations:

Abbreviations:
 
• D2

  type 2 deiodinase

 
• DNB

  DNA nano-ball

 
• ID

  intellectual disability

 
• SBP2

  selenocysteine-insertion sequence binding protein 2

 
• Sec

  selenocysteine

 
• SECIS

  selenocysteine-insertion sequence

 
• SNP

  single nucleotide polymorphism

 
• TH

  thyroid hormone

 
• UTR

  untranslated region

 
• WT

  wild-type

Acknowledgments

In the final stage of writing our report, Professor T. J. Visser suddenly and unexpectedly died. Professor Visser contributed largely to the design of the experiments and discussions thereof. We highly value his contributions to the field, and we miss a great scientist, mentor, and friend. While deceased contributors are rightfully recognized and acknowledged, they cannot be added posthumously to an article’s byline. We thank Ramazan Buyukcelik for DNA isolation and all the patients and their family members for participation in this study.

Financial Support: We acknowledge financial support from an Erasmus University Fellowship (to W.E.V.).

M. Rispens’ current affiliation is Coordan, Amsterdam, Netherlands. D. Venter’s current affiliation is the Department of Neuropathology, Royal Prince Alfred Hospital, Camperdown, New South Wales, Australia, and School of Medicine, The University of Wollongong, New South Wales, Australia.

Disclosure Summary: The authors have nothing to disclose.

References