Metabolomic Profiling of Body Fluids in Mouse Models Demonstrates that Nuclear Magnetic Resonance Is a Putative Diagnostic Tool for the Presence of Thyroid Hormone Receptor α1 Mutations

Abstract

Background:

Resistance to thyroid hormone alpha (RTHα) is a rare genetic disease due to mutations in the THRA gene, which encodes thyroid hormone receptor alpha 1 (TRα1). Since its first description in 2012, 46 cases of RTHα have been reported worldwide, corresponding to 26 different mutations of TRα1. RTHα patients share some common symptoms with hypothyroid patients, without significant reduction in thyroid hormone level. The high variability of clinical features and the absence of reliable biochemical markers make the diagnosis of this disease difficult. Some of these mutations have been recently modeled in mice.

Methods:

In our study, we used four different mouse models heterozygous for frameshift mutations in the Thra gene. Two of them are very close to human mutations, while the two others have not yet been found in patients. We characterized the metabolic phenotypes of urine and plasma samples collected from these four animal models using an untargeted nuclear magnetic resonance (NMR)-based metabolomic approach.

Results:

Multivariate statistical analysis of the metabolomic profiles shows that biofluids of mice that carry human-like mutations can be discriminated from controls. Metabolic signatures associated with Thra mutations in urine and plasma are stable over time and clearly differ from the metabolic fingerprint of hypothyroidism in the mouse.

Conclusion:

Our results provide a proof-of-principle that easily accessible NMR-based metabolic fingerprints of biofluids could be used to diagnose RTHα in humans.

Introduction

Thyroid hormones (TH: thyroxine [T4]; triiodothyronine [T3]) exert a broad influence on development and homeostasis (1). The nuclear receptors called thyroid hormone receptor alpha 1 (TRα1), TRβ1, and TRβ2 are T3-dependent transcription factors. They are encoded by the THRA and THRB genes, which are expressed in many cell types (2). Mutations of these genes cause two distinct genetic diseases, respectively, called resistance to thyroid hormone alpha (RTHα) and RTHβ, characterized by tissue resistance to TH in patients heterozygous for missense and frameshift mutations. Because THRA and THRB have different expression patterns and properties, the two genetic diseases have different outcomes (3 –5). In RTHβ, the feedback regulation of the hypothalamus/pituitary/thyroid axis is impaired resulting in an increase in serum T3 and thyrotropin (TSH) content. Patients with RTHβ have been found in more than 400 families (6), and sporadic cases are usually recognized when it appears that a treatment with pharmacological doses of thyroxine is unable to normalize the TSH level.

By contrast, RTHα has been reported in 2012 only, the first case being discovered by exome sequencing (7). Up to now, only 46 cases, all heterozygous, have been reported in 26 different families (3,8,9). All known mutations alter the ligand-binding domain of the receptor. While some missense mutations only reduce the affinity of TRα1 for T3, other missense and frameshift mutations modify or eliminate the C-terminal helix 12, preventing the recruitment of coactivators. This turns TRα1 into a constitutive transcription repressor, with dominant-negative properties, explaining the genetic dominance of the disease. The apparent rarity of RTHα probably stems from the difficulty of diagnosis. The genetic disease resembles congenital hypothyroidism, with variable and sometimes severe manifestations, including short stature and mental retardation, but without specific signs and symptoms or biochemical alterations. T4, T3, and TSH levels usually remain within the reference range, although a low T4/T3 ratio is frequent. The inability to quickly recognize patients with RTHα is regrettable because the developmental consequences of the disease can be greatly reduced by an early therapeutic intervention, notably when helix 12 is intact. Motor skills and body growth have been restored in a young patient treated with an excess of thyroxine to overcome tissue resistance (10). Two adult patients with normal adult stature also received thyroxine treatment during childhood by chance (11).

Metabolomic profiling has proven to be useful in diverse fields, including in diagnostics (12,13). Two technologies are used to analyze metabolic profiles: nuclear magnetic resonance (NMR) spectroscopy and mass spectrometry. NMR is less sensitive, but robust, and probably more suitable for future routine use in clinics (14). In this work, we exploit a series of mouse models of RTHα, which have recently been characterized (15) to assess the capacity of ¹H-NMR analysis of body fluids to recognize the presence of Thra mutations. Our data suggest that this method is a promising strategy for RTHα diagnostics in humans.

Materials and Methods

Animals

The research project was approved by a local ethics committee and subsequently authorized by the French Ministry of Research (license #6711), and carried out in accordance with the European Community Council Directives of September 22, 2010 (2010/63/EU), regarding the protection of animals. Four mouse models were used (Thra^S1/+ , Thra^S2/+ , Thra^L1/+ , Thra^L2/+ ) heterozygous for a Thra mutation created by CRISPR/Cas9 editing (15). Their genetic background had a major contribution of the C57Bl/6 strain. Wild-type male littermates without mutation were used as controls.

Hypothyroidism was induced by giving solid food containing propyl-thio-uracil to 7-week-old mice for a duration of 14 days to prevent TH production. We verified that this treatment results in sixfold reduction in the circulating level of T3. In one group of mice, T3 treatment was performed for the last four days of propylthiouracil (PTU) treatment using daily intraperitoneal injections (20 mg/kg/day T3 + 200 mg/kg/day T4). The study was performed on 29 C57BL/6J male mice (9 hypothyroid mice, 10 hypothyroid mice that received TH, and 10 control).

Sample collection

Eighty-eight male mice (12 Thra^S1/+ , 10 Thra^S2/+ , 10 Thra^L1/+ , 12 Thra^L2/+ , and 44 wild-type littermates) were followed for six months. Urine samples were collected at three and six months of age, and plasma samples were collected in heparin-coated tubes at six months. All samples were stored at −80°C until NMR analysis. Urine samples were thawed at room temperature and then centrifuged for five minutes at 5000 g. Two hundred microliters of the supernatant was mixed with 300 μL of water and 50 μL of buffer solution (1.5 M KH₂PO₄, 2 mM NaN₃, 0.1% trimethylsilyl propionate (TSP) pH 7.4 in D₂O). Five hundred fifty microliters of the mixture was then transferred into a 5 mm NMR tube. Plasma samples were thawed at room temperature. Coagulated particles were removed by centrifugation at 5000 g during five minutes. Two hundred microliters of the liquid phase was mixed with 100 μL of water and 300 μL of buffer (0.142 M Na₂HPO₄, 0.1% NaN₃, 0.1% TSP, pH7.4, in 20% D₂O).

NMR spectroscopy

All NMR experiments were performed on a Bruker Avance III spectrometer operating at 600 MHz (proton resonance frequency) equipped with a 5 mm TCI cryoprobe, and a cooled sample changer (4°C). The temperature was controlled at 27°C for urine samples and 37°C for plasma samples. Standard one-dimensional ¹H-NMR spectra were acquired with the “noesygppr1d” pulse sequence for each urine and plasma sample. A total of 128 transient free induction decays (FIDs) were collected for each experiment with a spectral width of 20 ppm. The acquisition time was set to 1.5 seconds with a relaxation delay of 2 seconds. The noisy mixing time was set to 10 ms. All FIDs were multiplied by an exponential function corresponding to a 0.3 Hz line-broadening factor, before Fourier transformation. ¹H urine spectra were referenced to the TSP signal (δ = −0.016 ppm at pH 7.4), and plasma samples were calibrated with respect to the α-glucose anomeric proton signal (δ = 5.23 ppm). Spectra were manually phased and corrected for baseline using Topspin 3.5 (Bruker). A data matrix was extracted from ¹H-1D-NMR profiles for multivariate statistical analysis using the Statistics toolbox of AMIX (Bruker). Spectra were integrated from 0.5 to 9.5 ppm at a step of 0.001 ppm, excluding the region of residual water at 4.72–4.92 ppm for urine and 4.4–5.05 ppm for plasma. Icoshift correction was applied to perform an alignment of peaks within manually set chemical shift ranges only for urine samples. To reduce the effect of dilution, the probabilistic quotient normalization method was implemented for urine samples (16). Plasma samples were normalized on their total sum of intensities. An additional series of quality control samples for each biofluid was included to evaluate the quality and the reproducibility of the NMR data acquisition.

In addition, 1D ¹H-NMR CPMG and 2D NMR experiments (¹H–¹³C HSQC, ¹H–¹H TOCSY, and ¹H J-resolved experiments) were recorded on a subset of samples to achieve structural assignment of the metabolite signals. Metabolite identification was performed using academic spectral databases [HMDB (17) and MMMDB (18)], as well as proprietary databases (NMR Suite v. 7.1, Chenomx, Inc., Edmonton, Canada; BBIOREFCODE, Bruker GmbH).

Statistical analysis

Unsupervised and supervised statistical multivariate analyses were conducted to build models for sample classification and to extract group-specific metabolic signatures. Orthogonal partial least-square discriminant analysis (OPLS-DA) (19,20) was performed for sample class discrimination using SIMCA 1 (Umetrics, Umea, Sweden) with variables scaling according to the Pareto method. The goodness-of-fit parameters R² and Q², which relate to the explained and predicted variances, respectively, were used to evaluate OPLS-DA model performance. Resampling the models under the null hypothesis was then performed 1000 times to validate model performance (21). Analysis of variance (ANOVA) of the cross-validated residuals' (CV-ANOVA) p-value was also used to assess the significance of the models. Univariate ANOVA (22) coupled with Hochberg–Benjamini correction of the p-value was used on clusters of variables, which were obtained from the automated binning procedure named statistical recoupling of variables (23).

Results

Metabolomic analysis segregates two types of frameshift mutations

We studied four mutations of Thra (15), which introduce a translation frameshift, eliminating the C-terminal helix, which normally interacts with transcription coactivators (Fig. 1A). E395fs401X and E395fs406X mutations are +1 frameshifts, introducing stop codons few nucleotides downstream to the mutated codon, and shortening the protein (Thra^S1/+ and Thra^S2/+ mice; S for small) as in some severe cases of RTHα found in patients. This results in a complete loss of transactivation capacity and a marked dominant negative activity in heterozygous mice. These mutations do not modify the circulating level of T4, and only slightly increase the T3 level. However, due to tissue resistance, they cause alterations of development reminiscent of hypothyroidism (15). The two other mutations, E395fs485X and K389fs479X, are +2 frameshifts, which have no known equivalent in human patients, and result in the production of an amphigoric protein of high molecular weight (Thra^L1/+ and Thra^L2/+ mice; L for large), which has little dominant negative activity. Extensive phenotype analysis (15) previously showed that Thra^S1/+ and Thra^S2/+ mice display several developmental traits present in RTHα patients, while Thra^L1/+ and Thra^L2/+ mice have an almost normal phenotype.

FIG. 1.

Experimental design. (A) Amino acid sequence of the C-terminus of TRα1. Two types of frameshift generate either a shortening (Thra^S1 , Thra^S2 ) or an extension (Thra^L1 , Thra^L2 ) of the amino acid sequence. In both cases, the C-terminal helix required for interaction for coactivators is lost. However, only shortened proteins possess a significant dominant negative activity and impair TH signaling in heterozygous mice. (B) Time line of the analysis. Urine was first collected in three-month-old animals. Urine and plasma were both collected from the same animals at the end of the experiment, three months later. NMR, nuclear magnetic resonance; TH, thyroid hormones; TRα1, thyroid hormone receptor alpha 1. Color images are available online.

Mice urine was collected at two time points (three and six months old), and blood at six months (Fig. 1B). Wild-type littermates without mutation were used as controls. ¹H-NMR spectra were recorded, providing for each sample of biofluid a complex metabolic profile of urine and plasma, where 47 and 32 metabolites can be unambiguously identified, respectively (Supplementary Fig. S1).

We first asked whether the four groups of mutant mice could be considered a single group in subsequent statistical analysis, or whether different mutations should be considered separately. We thus addressed whether metabolic profiling could differentiate Thra^S1/+ , Thra^S2/+ , Thra^L1/+ , and Thra^L2/+ mice. A four-class OPLS-DA of the metabolic profiles was performed to evaluate the impact of the different Thra mutations on urine metabolome (Fig. 2A). Although the four groups were not separated, we noted that mutants with short receptors (Thra^S1/+ and Thra^S2/+ ) were clustered in this statistical model as are mutants with long receptors (Thra^L1/+ and Thra^L2/+ ), and that the two groups tend to separate. Within each group, a new OPLS-DA was unable to discriminate between the two mutations (Thra^L1/+ from Thra^L2/+ mice, and Thra^S1/+ from Thra^S2/+ mice, in both urine and plasma; data not shown). As no clear discrimination was observed, we pooled the mutant mice according to the type of frameshift mutation they carry (Thra^L/+ group and Thra^S/+ group), and repeated OPLS-DA. This improved the statistical power of the analysis and led to a clear distinction between these two groups (Fig. 2B, C). Statistical significance of these models was assessed by high values of goodness-of-fit parameters R² and Q² (which explain the variance and predictive power of the model, respectively) and CV-ANOVA p-values <0.05 and model resampling under the null hypothesis (Supplementary Fig. S2). Altogether, these results show a clear distinction between Thra^L/+ and Thra^S/+ mice, as expected from previous phenotyping.

FIG. 2.

Multivariate analysis of metabolome. Distinction between Thra mutations. (A) Analysis of urine metabolome fails to discriminate between the four groups of three-month-old mutant mice. The OPLS-DA model (four classes) for three-month-old mice urine, discriminating Thra^S1/+ , Thra^S2/+ , Thra^L1/+ , and Thra^L2/+ (n = 40, 3 + 2 components, R²Y = 0.661, Q² = 0.355, ANOVA of the cross-validated residuals (CV-ANOVA) p-value = 0.005). (B) The group of Thra^S/+ can be segregated from the group of Thra^L/+ mice, based on the metabolome of urine, harvested from three-month-old and six-month-old animals; OPLS-DA model (two classes) for three-month-old mice urine (n = 41, 1 + 3 components, R²Y = 0.908, Q² = 0.669, CV-ANOVA p-value = 6.6 × 10⁻⁶) and for six-month-old mice urine (n = 43, 1 + 4 components, R²Y = 0.924, Q² = 0.57, CV-ANOVA p-value = 8 × 10⁻⁵). (C) Plasma analysis also segregates Thra^S/+ and Thra^L/+ mice and six-month-old mice blood plasma (n = 44, 1 + 5 components, R²Y = 0.829, Q² = 0.458, CV-ANOVA p-value = 0.04). Each point represents a sample. Tpred, Tpred1, and Tpred2 correspond to predictive components, which describe the variability between the samples. Tortho is an orthogonal component. ANOVA, analysis of variance; CV-ANOVA, ANOVA of the cross-validated residuals; OPLS-DA, orthogonal partial least-square discriminant analysis. Color images are available online.

Discrimination between Thra^S/+ and wild-type mice

We then asked whether the group of Thra^S/+ mice, which carry mutations relevant to human cases, could be recognized from a group of wild-type littermates. We performed three comparisons, using either urine (three or six months) or plasma. OPLS-DA of the metabolic profiles was performed to drive robust statistical models based on the discrimination between Thra^S/+ and wild-type mice. We observed a significant discrimination between Thra^S/+ samples and their associated wild-type controls in all three comparisons (Fig. 3). By contrast, similar comparisons were unable to discriminate the group of Thra^L/+ mice from the wild-type group, as expected from the very limited consequences of these mutations on phenotype (data not shown).

FIG. 3.

Multivariate analysis of metabolome. Distinction between Thra^S/+ animals and wild-type littermates. (A–C) OPLS models based on ¹H-NMR analysis of urine and plasma. (D–F) Model validation by resampling. Goodness-of-fit parameters are R² and Q². The statistical significance of these models is assessed by high values of goodness-of-fit parameters R² and Q² and by CV-ANOVA p-value <0.05. Resampling under the null hypothesis (1000 times) validates their discrimination. (G–I) Metabolic signature discriminating Thra^S/+ males and wild-type littermates. The colored spectral regions correspond to the statistically significant signals after significance to ANOVA tests followed by Benjamini–Hochberg multiple correction. Red asterisks indicate compounds with significant variation. Nonsignificant NMR variables are colored in gray. (A) OPLS-DA model for three-month-old mice urine (n = 62, 1 + 3 components, R²Y = 0.744, Q² = 0.453, CV-ANOVA p-value = 4.5 × 10⁻⁵). (B) OPLS-DA model for six-month-old mice urine (n = 62, 1 + 1 components, R²Y = 0.457, Q² = 0.217, CV-ANOVA p-value = 7 × 10⁻³). (C) OPLS-DA model for six-month-old mice plasma (n = 66, 1 + 4 components, R²Y = 0.768, Q² = 0.591, CV-ANOVA p-value = 9.24 × 10⁻⁸). Color images are available online.

Determination of a metabolic signature of Thra mutation in urine and blood plasma

The most significant variations of individual metabolites discriminating the Thra^S/+ group from the wild-type group were identified with univariate analysis of spectral variables from urine and plasma NMR profiles. Urine metabolic signatures of Thra^S/+ mice were characterized by increased levels of hippurate, formate, and fumarate, and decreased levels of Krebs cycle metabolites (acetate, succinate, citrate, and cis-aconitate), urea, allantoin, creatine, creatinine, isovalerylglycine, trimethylamine, dimethylamine, taurine, and oxoglutarate with respect to wild-type animals (Fig. 3G, H). Meanwhile, the plasma metabolic signature was characterized by higher levels of formate, phenylalanine, valine, isoleucine, allantoin, and lactate, and lower levels of Krebs cycle metabolites (acetate, succinate, and ketobutyrate), lipids, lipoproteins (very low-density lipoprotein [VLDL], low-density lipoprotein [LDL]), phosphocholine, glucose, creatine, threonine, glutamine, and pyruvate when compared with wild-type mice (Fig. 3I). The metabolic signatures of Thra^S/+ obtained from urine and plasma share similarities in the composition and variations of four metabolites: formate, creatinine, acetate, and citrate.

Concentrations of dimethylamine, trimethylamine, isovalerylglycine, choline, and N-acetylglucosamine were shown to be significantly lower in urine of Thra^S/+ mice, while polyunsaturated lipids were lower in the blood. Overall, this study provides metabolic signatures associated with RTHα in urine and plasma that present common and complementary information.

Comparison between the metabolic signatures of urine of three- and six-month-old mice provides useful information on the temporal stability of the observed metabolic changes. Overall, a large similarity in metabolite composition and variations was observed, 13 metabolites varying in the same way at both ages (Fig. 3G, H). Few differences were noticed, notably for formate and fumarate, for which differential concentrations only appeared in six-month-old mice. Variation in oxoglutarate content was also amplified in six-month-old mice. Dimethylamine, trimethylamine, and choline remained present in both signatures but became statistically significant, in univariate analysis, only after six months. To further ascertain the stability of the metabolic signature overtime, we verified that a set of variables of importance (>1) from the OPLS models obtained with urine samples from three-month-old mice could be used for classification of six-month-old mice urines (Supplementary Fig. S3).

Signature of Thra mutation is different from signature of hypothyroidism

RTHα patients share symptoms with hypothyroid patients. We thus investigated whether the metabolic signatures found for Thra mutations can be distinguished from a signature of hypothyroidism. Wild-type mice were made hypothyroid with propyl-thio-uracil treatment, and ¹H-NMR analysis was performed to obtain a signature related to hypothyroidism in urine and plasma. A strong discrimination was observed between hypothyroid and control groups (Supplementary Fig. S4). By investigating urine and plasma metabolic profiles, we noticed that 24 and 17 metabolites vary in hypothyroid condition with respect to the nontreated group (Fig. 4). We treated the hypothyroid mice to restore the circulating levels of T4 and T3, and analyzed the metabolomic response two days later. As expected, this treatment resulted in acute changes of the previously identified metabolites, with few additional changes (Supplementary Fig. S5).

FIG. 4.

Metabolic signature of hypothyroidism. Metabolic signature discriminating hypothyroid and wild-type littermates. The colored spectral regions correspond to the statistically significant signals after significance to ANOVA tests followed by Benjamini–Hochberg multiple correction. Nonsignificant NMR variables are colored in gray. (A) OPLS-DA loading plot of urine samples, discriminating hypothyroid group from control euthyroid group (n = 18, 1 + 1 components, R²Y = 0.982, Q² = 0.961, CV-ANOVA p-value = 4.76 × 10⁻⁹). (B) OPLS-DA loading plot of plasma samples, discriminating hypothyroid group from control euthyroid group (n = 19, 1 + 7 components, R²Y = 0.999, Q² = 0.95, CV-ANOVA p-value = 0.001). Color images are available online.

When the metabolic signatures reflecting the presence of a Thra mutation and hypothyroidism were compared, clear differences were obvious. Only six common metabolite variations were found in urine: increased level of hippurate, and decreased levels of acetate, citrate, succinate, dimethylamine, and trimethylamine (Fig. 5A). In plasma, only seven metabolites were shared between the two metabolic signatures. Both were characterized by a decrease in LDL, VLDL, acetate, citrate, pyruvate, choline, phosphocholine, and glucose (Fig. 5B).

FIG. 5.

Summary of key metabolites found in urine (A) and plasma (B) associated with Thra^S/+ and hypothyroid phenotypes. ↑: higher level compared with control group. ↓: lower level compared with control group. Metabolites in red vary significantly in Thra^S/+ samples. LDL, low-density lipoprotein; VLDL, very low-density lipoprotein. Color images are available online.

These results show that Thra mutation and hypothyroidism alter the urine and serum metabolomes in a very different manner. The metabolic profiles recorded in urine and plasma could therefore potentially contribute to the specific diagnosis of RTHα, which is a rare disease, much less frequent than hypothyroidism.

Discussion

The main objective of our study was to establish that metabolomic fingerprinting can be used as a diagnostic tool for a rare genetic disease, RTHα. Multivariate analysis of urine spectra provides a metabolic fingerprint that enables to recognize mice carrying a pathological mutation in the Thra gene, relevant to the human genetic disease. By contrast, the method did not identify the presence of weak alleles, which have little consequences on the development and physiology, and have never been reported in humans. The method did not pinpoint a single metabolite, the presence of which in urine or plasma would be the hallmark of RTHα. The statistical power rather relies on the capacity to evaluate the concentrations of a large number of metabolites simultaneously.

Previous metabolomic studies already analyzed the consequences of hypothyroidism in rat urine (24) and plasma (25) using mass spectrometry rather than ¹H-NMR. The serum metabolites identified by mass spectrometry are not found in our study, mainly because our analysis is not suitable for the identification of lipids, which represent a large fraction of the discriminating metabolites. A study of human plasma also points out that the plasma concentration of many lipids correlates with the thyroid status (26). By contrast, the overlap is significant for urine metabolites, where changes in allantoin, creatinine, urea, and hippurate contents were detected in both cases.

Considering that mouse Thra and human THRA genes display extensive sequence similarities, mouse lines with Thra mutations are highly relevant animal models for RTHα. We thus expect that ¹H-NMR can be used for RTHα diagnostics in humans. However, our experiments were performed using mice of the same age, sex, and genetic background, which were bred in a highly controlled environment. The clinical implementation of this method might thus be challenging, as body fluid composition is certainly more variable in human populations. Also, the interpretation of the metabolic signatures obtained in mice is not straightforward, and we cannot ascertain at this point that the underlying biochemical mechanisms are conserved in humans.

One encouraging observation is that, although ¹H-NMR is also instrumental to recognize hypothyroidism, thyroid hormone deprivation provides a very different signature than Thra mutations. Our analyses show that, while Thra mutations and hypothyroidism may have a similar impact on glucose metabolism, they differentially affect the plasma content in lipids and cholesterol. In particular, variations in the plasma content in fatty acids are in the opposite direction. This is expected if one considers that the liver, which secretes a large fraction of the lipids found in serum and urine, expresses Thrb at a much higher level than Thra. Liver is thus expected to be more sensitive to changes in TH status than to Thra mutations. Studying both urine and plasma is also useful because the two biofluids provide common and complementary information. For example, correlated variations suggest that changes in urine reflect variations initially present in plasma, for creatinine, citrate, and acetate, and not an alteration of the renal function.

Additional features of the metabolic signature can be tentatively interpreted in biochemical terms. Isovalerylglycine is a product of leucine degradation to acetyl-CoA, which is an important feeder into the Krebs cycle (27). A significant decrease in the urinary level of isovalerylglycine, and the decrease in the levels of citrate, succinate, and acetate may thus indicate a change in the degradation of leucine.

Finally, the lower concentrations of dimethylamine, trimethylamine, choline, and hippurate observed in the urine of mutant mice and hypothyroid mice may reflect intestinal alterations. Gut microbial metabolism of choline, present in diet, leads to the production of trimethylamine, which, once absorbed, is metabolized to generate dimethylamine. Hippurate is also synthesized by the gut microbiota and its decreased concentration in urine is a marker of bowel inflammation, which compromises the barrier capacity of the epithelium (28). Therefore, it is tempting to speculate that changes for these metabolites are direct or indirect consequences of the defect in the intestinal epithelium, reported in hypothyroid (29) and Thra mutant mice (30,31). Although providing a snapshot of merely a fraction of the metabolome, ¹H-NMR analysis of Thra mutant mice may thus be revealing changes in metabolic pathways and physiology, which are expected to be conserved in humans.

Footnotes

Acknowledgments

The authors thank Anne Lambert for her support in mouse breeding, and the AniRA-PBES facility (SFR Biosciences Lyon France). They acknowledge financial support from the Agence Nationale de la Recherche (ANR-15-CE14-0011-04 ThyroMUT2).

Author Disclosure Statement

No competing financial interests exist.

Supplementary Material

Supplementary Figure S1

Supplementary Figure S2

Supplementary Figure S3

Supplementary Figure S4

Supplementary Figure S5

References

Yen

, Ando

, Feng

, Liu

, Maruvada

, Xia

. 2006. Thyroid hormone action at the cellular, genomic and target gene levels. Mol Cell Endocrinol, 246:121–127.

Ortiga-Carvalho

, Sidhaye

, Wondisford

. 2014. Thyroid hormone receptors and resistance to thyroid hormone disorders. Nat Rev Endocrinol, 10:582–591.

van Gucht

ALM

, Moran

, Meima

, Visser

, Chatterjee

, Visser

, Peeters

. 2017. Resistance to thyroid hormone due to heterozygous mutations in thyroid hormone receptor alpha. Curr Top Dev Biol, 125:337–355.

Refetoff

, Bassett

, Beck-Peccoz

, Bernal

, Brent

, Chatterjee

, De Groot

, Dumitrescu

, Jameson

, Kopp

, Murata

, Persani

, Samarut

, Weiss

, Williams

, Yen

. 2014. Classification and proposed nomenclature for inherited defects of thyroid hormone action, cell transport, and metabolism. J Clin Endocrinol Metab, 99:768–770.

Weiss

, Refetoff

. 2000. Resistance to thyroid hormone. Rev Endocr Metab Disord, 1:97–108.

Dumitrescu

, Refetoff

. 2013. The syndromes of reduced sensitivity to thyroid hormone. Biochim Biophys Acta, 1830:3987–4003.

Bochukova

, Schoenmakers

, Agostini

, Schoenmakers

, Rajanayagam

, Keogh

, Henning

, Reinemund

, Gevers

, Sarri

, Downes

, Offiah

, Albanese

, Halsall

, Schwabe

, Bain

, Lindley

, Muntoni

, Khadem

, Dattani

, Farooqi

, Gurnell

, Chatterjee

. 2012. A mutation in the thyroid hormone receptor alpha gene. N Engl J Med, 366:243–249.

Moran

, Chatterjee

. 2016. Resistance to thyroid hormone alpha-emerging definition of a disorder of thyroid hormone action. J Clin Endocrinol Metab, 101:2636–2639.

Sun

, Wu

, Xie

, Wang

, Chen

, Wang

, Flamant

, Chen

. 2019. New case of thyroid hormone resistance alpha caused by a mutation of THRA/TRalpha1. J Endocr Soc, 3:665–669.

10.

van Gucht

, Meima

, Zwaveling-Soonawala

, Visser

, Fliers

, Wennink

, Henny

, Visser

, Peeters

, van Trotsenburg

. 2016. Resistance to thyroid hormone alpha in an 18-month-old girl: clinical, therapeutic, and molecular characteristics. Thyroid, 26:338–346.

11.

Moran

, Agostini

, Visser

, Schoenmakers

, Offiah

, Poole

, Rajanayagam

, Lyons

, Halsall

, Gurnell

, Chrysis

, Efthymiadou

, Buchanan

, Aylwin

, Chatterjee

. 2014. Resistance to thyroid hormone caused by a mutation in thyroid hormone receptor (TR)alpha1 and TRalpha2: clinical, biochemical, and genetic analyses of three related patients. Lancet Diabetes Endocrinol, 2:619–626.

12.

Newgard

. 2017. Metabolomics and metabolic diseases: where do we stand?. Cell Metab, 25:43–56.

13.

Beckonert

, Keun

, Ebbels

, Bundy

, Holmes

, Lindon

, Nicholson

. 2007. Metabolic profiling, metabolomic and metabonomic procedures for NMR spectroscopy of urine, plasma, serum and tissue extracts. Nat Protoc, 2:2692–2703.

14.

Brindle

, Antti

, Holmes

, Tranter

, Nicholson

, Bethell

, Clarke

, Schofield

, McKilligin

, Mosedale

, Grainger

. 2002. Rapid and noninvasive diagnosis of the presence and severity of coronary heart disease using 1H-NMR-based metabonomics. Nat Med, 8:1439–1444.

15.

Markossian

, Guyot

, Richard

, Teixeira

, Aguilera

, Bouchet

, Plateroti

, Guan

, Gauthier

, Aubert

, Flamant

. 2018. CRISPR/Cas9 editing of the mouse Thra gene produces models with variable resistance to thyroid hormone. Thyroid, 28:139–150.

16.

Dieterle

, Ross

, Schlotterbeck

, Senn

. 2006. Probabilistic quotient normalization as robust method to account for dilution of complex biological mixtures. Application in 1H NMR metabonomics. Anal Chem, 78:4281–4290.

17.

Wishart

, Feunang

, Marcu

, Guo

, Liang

, Vazquez-Fresno

, Sajed

, Johnson

, Li

, Karu

, Sayeeda

, Lo

, Assempour

, Berjanskii

, Singhal

, Arndt

, Liang

, Badran

, Grant

, Serra-Cayuela

, Liu

, Mandal

, Neveu

, Pon

, Knox

, Wilson

, Manach

, Scalbert

. 2018. HMDB 4.0: the human metabolome database for 2018. Nucleic Acids Res, 46:D608–D617.

18.

Sugimoto

, Ikeda

, Niigata

, Tomita

, Sato

, Soga

. 2012. MMMDB: mouse multiple tissue metabolome database. Nucleic Acids Res, 40:D809–D814.

19.

Boccard

, Rutledge

. 2013. A consensus orthogonal partial least squares discriminant analysis (OPLS-DA) strategy for multiblock Omics data fusion. Anal Chim Acta, 769:30–39.

20.

Worley

, Powers

. 2013. Multivariate analysis in metabolomics. Curr Metab, 1:92–107.

21.

, Zhou

, Ren

, Li

, Jiang

, Ma

. 2019. A selective review of multi-level omics data integration using variable selection. High-Throughput, 8:4.

22.

Eriksson

, Trygg

, Wold

. 2008. CV-ANOVA for significance testing of PLS and OPLS (R) models. J Chemom, 22:594–600.

23.

Blaise

, Shintu

, Elena

, Emsley

, Dumas

, Toulhoat

. 2009. Statistical recoupling prior to significance testing in nuclear magnetic resonance based metabonomics. Anal Chem, 81:6242–6251.

24.

, Tan

, Dong

, Zhu

, Li

, Lou

, Chai

. 2013. Metabolic profiling provides a system understanding of hypothyroidism in rats and its application. PLoS One, 8:e55599.

25.

Montoya

, Strauss

, Fabian

, Kamp

, Mellert

, Walk

, Looser

, Herold

, Krennrich

, Peter

, van Ravenzwaay

. 2014. Mechanistic analysis of metabolomics patterns in rat plasma during administration of direct thyroid hormone synthesis inhibitors or compounds increasing thyroid hormone clearance. Toxicol Lett, 225:240–251.

26.

Jourdan

, Linseisen

, Meisinger

, Petersen

, Gieger

, Rawal

, Illig

, Heier

, Peters

, Wallaschofski

, Nauck

, Kastenmuller

, Suhre

, Prehn

, Adamski

, Koenig

, Roden

, Wichmann

, Volzke

. 2014. Associations between thyroid hormones and serum metabolite profiles in an euthyroid population. Metabolomics, 10:152–164.

27.

Goudarzi

, Chauthe

, Strawn

, Weber

, Brenner

, Fornace

. 2016. Quantitative metabolomic analysis of urinary citrulline and calcitroic acid in mice after exposure to various types of ionizing radiation. Int J Mol Sci, 17:pii: E782.

28.

Williams

, Cox

, Walker

, Cobbold

, Taylor-Robinson

, Marshall

, Orchard

. 2010. Differences in gut microbial metabolism are responsible for reduced hippurate synthesis in Crohn's disease. BMC Gastroenterol, 10:108.

29.

Flamant

, Poguet

, Plateroti

, Chassande

, Gauthier

, Streichenberger

, Mansouri

, Samarut

. 2002. Congenital hypothyroid Pax8(−/−) mutant mice can be rescued by inactivating the TRalpha gene. Mol Endocrinol, 16:24–32.

30.

Plateroti

, Gauthier

, Domon-Dell

, Freund

, Samarut

, Chassande

. 2001. Functional interference between thyroid hormone receptor alpha (TRalpha) and natural truncated TRDeltaalpha isoforms in the control of intestine development. Mol Cell Biol, 21:4761–4772.

31.

Sirakov

, Boussouar

, Kress

, Frau

, Lone

, Nadjar

, Angelov

, Plateroti

. 2015. The thyroid hormone nuclear receptor TRalpha1 controls the Notch signaling pathway and cell fate in murine intestine. Development, 142:2764–2774.

Metabolomic Profiling of Body Fluids in Mouse Models Demonstrates that Nuclear Magnetic Resonance Is a Putative Diagnostic Tool for the Presence of Thyroid Hormone Receptor α1 Mutations

Abstract

Background:

Methods:

Results:

Conclusion:

Introduction

Materials and Methods

Animals

Sample collection

NMR spectroscopy

Statistical analysis

Results

Metabolomic analysis segregates two types of frameshift mutations

Discrimination between ThraS/+ and wild-type mice

Determination of a metabolic signature of Thra mutation in urine and blood plasma

Signature of Thra mutation is different from signature of hypothyroidism

Discussion

Footnotes

Acknowledgments

Author Disclosure Statement

Supplementary Material

References

Discrimination between Thra^S/+ and wild-type mice