Introduction

The discovery of thyroxine (T4), isolated from thyroid gland in 1914 by Edward C. Kendall [1] marked the start of thyroid hormone research. After this pioneering discovery, researchers showed that thyroid hormones control and regulate vital biological processes such as thermogenesis, growth, and metamorphosis [2–5]. The two most recognized thyroid hormones, the prohormone T4 and the biologically more active form 3,3’,5-triiodothyronine (T3, Fig. 1) are synthesized and secreted from the thyroid gland with the former produced in greatest amounts [3]. Thyroid hormones are enzymatically transformed by deiodinases in target tissues and released into circulation, where levels of T4 and T3 are maintained in a constricted range by negative feedback in the hypothalamic-pituitary-thyroid (HPT) axis [4, 6]. Importantly, T4 is deiodinated to form T3, but also ‘reverse’-T3 (3,3’,5’-triiodothyronine, rT3), and these two forms can be further deiodinated into three diiodothyronine-products (T2’s) that, in turn, can give rise to monoiodothyronines (T1’s) and thyronine (T0) through further deiodination events. Additional endogenous bioactive metabolites generated from thyroid hormones include the thyronamines (e.g. 3-iodothyronamine, T1AM), iodothyroacetic acids, such as 3,5-diiodothyroacetic acid (Diac), 3,3’,5-triiodothyroacetic acid (Triac) and 3,3’,5,5’-tetraiodothyroacetic acid (Tetrac) [2, 7]. Currently it is very difficult to capture the levels of all of these thyroid hormones. Enabling the scientific community to describe the thyroid hormone system state at an improved level would allow researchers to measure as many thyroid hormones and metabolites as possible, as metabolites are also highly bioactive [7–10]. Due to the characteristic lipophilic nature of thyroid hormones (e.g. log P for T4 is 4.12 [11]), they are chiefly transported in blood by carrier-proteins, and it is widely accepted that only the small fraction of unbound hormone (e.g. only 0.01-0.03 % of T4 is ‘free’) exerts the physiological effect [3]. Free serum concentrations of T3 (FT3) and T4 (FT4) are in the low picomolar range. These low concentration levels make it difficult to separate unbound and protein-bound hormones without disturbing the equilibrium. Thus, current methodologies are not likely to reflect a ‘true’ picture of thyroid hormone biochemistry [12, 13].

Fig. 1
figure 1

Molecular structures of the investigated thyroid hormones and associated metabolites. Three 13C6-analgoues were used as surrogate standards (cT4, cT3, and cT2) and a fourth as internal standard (crT3)

Full size image

Analytical methods based on mass spectrometry, especially isotope dilution mass spectrometry using selective reaction monitoring, provide more specific quantification technology than standard biological activity techniques (i.e., immunoassays) and thereby offer better precision, accuracy, and reproducibility [12–15]. Wang and Stapleton described the analysis of T4, T3, rT3, 3,3’-T2 and 3,5-T2 in serum [16], and later Kunisue et al. included 3-T1 [17]. Recently, Noyes et al. described sample preparation optimizations for analyzing T4 and T3 in fish serum [8]. Detection limits for these methods, even though not explicitly stated, ranged from 0.25 to 1.4 ng mL-1 [8, 16, 17]. Inspired by current LC-MS/MS methodologies analyzing two or more thyroid hormones [8, 16, 17], we developed a more sensitive method that includes eleven thyroid hormones and associated metabolites (Fig. 1).

The aim of the present study was to describe and evaluate a LC-MS/MS method enabling researchers to quantify absolute and total concentrations of eleven thyroid hormones and associated metabolites in plasma and serum from a range of vertebrates. Finally, we applied the methodology to tadpole serum and frog plasma under conditions relevant to thyroid hormone research.

Experimental

Chemicals

The investigated thyroid hormones and metabolites, surrogate standards, and internal standard are depicted in Fig. 1. Solvents (methanol and acetonitrile), ammonia hydroxide, formic acid, and urea were all of analytical grade and acquired from Fisher Scientific (Fair Lawn, NJ, USA). Ultrapure water was obtained from a MilliQ unit operated at 18.2 mΩ (EMD Millipore, Hayward, CA, USA).

Standard solutions

Individual certified reference material stock solutions in methanolic 0.1 M ammonium hydroxide at 100 μg mL–1 of T4, T3, rT3, 13C6-T4 (cT4), 13C6-T3 (cT3), and 13C6-rT3 (crT3) were obtained from Cerilliant Corp. (Round Rock, TX, USA). Individual stock solutions for 3,3′-T2 (ISOTEC, Miamisburg, OH, USA, purity >99 %), 3,5-T2 (Sigma-Aldrich, St. Louis, MO, USA, purity >99 %), 3-T1 (Santa Cruz Biotechnology, Santa Cruz, CA, USA, purity >95 %), T0 (Sigma-Aldrich, St. Louis, MO, USA, purity >98 %), T1AM (Santa Cruz Biotechnology, Santa Cruz, CA, USA, purity >98 %), 13C6-3,3′-T2 (cT2, ISOTEC, Miamisburg, OH, USA, purity >97 %), Diac (Toronto Research Chemicals, ON, Canada, purity >95 %), Triac (Santa Cruz Biotechnology, Santa Cruz, CA, USA, purity >96 %), and Tetrac (Santa Cruz Biotechnology, Santa Cruz, CA, USA, purity >99 %) were prepared at 1000 μg mL–1 in methanol. Equimolar mixture dilution series containing 11 thyroid hormones (T4, T3, rT3, 3,3’-T2, 3,5-T2, 3-T1, T0, T1AM, Diac, Triac, and Tetrac) were prepared in ultra pure water ranging from 0.10 to 100 fmol μL–1. A surrogate standard (SUR) diluted mixture containing 10.0, 2.00, and 4.00 fmol μL–1 cT4, cT3, and cT2, respectively, were prepared in ultra pure water. Finally, an internal standard (IS) mixture at 0.25 fmol μL–1 crT3 was prepared in 0.1 % acetic acid. A mixture of three antioxidants, L-(+)-ascorbic acid, R,R-dithiothreitol, and citric acid monohydrate, all obtained from Acros Organics (Bridgewater, NJ, USA) and purities above 99 %, were prepared daily at 25 mg mL–1 in water. In addition, 8 M urea solutions in 1 % ammonia hydroxide (by volume) were also prepared daily.

Serum and plasma samples

Serum was obtained from 10 American bullfrog (Rana (Lithobates) catesbeiana) tadpoles approaching metamorphic climax (Taylor and Kollros [18] stage 18) housed at the University of Victoria Outdoor Aquatics Unit. Animals were euthanized in 0.1 % (w/v) tricaine methanesulfonate (Syndel Laboratories Ltd., Vancouver, BC, Canada) buffered with 25 mM sodium bicarbonate in dechlorinated municipal water. Blood was collected by tail incision and allowed to coagulate at room temperature for 10–15 min. The blood from each animal was kept separate and never pooled. The coagulated blood was centrifuged (10,000 × g for 10 min at 4 °C; Beckman Coulter microfuge 16, Brea, CA, USA) and the serum collected in cryo tubes (Eppendorf, Hauppauge, NY, USA) followed by immediate flash freezing in liquid nitrogen and storage at –80 °C. The serum samples were later shipped on dry ice to UC Berkeley.

Plasma was obtained from 20 2-y-old African clawed frog (Xenopus laevis), male and female, kept in animal colonies maintained at UC Berkeley. In short, blood was obtained by cardiac puncture with heparinized needles and transferred to heparinized test tubes. Plasma was obtained by centrifugation (2000 × g for 3 min at 4 °C, Sorvall RT 6000D; DuPont, Wilmington, DE, USA) and transferred to individual cryo tubes (Eppendorf, Hauppauge, NY, USA) and stored at –20 °C.

Extraction of thyroid hormones

Thyroid hormones were extracted by placing 50 μL serum or plasma aliquots in a polypropylene Eppendorf tube. Fifty microliters of the SUR, corresponding to 391 pg cT4, 66 pg cT3, and 106 pg cT2) and 2.50 mg of each antioxidant (100 μL, 25 mg mL–1), were added, vortexed, and the sample mixture was allowed to equilibrate for 1 h on ice. Thereafter, 400 μL urea solution was added, vortexed, and incubated for another hour at 50 °C. The incubated samples were purified using polymeric solid-phase extraction cartridges (30 mg Plexa, 130 Å, mixed-mode; Agilent Technologies, Santa Clara, CA, USA). Solid-phase extraction cartridges were placed on a vacuum manifold (Supelco, Bellefonte, PA, USA) and preconditioned using 1.0 mL methanol, 1.0 mL ultrapure water, and 1.0 mL 1 % ammonia hydroxide. Thereafter, samples were applied and extracted, and cartridges were immediately washed with 2.0 mL-aliquots of 1 % ammonia hydroxide, ultrapure water, 1 % formic acid, and again ultrapure water. Using the vacuum pump, a 1-min drying period was used, followed by a final wash with 0.50 mL 30 % methanol, to remove more lipophilic residues. The cartridges were again dried for another minute by applying vacuum. Analytes were eluted into polypropylene eppendorf tubes with 1.50 mL methanol. The eluate was reduced to less than 50 μL using a Vacufuge operated at 30 °C and 1400 rpm (Eppendorf, Hauppauge, NY, USA). The residual extract was dissolved in 400 μL IS 0.25 fmol μL–1 mixture (dissolved in 0.1 % acetic acid), vortexed, and transferred to an LC-vial.

Liquid chromatography-tandem mass spectrometry

The analysis was performed on an Agilent 6460 triple-quadrupole system with a hyphenated Agilent 1260 HPLC system (binary pump, degasser, and autosampler; Agilent Technologies). A hyphenated stationary phase-system was used. The first column, 2.1 × 40 mm, 5 μm, 120 Å phenyl-phase, was protected with a 0.5 μm stainless steel frit filter and guard column (2.1 × 20 mm, 5 μm phenyl-phase) and connected to a second column, 2.1 × 40 mm, 5 μm, 120 Å C18-phase; all from Higgins Analytical Inc. (Mountain View, CA, USA). The injection volume was 100 μL. Mobile phase A was 0.1 % acetic acid (by volume) in ultrapure water, and mobile phase B was 0.1 % acetic acid (by volume) in acetonitrile. Initially, the flow rate was 500 μL min–1 and from 4.0 min it was rapidly reduced, reaching 200 μL min–1 2 s later. The flow rate was maintained at 200 μL min–1 until 16.0 min into the program and again ramped up to 500 μL min–1 over 30 s and maintained for the remainder of the run (28.0 min). The binary mobile phase started at 0 % B and ramped to 40 % over initial 4.0 min, and thereafter increased to 62.5 % B, reached at 10.0 min. In the subsequent 30 s, the mobile phase was ramped to 100 % B (10.5 min) and kept for another 12 min to remove lipophilic components (22.5 min), returning to 100 % A at 23.0 min and kept for 5 min (28.0 min) for re-equilibration. The MS/MS instrument was equipped with Agilent Jet Stream electrospray interface operated in fast-polarity-switching selective-reaction monitoring mode (SRM) using a preinstalled hexabore resistive capillary (Agilent Technologies) and high-purity (>99.999 %) nitrogen as sheath and collision gas. Optimal instrument-specific parameters were: gas temperature 300 °C, gas flow 8 L min–1, nebulizer 50 psi, sheath gas temperature at 300 °C, and flow at 10 L min–1. Nozzle voltage and ionization spray voltages were ±3.00 kV and ±6.00 kV, respectively, in both ionization modes. Compound-specific parameters were optimized for each hormone by infusing neat 1000 fmol μL–1 standards dissolved in mobile phases A and B (1:1). The mass spectrometer acquired SRM-data on two MS/MS-ion transitions for each target compound, unless other stated (Supplementary Table S-1). The column effluent was only infused into the mass spectrometer from 5.5 to 16.0 min to minimize contamination. Acquisition and handling of data were performed using an Agilent MassHunter Workstation ver. B.05.00.

Method evaluation

Calibration curves were made for all thyroid hormones in neat solvents and in fortified Xenopus laevis plasma (matrix-matched calibration). Neat standard seven-point equimolar calibration curves (n = 3) were prepared in 0.08 % acetic acid using 0, 5, 10, 50, 100, 500, and 1000 fmol mL–1 thyroid hormone, and all vials contained a fixed amount SUR; 400, 200, and 1000 fmol mL–1 cT2, cT3, and cT4, respectively, and 800 fmol mL–1 IS (crT3). Matrix-matched 10-point calibration curves (n = 3) were established from a X. laevis plasma pool. The homogenized plasma pool was subdivided into 50-μL aliquots and spiked with thyroid hormones prior to applying the optimized sample preparation. Thyroid hormones were spiked at 0, 25, 50, 100, 150, 200, 250, 500, 750, and 1000 fmol in each 50-μL sample, and 50-μL SUR mixture was added (i.e., 200, 100, and 500 fmol cT2, cT3, and cT4, respectively). This plasma contained endogenous levels of T4 (3.00 pmol mL–1) and Tetrac (0.03 pmol mL–1).

The method procedure was further evaluated using spike-recovery experiments at three concentration levels (0.50, 5.0, and 20 pmol mL–1) in X. laevis plasma for assessing method accuracy. Precision was investigated as reproducibility (or day-to-day variation) and repeatability [19]. Reproducibility was obtained analyzing a X. laevis plasma pool spiked with 25 pmol mL–1 on different days and included freeze–thaw cycles (n = 9). Repeatability was found using coefficient of variation (CV%) from matrix-matched calibration samples (n = 3). Instrument detection limit (IDL), given as absolute amount injected on-column, was determined using analyte signal-to-noise (S/N) of 3 from observed peaks at lowest concentration level in neat solvents. Likewise, method detection limit (MDL) was determined using analyte signal-to-noise (S/N) of 3 from observed peaks at endogenous or lowest fortified level in sample matrix. Method quantification limit (MQL) was extrapolated as 10/3.3 times MDL [20].

A quality assurance and control scheme (QA/QC) was established and covers a compilation of LC-MS/MS data (i.e., for each compound retention time and ratios between peak areas for quantifier and qualifier ion transitions were monitored). Prior to quantification, the following identification criteria for each substance in samples compared with a neat reference standard were defined: (a) retention times were not allowed to deviate more than 1.0 %; and (b) deviation on ion transition ratios were below 20 %. Moreover, procedural blanks and fortified samples were included in every sample preparation batch, and surrogate standard recoveries were monitored in every sample. Compared with neat analytical standards, samples with deviations higher than 30 % on peak area for IS or lower than 50 % SUR recovery were discarded from data set, and reanalyzed or if possible reprocessed. These criteria were rarely triggered; however, if triggered, acceptable results for reanalyzed extracts were normally obtained when common LC-MS/MS maintenance was applied, such as frit and guard column replacement.

Results and discussion

LC-MS/MS instrumentation

We established and optimized a methodology for measuring 11 thyroid hormones and metabolites simultaneously (Fig. 2). Hormones were analyzed in positive ionization mode with the exception of the iodothyroacetic acids (Supplementary Table S-1). Analyte–surrogate relationships are also displayed in Supplementary Table S-1. All analytes yielded two SRM-ion transitions; however, it was only possible to obtain a single high abundance ion-transition for T0 and T1AM. Our MS/MS-fragmentation is in agreement with other studies that also found similar ion fragments caused by loss of carboxylic acid for T4, T3, rT3 [8, 21–23], 3,5-T2, and 3,3′-T2 [16, 24]. For T1AM, we observed a MS/MS-fragment at m/z 212, corresponding to loss of aromatic hydroxyl and iodine, which agrees with other findings [25, 26]. In negative ionization mode, analyzing Triac and Diac, we observed an initial in-source fragmentation caused by loss of carboxylic acid (i.e., the negative parent ion of Triac should be at m/z 620.7; however, we observed it at m/z 576.7). We selected these in-source fragmented parent ions for Triac and Diac (Supplementary Table S-1) and scanned for MS/MS iodine fragments (m/z 126.9). Also in negative mode, we did not observe a similar in-source fragmentation for Tetrac; thus m/z 746.6 was selected as parent ion, and loss of carboxylic acid (m/z 702.7) and generation of iodine (m/z 126.9) were used as fragments for SRM (Supplementary Table S-1). A number of studies have used negative ionization for T4, T3, rT3, 3,5-T2, and 3,3′-T2 [16, 17, 27–29]; however, we did not find negative mode as sensitive for these substances. Interestingly, we also observed T3, rT3, 3,5-T2, and 3,3′-T2 yielded better LC-MS/MS signals compared with T4 at the optimized instrument conditions (Fig. 2) and likely depending on compound differences in dissociation constants [30].

Fig. 2
figure 2

Reconstructed ion chromatograms of a neat 50.0 fmol mL–1 equimolar analytical standard in 0.080 % acetic acid. Surrogate standard (SUR) concentrations are 400, 200, and 1000 fmol mL–1 for cT2, cT3, and cT4, respectively. Concentration of the internal standard (IS), crT3, is, in this case, 800 fmol mL–1

Full size image

We evaluated analyte signal intensities and chromatographic performance in combination with various mobile phases and HPLC columns (i.e., using formic acid, ammonium formate, and acetic acid as additives for water-methanol or water-acetonitrile mobile systems). We found that combining two 40-mm columns with different stationary phases (phenyl and C18) in combination with 0.1 % acetic acid in a water-acetonitrile mobile phase yielded sufficient separation (Fig. 2). Applying only 40 mm of either stationary phase gave insufficient peak separation, whereas 80 mm phenyl phase caused a weak peak broadening. Moreover, using the combined setup, it was not possible to obtain complete baseline separation between T3 and rT3, and that of cT3 and crT3 (Fig. 2). As these substances share identical SRM ion transitions, misinterpretations may arise (Supplementary Table S-1). Yet, we observed no caveats when analyzing tadpole serum or frog plasma, as the separation of cT3 and crT3 was carefully assessed in each sample.

Injection volume and use of methanolic extracts in vials were also investigated. Using up to 20–25 % methanol in 0.1 % acetic acid did not affect chromatographic properties such as peak retention time, shape, and resolution. After appropriate modification of the autosampler injection loop according to the manufacturer’s instructions, we evaluated injection volumes using neat standards of thyroid hormones dissolved in 20 % methanol and 0.08 % acetic acid. Using injection volumes from 5 to 350 μL resulted in similar chromatography and greater signal-to-noise ratios with increasing volume (data not shown). However, when injecting matrix samples, high variation on column backpressure was observed. Ultimately we reduced the injection volume to 100 μL and found near identical backpressure profiles, and IS peak retention times and intensities comparing hundreds of sample injections. One probable cause for the high system stability was the use of initial high mobile-phase flow rate, which separates polar and large molecular residues (e.g., protein fragments) from the guard-column trapped targeted analytes. Unfortunately, it was not possible to apply an additional high-pressure switching valve between guard and analytical columns, as this could have enhanced system stability.

Stability of thyroid hormones

The stability of thyroid hormones and metabolites was initially investigated using an accelerated study as suggested in OECD guideline 111 [31]. Thyroxine (T4) was selected as a model compound and spiked in pH-adjusted ultrapure water and measured in the water-phase over a 5-d period at 50 °C. T4 dissipated from solution with half of the thyroid hormone lost in 2 d under both acidic and circumneutral pH conditions (Fig. 3). Washing the inside of polypropylene test tubes with methanol did not reveal any adsorbed residues. A number of protocols for extracting thyroid hormones from plasma use acids in combination with heat [8]. This prompted a more meticulous stability study to provide essential knowledge on these substances behavior (Supplementary Fig. S-1). Up to 40 % of thyroid hormone concentration could be lost during 1 h at 50 °C under strongly acidic conditions (i.e., pH = 4), even with antioxidants present. However, under basic conditions (i.e., pH = 11), no degradation was observed in this timeframe (Supplementary Fig. S-1). More studies on the stability of these molecules are needed to elucidate the reaction mechanism and its pH-dependence. As a consequence of these findings, we opted to use a strong basic environment for sample preparation.

Fig. 3
figure 3

Thyroxine (T4) degradation experiments were performed at 50 °C in polypropylene capped tubes spiked with 10 pmol T4 per milliliter media (n = 3). Samples were measured at 0, 2, and 5 d - displayed as first, second, and third column at each pH, respectively. Dunnett’s multiple comparisons test two-way ANOVA statistics was applied by comparing d 2 and 5 to d 0 at each condition (****, p < 0.0001 and **, p < 0.01)

Full size image

Photolysis may transform some thyroid hormone-related compounds (especially T4 [32, 33]) and may lead to erroneous results. Under normal laboratory working conditions (mainly fluorescent light sources), we observed photolytic deiodination of e.g., rT3 transformed into 3,3′-T2 and 3-T1 in clear glass tubes and no transformation when amber glass vials were used (data not shown). Hence, we suggest shielding test tubes and using amber materials to minimize photolytic processes during sample manipulation. Even using a SUR will probably not correct for these or other transformations appropriately. Adding an additional isotopic analogue, as suggested by Tai and co-workers [34], may circumvent transformations.

Sample preparation

Development of the method to extract thyroid hormones from blood plasma and serum mainly covered optimization of the solid-phase extraction. Nevertheless, we initially evaluated ultrafiltration membranes with 3000 and 30,000 dalton molecular weight cutoff values, as a potential rapid sample preparation procedure to eliminate proteins and protein-bound hormones. However, these filtration devices yielded high matrix effects and poor recoveries (data not shown). Therefore, we investigated several solid-phase materials, viz. hydrophilic-lipophilic balanced mixed-mode materials (Oasis HLB from Waters and Bond Elut Plexa from Agilent Technologies), weak and strong anion-exchangers (Oasis WAX and MAX from Waters and PAX from Agilent Technologies), weak cation-exchangers (Oasis WCX from Waters, Milford, MA, USA), and finally C18 (50 and 100 mg HyperSep C18 from Thermo Scientific, Waltham, MA, USA). We used recovery and matrix effect assessment [35–37] to investigate the effect of pH on loading media (pH 4 and 11) and consecutive wash (water) and elution profiles of methanol and acetonitrile in combination without or with (0.1–5 %) formic acid or 0.1 % ammonia hydroxide. In short, the two hydrophilic-lipophilic balanced mixed-mode polymeric materials (HLB and Plexa), the polymeric strong anion-exchanger PAX, and the C18-material displayed best absolute recoveries. Based on these results, we selected the 30 mg Plexa and the 100 mg C18 material as the best options for further optimization, and investigated elution profiles at acidic and basic loading conditions using fortified 50 μL-plasma pool aliquots as matrix (Supplementary Fig. S-2 and Fig. 4, respectively).

Fig. 4
figure 4

Sequential wash and elution profiles from two different SPE materials (100 mg C18 [circles] and 30 mg Plexa mixed-mode polymer [triangles]) at basic conditions. Fortified (100 pmol mL–1) base pretreated bovine plasma was used (50 μL) as matrix. Primary axes values corresponds to percent methanol in 500 μL aliquots (i.e., ‘10’ is a 10 % methanol solution in water). PT (pass-through media), W1 (wash 1, 2 mL 1 % NH4OH), W2 (wash 2, 2 mL ultrapure water), W3 (wash 3, 2 mL 1 % formic acid), W4 (wash 4, 2 mL ultrapure water), M is a single 500 μL aliquot methanol. Corresponding study at acidic conditions is shown in Supplementary Figure S-2

Full size image

Noyes et al. found that washing a hydrophilic-lipophilic balanced mixed-mode polymeric SPE cartridge (60 mg) with 30 % methanol after sample enrichment in acidic environment would eliminate matrix components and still retain T3 and T4 [8]. We found it was possible to use greater than 50 % methanol for washing and still retain T3 and T4; however, as expected, the more polar analytes T0 and T1AM migrated already with 10-20 % methanol. Under basic pH conditions, we found that washing 4 × 2 mL water (with alternating pH) and 30 % methanol did not cause losses of the thyroid hormones with the expectation of T0 and the iodothyroacetic acids (Fig. 4). We also (surprisingly) observed leaching of rT3, 3,3′-T2, 3-T1, and T0 from the C18-material. Consequently, given these and our findings concerning the acid instability, we decided to use the Plexa cartridges at high pH in combination with multiple wash-steps (4 × 2 mL water at alternating pH and 0.5 mL 30 % methanol) prior to elution using 1.50 mL methanol.

Critical parameters

The intention of using antioxidants is mainly to prevent degradation of targeted analytes. Addition of antioxidants to the sample has previously been investigated [34]. Tai et al. found inconsistent results, especially for T3, when omitting antioxidants in the sample preparation [34]. We found similar results (data not shown) and, thus, decided to follow previous recommendations and add 2.5 mg of each antioxidant (ascorbic acid, citric acid, and dithiothreitol) in our sample preparation.

Equilibration of thyroid hormones with surrogate standards is a time-dependent process and is critical to accurate measurements. Equilibration times, for similar conditions used in the present study, have been shown previously for T4 and T3 and are less than 1 h [8, 29, 34]. Consequently, we decided to use 1 h equilibration time (covered on ice), as suggested by Noyes and co-workers [8], with the assumption that all thyroid hormones have similar equilibration times.

In blood, the majority of thyroid hormones are bound to carrier proteins. Therefore, releasing the protein-bound fraction of thyroid hormones is an essential step to enhance recovery. We found using protein precipitation approaches, such as crash solvents (e.g., methanol, acetone, and acetonitrile) and salting out effects [17], yielded low absolute recoveries and inconsistent results, likely due to coprecipitation of targeted analytes [38]. Accordingly, we focused on denaturing the proteins, thereby unfolding and bringing the macromolecules into solution. Typically, other researchers have done this by adding strong acids, such as trichloroacetic acid and hydrochloric acid [8, 23, 29]. As per our stability study (Fig. 3 and Supplementary Fig. S-1), we aimed for using strong basic conditions (1 % ammonia hydroxide) in combination with the chaotropic agent urea and heat. Using this combination, in addition to solid-phase extraction cartridges, we observed very good results, with repeatability deviations of less than 5 % for most analytes (Table 1).

Table 1 Method evaluation parameters and results for determining thyroid hormones in plasma. All values are from fortified Xenopus laevis plasma with the exception of instrument detection limit (IDL), given as an absolute amount injected on-column
Full size table

We investigated reconstituting the purified extract in 10 mM ammonia hydroxide instead of 0.1 % acetic acid (mobile phase A) and if there was any thyroid hormone dissipation in reconstituted extracts stored at 4 °C. The idea behind this was that the analytes would adsorb less to the surfaces of glass HPLC vials, as the analytes would be negatively charged and repulsed by borosilicate surfaces at high pH. We observed no differences in chromatographic performance when injecting analytes dissolved in 10 mM ammonia hydroxide compared with 0.1 % acetic acid. Analyzing the same samples (dissolved in 10 mM ammonia hydroxide or 0.1 % acetic acid) stored for 1 wk at 4 °C resulted in analyte signal losses of approximately 30 % at both conditions and demonstrated the importance of immediately analyzing processed extracts.

Method evaluation

Calibration curves prepared in neat solvents yielded R2 values of 0.9825 to 0.9995 in the concentration range 0–1000 fmol mL–1 for all thyroid hormones and metabolites (Supplementary Table S-2). Matrix-matched calibration curves (in X. laevis plasma) at fortified levels (0.50–20 pmol mL–1) also displayed impressive R2 values above 0.998 for all but three analytes, viz. T0, Triac, and Tetrac (Table 1). Consequently, we applied matrix-matched calibrations using X. laevis plasma for all samples. We found instrument detection limits (IDL) in the range 0.07–1.00 fmol for all 11 analytes, i.e., T4 was 0.19 fmol (0.15 pg) and T3 was 0.14 fmol (0.09 pg, Table 1).

Method detection limits (MDL), when using 50 μL sample aliquots, ranged from 0.01 to 0.38 pmol mL–1, i.e., T4 was 0.09 pmol mL-1 (corresponding to 70 pg mL–1) and T3 was 0.04 pmol mL–1 (29 pg mL-1, Table 1). Compared to other methodologies for analyzing only T4, the developed methodology is at least twice as sensitive [23, 39]. Compared to methods analyzing multiple thyroid hormones, the developed method is 5- to 20-fold more sensitive [16, 17]. In the present study, it was not possible to obtain plasma or serum without T4, and removing this analyte by carbon filtration may also eliminate other matrix components. Consequently, for this analyte we used neat standards to determine MDL and MQL.

Accuracy assessment, using relative recoveries, yielded acceptable results in the investigated concentration range (0.50, 5.00, and 20.0 pmol mL–1, Table 1). T0, T1AM, Triac, and Tetrac, however, did not yield usable signals at 0.50 pmol mL–1. T4 yielded high recoveries at this level likely due to an endogenous content (3.00 pmol mL–1) of this key analyte. We did not apply recovery correction for any analytes when calculating concentration in samples. A more in-depth accuracy assessment using absolute recovery studies confirmed that only about 16 % of T0 is recovered from plasma, and about 40 % of Triac and Tetrac (Fig. 5). These low recoveries are likely due to washing off some proportion of these analytes from the solid-phase extraction material (Fig. 4). Even so, key thyroid hormones demonstrated very good absolute recoveries of around 80 % or higher and, even more importantly, the surrogate standards produced similar absolute recoveries (Fig. 5). It is not evident from any studies in the literature if absolute recoveries of thyroid hormones and metabolites have been assessed in previous methodologies. The observed matrix effect in X. laevis plasma was on average –20 % (ranging from –7 % up to –30 %), indicating weak to moderate ion suppression [35].

Fig. 5
figure 5

Absolute recovery profiles applying the optimized procedure using 50 μL X. laevis plasma (n = 3, also applied in Table 1) fortified at 100 pmol mL–1 with all analytes (blue) and 20 pmol mL–1 surrogate standards (red)

Full size image

Application

The optimized methodology was applied in animal studies. To the best of our knowledge, this is the first time that such thyroid hormone and metabolite profiles have been presented. We analyzed plasma from six individual 2-y-old X. laevis (male) and found T4 at 8.73 ± 1.16 pmol mL–1 (6.79 ± 0.90 ng mL–1) with minimum and maximum values 7.75 and 10.5 pmol mL–1, respectively. The average surrogate recoveries were 70.7 % for cT4, 89.4 % for cT3, and 83.9 % for cT2, and we did not detect any other analyte in these samples. These plasma levels for T4 are in accordance with another X. laevis study where NF stage 62 [40] tadpoles yielded 10.7 ± 0.8 ng mL–1 at this metamorphic climax stage [23]. Metamorphic climax is the period in which maximal levels of T4 and T3 are typically reached in the frog.

In serum from 10 R. catesbeiana tadpoles that were approaching metamorphic climax (TK stage 18 [18]), we observed seven of 11 thyroid hormones and associated metabolites; T4 7.68 ± 2.45 (100 %), T3 1.15 ± 0.22 (100 %), rT3 0.36 (20 % with one sample below MQL), 3,3′-T2 0.64 ± 0.10 (30 %), 3,5-T2 0.66 ± 0.06 (80 %), T1AM 3.11 ± 1.18 (30 %), and T0 0.74 pmol mL–1 (10 %) with detection frequencies in parentheses. A representative result is shown in Fig. 6. We also observed traces of Triac and Tetrac (<MDL). The average surrogate recoveries in these samples were 73.2 % for cT4, 69.4 % for cT3, and 77.1 % for cT2.

Fig. 6
figure 6

Reconstructed ion chromatograms of a Rana (Lithobates) catesbeiana tadpole (TK stage 18) serum sample. Surrogate standards (SUR) were added as 200, 100, and 500 fmol of cT2, cT3, and cT4, respectively. Internal standard (IS), crT3, was added to reconstitute the extract as 100 fmol in 400 μL 0.10 % acetic acid. Serum concentrations corresponds to 1.36, 0.37, 0.33, 0.77, and 5.74 ng mL–1 for T1AM, 3,5-T2, 3,3’-T2, T3, and T4, respectively. Triac and Tetrac were below detection limits

Full size image

Conclusions

We developed, optimized, and evaluated a highly selective and sensitive analytical methodology, enabling scientists to quantify 11 thyroid hormones and metabolites in 50 μL of blood serum or plasma. Method detection limits were 70, 29, 57, 34, 46, 150, 41, 37, 44, 170, and 6.7 pg mL–1 for T4, T3, rT3, 3,3′-T2, 3,5-T2, 3-T1, T0, T1AM, Diac, Triac, and Tetrac, respectively. We observed that this class of hormones inherently displays a photolytic and pH-dependent instability at acidic conditions. Use of antioxidants and a protein-chaotropic agent enhanced accuracy and precision of hormone extraction from blood serum or plasma. Washing solid-phase extraction cartridges extensively in combination with an online LC-MS/MS trap-and-flush event resulted in very low background noise and only weak matrix effect. It is likely that reducing sample extract volume and increasing sample size can improve detection limits. Applying the optimized method to animal samples, we observed an expected higher number and concentration of thyroid hormones and metabolites in tadpoles at metamorphic climax compared to the frog, as these hormones are vital in development, growth, and metamorphosis.