Rapid and Comprehensive Impurity Profiling of Synthetic Thyroxine by

Article pubs.acs.org/ac

Rapid and Comprehensive Impurity Profiling of Synthetic Thyroxine by Ultrahigh-Performance Liquid Chromatography−High-Resolution Mass Spectrometry Volker Neu,† Chris Bielow,‡ Iris Gostomski,§ Reiner Wintringer,§ Ralf Braun,⊥ Knut Reinert,‡ Peter Schneider,∥ Hermann Stuppner,∥ and Christian G. Huber*,† †

Department of Molecular Biology, Division of Chemistry and Bioanalytics, University of Salzburg, Hellbrunner Straße 34, 5020 Salzburg, Austria ‡ Department for Computer Science and Mathematics, Algorithmic Bioinformatics, Free University of Berlin, Takustrasse 9, 14195 Berlin, Germany § Department of Chemistry, Instrumental Analysis and Bioanalysis, Campus Building B 2.2, 66123, Saarland University, 66123 Saarbrücken, Germany ⊥ Peptido GmbH, Am Kraftwerk 6, 66450 Bexbach, Germany ∥ Institute of Pharmacy, Pharmacognosy, Member of CMBI, University of Innsbruck, Innrain 80-82, 6020 Innsbruck, Austria S Supporting Information *

ABSTRACT: Rapid and efficient quality control according to the public authority regulations is mandatory to guarantee safety of the pharmaceuticals and to save resources in the pharmaceutical industry. In the case of so-called “grandfather products” like the synthetic thyroid hormone thyroxine, strict regulations enforce a detailed chemical analysis in order to characterize potentially toxic or pharmacologically relevant impurities. We report a straightforward workflow for the comprehensive impurity profiling of synthetic thyroid hormones and impurities employing ultrahigh-performance liquid chromatography (UHPLC) hyphenated to high-resolution mass spectrometry (HRMS). Five different batches of synthetic thyroxin were analyzed resulting in the detection of 71 impurities within 3 min total analysis time. Structural elucidation of the compounds was accomplished via a combination of accurate mass measurements, computer based calculations of molecular formulas, multistage high-resolution mass spectrometry (HRMSn), and nuclear magnetic resonance spectroscopy, which enabled the identification of 71 impurities, of which 47 have been unknown so far. Thirty of the latter were structurally elucidated, including products of deiodination, aliphatic chain oxidation, as well as dimeric compounds as new class of thyroid hormone derivatives. Limits of detection for the thyroid compounds were in the 6 ng/mL range for negative electrospray ionization mass spectrometric detection in full scan mode. Within day and day-to-day repeatabilities of retention times and peak areas were below 0.5% and 3.5% R.SD. The performance characteristics of the method in terms of robustness and information content clearly show that UHPLC-HRMS is adequate for the rapid and reliable detection, identification, and semiquantitative determination of trace levels of impurities in synthetic pharmaceuticals. called “grandfather product” with long experience in dosages and effects on patients.4 It can be produced by employing published7−9 or proprietary multistep total chemical synthesis. In order to properly meet the strict regulations of international drug authorities like FDA or EMEA, the purities of intermediates and final drug products must be assessed rigorously during their manufacture, formulation, and storage.10−12 According to the regulations of the European Pharmacopoeia,13,14 compounds showing a peak area of more than 0.05% relative to T4 in a chromatographic analysis need to

T

he thyroid gland produces two iodinated thyroid hormones called thyroxine (T4) and liothyronine (T3). Very small quantities of excreted thyroid hormones already cause substantial effects on the organism. Therefore, thyroid hormones are essential for the proper development and the differentiation of cells of the human body. Only 10% of the hormone secreted from the thyroid gland is the biologically active form T3, while the residual 90% are released as T4 prohormone, which is converted into T3 by deiodinases in the target tissue.1,2 If the human organism lacks thyroid hormones, a condition called hypothyroidism causes characteristic physiological effects ranging from weight gain, muscle cramps, and weakness to psychic problems like depressions.1,3−5 Orally administered thyroxine has been used for the treatment of hypothyroidism for many years6 and hence, it ranks as a so© 2013 American Chemical Society

Received: December 21, 2012 Accepted: February 8, 2013 Published: February 8, 2013 3309

dx.doi.org/10.1021/ac303722j | Anal. Chem. 2013, 85, 3309−3317

Analytical Chemistry

■

be characterized and eventually structurally elucidated. Despite the availability of HPLC methods for the profiling of synthetic thyroxin in terms of enantiopurity or content of T3 as main impurity,11,15 the structural elucidation of low abundant impurities remains challenging. To our best knowledge only four studies describe the impurity profiling of synthetic thyroxine by applying thin-layer chromatography-electron ionization mass spectrometry16 or high-performance liquid chromatography hyphenated to mass spectrometry (HPLCMS).10,17,18 But in contrast to the presented work, the concentration of degradation products in thyroxin samples was artificially increased through pretreatment by different stressing procedures including thermal stressing, irradiation, and extreme pH, which is mandated in stress tests required by the regulatory agencies. Moreover, all proposed structures of impurities were deduced from low-resolution mass spectrometric data, except for one study,19 leading to uncertainties in structural assignment. Nevertheless, the studies revealed a very high sample complexity, which suggests that efficient chromatographic separation combined with high-resolution mass spectrometric analysis should be beneficial when aiming at a comprehensive analysis of untreated synthetic thyroxin. Very rapid chromatographic separations without suffering resolution can be achieved by transferring established HPLC methods to the domain of ultrahigh-performance liquid chromatography (UHPLC).20 The main distinguishing feature between UHPLC and conventional HPLC is the use of columns packed with particles smaller than 2 μm in diameter. According to Halász, this reduction of the particle size allows an increase of flow rates without significant decrease in column performance, however at the cost of relatively high column backpressures up to 1000−1500 bar, which requires special instrumentation.21 Similarly, high-resolution mass spectrometry (HRMS) has tremendously benefited in the past decade from the introduction of cost-efficient, high-resolution mass spectrometers,22 which allow the measurement of molecular masses on a chromatographic scale with accuracies in the 1−2 ppm range. This high mass accuracy aids significantly in the derivation of molecular formulas for detected compounds, as well as in the interpretation of spectra obtained from multistep high-resolution mass spectrometry experiments (HRMSn), which significantly increases the confidence of structural elucidation. This study aims at the development and implementation of a systematic workflow using UHPLC-HRMSn and bioinformatic data evaluation that allows the comprehensive analysis of unstressed synthetic thyroxin samples with respect to as many impurities as possible. Method development started from a conventional HPLC-MS method which was used for the analysis of potential degradation products of thyroid hormones.18 Optimization involved transfer to UPLC conditions using a sub-2 μm stationary phase and the so-called “gradient volume principle” for gradient optimization.23 Structural elucidation of detected impurities should be achieved by a combination of accurate mass measurements, computer based calculations of molecular formulas and HRMSn-experiments taking advantage of the high-resolution capabilities of a linear ion trap (LTQ)-Orbitrap mass spectrometer. The optimized method should finally allow a comparative analysis of different batches of synthetic thyroxin with respect to their impurity pattern.

Article

EXPERIMENTAL SECTION

Chemicals, Materials, and Samples. Acetonitrile (Optigrade for LC−MS) was obtained from Promochem (Wesel, Germany). Methanol (LC-MS CHROMASOLV) was purchased from Fluka (Buchs, Switzerland). Deionized water (18.2 MΩ cm) was prepared using the Milli-Q system from Millipore (Billerica, MA, USA). Formic acid (98−100% purissimum) was obtained from Sigma-Aldrich (Steinheim, Germany) and sodium hydroxide (pro analysi) was from Merck (Darmstadt, Germany). Five batches of thyroxine sodium pentahydrate as well as standard compounds were provided by Peptido GmbH (Bexbach, Germany). For method development, a mixture of ten standard compounds was used comprising thyroxine (T4) as well as 4-(4-hydroxyphenyl)-3,5-diiodo-L-tyrosine (diodothyronine, T2), 4-(4-hydroxy-3-iodophenyl)-3-iodo-L-tyrosine (reverse diiodothyronine, rT2), 4-(4-hydroxy-3-iodophenyl)3,5-diiodo-L-tyrosine (triiodothyronine, liothyronine, T3), 4-(4hydroxy-3,5-diiodophenyl)-3-iodo-L-tyrosine (reverse triiodothyronine, rT3), 4-(4-hydroxyphenoxy)-3,5-diiodophenylacetic acid (diiodothyroacetic acid, DiAc), 4-(4-hydroxy-3-iodophenoxy)-3,5-diiodophenylacetic acid (triiodothyroacetic acid, TriAc), 4-(4-hydroxy-3-iodophenoxy)-3,5-diiodophenylpropionic acid (triiodothyropropionic acid, TriProp), 4-(4hydroxy-3,5-diiodophenoxy)-3,5-diiodophenylacetic acid (tetraiodothyroacetic acid, TetraAc), and 4-(4-hydroxy-3,5-diiodophenoxy)-3,5-diiodophenylbenzoic acid (tetraiodothyrobenzoic acid, TetraBA) in a concentration range of 3−5 μg/mL (structures are shown in Figure S-1 of the Supporting Information). In order to simplify the nomenclature of thyroxine derivatives, the recurring structural element specified by “4-(4-hydroxy-3,5-diiodophenoxy)-3,5-diiodophenyl” will be named as “tetraiodothyro”. Derivatives containing less than four iodine atoms contain the analogous structural element “triiodothyo-“ or “diiodothyro”. For method validation, a mixture of six compounds was used comprising T2, T3, T4, DiAc, TriAc and TetraAc in a concentration of 20 μg/mL and 200 μg/mL, respectively. All working solutions were prepared by dissolving the samples in water/methanol (50:50, v/v) + 400 mg/L NaOH. The injection volume was 6.0 μL both for the standard mixtures and for the 2 mg/mL T4 sample solutions. Instrumentation and Columns. Chromatographic separations were performed in Thermo Hypersil GOLD columns, 100 × 2.1 mm i.d. column, packed with 1.9 μm C18-silica particles, which were kindly provided by G. Böhm from Thermo Fisher Scientific. HPLC-UV analysis was performed using an Accela UHPLC System, equipped with a pump capable of handling column backpressures up to 1000 bar, and controlled by LC-Devices version 2.01 from Thermo Fisher Scientific (Bremen, Germany). The Accela system was directly coupled to a linear ion trap-Orbitrap mass spectrometer (LTQOrbitrap XL) from Thermo Fisher Scientific equipped with an Ion Max electrospray ionization (ESI) source and controlled by XCalibur version 2.0.7. Computer based calculations of molecular formulas were performed by using the molecular formula generator included in XCalibur. The processing algorithm for automated screening of synthetic thyroxin samples was implemented by C. Bielow from the Free University of Berlin. Chromatographic and Mass Spectrometric Conditions. The eluents used for the gradient separations were 3310



Article

water containing 0.10% (v/v) formic acid (eluent A) and acetonitrile containing 0.10% (v/v) formic acid (eluent B). Measurements were performed at room temperature and 60 °C, respectively. UV detection was performed at 230 or 290 nm. The LTQ-Orbitrap XL mass spectrometer was tuned and calibrated in the negative ion mode following the procedure described by the manufacturer. HPLC-MS fine-tuning was performed by adding a makeup flow of 10 μL/min of T3 (20 μg/mL) via a T-piece to an HPLC-flow of 990 μL/min of water/acetonitrile (50:50; v/v) + 0.1% formic acid. The optimized mass spectrometric parameters were as follows: mass range in full scan mode, m/z 200 − 2000; resolution: 15.000; source voltage, 4.0 kV; capillary voltage, −35 V; capillary temperature, 350 °C; sheath gas flow, 90; aux gas flow, 25; tube lens, −150 V; multipole 00 offset, 3.50 V; lens 0 voltage, 6.00 V; multipole 0 offset, 5.25 V; lens 1 voltage, 13.00 V; gate lens offset, 76.00 V; multipole 1 offset, 6.00 V; front lens, 5.25 V. Nuclear Magnetic Resonance Spectroscopy Measurements. For obtaining sufficient amounts of impurities, a 5 g sample of thyroxine was thermally stressed for 29 days at 100 °C. Twenty mg aliquots of the stressed sample we subjected to semipreparative fractionation on an analytical HPLC system (Model 1050, Agilent Technologies, Waldbronn, Germany) utilizing a 150 × 10 mm i.d. Synergy Polar-RP column (Phenomenex, Aschaffenburg, Germany) and elution with 30% acetonitrile in 0.10% formic acid for 5.0 min, followed by a gradient of 30−70% acetonitrile in 0.10% formic acid in 37 min and 80% acetonitrile in 0.10% formic acid for 8.0 min at a flow rate of 5.0 mL min−1. Fractions were collected by peak, gently evaporated to dryness, and redissolved in 0.70 mL deuteromethanol. NMR spectra (512 scans each) were recorded in 5 mm thin wall, 7 in. long NMR tubes (Sigma Aldrich) on a 600 MHz NMR spectrometer (Model Avance AV 600 II+, Bruker Biospin, Rheinstetten, Germany) equipped with a PA TXI 600 S3 H−C/N-D-05 Z BTO sample head. Spectra were evaluated using the Topspin 2.1 software from Bruker.

Figure 1. Method transfer to UHPLC conditions through application of the gradient volume principle. Column, Hypersil GOLD, 100 × 2.1 mm i.d., 1.9 μm; mobile phase, (A) H2O + 0.1% FA, (B) ACN + 0.1% FA; gradients, flow rates, and column backpressures, (a) 30% B for 5.0 min, 30−70% B in 37 min, 225 μL/min, 280 bar, (b) 30% B for 2.5 min, 30−70% B in 18.5 min, 450 μL/min, 520 bar, (c) 30% B for 1.25 min, 30−70% B in 9.25 min, 900 μL/min, 970 bar, (d, e) 30−80% B in 2.1 min, 1000 μL/min, 630 bar; temperature, 60 °C; detection, (a-d) UV at 230 nm, (e) ESI-LTQ Orbitrap MS in positive and negative ion mode, m/z 200−2000; sample, 6.0 μL of the standard mix containing 3−5 ppm of (1) T2, (2) r-T2, (3) T3, (4) r-T3, (5) T4, (6) DiAc, (7) TriAc, (8) TriProp, (9) TetraAc, (10) TetraBA, for abbreviations see Experimental Section or Supporting Information.

■

RESULTS AND DISCUSSION UHPLC-HRMS Method Development. A mixture of ten compounds comprising T4, the biologically relevant T3 and T2, as well as their structural isomers r-T2 and r-T3, and five further derivatives known from medical research24 were used for method development as a model of synthetic thyroxin and impurities (see also Figure S-1). As shown in Figure 1a, all components of the standard mixture were well separated to baseline using a 60-min HPLC-MS method employing a small bore column packed with 1.9 μm C-18 particles and 0.10% formic acid as mobile phase additive,18 which demonstrates a good overall selectivity of the Hypersil GOLD column for the analysis of thyroid hormone derivatives. Figure 1b and c illustrate the chromatograms obtained after method acceleration by halving the gradient time while doubling the flow rate,23 resulting in separations with equivalent chromatographic efficiency at pressures up to 970 bar. Further method acceleration was accomplished upon omitting the initial isocratic step, reducing the gradient time by another factor of 4.4, ending with a higher acetonitrile concentration, and increasing the column temperature to 60 °C. Figure 1d illustrates the separation of all compounds of the standard mix in less than 2 min, in which the critical peak pair (peaks 9 and 10) is still separated to baseline with a resolution of R = 1.76. Interfacing to LTQ Orbitrap mass spectrometry

revealed that thyroxine and its amino acid derivatives (peaks 1− 5) were detectable upon electrospray ionization both in positive and negative ion mode, while the carboxylic acid derivatives (peaks 6−10) were seen only in the negative ion mode (Figure 1d). In consequence, electrospray ionization in the negative mode was applied in all following analyses, in spite of the use of an acidic eluent, a technique known in the literature as “wrongway-round ionization”.25,26 Method Evaluation for Impurity Profiling. The performance of the developed method in terms of intra and day-to-day repeatabilities of retention times and UV-peak areas were measured using a mixture comprising T2, T3, T4, DiAc TriAc and TetraAc and the chromatographic method shown in Figure 1d. Table 1 collects the relative standard deviations of retention times, peak areas with UV detection, peak areas with mass spectrometric detection, and relative mass deviation, measured by 20 runs each performed on three consecutive days. It can be seen that all parameters of the high-speed method are well within the limits characteristic for robust 3311



Article

injections, and 600 injections of real samples with high column load of 12 μg of sample per injection. We consider this as entirely acceptable values making this type of column attractive for routine analysis and industrial applications. Impurity Pattern of Synthetic Thyroxine. Synthetic thyroxine samples of five different batches were analyzed in triplicate using the 3-min UHPLC-HR-MS method. Figure 2

Table 1. Quality Parameters for the Separation and Detection of Thyroxine and Thyroxine Derivatives relative standard deviation (RSD), N = 20 or 60 parameter retention timea peak area (UV, 290 nm)a peak area (MS)a mass accuracya

day 1

day 2

0.50 % 0.42 %

0.46 % 0.28 %

day 3 0.38% 0.53%

days 1−3 0.50% 0.61%

2.48 % 2.10 % 1.63% 3.43% average relative mass deviation (ppm), N = 20 or 60 0.61 ppm 0.36 ppm 0.61 ppm 0.48 ppm

a

Parameters were deduced from 20 analyses each performed on three consecutive days using the method of Figure 1d and e. Detailed data are collected in Table S-1 and Table S-2 of the Supporting Information.

HPLC-MS methods. Repeatabilities within 0.5% for retention time, 0.61% and 3.4% for peak areas with UV and mass spectrometric detection, respectively, and below 0.65 ppm for relative mass deviation classify the developed method as well suitable for impurity profiling. Values for individual compounds can be deduced from Table S-1 and S-1 of the Supporting Information. UV detection at 290 nm, showing a 3-day relative standard deviation of 0.61% in peak areas, was more robust than detection at 230 nm (1.29% R.S.D.), which is most probably the result of more difficult integration due to baseline drift at the lower wavelength. Method sensitivity was determined by analysis of T3 using UV- and negative electrospray ionization (negESI)-MS detection. The limits of detection (LOD) were 5.91 ng/mL for neg-ESI-MS detection in full scan mode and 0.23 μg/mL for UV detection at 230 nm, corresponding to a more than 25-fold lower LOD for mass spectrometric detection. T4 and T3 showed comparable signal responses in MS as concluded from a maximum difference in peak area and peak height of about 15% when analyzing equimolar solutions in the lower ppm range. The LOD of UV detection for T3 was calculated from an external calibration graph using the 3.3-fold standard deviation of the procedure (sy/b) criterion after performing triplicate analysis over 5 equidistant calibration points from 1.0 to 5.0 ppm. Statistical parameters of the calibration were: slope b = 15301 μAU*s/ppm, correlation coefficient R2 = 0.998, and residual standard deviation sy = 1037 μAU*s. The LOD of MS detection for T3 was calculated directly from the base peak chromatogram of full scan MS data at a signal-to-noise ratio of 3 to 1 for the [M-H]− ion of T3. According to the ICH guideline Q3A R2 (2006) for quality control of drugs such as thyroid hormones, all impurities with a content higher than 0.05% relative to the main compound need to be identified.14 Assuming equal response factors and following a rough calculation, a concentration of approximately 1 μg/mL corresponds to this 0.05% peak area threshold when analyzing thyroid hormone samples with a concentration of 2 mg/mL thyroxine as main compound. This concentration is more than a factor of 165 higher than the detection limit for impurities in synthetic thyroxine. Finally, the stability of the chromatographic column under the optimized UHPLC conditions (60 °C, 620 bar) was studied with three Hypersil GOLD columns from three different batches by repetitive injections of the standard mixture in between analyses of real samples or blank injections. At least 2000 injections per column could be performed under UHPLC conditions, comprising 1000 water injections, 400 standard

Figure 2. Comparison of (a) UV and (b) mass spectrometric detection of impurities in different batches of synthetic thyroxine. Conditions as in Figure 1e.

gives an impression of differences in information content between UV and mass spectrometric detection and compares corresponding base peak chromatograms of five different thyroxine batches with respect to their impurities. Obviously, MS detection reveals a much more detailed impurity pattern of synthetic thyroxin as corroborated by 23 chromatographic peaks detected in addition to the main peak of thyroxine. In spite of this complexity, the chromatograms were readily comparable and showed a very high degree of similarity. More than a single mass was found under most of the peaks labeled in Figure 2. Upon elimination of artifact peaks resulting from adduct formation with sodium, formic acid, formate, or other thyroid compounds in the gas phase, fragmentation (mostly CO2 or I-elimination), the total number of detected masses was reduced from 122 to 72 candidate masses, of which between 59 to 69 candidate masses were found in each of the single analyses (see Table S-3 of Supporting Information). From Masses to Structures. For structural assignment, molecular formulas were generated for each mass by using the molecular formula generator included in XCalibur 2.1. Settings applied for the search were restriction of the elements to C, H, O, N, I, and Na and a maximum mass deviation of 10 ppm. The generator ranks possible molecular formulas compatible with an experimental mass on the basis of mass deviation. It additionally calculates the corresponding number of rings and double bonds (RDBs), which are half-integer because of the detection of deprotonated molecules and which are highly useful in matching the degree of unsaturation between molecular formulas and proposed structures. Moreover, the 3312



Article

Figure 3. Fragmentation pattern for the structural elucidation of the thyroxine dimer 2-(2-hydroxy-tetraiodothyroacetamido)-tetraiodothyropropionic acid. For fragment assignments see Table 2.

Table 2. Masses, Molecular Formulas, and Assignment of Fragments Used to Deduce the Structure of the Dimer 2-(2-Hydroxytetraiodothyroacetamido)-tetraiodothyropropionic acid found mass

molecular formula

RDB

Δ ppm

neutral loss

RDB

Δ ppm

e− config.

rel. signal height

−30.444 1.768 −1.667 −6.359 1.011

even even even even even even

100 0.06 1.54 0.04 0.15 0.23

2

1521.3243 1477.3359 967.7000 803.6760 758.6582 729.6736

C29H16O8NI8 C28H16O6NI8 C28H14O6NI4 C16H10O5NI4 C15H7O4I4 C14H8O2NI4

18.5 17.5 20.5 10.5 10.5 9.5

Structure Determining MS Fragments at CID-Energy of 20%: 0.3 1.215 CO2 2 −0.54 CO2 + H2O + 4 I −1 2.056 C13H6O3I4 9 6.994 C14H9O4NI4 9 −0.471 C13H6O2I4 + CO2 + CO + H2O 10

resonance spectroscopy to confirm the proposed structures. Datasheets collecting all chromatographic and spectroscopic data of the individual compounds are provided as Supporting Information. A graphical summary of all steps performed to derive the structures of impurities is shown in Figure S-2 of the Supporting Information. Furthermore, Figure 3 and Table 2 exemplarily present the fragmentation reactions and data utilized to obtain the structure of a previously unknown dimer formally resulting from the generation of a peptide bond between a thyroxine molecule and 2-hydroxy-tetraiodothyroacetic acid as a known impurity.19,27 In total, 65 fragments were detected in MS2 and MS3 spectra of 2-(2-hydroxy-tetraiodothyroacetamido)-tetraiodothyropropionic acid. Loss of CO2 and H2O suggest the presence of a carboxyl and a hydroxyl group in the molecule. Bond breaking on either sides of the NH-group (reaction pathway 1 and 2) in the peptide bond results in four fragments that suggest that both units of the dimer contain the “core structure” of two phenyl rings and four iodine atoms (C6H4O2I4). Moreover, the lactic acid amide derivative formed in reaction pathway 1 identifies one of the monomers as lactic acid derivative. Finally, elimination of water,

experimental and theoretical isotope patterns of assigned molecular formulas were manually matched, and unlikely molecular formulas were discarded. Finally, molecular formulas of the highest ranks were “transformed” to chemically plausible molecular structures by starting from thyroxine as precursor molecule, assuming the (partly deiodinated) core structure of the ether combining the two aromatic rings as conserved, and allowing deiodination, oxidation, and elimination of CO, CO2, and NH3 in the aliphatic chain, as well as condensation of molecules in the sample as main reactions for forming new derivatives. Aliphatic chain in this context specifies the C3 carbon backbone constituting the amino acid or carboxylic acid in the derivatives. Subsequently, all candidate structures derived from molecular formulas were confirmed by performing HRMSn experiments. The number n of MS stages was chosen such as to fragment the molecules into substructures small enough to allow unequivocal assignment of the molecular structure, which in some cases required even MS5 experiments. Finally, a larger sample of thermally stressed thyroxine was semipreparatively fractionated and the fractions were subjected to analysis by nuclear magnetic 3313



Article

Table 3. Selection of Unknown Impurities Found in Synthetic Thyroxine

examples of representative structures for the different structural classes that demonstrate the structural diversity within the impurity pattern of synthetic thyroxine. The full set of structures of impurities is provided in Tables S-4a-f of the Supporting Information. Class I collects products deriving from oxidative aliphatic chain degradation with and without concomitant deiodination (Tables S-4a), such as 2-hydroxytetraiodothyroacetaldehyde or 2-aminotetraiodothyroacetic acid. In these products, the number of carbon atoms is

carbon dioxide, and carbon monoxide (reaction pathway 3) allows the confirmation of the bridging via a peptide bond. The structures of 46 other impurities were derived analogously from intact masses and accurate masses of fragments generated by MSn with different collision energies (data sheets for each compound are supplied as Supporting Information). In order to facilitate a more systematic overview, identified derivatives were grouped into six different classes having common structural features. Table 3 presents eight 3314



Article

Figure 4. Quantitiative distribution of 51 impurities (47 new plus 4 already detected but not assigned in ref.19) found in five synthetic thyroxine batches. Given is the abundance of impurities with respect to mass and contribution to the total impurity pattern (sum of relative signal abundances of all impurities = 100%). Known impurities from previous publications are not included. Compound classification as in Table 3.

aromatic ring (Table S-4d). Within this compound class, 4-(4hydroxy-3,5-diiodophenyl)-tetraiodothyronine (T4-(4hydroxydiiodophenyl)ether) appeared with highest abundance. Another example of detected compounds is the 4-(4-hydroxy3,5-diiodophenyl)-triiodothyronine (T3-(4hydroxydiiodophenyl)ether), which can be present in different isomers that are undistinguishable by MSn. Class V, finally, contains structures containing four aromatic rings that originate from dimers of thyroxine and thyroxine derivatives, examples being the dioxane-bridged T3-T3 dimer or 2-(2-amino-3formamido-tetraiodothyropropamido)-tetraiodothyropropionic acid (Table S-4e and f). This class can be divided into two subclasses namely dimers formed by so-called tail−tail coupling (Table S-4e), meaning a condensation via core structures by substitution of one iodine atom, and dimers formed by head−

generally equal or lower than that of intact thyroxine. Class II summarizes thyroxine derivatives that contain the same number of carbon atoms as thyroxin itself, which means that deiodination, deamination, or oxidation of the aliphatic chain are the predominant formation reactions, resulting in 3hydroxy-r-T3 or 2-hydroxy-tetraiodothyropropionic acid, for example, or products of “sole” deiodination, such as T2, T3, and r-T3 (Table S-4b). Molecules included in class III were formed by oxidative aliphatic chain elongation and oxidation and/or deiodination, yielding molecules with more than 15 carbon atoms (Table S-4c), such as 3-formamidotetraiodothyropropionic acid. Class IV includes molecules containing three aromatic rings that can be formed upon condensation reactions of thyroxine or thyroxine derivatives with compounds containing only one 3315



Article

grams. Relative abundances of all impurities are given in the rightmost column of Table 3 and Table S-4. The distribution of the different impurities in terms of molecular mass and relative abundance is illustrated in Figure 4, which suggests that the most abundant impurities are resulting from aliphatic chain branching, elongation, and oxidation, respectively, or from dimerization. The maximum difference in the number of identified compounds was 16 between two different batches (Table S-3), and the abundance of these was generally below 0.1%. The variation of relative signal abundance between all 5 batches for all different compounds ranges between 12 and 170% (average value of 65%), which reveals also a high similarity of impurity pattern from a quantitative perspective. As far as the impurity pattern of unstressed thyroxin is concerned, 47 of 71 compounds found in total have not so far been described in the literature. And although these 47 compounds constitute only 26% of the total signal abundance of the impurity pattern, they represent the most heterogeneous part of the pattern from structural perspective. Apart from the monomeric structures which mainly derive form aliphatic chain modifications due to degradation, oxidation, or extension, a remarkable number of relatively abundant derivatives resulting from dimerization has been found, which represent a previously unknown class of impurities of synthetic thyroxine.

tail or head−head coupling, meaning a coupling via aliphatic chain and core structure or via two aliphatic chains (Table S4f). The formation of tail−tail dimers under substitution of one iodine atom opens up many possibilities for positional isomers, especially if the two derivatives forming the dimer are not containing four iodine atoms. Since most of the fragmentation spectra of these dimers did not offer information about the constitution of the core structure, most of them could not be discriminated by high resolution mass spectrometry. Nevertheless, in accordance with the excess of thyroxine compared to impurities in the bulk material, most derivatives of this compound class originate from the coupling reaction of two thyroxine molecules, followed by a subsequent deiodination, or vice versa. The remaining derivatives are supposed to be formed by the coupling of thyroxine with other high-abundant impurities like tetraiodothyrobenzoic acid or 2-hydroxytetraiodothyroacetic acid. Of all impurities found, T3, r-T3, T2, TetraBA, TetraAc, and T4-(4-hydroxydiiodophenyl)ether have already been known from standard mixture analysis and are described in the literature10,19 or the European or U.S. Pharmacopoeia.28,29 Tetraiodothyroacetamide, tetraiodothyrobenzaldehyde, 2-hydroxytetraiodothyroacetic acid, and triiodothyrobenzoic acid were found in thermally stressed T3,10 but due to low mass resolution available in this study, only tetraiodothyroacetamide was assigned to the correct molecular formula and structure. The misleading assignments were corrected in a study that analyzed synthetic and thermally stressed thyroxin using HPLC-ESI-time-of-flight-MS.19 Moreover, 24 out of 71 impurities found within this study were detected, but seven of those were either not at all or not correctly assigned due to insufficient mass accuracy and the lack of MSn experiments; 3 impurities found have been described in our previous study dealing with thermally stressed thyroxine.27 In consideration of the remaining 47 impurities, about two-thirds of the impurity pattern detected here have not been described earlier at all. Information about the Relative Abundance of Impurities. Due to the lack of reference compounds, absolute quantification of the impurities is impossible and hence, quantification has to rely on relative peak intensities observed in the impurity pattern. Quantification relative to the mass spectrometric signal of thyroxine is not practical because it is saturated, while relative quantification at higher dilution is not viable because the low-abundant impurities would fall below the limit of detection. As a feasible workaround, we propose to relate the intensity of the observed impurities to the sum of all intensities of the impurity pattern, which can give some indication of the abundance of observed impurities. A comparison of the total number of identified impurities revealed that a similar qualitative impurity pattern was detected in all five batches (Table S-3). Guidelines of the regulatory agencies regarding (in)acceptable levels of impurities or levels, at which structural elucidation is mandatory, are generally based on relative peak areas in chromatograms. Because mass spectrometric detection as utilized in our study is much more sensitive than UV detection, we estimate here the relative abundances on the basis of peak heights of the individual impurities relative to the sum of peak heights of all thyroxine derivatives (except thyroxine itself) in extracted ion current chromatograms (m/z ± 10 ppm). This practice is justified by the chromatogram in Figure 1e demonstrating that similar amounts of thyroxine derivatives show comparable signal intensities in the reconstructed total ion current chromato-

■

CONCLUSIONS

■

ASSOCIATED CONTENT

Using synthetic thyroxine as model system, this study clearly demonstrates the high potential and the limitations of a workflow involving UHPLC interfaced to HRMSn for the comprehensive profiling of impurities in synthetic pharmaceuticals. Long-term stability and robustness of the setup have been investigated in detail and were shown to be suitable for routine and high-throughput analysis. An in-depth evaluation of the chromatographic and mass spectrometric data using manual as well as bioinformatic tools has demonstrated the high information content of the workflow, which facilitated the derivation of more than 47 previously unknown, well-founded proposed structures. Structural elucidation also of very low abundant compounds was shown to be accessible via a combination of accurate mass measurements, computer based calculations of molecular formulas and high-resolution fragmentation experiments. The formation of dimers as a new class of thyroid hormone derivatives could be verified by HRMSn. By sorting information derived from HRMSn data and the formulation of plausible reaction mechanisms for fragments observed, a kind of retro-synthesis for reliable structure elucidation is feasible also of rather complex structures. Uncertainties in structural elucidation via HRMSn data can arise from isomeric compounds, especially regarding the position of iodine atoms in core structure of the molecule, and the presence of stereocenters which cannot be resolved. Nevertheless, 71 impurities could be found in total in synthetic thyroxin samples. This clearly demonstrates the potential of the developed screening method in terms of sensitivity and complexity, which represents a significant step forward in the tools available to gain insights in the impurity pattern of synthetic pharmaceuticals.

S Supporting Information *

Additional Information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org. 3316



■

Article

(21) Halasz, I.; Schmidt, H.; Vogtel, P. J. Chromatogr. 1976, 126, 19− 33. (22) Hu, Q.; Noll, R. J.; Li, H.; Makarov, A.; Hardman, M.; Graham Cooks, R. J. Mass Spectrom. 2005, 40, 430−443. (23) Engelhardt, H.; Elgass, H. Chromatographia 1986, 22, 31−39. (24) Moreno, M.; de Lange, P.; Lombardi, A.; Silvestri, E.; Lanni, A.; Goglia, F. Thyroid 2008, 18, 239−253. (25) Tso, J.; Aga, D. S. Anal. Chem. 2011, 83, 269−277. (26) Mansoori, B. A.; Volmer, D. A.; Boyd, R. K. Rapid Commun. Mass Spectrom. 1997, 11, 1120−1130. (27) Neu, V.; Bielow, C.; Schneider, P.; Reinert, K.; Stuppner, H.; Huber, C. G. Anal. Chem. 2013, DOI: 10.1021/ac303404e. (28) U.S. Pharmacopeial Convention. Levothyroxine Sodium. In U.S. Pharmacopeia 34; U.S. Pharmacopeial Convention: Rockville, MD, 2010, pp 3301−3301. (29) European Directorate for the Quality of Medicines. Levothyroxine Sodium. In European Pharmacopoeia 7.5; EDQM Council of Europe: Strasbourg, France, 2012, pp 4651−4652.

AUTHOR INFORMATION

Corresponding Author

*Address: Prof. Dr. Christian Huber Department of Molecular Biology, Division of Chemistry and Bioanalytics, University of Salzburg, Hellbrunnerstraße 34, 5020 Salzburg, Austria. Tel: +43(0) 662-8044-5704. Fax: +43(0)662-8044-5751. E-mail: c. [email protected]. Notes

The authors declare the following competing financial interests: Ralf Braun is executive director of Peptido GmbH, one of the largest worldwide producers of synthetic thyroxine. The contribution of Iris Gostomski to this work was financially supported by Pepido GmbH, Bexbach, Germany.

■

ACKNOWLEDGMENTS The authors thank Günter Böhm, Kornelia Weidemann, and Jürgen Srega from Thermo Fisher Scientific for supplying the UHPLC instrument and the Hypersil Gold columns as well as for instrumental support regarding the LTQ-Orbitrap instrument used in this study.

■

REFERENCES

(1) Roy, G.; Mugesh, G. J. Chem. Sci. (Bangalore, India) 2006, 118, 619−625. (2) Zhang, J. S.; Lazar, M. A. Annu. Rev. Physiol. 2000, 62, 439−466. (3) Stipcevic, T.; Pivac, N.; Kozaric-Kovacic, D.; Muck-Seler, D. Coll. Antropol. 2008, 32, 973−976. (4) Hueston, W. J. Am. Fam. Physician 2001, 64, 1717−1724. (5) Force, A. T. T. Endocrine Practice 2006, 8, 457−469. (6) Wishart, D. S.; Knox, C.; Guo, A. C.; Cheng, D.; Shrivastava, S.; Tzur, D.; Gautam, B.; Hassanali, M. Nucleic Acids Res. 2008, 36, D901−906. (7) Chalmers, J. R.; Dickson, G. T.; Elks, J.; Hems, B. A. J. Chem. Soc. 1949, 3424−3433. (8) Salamonczyk, G. M.; Oza, V. B.; Sih, C. J. Tetrahedron Lett. 1997, 38, 6965−6968. (9) Hillmann, G. Z. Naturforsch. 1956, 11b, 419−420. (10) Andre, M.; Domanig, R.; Riemer, E.; Moser, H.; Groeppelin, A. J. Chromatogr., A 1996, 725, 287−294. (11) Gika, H.; Lämmerhofer, M.; Papadoyannis, I.; Lindner, W. J. Chromatogr., B 2004, 800, 193−201. (12) Kannamkumarath, S. S.; Wuilloud, R. G.; Stalcup, A.; Caruso, J. A.; Patel, H.; Sakr, A. J. Anal. At. Spectrom. 2004, 19, 107−113. (13) European Agency for the Evaluation of Medicinal Products: Guideline on Control of Impurities of Pharmacopoeial Substances: Compliance with the European Pharmacopoeia General Monograph “Substances for Pharmaceutical Use” and General Chapter “Control of Impurities in Substances for Pharmacuetical Use”, London , 2004. (14) European Agency for the Evaluation of Medicinal Products: Guideline on Control of Impurities of Pharmacopoeial Substances: Compliance with the European Pharmacopoeia General Monograph “Substances for Pharmaceutical Use” and General Chapter “Control of Impurities in Substances for Pharmacuetical Use”, London , 2004. (15) Toussaint, B.; Schimmel, H.; Klein, C. L.; Wiergowski, M.; Emons, H. J. Chromatogr., A 2007, 1156, 236−248. (16) Won, C. M. Pharm. Res. 1992, 9, 131−137. (17) Kazemifard, A. G.; Moore, D. E.; Aghazadeh, A. J. Pharm. Biomed. Anal. 2001, 25, 697−711. (18) Gostomski, I.; Braun, R.; Huber, C. G. Anal. Bioanal. Chem. 2008, 391, 279−288. (19) Gostomski, I., Synthesis of Stationary Phases and Method Development for the HPLC Analysis of Active Pharmaceutical Ingredients. Ph.D. Dissertation, Saarland University, Germany, 2009. (20) Mazzeo, J. R.; Neue, U. D.; Kele, M.; Plumb, R. S. Anal. Chem. 2005, 77, 460 A−467A. 3317


Rapid and Comprehensive Impurity Profiling of Synthetic Thyroxine by

Recommend Documents