Serena
Ambrosini
a,
Sudhirkumar
Shinde
b,
Ersilia
De Lorenzi
*a and
Borje
Sellergren
b
aDepartment of Drug Sciences, University of Pavia, Via Taramelli 12, 27100, Pavia, Italy. E-mail: ersidelo@unipv.it; Fax: +39 0382 422975; Tel: +39 0382 987747
bINFU, Faculty of Chemistry, Technical University of Dortmund, Otto Hahn Strasse 6, 44221, Dortmund, Germany. E-mail: B.Sellergren@infu.uni-dortmund.de; Fax: +49 231 7554234; Tel: +49 231 7554082
First published on 28th October 2011
Two molecularly imprinted polymers (MIPs) that we recently described to be class-selective for glucuronides have been successfully exploited for the molecularly imprinted solid-phase extraction (MISPE) of testosterone glucuronide (TG) from its parent drug (T) in urine. Both sorbents targeted the glucuronate fragment but feature different functional groups for binding the carboxylate anion, MIP1, a neutral 1,3-diarylurea group, and MIP2, a cationic imidazolium functionality. MISPE-HPLC-UV methods developed using both sorbents allowed the extraction of TG from its parent compound in urine samples spiked at 150, 300 or 600 ng mL−1 for TG and at 50 ng mL−1 for T. By comparing the performance of the two sorbents it came out that MIP1 is a more suitable SPE packing than MIP2, since it isolated the glucuronide with a higher precision (RSD 2–5%, n = 3) and with an enhanced enrichment factor (EF = 4.2). On the basis of these results, the imprinted receptor MIP1 can be applied for the direct extraction of TG in doping and clinical analysis and to selectively capture any other relevant glucuronated metabolite avoiding tedious deconjugation steps prior to quantification.
Nevertheless, the high hydrophilicity of glucuronated metabolites makes their analysis a difficult task. In doping control TG is still determined through indirect methods. The glucuronide has to be hydrolyzed to the corresponding free testosterone, which is subsequently isolated by liquid–liquid extraction (LLE), derivatized to form trimethylsilyl ethers and subjected to gas chromatography-mass spectrometry (GC-MS) analysis.3,9–13
The direct analysis of TG through the LC-MS technique after solid-phase extraction (SPE) is highly preferable, as it does not require the time-consuming and laborious steps of hydrolysis and derivatization that may produce inaccurate results. However, the main issue that must be overcome in the routine use of such methods is matrix effect. The literature available on the analysis of TG and steroid glucuronides gives evidence that signal suppression occurs when the glucuronated metabolite is not well separated from its parent drug or from other polar matrix interferents, thus limiting the precision, accuracy and sensitivity of the method.3,6,14,15 The key point and the present bottleneck to develop a LC-MS method for direct quantification of TG imply the use of an optimum sample preparation. To this regard, the application of molecular imprinted polymers (MIPs) to SPE—namely, molecularly imprinted solid-phase extraction (MISPE)—may offer a solution.
The synthesis of MIPs entails the polymerization of functional monomer(s) in the presence of a template molecule. Removal of the template molecule leaves behind cavities within the cross-linked polymer that are complementary to the original template and/or its analogous structures. These polymeric materials are intrinsically characterized by high levels of affinity and selectivity and hence they allow a customized sample pretreatment that minimizes background to be performed.16–20
One of the major limitations of MISPE is the poor recognition property of MIPs in aqueous media (e.g., biological samples). An additional step (LLE or SPE) before the selective MISPE protocol may need to be performed in order to transfer the analyte to an organic solvent where the MIP shows specific interactions for the target compound. Current trends are towards the development of water-compatible MIPs that allow the direct loading of aqueous samples.21–24 In this case, during the loading step the MIP behaves as a reversed-phase sorbent, capturing the analyte and matrix components by means of non-specific interactions. Then the percolation, prior to the elution step, of an optimized washing solvent discloses the imprinting effect and leads to the selective removal of matrix interferents.
We have recently described for the first time25 the synthesis of MIPs for the class-selective recognition of glucuronides. The MIPs were prepared by following the substructure approach along with the stoichiometric imprinting. The former was based on the use of glucuronic acid as template, preventively derivatized either to make it compatible with the organic solvents used during polymerization or to simulate any drug-like molecular structure linked to glucuronic acid. With the aim of improving the performance of MIPs in aqueous systems, either a urea- or a cationic imidazolium-based monomer that stoichiometrically interacts with the carboxylic function of the template was employed. Among the eleven MIPs produced, two have shown a high specificity for the glucuronated form of all the tested compounds—namely, testosterone, cotinine and mycophenolic acid. The diversity of the chemical structure of these compounds proves the class-selectivity against glucuronides of the new imprinted receptors.
As reported in this work, these two MIPs successfully accomplish the direct extraction of testosterone glucuronide from its parent drug in urine sample and importantly circumvent the cumbersome deconjugation step prior to analysis.26
The MISPE protocol was first optimized in artificial urine and then successfully implemented to real urine. Although the recovery of testosterone glucuronide was not quantitative, it was highly reproducible, leading in turn to a good analytical precision. Hence the MISPE protocol setup for the urea-based MIP1 has proven to be an efficacious sample clean-up procedure for the direct analysis of TG. Moreover, since glucuronidation is a common phase II reaction of human metabolism, this imprinted sorbent would positively impact all those areas of analysis involving glucuronide detection (e.g., clinical, environmental and forensic analysis).3,27–32
Fig. 1 Chemical structure of the template TAG and DGNa+ and of the functional monomers FM1 and FM2. |
1,2,2,6,6-Pentamethylpiperidine (PMP), ethylene glycol dimethacrylate (EDMA), and pentaerythritol triacrylate (PETRA) were purchased from Sigma-Aldrich (Steinheim, Germany). The initiator 2,2′-azobis(2,4-dimethylvaleronitrile) (ABDV) was supplied from Wako (Neuss, Germany). All chemicals for the synthesis of the template, functional monomers and polymers were of reagent grade and used as received with the exception of EDMA. EDMA was washed consecutively with 10% NaOH, water, and brine; it was then dried over Na2SO4; and distilled under reduced pressure. All solvents used for synthesis were purchased from Sigma-Aldrich and purified according to the guidelines in Purification of Laboratory Chemicals. Anhydrous solvents for polymer preparation were stored over appropriate molecular sieves.
Testosterone (T) was bought from Sigma-Aldrich (Steinheim, Germany) whereas testosterone-β-D-glucuronide (TG) was purchased from Steraloid Inc. (Newport, USA). Chemical structures of T and TG are reported in Fig. 2. The salts NaN3, NaH2PO4 and Na2HPO4 were obtained by Merck (Darmstadt, Germany). NaCl and urea were purchased from BDH Laboratory Supplies (Poole, England). Methanol (MeOH), acetonitrile (MeCN), acetic acid (AcOH), HCOOH and H3PO4 for analysis were of HPLC grade and were supplied from Carlo Erba (Milan, Italy). LC-grade water was prepared using a Millipore Direct-QTM system (Bedford, MA, USA).
Fig. 2 Chemical structures of T and TG. |
The washed polymer fragments were crushed using a mortar and then sieved to obtain the desired particle size (36–50 μm). The polymers were sedimented several times using MeOH:H2O (80:20, v/v) to eliminate smaller particles. Control polymers (non-imprinted polymers, NIPs) were prepared in the absence of template and of the base when this was added to the pre-polymerization solution of the corresponding MIP.
All loading fractions and washing fractions with an aqueous content higher than 50% were directly injected into the HPLC, whereas the washing fractions with an aqueous content lower than 50% and all the elution fractions were collected, evaporated under nitrogen, reconstituted with 250 μL of H2O, and finally injected (100 μL).
The specificity of the extraction was evaluated by applying the same MISPE procedure on the respective NIP cartridge. The enrichment factor (EF) was calculated as the ratio of analyte concentration in the reconstituted extract and that in the loaded sample.
The MISPE protocol applied to real urine samples involved the same conditions as those optimized in artificial urine, except for the washing and/or the re-equilibration steps. In the case of MIP1 the washing volume was increased from 1 mL to 2 × 1 mL to obtain a cleaner extract. All packed cartridges were re-equilibrated by percolating 2 × 1 mL of MeCN:HCOOH (99:1, v/v) followed by 2 × 1 mL of MeOH, to guarantee the complete washing of the sorbent from any residue of the glucuronide and any other interferent present in the urine samples.
The evaluation of the eluted fractions obtained from MISPE in real urine samples was carried out setting column temperature at 45 °C and employing a gradient program slightly different from that described above. In detail a gradient profile of 0–7 min A (10–50%), 7–11 min A (50%), 11–12 min A (50–90%), 12–15 min A (90%), 15–16 min A (90–10%), and 16–19 min A (10%) was used for fractions collected from the MIP1 cartridge, whilst a gradient elution of 0–7 min A (10–45%), 13–14 min A (45–90%), 14–17 min A (90%), 17–18 min A (90–10%), and 18–21 min A (10%) was applied for those obtained from the MIP2 cartridge. In order to quantify the extracts, blank urine samples were preconcentrated in both MIP cartridges following the optimized MISPE protocols, checked for not containing the analytes at the method detection limit and spiked with working standard solutions of TG and T. Linear calibration curves in the range between 15 ng mL−1 and 2400 ng mL−1 for TG and between 15 ng mL−1 and 200 ng mL−1 for T were built to quantify the extracts obtained from the MIP1 cartridge, whereas linear calibration ranges between 15 ng mL−1 and 500 ng mL−1 for TG and between 15 ng mL−1 and 90 ng mL−1 for T were built to calculate the recovery of the analytes after MISPE on MIP2. All calibration curve equations and correlation coefficients are reported in the ESI (Fig. S1†).
The limit of detection (LOD) and the limit of quantification (LOQ) were the minimum detectable amounts of analyte in blank urine extracted and then spiked giving a signal-to-noise ratio of 3 or 10, respectively. For both TG and T the LOQ was 15 ng mL−1, whereas the LOD was 10 ng mL−1.
The identity of glucuronide and the parent drug were confirmed by the UV spectrum and by comparing the chromatograms obtained after the injection of the extract of spiked urine with those obtained after the injection of the extracts of the blank urine and of the blank urine spiked only after performing the MISPE protocol (see ESI, Fig. S2 and S3†).
As anticipated in the Introduction, a urinary concentration of TG > 200 ng mL−1 serves as a marker for possible testosterone abuse among males.4 The free steroid (T) occurs at extremely low levels in urine, given that less than 3% of testosterone is excreted as free fraction.37 Therefore, the MISPE experiments were carried out by using artificial urine spiked at 300 ng mL−1 and 50 ng mL−1 for TG and T, respectively. A TG concentration of 300 ng mL−1 was selected in order to reproduce altered profiles of TG in urine, whilst a T concentration of 50 ng mL−1 was chosen so as to avoid UV detection problems and to test the recognition properties of the MIP when the parent drug concentration is quite lower than that of the glucuronide.
The elution step was established in the first instance. With the use of an organic solvent in the presence of an acid, such as formic acid (1%), we expected to effectively disrupt the interactions between the analyte and the sorbent functional groups and we therefore tested mixtures of MeCN and formic acid as eluents. The application of 1 mL of MeCN:HCOOH (99:1, v/v) allowed the quantitative recovery of both T (MIP1: R 94%, RSD 6%, n = 2; MIP2: R 103%, RSD 6%, n = 2) and TG (MIP1: R 102%, RSD 6%, n = 2; MIP2: R 102%, RSD 11%, n = 2) from the two MIP cartridges. Hence, this elution solvent was selected for further experiments.
As described in our previous work,25MIP1 and MIP2 show an excellent specific recognition of glucuronide from its parent drug in MeCN:H2O:AcOH (97.99:2:0.01, v/v/v) or MeCN:H2O (30:70, v/v), respectively. Therefore these were the first solvent mixtures tested to selectively remove the parent compound T from the corresponding imprinted cartridges.
With regards to MIP1, 0.3 mL of MeCN:H2O:AcOH (97.99:2:0.01, v/v/v) was the minimum volume required to totally wash out T and extract the glucuronide (41%). The percolation of a higher washing volume, e.g. 1 mL, does not significantly affect TG recovery while it further reduces the non-specific interactions. This is clearly demonstrated by the extract obtained from NIP1, where TG recovery decreases from 20% to 7%. So as to reduce the loss of TG, 1 mL of MeCN with the addition of AcOH (0.01%) was tested. A higher glucuronide recovery was obtained exclusively from the non-imprinted cartridge (36%), thus giving evidence that the presence of 2% H2O is essential to disclose the imprinting effect. The hydrogen bond capability of H2O, added in such a small percentage, is probably enough to disrupt most non-specific binding effects, whilst specific ones are not influenced at all.38
As far as the optimization of the washing step on MIP2, the percolation of only 0.2 mL of MeCN:H2O (30:70 v/v) causes a dramatic loss of both glucuronide and T. Moreover, the latter is still detected in the elution fraction (data not shown). So as to minimize the loss of TG, it was decided to test sequential washings, 4 × 0.2 mL or 4 × 1 mL of MeCN:H2O (15:85, v/v). We anticipated that a higher water content in the washing fraction could strengthen the binding of both glucuronide and the parent drug to the polymer matrix due to hydrophobic effects. Eventually, a sequential washing of 4 × 1 mL allowed a complete removal of T and a good recovery of TG (62%) from the MIP2 cartridge. In contrast, NIP2 releases most of the glucuronide during the loading and washing steps resulting in a much lower recovery of TG (7%).
The final step of the MISPE optimization was to identify the highest sample volume which could be loaded on the SPE cartridges without significant breakthrough. On the MIP1 cartridge, TG starts to break through in the collected loading fractions after loading more than 3 mL of spiked artificial urine. In contrast, glucuronide breakthrough occurs on MIP2 by applying only 0.4 mL of spiked artificial urine. Based on these results, a loading volume of 3 mL or 0.4 mL was established for MIP1 and MIP2, respectively.
MIP1 | NIP1 | MIP2 | NIP2 | |
---|---|---|---|---|
Replicate 1 | 32 | 9 | 62 | 7 |
Replicate 2 | 35 | 6 | 58 | 6 |
Replicate 3 | 37 | 9 | 60 | 39 |
Mean | 35 | 8 | 60 | 17 |
RSD % | 6 | 17 | 3 | 108 |
Despite low recovery, the high loading volume (3 mL) on MIP1 led to an enrichment factor for TG of 4.2, which is markedly higher than that achieved on MIP2 (EF = 1.0). This is also evident from the representative chromatograms reported in Fig. 3.
Fig. 3 Chromatograms of spiked artificial urine after the optimized MISPE protocol on (A) MIP1 and (B) MIP2. |
From this series of experiments it can be concluded that MIP1 is better suited as an SPE sorbent for extraction of TG from real urine samples.
Spiked level (TG–T, ng mL−1) | MIP1 | MIP2 |
---|---|---|
600–50 | 32 (2) | 38 (10) |
300–50 | 28 (5) | 47 (21) |
150–50 | 34 (3) | 32 (19) |
By comparing the data achieved in artificial urine (Table 1), it is evident that accuracy and precision of the MISPE procedure on MIP2 drastically decrease when moving to a real urine sample, whereas they do not significantly vary for the MIP1 cartridge. Repeating the extraction procedure on the corresponding NIP cartridges led to significantly lower recoveries of both the parent drug (not detected) and the glucuronide (≤3%) (see ESI, Fig. S6 and S7†). This further confirms the specificity of the extraction provided by the MIP cartridges.
Stability and regeneration of the MISPE cartridges were not investigated. Nevertheless, the same polymer material was used throughout the studies—i.e., MIP and NIP used for the HPLC characterization25 were unpacked from the chromatographic columns and repacked into the SPE cartridges. This at least indicates that polymers, and MIP1 in particular, retain specificity and precision for a long-time usage under stressed conditions.
Fig. 4 Chromatograms of spiked urine with TG (300 ng mL−1) and T (50 ng mL−1) after MISPE on (A) NIP1 and MIP1 and (B) NIP2 and MIP2 (NIP, dashed line; MIP, solid line). |
Although the recovery of testosterone glucuronide is low (28–34%), the MISPE protocol allows the direct determination of the conjugated metabolite with an excellent precision (RSD 2–5%, n = 3). This MIP recognizes the glucuronide moiety of the analyte, and therefore it would be useful to directly quantify not only testosterone glucuronide, but also those compounds which undergo the glucuronidation process, thus offering an important new tool for metabolite detection.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c1an15606c |
This journal is © The Royal Society of Chemistry 2012 |