Antipsychoticdrug poisoning monitoring of clozapine in urine by using coffee rin

ZHU Qingxia, YU Xiaoyan, WU Zebing, LU Feng, YUAN Yongfang*

(a Department of Pharmacy, Shanghai Ninth People’s Hospital, Shanghai JiaoTong University School of Medicine, Shanghai, 201999, China; b Department of Pharmaceutical Analysis, School of Pharmacy, Second Military Medical University, Shanghai, 200433, China)

Abstract: Antipsychotics are the drugs most often involved in drug poisoning cases, and therefore, therapeutic drug monitoring (TDM) for is necessary safe and effective medication administration of these drugs.In this study, a coffee ring effect-based surface-enhanced Raman spectroscopy (CRE-SERS) method was developed and successfully used to monitor antipsychotic poisoning by using urine samples for the first time. The established method exhibited excellent SERS performance since more hot spots were obtained in the “coffee ring”. Using the optimized CRE-SERS method, the sensitivity was improved one order more than that of the conventional method with reasonable reproducibility. The antipsychotic drug clozapine (CLO) spiked into urine samples at 0.5-50 g·mL-1 was quantitatively detected, at concentrations above the thresholds for toxicity. The CRE-SERS method allowed CLO and its metabolites to be ultimately distinguished from real poisoning urine samples. The coffee-ring effect would provide more opportunities for practical applications of the SERS-based method. The frequent occurrence of drug poisoning may have created a new area for the application of the CRE-SERS method. It is anticipated that the developed method will also have great potential for other drug poisoning monitoring.

Key words: antipsychotic drug poisoning monitoring; clozapine; surface-enhanced Raman spectroscopy; coffee ring effect; urine metabolites

1 Introduction

Clozapine (CLO) is an antipsychotic drug indicated for the treatment of resistant schizophrenia and is considered to be the most effective medication for this disease[1,2]. It belongs to the class of benzodiazepines, which have a rapid absorption and slow excretion in vivo. Drug poisoning which is a potential cause of death frequently occurs in patients medicated with CLO. According to a 13-year retrospective study[3]of severe poisoning and influence factors in acute human CLO intoxication, more than half of poisoning cases (59%) happened intentionally as suicide attempts or in the commission of a crime. Furthermore, 36% occurred accidentally because of confused or inadvertent ingestions, which is common in the elderly or in patients having psychiatric conditions with long-term prescription of multiple drugs.

Recommendations to monitor the real-time physiological load of therapeutic drugs in patients, referred to as therapeutic drug monitoring (TDM), have been increasing with the move towards precision medicine[4,5]. Drugs that typically warrant TDM include antipsychotics[6,7], chemotherapeutics[8,9], and antimicrobials[10,11], etc. Unlike other antipsychotic medicines, a safe therapeutic dosage range for CLO has not been clearly established, since large variations in individual responses result in a wide range of doses being prescribed. Although it is possible to measure plasma levels of other antipsychotics, CLO is the only antipsychotic where routine monitoring of levels is recommended[10,11]. After oral administration, only part of the CLO dosage reaches the systemic circulation unchanged and most of it is metabolized into N-desmethylclozapine (NDMCLO) and clozapine-N-oxide (CLONOX). Simultaneous detection of CLO and its main metabolites is important for poisoning death cases because, the result would be conducive for forensic calculation of the ingestion amount and time[13,14]. Moreover, for patients who require long-term administration of medication or those taking multiple antipsychotic medications concurrently, the TDM analysis of CLO would facilitate the rapid determination of drug poisoning, which would enable timely dosage adjustment and individualized drug administration.

Considering the demands of TDM and forensic analyses, methods used for drug poisoning detection should be rapid, inexpensive, portable, near real-time, and achieve sensitive determination of drug concentrations. The currently available methods include HPLC and some MS coupled methods[6,15,16]. These standard analytical chemistry techniques are sensitive and allow rapid quantification even when long-term combination therapy is used. However, they are expensive and require complex equipment as well as long analysis time. They can hardly be used to achieve real-time detection since samples must be transported to a facility for analysis and often end up in a queue[10]. New technologies for the implementation of TDM are highly desirable, since they could promote its use in drug poisoning monitoring.

Since its invention in the 1970s, surface enhanced Raman spectroscopy (SERS) has attracted attention because of its high sensitivity and rapid detection[17]. SERS is based on Raman scattering, the inelastic scattering of light unique to molecular vibrations inherent in the molecule. Although Raman scattering is a relatively rare event, noble metal nanoparticles provide localized plasmonic enhancement to the intensity of Raman scattering. Currently, SERS has been investigated as an alternative to traditional analytical techniques in procedures such as food safety evaluation[17], adulteration screening[18,19], environmental pollutant testing[20]and TDM analysis[21]. However, reports of SERS in drug poisoning monitoring from urine are still lacking.

The “coffee ring effect” (CRE) is widely known as a powerful tool for self-structuring of nanomaterials[22-24]including colloidal metal particles in suspension and metallic particles immobilized on solid substrates. However, the CRE has not been well exploited in those applications and its use based on SERS (“CRE-SERS”) has yet to be reported.

In this study, CRE-SERS was developed and applied for the first time to monitor CLO poisoning in urine samples. Here, we first report the findings of our CRE-SERS method combined with thin-layer chromatography (TLC) for CLO poisoning monitoring. Different states between the sample spot and the “coffee ring” formed by silver colloids were analyzed to identify the best CRE-SERS parameters. Using the optimized CRE method, the sensitivity was improved by one order with an acceptable reproducibility. Subsequently, CLO spiked into urine samples at 0.5-50g·mL-1was quantitatively detected, allowing the analysis of concentrations beyond the thresholds for toxicity (0.6-2g·mL-1)[1,3,13]. Finally, CLO and its metabolites in real urine samples were distinguished using the established method.

2 Experimental

2.1 Chemical reagents and materials

Clozapine(CLO) standard was purchased from the National Institute for Food and Drug Control, China. While the N-desmethylclozapine (NDMCLO) standard and the reference solution of clozapine N-oxide (CLONOX) were obtained from J&K Scientific (Beijing, China). Clozapine tablets (Spec: 25 mg×100 tablets) from Shanghai Xinyi Pharmaceutical Co., Ltd were used for the drug poisoning experiment. Silver nitrate (AgNO3), sodium citrate, sodium thiocyanate (NaSCN) and all organic solvents of analytical grade were bought from Fisher Scientific and used without further purification. Ultrapure water used throughout the experiments water was obtained using a Smart-DUV (18MΩ·cm resistivity) filter (Shanghai Hitech Instruments Co., Ltd., China). Other reagents were all of analytical grade. Commercial TLC plate (Yantai E.S.T. Silicone Tech Co., Ltd., China) consisted of high-performance silica gel 60-F254 plates (silica gel particle size: 8±2m≥80%, layer thickness: 0.2mm±0.03mm) with glass back plates was used. The plate containing fluorescing additive, F254, was used for easy spot visualization.

2.2 Apparatus and instruments

A centrifugal machine (HERAEUS, FRESCO 17,Thermo Scientific, USA) was applied during the protein precipitation process of urine sample. An ultraviolet analyzer (WFH-203B, Shanghai Jing Branch Industrial Co., Ltd., China) was used to locate the separated spots in the thin-layer plate. Separated spots were located using ultraviolet analyzer with 254nm wavelength (WFH-203B, Shanghai Jing Branch Industrial Co., Ltd., China). Scanning electron microscope (SEM) images were taken on a ZEISS EVO MA-10 (Carl-Zeiss, Germany). Ultraviolet-visible (UV-vis) absorption spectra of silver colloids were obtained with a double beam UV-Vis spectrophotometer (TU1901, Beijing Purkinje General Instrument Co., Ltd., China). Raman spectra were recorded by a portable Raman spectrometer (BWS415, B&W Tek Inc., USA) at 785nm, a resolution of 5 cm-1and a 20×long working distance microscope objective. TLC-SERS result of real urine samples were verified by using an Agilent Technologies 1290 Infinity-6538 UHD Accurate-Mass UPLC-QTOF/MS (Agilent Technologies, Germany).

2.3 Animals and treatment

Male Sprague-Dawley (SD) rats weighing 180-220 g were supplied by Sino-BritishSippr/BK Lab Animal Ltd. (Shanghai, China). The rats were kept in an air-conditioned animal quarter at 22±2 ℃ and 50±10% relative humidity for 7days before starting the experiments. Standard food (laboratory rodent chow, Shanghai, China) and water were ad libitum. After oral administered with a single 250 mg/kg body mass dose of clozapine tablets, they were housed in metabolism cages and urine was collected separately from the faeces over a 24h period. Blank urine samples were collected before drug administration to check whether they were free of interfering compounds. All samples were directly analyzed and then stored at -20 ℃. This animal experimental protocol was approved by the Ethical Committee (approval ID: 2013-0016) of North Shanghai 9thPeople’s Hospital, Shanghai Jiao Tong University School of Medicine (Shanghai, People’s Republic of China).

2.4 Sample preparation

CLOa nd NDMCLO stock solutions were prepared by dissolving standard substance in methanol with a concentration of 1mg/mL, while CLONOX sample solution was diluted to this concentration from the purchased commodity. Urine sample of 200L was added to a 1.5 mL centrifuge tube. Purification was performed adding 800L methanol followed by vortexing (5 min) and centrifugation at 10000 rpm for 10 min. The supernatants were collected in another tube and gently evaporated to dryness under a stream of nitrogen. 200L methanol was added to repeat the purification process, and then the supernatant was obtained as pre-treated urine sample for further analysis.

2.5 Nanoparticle preparation

Silver colloids were synthesized by the classical Lee-Meisel method[25]. Briefly, 45 mg of silver nitrate were dissolved in 250 mL of distilled water and the solution was boiled. A solution of 1% (w/v) sodium citrate (5 mL) was added to the solution under vigorous magnetic stirring and kept boiling for 1 h.

2.6 Analytical methods

Analyte stock solution of 1L was spotted to a silica gel 60-F254 plate, let dry, and then eluted with CHCl3:CH3OH 8.5∶1.5 (v/v). After the eluent on the HPTLC plate evaporated naturally, the separated spots were visualized and marked under ultraviolet illumination at 254 nm. SERS analyses were directly performed on the plate after local deposition of 4L silver colloid suspension directly on the marked spot. SERS spectra for the spots were acquired using a Raman spectrometer with a suitable power (200 mW) and an integration time of 5 s. All the measurements were repeated more than three times. Data were pretreated with the Savitzky-Golay polynomial fitting (9-point smoothing) and baseline correction, with Matlab 7.0 (Math works, Massachusetts, USA) and Origin 7.5 software.

2.7 Density functional theory calculation

Density functional theory (DFT) calculations provides a relatively efficient tool to characterize properties of molecules, bulk materials and their surfaces[26,27].To identify different Raman vibrational modes, DFT was applied to calculate the Raman spectra of CLO, NDMCLO and CLONOX. In the computational studies of this paper, optimized structures of the three analytes in the gas phase were obtained using the B3LYP exchange-correlation functional and 6-311+G* basis set. Calculations were performed using the Gaussian 09W package on the Linux system[30]. These optimized structures had the lowest energy state, thus had the most stable molecular structures.

3 Results and discussion

3.1 Separation of CLO and its metabolisms using TLC

Urine is known to consist substances such as water, urea, uric acid, creatinine, and ammonia, which might interfere with the SERS detection of target analytes except for water which has a weak Raman scattering effect. Therefore, a TLC analysis was conducted to extract CLO and its two metabolites from the complex matrices. After mobile phase optimization, a chloroform and methanol system was selected. The stock solutions of the three references (0.5 mg·mL-1), blank urine, and urine samples spiked with CLO, NDMCLO, and CLONOX at a concentration of 0.25 mg·mL-1were deposited onto a TLC plate (10 cm × 10 cm, height × weight, respectively) loaded at a distance of 10 mm apart and 1 cm from the bottom. Then, the plate was eluted in a TLC developing chamber previously saturated with chloroform and methanol (=8.5∶1.5,v/v). As shown in Fig. 1, the three analytes were extracted from the urine matrices and were distinguishable. The retardation factor (Rf) values of CLO, NDMCLO, and CLONOX were 0.68, 0.27, and 0.15, respectively. Moreover, the relative standard deviation (RSD) values of the Rf values were all <10% which indicated the reasonable reproducibility of the TLC. The well-distinguished Rfvalues might be useful for SERS detection when the chromatographic spot is invisible because of a low analyte concentration. Thus, this mobile phase was used in the subsequent analysis.

3.2 Optimization of CRE-SERS method

When a droplet of solutions containing nonvolatile solutes dries on a solid surface, it leaves a dense and ring-like deposit of the solutes along the perimeter. This phenomenonreferred to as the “coffee-ring effect (CRE)”[22], is familiar to anyone who has observed a dried drop of coffee. During the drying process, the droplet edges become pinned to the substrate, and capillary flow outward from the center of the drop moves suspended particles to the edge as evaporation proceeds. After evaporation, the suspended particles are left highly concentrated along the original drop edge. The CRE occurs in systems with diverse constituents ranging from large colloids[28-29]to nanoparticles[30]. In this study, silver colloids prepared by the classical Lee-Meisel method were applied as SERS substrates due to the simple synthesis in batches at relatively low cost. UV-vis spectroscopy and SEM imaging were employed to characterize the dispersibility and morphology of the prepared silver colloids (Fig. S1). An absorption maximum at 420 nm with a full width at a half-maximum of approximately 90 nm was observed, indicating that relatively monodispersed silver nanoparticles existed with a diameter of 50-60 nm[20]. This was in agreement with the SEM result. Silver nanoparticles aggregating in a narrow area of the “coffee ring” lead to an increase in the number of hot spots[21], which would be propitious for SERS enhancement. Thus, SERS performance of different positions along the diameter of coffee ring formed by silver colloids, and its relative distance to the TLC spots were studied to optimize the CRE-SERS method.

Fig. 1 TLC analysis test samples: TLC of (1) blank urine sample, (2) urine sample spiked with references, and references of (3) CLO, (4) NDMCLO, and (5) CLONOX

The CLO reference solution (1mg·mL-1) was used in the optimization process. After elution by the optimized mobile phase, the naturally evaporated TLC plate was placed under UV illumination at 254 nm to mark the CLO spot. Due to the strong capillary action between the TLC plate and the solvent, a “coffee ring” can be observed immediately (within 5 seconds) after a drop of silver sol. During the study, it was found that the size and shape of the coffee ring formed by 4L silver sols of the original concentration were basically the same at a temperature of 20 ℃ and humidity of 40%. The three coffee rings obtained by parallel operations were shown in Fig. S2. The diameters were 6.02, 6.18, and 6.10 mm, respectively (RSD=1.31%). It indicated that the formation of the coffee ring can be reproduced. SERS detection was performed directly after 4L of silver colloid suspension was locally deposited onto the marked spot (Fig. 2a). Then, the SERS spectrum of CLO with rich signals was obtained. To enhance the visualization of the CRE, TLC images of the silver colloid and CLO spots, as well as their corresponding densitograms obtained by a TLC scanner with Goodlooking software, were enlarged (Fig. 2b). SERS detection was conventionally conducted with laser illuminating at the center of the sample spot since the concentration was the highest at the center[17-18], as shown in its densitogram. From the TLC densitogram of the silver colloids, the highest concentration position was located somewhere between the center and edge and could be observed more obviously from their TLC image. Position c had the highest density of nanoparticle distribution due to the ring-like deposition, while that of position b and a decreased successively.

Fig.2 Optimization process of coffee ring effect based surface-enhanced Raman spectroscopy CRE-SERS method: (a) Schematic illustration of CRE-SERS detection. (b) TLC and corresponding densitograms of silver colloid spot and sample spots under UV 254 nm. Arrows of a-d represent different positions along the “coffee ring” diameter. (c) TLC images of different covering states between silver colloid and sample spots. Images a-d represent sample spot centers through different positions along the “coffee ring” diameter, corresponding to positions a-d, respectively in Fig. 2B. Red dot represents 785 nm laser beam. (d) Relationship between exposure time and relative intensity of peak at 1040 cm-1 under the four covering states.

Image a, which represents the conventional detection method used in previous studies[18,36], showed the highest reproducibility. However, the spectral intensity of image a was the lowest, and the SERS intensity of the three unconventional states were all better than that of image a. Thus, the conventional detection method (image a) did not exhibit a superior sensitivity. Image c presented an opposite result, showing the highest SERS enhancement as well as the worst reproducibility. The nanoparticles aggregating at position c increased the hot-spots, which was reflected in the higher spectral intensity. Aggregates with multiple particles yielded large enhancements due to the enormous electromagnetic field enhancement at junction sites[37]. However, aggregated metal colloids tend to coagulate, which makes them less stable, leading to poor RSD values. Therefore, position c presented the highest intensity as well as the worst reproducibility compared to other positions along the diameter. The most sensitive state, similar to that of image c, would not be preferred in quantitative applications due to the poor reproducibility. Although image d showed respectable reproducibility and higher spectral intensity than image a did, the result of the analysis of image b appeared much better. Under the detection state similar to that in image b, the SERS spectra presented good reproducibility during the exposure time of 0-160 s. Furthermore, the SERS spectral intensity increased with increasing exposure time and ultimately reach a plateau at the 100th second. The hit-quality index (HQI) value of the spectra during the stable period (the 100th second) was calculated to be more than 0.99 and the corresponding RSD value of the spectra intensity was 10.35% lower than 20%, which indicated reasonable reproducibility as far as SERS is concerned. It also can be concluded that the spectral intensity at the plateau was more than two times higher than that of image a was, which indicates that the CRE-SERS method exhibited a better SERS performance than the conventional detection method did (image a). Therefore, the state shown in image b was selected as the optimal detection condition and the spectra at the stable period (plateau) were used in our subsequent detection. Further quantitative analyses were conducted using the optimized CRE-SERS method.

3.3 SERS spectral comparison of CLO and its metabolites

The SERS spectra of CLO and the two metabolites (NDMCLO and CLONOX) were recorded using the above optimized CRE-SERS method. Fig. 3 shows the structures as well as the SERS and corresponding 2nd derivative spectra. The three analytes showed a remarkable similarity in most regions of the SERS spectra because of their extremely similar structure. A careful comparison of all the features in the SERS and 2nd derivative spectra revealed differences. CLO could be differentiated from NDMCLO and CLONOX based on its characteristic peaks at 420, 621, 1181, and 1226 cm-1. NDMCLO presented distinguishable SERS features at 427, 754, 1016, and 1266 cm-1, and although the last peak was not obvious in its SERS spectrum, it was recognized easily from the 2nd derivative spectrum. The peak at 708 cm-1, assigned to N-H rocking, was an obvious feature in the SERS spectra of CLO and NDMCLO. However, the peak showed a blue shift to 714 cm-1in the SERS spectrum of CLONOX, which might be attributable to N-O stretching. Thus, 714 cm-1could be considered a unique feature of CLONOX. Moreover, additional characteristic signals for identification were found in its 2nd derivative spectrum such as 628 and 671 cm-1. All the assignments of the SERS peaks were from the DFT calculations, and the computational details are shown in Table S1. It could be concluded that CLO and the two metabolites all presented good SERS signals in the CRE-SERS method and were distinguished by their characteristic peaks. To evaluate the sensitivity of the established CRE-SERS method, a series of concentrations of CLO, NDMCLO, and CLONOX solutions ranging from 10 to 0.01g·mL-1were measured using both the CRE-SERS detection and conventional SERS methods (Fig. S3). The LOD of the CRE-SERS method for CLO was 0.1g·mL-1, which showed a one order magnitude of increased enhancement compared to that of the conventional SERS method which was 1g·mL-1. Similarly, for NDMCLO and CLONOX, the CRE-SERS method showed better sensitivity than the conventional SERS method did. The LODs for NDMCLO and CLONOX were 0.04 and 5g·mL-1, respectively. The established method showed good specificity and acceptable sensitivity for CLO and its metabolites. Thus, it can be applied in the detection of real urine samples.

3.4 Interference experiment

Since the CLO poisoning cases might occur in situations where other medications have been simultaneously consumed, aninterference experiment was conducted to evaluate the selectivity of the established method. Though there were too many kinds of simultaneously consumed drugs, the interference drugs can be divided into two types from the methodological perspective, drugs interference with the TLC analysis of CLO and no drugs interference with the TLC analysis of CLO. Chlorpromazine was selected as the representative drug that showed chromatographic interference in the identification of CLO, while caffeine was selected as the other type. Then, the CLO, chlorpromazine, and caffeine references were spiked into a blank urine sample. Fig. S4A depicts the TLC result of the simulated positive urine sample and corresponding references. All the three references were efficiently separated from urine matrices. Although chlorpromazine presented a Rfvalue similar to that of CLO, which might interfere with the TLC analysis result, they can be easily differentiated based on their SERS spectra. As shown in Fig. S4B, the SERS spectra of chlorpromazine and caffeine showed high specificity with abundant characteristic signals. Chlorpromazine and CLO were distinguished based on their SERS bands at 439, 550, 678, 1106, and 1461 cm-1, and 472, 527, 621, 709, 1181, 1227 and 1600 cm-1, respectively. Caffeine can be easily distinguished from CLO by not only the different Rfvalues but also the specific SERS peaks. The result of interference experiment indicated that the method afforded good selectivity, which was attributable to the angle of not only spectroscopy but also chromatography.

Fig. 3 Structures and spectral comparison of clozapine and metabolites: (a) Structures, (b) SERS spectra and (c) 2nd derivative spectra of (a) clozapine (CLO), (b) N-desmethylclozapine (NDMCLO), and (c) clozapine-N-oxide (CLONOX).

3.5 Quantitative analysis of CLO using CRE-SERS method

3.5.1Selection of internal standard

Quantitative analysis using the SERS method is widely known to be challenging[38]. However, considerable efforts have been made[39,40]to obtain convincing results. Designed and optimized SERS substrates with high sensitivity and reproducibility are currently prevalent[17，21，23]. Different substrates cause significant variations in the absolute SERS intensities, which may dramatically affect the accuracy and precision of quantitative results. An internal standard may be one of the choices for regulating the SERS peaks intensity when conducting quantitative analysis. The internal standard must exhibit stable and unique Raman scattering in the Raman silent region of the analytes. Specifically, the internal standard should have low effects on the characteristic SERS peaks of the sample. In this study, three kinds of internal standards were investigated, a solvent (methanol)[41], sodium thiocyanate (NaSCN)[35，42], and melamine[43](Fig. S5). A characteristic SERS band of 2120 cm-1of NaSCN was present in the silent region of CLO, which was not observed in that of methanol or melamine. Moreover, NaSCN as an internal standard had little effect on the SERS of the sample, which could eliminate the effects of the matrices and correct the unstable laser intensity[35]. Thus, NaSCN was suitable for the quantitative analysis of CLO.

3.5.2Quantification of CLO in urine

In this study, the internal standard (2.5 mg·mL-1NaSCN) solution was mixed with the silver sols (1∶60, v/v) before the quantification of CLO using the CRE-SERS method. The SERS intensity ratio of peaks 1040 and 2120 cm-1from CLO and NaSCN, respectively (I1040/I2120) was used for the quantitative analysis. The SERS reproducibility of the mixed sols was investigated, as shown in Fig. S6. The relationship between the standing time of the mixed sols and the I1040/I2120revealed that the intensity ratio was fairly stable during the standing time of 60 min. This observation indicates that the mixed treatment with silver sols used in this study was feasible.

A series of CLO concentrations in the urine samples ranging from 0.5 to 50g·mL-1were measured using the established CRE-SERS method coupled with TLC. CLO was separated from the simulated positive urine samples using a TLC plate (eluent, chloroform/methanol, 8.5∶1.5, v/v). The separated spots were then located using UV illumination. Finally, the SERS spectra were recorded in the presence of silver colloids using the internal standard. Fig. 4a depicts the measured spectra for a representative selection of the CLO concentrations (0.5, 1, 3, 7, 30, and 50g·mL-1). The relative intensity of the peak at 1040 cm-1increased concentration-dependently while that at 2120 cm-1remained constant. The corresponding linear relationship is presented as an inset in Fig. 4b. The calibration curve presented a good linearity with a correlation coefficient (R2) of 0.9997 and small error with an average RSD of 9.28% in the range of 0.5-50g·mL-1. The linear range covered the toxicity thresholds for CLO (0.6-2g·mL-1). Moreover, the methodological evaluation using three urine samples spiked with CLO showed that the recovery rates were 86.98-119.64%, and the RSD was 9.60% (Table S2). We also found that a power function could be fitted to the data with an R2of 0.9820 in the range of 11-5500g·mL-1. The result indicated that the calibration curve had the potential for use in the quantitative analysis of CLO in urine samples while predicting the concentration of the unknown sample efficiently.

Fig. 4 Calibration curve of CLO using CRE-SERS methods: (a) SERS spectra of NaSCN (a), CLO (i), and different concentrations of CLO with NaSCN as internal standard. Gray dotted lines represented selected bands to establish the calibration curve using the intensity ratio of each spectrum (b-h: 50, 30, 7, 3, 1, and 0.5 g·mL-1, respectively). (b) SERS intensity ratio of peaks 1040 and 2120 cm-1 was used to generate a calibration curve. A power function was fit to the data in the range of 11-5500 g·mL-1 with R2 of 0.9820. The inset shows the signal at lower concentrations of 0.5-50 g·mL-1 with R2 of 0.9997. Error bars shows SD.

The established method was used to analyze real CLO poisoning in rat urine samples. The TLC images of different concentrations of urine samples and the corresponding SERS spectra are depicted in Fig. 5. As shown in Fig. 5a, although spots a-c with Rfvalues of 0.68, 0.27, and 0.15 corresponding to that of CLO, NDMCLO, and CLONOX, respectively were not obvious on the chromatography strip of sample 3 (the original concentration), the concentrated samples presented obvious spots at these positions. To further identify these spots, the SERS spectra of spots a-c were analyzed as shown in Fig. 5b. The SERS spectrum of position a with the same Rfvalue (0.68) as that of CLO presented signals at 420, 472, 621, 709, 1040, 1181, 1227, 1573, and 1602 cm-1, which were consistent with the main Raman features of the SERS spectrum of the CLO reference. Peaks at 420, 621, and 1181 cm-1could be used to differentiate CLO from the other metabolites as depicted in Fig. 3. At the same time, the SERS spectrum of spot b showed similar characteristic Raman features (427, 594, 754, 1010, and 1266 cm-1) to those of NDMCLO. No similar result was obtained from the SERS spectrum of spot c. These observations indicate that CLONOX, CLO, and NDMCLO could be detected from the poisoned rat urine. Urine matrices existed obviously at positions with the same Rfvalue as that of CLONOX (Fig. 5A), which would interfere its identification. Moreover, the higher LOD value might be another disadvantage for the detection of CLONOX. Nevertheless, the specificity and accuracy in distinguishing CLO (rather than other components) were ensured by its characteristic Raman features as well as the simultaneous identification of NDMCLO. Then, the concentration of CLO was calculated to be 20.33g·mL-1according to the calibration curve (Fig. 5B). The result was consistent with the LC-MS/MS result (Table S3).

4 Conclusion

In summary, we successfully developed a TLC-coupled CRE-SERS method for the qualitative and quantitative analysis of antipsychotic drug poisoning monitoring using urine samples. The CRE-SERS method exhibited excellent SERS performance since more hot spots were obtained. The optimized CRE-SERS method exhibited an improved sensitivity with good reproducibility. The optimized method distinguished the antipsychotic drug CLO and its metabolites from real drug poisoning urine samples. Furthermore, the concentration of CLO was quantified, and the tested linear range (0.5-50g·mL-1) covered the toxicity threshold of CLO (0.6-2g·mL-1), potentially enabling monitoring of this life-saving therapeutic. The frequent occurrence of drug abuse-induced poisoning may have created a new area for the application of CRE-SERS. Further efforts are still needed to develop efficient methods for the quantification of drug poisoning serum, plasma, or whole blood samples. We also anticipate that the improved method established in this study is a potential technique for other drug poisoning monitoring.

Fig. 5 TLC images of different urine sample concentrations and SERS spectra analysis: (a) TLC image of real CLO poisoned rat urine samples. Samples 1 and 2 represent four and twice times concentrations of sample 3, which represents original concentration of urine sample. Spots a-c represent positions with Rf values of 0.68, 0.27, and 0.15, respectively corresponding to those of CLO, NDMCLO, and CLONOX. (b) SERS spectra of spots a-c in Fig. 5A and CLO, NDMCLO, and CLONOX references.

Acknowledgement

This study was financially supported by Shanghai Municipal Commission of Health and Family Planning (Grant no. 2016ZJP002), Shanghai Jiao Tong University Schoolof Medicine (Grant no. JDYX2016ZD004), and the National Natural Science Foundation of China (Grant no. 81573598).