Green and sustainable drug analysis – Combining microsampling and microextraction of drugs of abuse

A novel method combining microsampling of whole blood containing drugs of abuse with 96-well liquid solution microextraction ( Parallel Artificial Liquid Membrane Extraction , PALME) and analysis by ultra-high-performance liquid chromatography-tandem mass spectrometry (UHPLC-MS/MS) was evaluated. Different donor solutions, supported liquid membranes and acceptor solutions were tested, and the most promising set-up was validated. The method included common classes of drugs of abuse, such as opioids, amphetamines, cocaine, benzodiaze-pines and z-hypnotics. Extraction recoveries above 70% were found for 13 of the compounds, while the last four compounds had recoveries of 10 – 58%. THC was not extracted with the current method. A linear calibration model was found for all drugs but morphine. Limits of quantitation were between 1 and 5 ng/mL and inter-day precision and accuracy was within 20% for all compounds except for morphine and zopiclone that had a CV of 25% at LOQ. All matrix effects were within 78 – 123%. Samples were stable for 14 days except for zopiclone and zolpidem. With low-cost, high sample throughput, semi-automated miniaturized sample preparation in combination with dried blood microsamples, and an Eco-Scale score of 78, the proposed method fulfills green chemistry principles.


Introduction
In green chemistry a set of principles is applied, that leads to the reduction or elimination of the use or generation of hazardous substances (Anastas and Warner 1998).In the field of analytical chemistry the key factors to adhere with green chemistry principles are miniaturization and automation of the sample preparation process (Abdel--Rehim et al., 2020).Both sorbent-based and liquid-based microextraction approaches for sample preparation have been highlighted as promising tools for greener bioanalysis (Gjelstad 2019;He and Concheiro-Guisan 2019;Abdel-Rehim et al., 2020).For liquid-based microextractions, membrane based methods such as hollow-fiber liquid solution microextraction (HF-LPME), 96well LPME, or electromembrane extraction, are suggested as the most appropriate for handling complex biological samples.These techniques require less than 10 μL of organic solvents compared to traditional LLE that operates in the mL range (Gjelstad 2019).
Analysis of drugs of abuse is an important branch of bioanalysis.This type of analysis is necessary in many fields: to control drivers suspected of driving under the influence of drugs, perform workplace drug testing, follow up of prisoners, pain patients or patients in opiate maintenance treatment programs or for epidemiological studies in different populations.To decide if a driver is under the influence of drugs (Busardo et al., 2018) or if a patient has taken their opiate maintenance medicine in compliance with the prescribed amount of medication, analysis of blood is necessary.To preserve blood it should be cooled for short term storage, and frozen for long time storage.With a wider perspective for sustainable bioanalysis, reduction of the energy used for storage and cost of operation could be included, and the cost of facilitating the testing process should include the ones related to specimen collection (Jarvis et al., 2017).The use of dried blood microsamples, which can be shipped and stored, at least temporarily, at room temperature and classified as non-hazardous could drastically reduce costs and energy usage, as well as enabling drug monitoring closer to the point of care (Lei and Prow 2019).
Conventional blood microsamples, dried blood spots (DBS), was first proposed for glucose monitoring (Bang 1913), and have been used worldwide for neonatal drug testing since Guthrie and Susi demonstrated the applicability for phenylketonuria screening (Guthrie and Susi 1963).Dried blood spots are known to have issues with homogeneity of the samples due to different hematocrit content (Denniff and Spooner 2010), and several devices have been made to overcome this problem (Nys et al., 2017).One of the most promising is a device called volumetric absorptive microsampling (VAMS), that has been found to overcome area bias and homogeneity issues (Denniff and Spooner 2014).Summaries of methods using VAMS have been published (Kok and Fillet 2018;Protti et al., 2019); methods for therapeutic drug monitoring of antiepileptic drugs, immune suppressants or antibiotics have been the most common.Only a few methods for VAMS sampling of drugs of abuse have so far, to our knowledge, been published (Mercolini et al., 2016;Protti et al., 2017;Protti et al., 2018;Mandrioli et al., 2020).
To combine microsampling and microextraction for analysis of drugs of abuse would provide green chemistry drug analysis.A recent review summarized microextraction methods for drugs of abuse (He and Concheiro-Guisan 2019), and several other recent articles confirm the interest for liquid microextraction techniques for drugs of abuse analysis (Fisichella et al., 2015;Lin et al., 2017;Vårdal et al., 2017;Vardal et al., 2018;Vårdal et al., 2018;Fernández et al., 2019;Vardal et al., 2019;Ares-Fuentes et al., 2020;da Cunha et al., 2020).A recent paper presents a method combining microsampling with VAMS and miniaturized pretreatment by dispersive pipette extraction (DPX) (Mandrioli et al., 2020), this gives a very fast sample preparation using only 125 μL of organic solvents.Another interesting article (Ask et al., 2018) described the use of 96-well LPME (PALME) for the extraction of model basic and acidic drug substances from conventional DBS.The one step extraction process was found to promote release of the analytes from the DBS into the desorption solvent, and recoveries between 63 and 85% were found for the six model analytes, amitriptyline, quetiapine, ketoprofen, fenoprofen, flurbiprofen, and ibuprofen.A simple workflow and highly efficient sample clean-up with respect to phospholipids was in addition demonstrated.The low cost of PALME, less than 1€ per sample (Gjelstad 2019), is another benefit of this technique.
The aim of our study was therefore to combine VAMS microsampling of whole blood containing drugs of abuse with microextraction using PALME, exploring the possibilities for a robust and efficient green chemistry analysis method.

Standard, calibrator and internal standard solutions
Stock solutions of each analyte were made in methanol or acetonitrile; zopiclone and zolpidem were dissolved in acetonitrile and protected from light to prevent degradation.Stock solutions were stored at 4 • C. Working solutions in acetonitrile:water, 30:70 (v/v) were prepared from the stock solutions for all compounds but zopiclone, which was prepared separately in acetonitrile.The working solutions were kept at 4 • C, and calibrators and quality control (QC) samples were prepared by dilution of the working solutions with Milli-Q water before addition to whole blood.Internal standard solution were prepared by mixing appropriate amounts of stock solutions in methanol or acetonitrile and diluting with water:ethanol 2:1 (v:v).Concentrations for calibratiors, QC-samples and internal standards are given in Supplementary Tables 1-3.

Whole blood samples
Human whole blood with sodium fluoride and heparin added as preservatives was supplied by Oslo University Hospital Blood Bank.Blood was stored at − 20 • C in brown glass bottles, and thawed prior to preparation of calibration standards and quality controls (QC).Calibrators and QC-samples were prepared using Mitra® devices from Neoteryx, by dipping the VAMS-tip into spiked whole blood making sure they were not submerged and waiting an additional 2 s after the tip was fully red as recommended by the manufacturer.Samples were dried between minimum 2 h and 2 days before extraction.
Dried blood samples were collected as part of a so-called road side study of drug use among random drivers, in collaboration with the Norwegian Mobile Police Service and Oslo Police District.The study was performed in the south-eastern part of Norway (Furuhaugen et al., 2018).After making a finger prick with a lancet the blood samples were collected by Oslo University Hospital staff using 20 μL Mitra® devices, Neoteryx, using volumetric absorptive microsampling, VAMS®, technology.Participation was voluntary, and the study was approved by the Norwegian Regional Committee for Medical and Health Research Ethics (approval no.2015/2092).

PALME equipment and procedure
The PALME equipment was assembled from a polypropylene 96-well donor plate with 0.50 mL wells, Agilent, and a 96-well acceptor plate with polyvinylidene fluoride (PVDF) filters with pore size 0.45 μm from Millipore.The PVDF filters served as support for the supported liquid membrane (SLM).The 20 μL VAMS-tips were put in individual donor wells, and in the final procedure 185 μL ammonium carbonate buffer pH 9.3 and 20 μL internal standard was added.The SLM was prepared by pipetting 4.0 μL of the mixture 6-undecanone:dihexyl ether (1:1) with 1% trioctylamine (w/w/w) onto the PVDF-filters and the acceptor wells were filled with 100 μL acceptor solution containing DMSO mixed with 200 mM HCOOH (75:25, v/v).The donor and acceptor plates were clamped together and sealed with a Platemax pierceable aluminum sealing film from Axygen to prevent evaporation during extraction.The PALME setup was then placed on a Vibramax 100 platform shaker (Heidolph Instruments) providing an agitation of 900 rpm and extracted for 60 min.An aliquot of 60 μL from each well was transferred to a Nunc 96-well Polypropylene MicroWell plate (Thermo Fisher Scientific) with 450 μL conical wells and diluted with 40 μL deionized water before UHPLC-MS/MS analysis.A Liquidator 96-channel Manual Pipetting system, Mettler Toledo, was used to efficiently add all solutions except for the SLM and internal standard, which were added by an Xstream Eppendorf Multipipette, VWR.

Chromatographic analysis
Ultra-high performance liquid chromatography tandem mass spectrometry (UHPLC-MS/MS) was performed with an I-class Acquity UHPLC with flow-through needle injector and a Xevo TQS triple quadrupole mass spectrometer, both from Waters.Separation was achieved with a Kinetex® Biphenyl column (100 × 2.1 mm I.D., particle size 1.7 μm, pore size 100 Å) from Phenomenex.The column temperature was 60 • C, and the injection volume was 5 μl.A pre-injection wash of 2 s and a post-injection wash of 6 s were used, with methanol:water, 90:10 (v/v) as the purge solvent.Mobile phase A consisted of 10 mM ammonium formate buffer pH 3.1 and mobile phase B was methanol (B).Initial conditions were 2% B held for 0.2 min, followed by an increase to 25% B at 0.3 min, and a linear rise to 70% B at 4 min.The composition was changed to 99% B at 5 min, and held until 6.2 min, before returning to initial conditions at 6.3 min.The total cycle time was 8 min.
MS/MS acquisition was performed in the multiple reaction monitoring (MRM) mode with positive electrospray ionization.The capillary voltage was set to 1.35 kV and the ion source temperature to 150 • C. Argon was used as collision gas and maintained at a pressure of 0.0043 mbar in the collision cell.Nitrogen was used as desolvation gas, delivered at a flow of 1000 L/h with a desolvation temperature of 500 • C, while cone gas (nitrogen) flow was 300 L/h.Instrumental parameters (MRM transitions, cone voltages and collision energies) used for the measurements are provided in Supplementary Table 4 (analytes) and Supplementary table 5 (internal standards).System operation and data acquisition were controlled using MassLynx 4.1 software, and processed with the TargetLynx quantification program from Waters.

Method development
Based on previous literature and compound characteristics a 3 × 3 × 3 combination of donor solutions, SLMs and acceptor solutions were set up and extracted from whole blood giving 27 unique combinations, Table 1.Donor solutions were tested at pH 7.5, 9.3 and 14.Dodecylacetate and combinations of undecanone and dihexyl ether with additives (TOA and DEHP) were tested as SLMs and acceptor solutions consisted of formic acid with or without DMSO.
The use of DBS or VAMS provides a small amount of blood for extraction, and the 20 μL tips were chosen over the 10 μL to provide more material.The size did not hinder free movement in the well during agitation.

Method validation
Validation was performed on VAMS samples prepared from spiked whole blood according to internal guidelines that are very similar to SWGTOX-guidelines (Scientific Working Group for Forensic Toxicology 2013), evaluating the following parameters: limit of detection (LOD), limit of quantitation (LOQ), calibration model, accuracy and precision, recovery, matrix effects (ME), interferences and carry-over.

Limits of detection and quantification
The limits of detection was determined by analyzing a dilution series in whole blood sampled with VAMS.The samples were extracted in triplicate, and LOD was found as a signal-to-noise ratio (S/N) ≥ 3. The LOQs were determined from accuracy and precision data for the lowest QC level where precision and accuracy were ±20%.

Calibration model
The calibration model was evaluated based on five replicates of six standards measured in different runs, as recommended by SWGTOX (Scientific Working Group for Forensic Toxicology 2013).All data points from the five runs were plotted together and evaluated using the curve estimation function in IBM SPSS Statistics for Windows version 23.0.

Accuracy and precision
Spiked and extracted VAMS sample calibrators and QC-samples were used to determine intra-and inter-day accuracy and precision.For intraday accuracy and precision, eight replicates of the QC samples were extracted and analyzed, while three replicates extracted and analyzed at eight consecutive days were used for inter-day accuracy and precision.Accuracy was given as % bias -the difference between measured analyte concentration and theoretical value, while precision was given as the coefficient of variation for the measured values.

Extraction recovery and matrix effects
Extraction recovery and matrix effects were quantified (Matuszewski et al., 2003) at the QC 2 and QC 4 levels.Six different blood batches were spiked before extraction and collected with VAMS, and compared to the same blank blood collected with VAMS and spiked after extraction.Recovery was calculated as the ratio of peak height of analyte added before extraction to peak height of analyte added after extraction.Matrix effects were found by comparing post-extraction spiked VAMS samples with neat standard solutions, and the ME was considered acceptable between 80 and 120%.In all cases internal standard was added after extraction.

Stability and long term extractability
Stability and long term extractability of freshly dried and extensively dried samples was evaluated by comparing four replicates of VAMS samples dried for 1 day, 2 days and 2 weeks with samples dried for 2 h.

Interference studies
Interferences from other commonly encountered analytes were evaluated by analysis of three mixtures containing antipsychotics, antidepressants, antiepileptic and cardiac drugs, Supplementary Table 6, spiked in whole blood containing drugs of abuse in a concentration equivalent to QC 2. VAMS samples were prepared in triplicate from each mixture and dried overnight before analysis.

Carry-over
Carry-over was examined by injecting extracted blank samples after the highest calibrator and a sample with concentration of three times the highest calibrator.Carry-over was defined as blank samples displaying peak height of 20% or more of the peak height at LOQ.

Method development
To find conditions suitable for all compounds can be a challenge in all extraction processes.In LPME microextraction the compounds are transported through the liquid membrane via a pH gradient.The compounds should be uncharged in the donor, and charged in the acceptor to prevent back-extraction.In the original PALME paper, serum with 20 mM sodium hydroxide comprised the donor solution, dihexyl ether the SLM and 20 mM formic acid the acceptor solution (Gjelstad, Rasmussen et al., 2013).Stimulating new psychoactive substances have been extracted from plasma and whole blood using 40-80 mM sodium hydroxide in the donor solution, a 5 μL dodecyl acetate with 1% trioctylamine (w/w) SLM and 20 mM formic acid acceptor solution (Vårdal  (Ask et al., 2016;Olsen et al., 2018).
For more polar basic compounds or basic compounds with low pKa values the extraction can be more difficult.Pilařová et al. proposed a method for more efficient extraction of more polar analytes from plasma with 25 mM phosphate buffer pH 7.0 in the donor, 2.5 μL 15% bis(2ethylhexyl)phosphate (DEHP) in 2-nonanone as SLM, and 150 mM TFA in the acceptor solution (Pilarova, Sultani et al., 2017).The pH was chosen so that DEHP could form ion pairs with the polar analytes to transport them into the SLM.Vårdal et al. described the extraction of benzodiazepines and z-hypnotics from whole blood added with 50 mM phosphate buffer, pH 7.5 as the donor, a 4.0 μL SLM with 2-undecanone and dihexyl ether (1:1, w/w) with 1% trioctylamine (w/w) and an acceptor solution with DMSO mixed with 200 mM HCOOH (75:25, v/v) (Vårdal et al., 2018).The highly acidic acceptor solution and added organic solvent was necessary to extract the benzodiazepines, with their low pKa, over to the acceptor phase.
A general drugs of abuse screening consists of both polar compounds such as morphine and hydrophobic compounds such as methadone or THC, and problems with passing into the SLM for polar compounds, or passing out of the SLM for unpolar compounds can be encountered.Different substances will in addition have different optimum pH.While e.g.diazepam will be uncharged for all pH conditions above pH 5, methamphetamine with pKa 10.4 would be partly ionized until pH is above 12.At very high pH several opioids as well as THC, clonazepam and nitrazepam will exist as negative ions.
Fig. 1 shows results for ten compounds and six of the 27 conditions tested.THC was not extracted under any of the conditions, and had to be omitted from the method.Condition 2 has the same set up as the method previously published for benzodiazepines (Vårdal et al., 2018), and condition 22 corresponds to the set up used for stimulating NPS and antidepressants (Vårdal et al., 2017;Olsen et al., 2018;Ask et al., 2019).Conditions 2, 11 and 19 have DMSO in the acceptor, which is beneficial for the benzodiazepines and for non-polar drugs, while conditions 7 and 16 have DEHP in the SLM, which could possibly increase the partitioning of more polar drugs.Log P and pKa values are provided in Supplementary Table 4.
Fig. 1 shows that while some compounds, like zolpidem and methadone, are extracted fairly well under most conditions other compounds, like morphine and clonazepam, have very large variations between different conditions.For the benzodiazepines, exemplified here by diazepam and clonazepam, DMSO in the acceptor solution is very beneficial as previously demonstrated (Vårdal et al., 2018).At very high pH clonazepam exists as a negative ion, which might explain the very low signal for condition 19 for clonazepam.Adding DEHP was beneficial for the polar drugs like morphine, codeine and amphetamine, however this was very detrimental for the benzodiazepines.As for ordinary LLE it can be difficult to find optimum conditions for all compounds at the same time.Condition 11 was found to give acceptable result for most compounds, although it was not very favorable for morphine, and was chosen for validation.

Method validation
Table 2 shows the validation results, and a chromatogram of QC 1 is shown in Fig. 2.

Limits of detection and quantification
LODs ranging from 0.05 to 1.7 ng/mL were obtained.The LOQs were equal to the lowest calibration standard, with values ranging from 1 to 5 ng/mL (Table 2).For morphine the LOQ is set to 3 ng/mL, due to low signal intensity.

Calibration model
Based on inspection of residuals 1/x weighing was chosen.Both linear and quadratic curves were evaluated, and a linear curve model was found to best fit the data for all compounds but morphine, where a quadratic curve was chosen.The resulting correlation coefficients, R 2 values, are given in Table 2, and varied between 0.990 and 0.997.

Accuracy and precision
Table 2 shows the results for inter-and intra-day precision and accuracy; results were within 20% for all compounds except for morphine and zopiclone.Morphine had CVs of 21% and 25%, respectively, for inter-and intra-day precision at QC1, and Zopiclone had a CV of 25% for inter-day precision.

Extraction recovery and matrix effects
Extraction recoveries above 70% were found for 13 of the compounds.Two of the more polar analytes, codeine and amphetamine, had recoveries of 58 and 53%, respectively.Both morphine and zopiclone have several charged species in the pH-range from 6 to 10, and have maximum amount of uncharged species at pH 9.7 and 7.4 respectively, morphine is in addition the most polar compound in this study, which could explain why extraction recoveries of 45% for zopiclone and only 10% for morphine were found.Previous investigations have found that PALME gives very clean extracts (Ask et al., 2016), and all matrix effects corrected with internal standards were within the recommended values of 80-120%, except for QC2 for oxycodone which were 78% and QC 2 for zopiclone which were 123%.

Stability and long term extractability
Stability and long term extractability of freshly dried and extensively dried samples were evaluated by comparing four replicates of VAMS samples dried for 1 day, 2 days and 2 weeks with samples dried for 2 h.Table 3 shows the difference in measured concentration for VAMS calibrators prepared 1, 2 and 14 days in advance compared to VAMS calibrators prepared 2 h prior to extraction.The values are systematically lower for the more thoroughly dried samples.The difference between 1 day and 14 days is however small for most compounds.Preparing calibrators and QC samples at least 24 h before extraction is therefore recommended.
Previous studies of drugs of abuse sampled with VAMS have found cocaine and metabolites and oxycodone and metabolites to be stable for up to 2 months (Protti et al., 2018;Mandrioli et al., 2020), while cannabinoids was tested and found stable up to 1 month (Protti et al., 2017).Stability is however compound dependent, and Parker et al. found an approx.30% decrease after 2.5 months and 50% decrease after four months for the antibiotic fosfomycin (Parker et al., 2015).Freezing may however improve storage time, as the compound was stable in VAMS samples at − 20 • C for four months.Similar improvement in storage stability has been found for other compounds with stability issues sampled with DBS (Barco et al., 2014).
For other drugs of abuse not previously sampled with VAMS, DBS results could be used as an indication of stability, although the difference between VAMS samples polymeric material and cellulose used in DBS cards should be kept in mind.Up to one month stability in DBS at room temperature has been found for benzodiazepines and cocaine (Alfazil and Anderson 2008), fentanyl analogues (Seymour et al., 2018) and methadone (Saracino et al., 2012).Simões et al. found long term stability (eight months) for codeine, methamphetamine, MDMA, EDDP and methadone.Amphetamine, MDA and cocaine were seen to decrease, while benzoylecgonine, morphine and 6-AM increased with time.For morphine and benzoylecgonine degradation of 6-AM or cocaine could contribute, in addition an increase of the extraction efficiency from the paper support was suggested as a possible explanation.For all compounds storage at − 10 • C yielded stable DBS results for up to 8 months.Our results indicate sufficient compound stability when comparing 14 days with 24 h for most compounds, similar to what has previously been published.For morphine a higher concentration was seen for the VAMS samples extracted after 2 weeks than 24 h, similar to the result published by Simões.For zopiclone and zolpidem the results for 14 days could indicate issues with stability.Especially for zopiclone stability in blood in general is known to be limited (Nilsson, Kugelberg et al. 2010, 2011).Further investigation of long time stability of drugs of abuse sampled with VAMS is therefore warranted.

Interference studies
A total of 45 compounds were tested for interferences.Morphine and zolpidem were found to have a change in concentration equal to 24% and − 25%, respectively.As this change in concentration was within ±30%, which is the variation allowed between two parallels of a sample before reanalysis in our routine laboratory at Oslo University Hospital, this was deemed acceptable.

Carry-over
No carry-over was found, even for blanks run after an extracted sample with concentration 3 times the highest calibrator.

Application to road side samples and spiked external quality control samples
The method was applied to 18 VAMS samples collected as a part of a road side study collecting oral fluid.Analysis of dried samples was performed with the documented method, however all samples were negative.This result was confirmed by negative results for the corresponding oral fluid samples.As unfortunately no positive samples were found in the road side study and other real samples were not available, four old external quality control samples were analyzed, and compared to a rerun with the accredited whole blood method (Kristoffersen et al., 2018).The samples contained morphine, codeine, amphetamine, MDMA, alprazolam and clonazepam.Morphine, codeine and MDMA  had values within 20% of the routine method, amphetamine and clonazepam was 24% higher, while alprazolam was 36% higher.All compounds could be found after VAMS sampling, confirming the applicability of the method, although it could seem there is a tendency for higher concentrations with VAMS.A thorough evaluation of venous samples compared to VAMS samples collected from the same donor should be performed to truly validate the applicability.

Evaluation of method greenness
The greenness of the proposed methodology was finally evaluated and compared with a representative reference method (Jørgenrud et al., 2021).The latter was based on supported liquid extraction, followed by evaporation and sample reconstitution in a LC-MS compatible liquid.Comparison was based on the analytical Eco-Scale for semi-quantitative assessment, based on penalty points to parameters of an analytical process that are not in agreement with ideal green analysis (Gałuszka et al., 2012).The data are summarized in Table 4.The final LC-MS analysis, which was similar for both methods, counted 12 penalty points.Sample preparation with the reference method counted 29 penalty points, whereas the PALME method counted ten points.In terms of penalty points, the difference was substantial.In addition, the total waste during sample preparation was ten times less with PALME than with SLE, and more energy will be used to keep the blood samples frozen or refrigerated.With PALME combined with UHPLC-MS, the Eco-Scale score (100 -penalty points) was 78, and this represents excellent green analysis.

Conclusion
The combination of microsampling with VAMS and microextraction by PALME was explored for the first time.Most major drug classes relevant for drugs of abuse determination were included, and validation results within common criteria for linearity, reproducibility and matrix effects were found, demonstrating the applicability of the method.Most compounds were stable at room temperature for 2 weeks.Combining a low-cost, high sample throughput, semi-automated miniaturized sample preparation using only 4 μL organic solvent for the SLM per sample with VAMS samples, the method fulfilled green chemistry principles for analytical chemistry and should have great potential for further application.

Table 1
Conditions used to explore different extraction set-ups, all combinations were combined giving 27 different conditions.

Table 2
Validation data: calibration range, correlation coefficient, LOQ, extraction recovery, intra-and interday precision and accuracy, matrix effects corrected with internal standard and the corresponding coefficients of variation.

Table 2
(continued ) a A quadratic curve fit was found to best represent morphine.