Capillary Microsampling of Mouse Blood in Early Pre-Clinical Studies: A Preferred Alternative to Dried Blood Spot Sampling

Pre-clinical pharmacokinetic (PK) studies establish the key parameters of clearance, volume of distribution and oral bioavailability for novel chemical entities in drug discovery. Following favourable in vitro screens for metabolism and permeability, early in vivo evaluation ensures that chemical series and potential drug candidates with poor in vivo profiles are deprioritised early, focusing project resources on chemical series and lead compounds with favourable pharmacokinetics. The selection of pre-clinical species may vary; the rat is often used in early PK studies. However, mice are preferred as disease models of oncology; either as genetically engineered animals or as athymic mice bearing human tumour xenografts [1,2].


Introduction
Pre-clinical pharmacokinetic (PK) studies establish the key parameters of clearance, volume of distribution and oral bioavailability for novel chemical entities in drug discovery. Following favourable in vitro screens for metabolism and permeability, early in vivo evaluation ensures that chemical series and potential drug candidates with poor in vivo profiles are deprioritised early, focusing project resources on chemical series and lead compounds with favourable pharmacokinetics. The selection of pre-clinical species may vary; the rat is often used in early PK studies. However, mice are preferred as disease models of oncology; either as genetically engineered animals or as athymic mice bearing human tumour xenografts [1,2].
Using mice in routine PK studies can be challenging as total blood volume, often less than 1.5 ml, limits the size and number of samples that can be collected. In the past, the poor sensitivity of mass spectrometers required sample concentration of large volumes of plasma and discrete sampling strategy was then required to generate concentration-time profiles for a full PK, with an individual mouse used for each time point, minimising the physiological effect of taking large blood volumes. This approach has an inherent disadvantage as many individuals are used to create one concentration -time profile, generating average values with inter-animal variability and dosing error.
Direct analysis of blood instead of plasma has the advantage of being directly related to physiologically parameters, such as blood flow, used in the interpretation PK data.
Serial bleeding, an alternative method of sampling, provides a complete concentration-time profile of a compound by sampling just one animal over the time course. Large volume sampling cannot be applied to a full pharmacokinetic time course in mice as the volume of blood required would exceed animal welfare guidelines. Applying a low volume sampling strategy to serial bleeding has the advantage of providing an estimation of inter-animal variability throughout the time course as well as reducing study costs through a significant reduction in mice [3]. Introducing serial sampling has the benefit of conforming to the 3Rs (reduce, refine, and replace), an obligation for every project license under the Animals Scientific Procedures act 1986 [4]. A typical pharmacokinetic study using eight time points per dosing route (intravenous and oral drug administration with vehicle controls) would require sixty-four mice for traditional large volume plasma samples, whereas a blood microsampling methodology would use only eight.
A number of different methods for low volume sampling of matrix have been utilised, a popular one in recent years being the use of dried blood spots (DBS). Blood is collected and spotted onto filter paper and allowed to dry prior to punching disks of equal size for analysis. This technique has gained acceptance across many laboratories due to the ease of use, storage and extraction [5][6][7]. For the majority of analytes there is no requirement for centrifugation or storage on ice during the sample collection process and the cards can be stored at room temperature and can be shipped easily. The critical assumption in using DBS lies with sample homogeneity on the card or paper, and it is known that haematocrit values (a measure of the cell volume in blood) can impact the quality of the data [8]. Blood spot size varies with viscosity so a higher haematocrit value will reduce the spread of the blood on the card, a disk of the same size can therefore contain variable amounts of matrix and analyte. Confidence in reproducible analyte recovery is therefore contingent on even distribution of blood on the cards or on the punching of the whole spot which can vary in size. Capillary microsampling (CMS) is a relatively new technique, championed by Ove Jonsson [9]. Samples are collected and stored in capillaries prior to extraction. The extraction process involves addition of a washout liquid to the capillaries, protein precipitation of the capillary washout, and finally centrifugation to yield supernatant for LC-MS/MS analysis.
The primary aim of this investigation was to compare the PK results obtained from two distinct sampling methods: dried blood spot sampling which we have routinely used in our laboratory for over five years and capillary microsampling which we are testing in this study. A secondary aim was to integrate the in vivo and bioanalytical workflows, with a specific focus on the use of automated liquid handling systems. Small molecules (MW range: 341.8-569.6 Da) were chosen from multiple pre-clinical projects based on diverse physiochemical and pharmacokinetic properties (Table 1). Overall methodology and results in the form of PK profiles and parameters were compared for each sampling method.

Materials and Method
Chemical and solutions UPLC/MS grade water, formic acid and methanol were obtained from Biosolve, France. Dimethyl sulfoxide (DMSO), saline and Tween 20 were purchased from Sigma Aldrich, UK. Test compounds ICR1, ICR2 and ICR3 and ICR4 were synthesised at The Institute of Cancer Research (ICR, London, UK). Molecular weight and physicochemical properties are presented in Table 1.

In vivo PK studies
All in vivo studies were performed in accordance with UK Home Office regulations, ICR ethical review processes and Cancer Research Institute guidelines [10].
Due to volume collection limitations, sufficient blood could not be obtained from one animal for both sampling methods, therefore two separate groups of animals were used. Female BALB/C mice, weighing 18-20 g, were obtained from Charles River Laboratories (Margate, Kent) and acclimatised for a minimum of 5 days prior to use. Each experiment required four mice for each sampling method (one control, three drug treated). Compounds were dissolved in DMSO and prepared for dosing by addition of saline and tween 20. Mice were warmed at 37°C for ten minutes in a Datesand mini-thermacage prior to dosing. Following intravenous administration, via the lateral tail vein, samples were collected by bleeding from the opposing vein to the dosage site at 5, 15, 30 min and 1, 2, 4, 6 and 24 h.

Sample collection
CMS samples were taken using capillary tubes (10 µl) and snap frozen by placing in a 2 ml 96 deep well plate on dry ice. The plate was sealed after sample collection and stored at -80°C. It is essential that the samples are frozen prior to extraction as this reduces possible non-specific binding of the analyte to capillaries and plates. Samples extracted without a freeze-thaw cycle did not produce reliable, linear calibration lines (data not shown).
DBS samples were collected via Jaytec capillaries (exact volume 20 µl) and spotted onto Whatman B DBS cards using a capillary tube pipette to dispense the 20 ul of blood.

Analytical stock preparation
Separate weightings were used to prepare two stock solutions for each compound (1 mM DMSO). Liquid handling automation (Hamilton Star, Switzerland) was used to produce working solutions for quality controls and calibration standards in DMSO. The top five standard concentrations were prepared by spiking directly from the stock solution into a well plate with DMSO. These working solutions were serially diluted to produce the lower standard concentrations. Similarly, quality control solutions were prepared from stock solutions and serial dilutions with DMSO. All DMSO stock solutions were stored at ambient conditions.

Calibration and quality control preparation
Fresh tail vein blood, pooled from two un-treated female BALB/C mice, was used to prepare eight calibration standards (1 nM to 25 μM) and four quality control (QC) samples (12.5 nM to 3750 nM) by manually spiking DMSO working solution (2 μl) into blank blood (40 μl) and mixing gently before spotting on DBS cards (20 μl) and for CMS collection (10 μl). The calibration standards and QCs for both methods were prepared on the day of study to allow for DBS cards to dry overnight in ambient conditions and for capillaries to be stored alongside samples at -80°C.
Capillary samples were defrosted and washout solution (125 μl, 30% MeOH: 70% H 2 O) was added to each plate well using the liquid handling automation. Washout volume may be varied; enough is required to wash the blood out of the capillary and give a homogenous blood-water mix, however, over dilution may be problematic for detection on some analytical platforms. In this method 125 μl was deemed to be sufficient to provide a homogenous mix with enough volume to permit a second, lower volume extraction if required.
Washout composition can also be varied to suit specific analytes. After testing over forty compounds with a range of AlogP values it was determined that some organic solvent, such as methanol or acetronitrile, is necessary in the washout as this will help solubilise and stabilise the more lipophilic compounds. Too much organic will cause protein precipitation in the capillaries and should be avoided. A washout content of 30% organic was shown to be optimal for the analytes under investigation. The overall process of collection and processing of the CMS is described in Figure 1.
Plates were sealed and thoroughly mixed, particularly by hand in the longitudinal direction of the capillaries, as this is required to fully wash out the blood. Mixing was considered complete when the bloodwater mix was homogenous. An aliquot of 75 μl blood -washout mix was transferred to another deep well plate, again using automated liquid handling, and protein precipitation was performed using a 1:3 ratio of methanol (225 μl) (containing internal standard at 500 nM), aliquoted by automation.
DBS cards were punched into 6 mm disks centred on the blood aliquots and transferred to a deep-well plate and extracted with 200 μl methanol containing internal standard at 500 nM.
Plates were mixed and centrifuged for 30 min at 1800 g. Supernatant (50 μl) was diluted with MeOH/H 2 O (50 ul) ready for LC-MS/MS analysis.

Recovery following CMS and DBS extraction
Control extracts were spiked with compound and compared to extracted samples for both sampling methods.

LC-MS analysis:
Separations were performed on Waters Acquity LC system controlled through Analyst software. Samples were injected (8 μl) on a Phenomenex Kinetex C18 column 2.6 u (50 * 2.1 mm) 100 Å with a Krudcatcher column guard maintained at 45°C. The flow rate was 0.6 ml/min and a six minute gradient used mobile phase A (0.1% formic acid) and B (MeOH) (95/5 to 0/100 linear over 4 min 0/100 to 95/5 from 5-5.2 min). An ABSciex QTRAP4000 triple quadrupole mass spectrometer was used in positive mode with an APCI source (450°C curtain gas 35 psi). Multiple reaction monitoring was used to detect the analyte and internal standard Table 2 for MS parameters. APCI was preferred over ESI to avoid potential ion suppression from the dosing vehicle surfactant. Standard curves were run in singlicate at the beginning of the run and quality controls in duplicate at the end of the curve and at the end of the analytical run.

Acceptance criteria for analytical runs
The calibration lines were analysed by linear regression and optimal weighting. No analytical validation was carried out for these studies. Concentrations were deemed acceptable if the nominal value of the analyte was within 15% of nominal concentration and 20% for the lower limit of quantification.
Six out of 8 quality controls had to be within 15% of nominal values at least one at each level.

Pharmacokinetic data analysis
Phoenix WinNonLin (Pharsight, USA) was used to calculate area under concentration-time curves (AUC last ) in order to determine PK parameters of clearance, half-life and volume of distribution.

Result
The recoveries from CMS and DBS extracts were complete for ICR2 and ICR3 while ICR1 and ICR4 were recovered better from CMS and DBS extracts respectively ( Table 3). The recovery for ICR1 and ICR4   were above 50% and consistent across the concentration range which was considered acceptable for these early studies given the fact that standard curves quality controls and samples were extracted from blood. No signal was detected in any of the control samples from animals treated with vehicle alone (data not shown). The range of quantification varied between compounds with saturation observed at 25000 nM for all compounds but ICR1 in the CMS assay and ICR4 in the DBS assay. The quality controls in Table 4 demonstrate that the analytical runs were within the acceptance criteria set for these early studies.
Despite the differences in recovery, sensitivity and dynamic range following both sampling methods, the average concentration-time profiles for each of the four compounds were superimposable for all analytes (Figure 2).The current sensitivity of our LC-MS/MS systems was sufficient to obtain meaningful PK profiles for early evaluation of novel chemical entities.
Systemic exposure (AUC) was used to calculate pharmacokinetic parameters of blood clearance, half-life and volume of distribution (Table 5). There was no significant difference between AUCs, half-lives and volume of distributions derived from CMS and DBS data (Mann-Whitney test P<0.05). The variation in exposure was less than 5% for 3 compounds and 16% for ICR3. Variations in volume of distribution were more pronounced but less than 30% difference was observed between the two sampling methods often mirroring variation within a single sampling method. This was therefore considered acceptable in these early PK studies.
We concluded that concentration-time profiles and PK data generated from our CMS and DBS studies were comparable for the investigated compounds in-vivo. The full sample plates were delivered to the bioanalytical laboratory where automated liquid handling was used to add washout solution and to protein precipitate the samples ready for LC-MS/MS analysis, simplifying the analytical workflow. The use of laboratory automation had the benefits of increasing precision, saving analyst time and improving data quality. Although automation of DBS is possible with the purchase dedicated systems or LC-MS/MS add-ons, the equipment required is not always compatible with the layout of existing multipurpose liquid handling systems.
The transition from DBS to capillary microsampling required minimal training as sample collection, handling and extraction are very simple. Sampling templates, guiding correct placement of samples into 96 well plates, were created in order to facilitate an integrated in vivo analytical method. Different storage conditions for CMS compared to DBS had a significant impact in the use of laboratory storage space. DBS cards were stored at ambient temperature in racks with a full PK having a similar footprint to an A4 piece of paper. In contrast, CMS samples were stored in freezers (-80°C) in the 96 deepwell plate format, often just for 24 h, until extraction. This maintained a clear and tidy lab space and protected samples from contamination and damage.
Blood sampling using CMS was speedy and efficient compared to the DBS method. In part this was due to the reduced quantity of blood collected, reducing the time taken to fill the sample capillary with no additional card spotting. Using smaller volumes also raised the possibility of adding additional time points into the study, either repeating time points or lengthening the study.
Care must be taken to avoid transferring other contaminants when spotting blood onto DBS cards, particularly in a busy laboratory setting where dosing solutions may be present. In contrast, capillaries are sampled and immediately placed in a 96 well plate away from sources of contamination.
Extracting blood directly from capillaries eliminates haematocrit issues experienced with DBS. While it may be possible to take the entire blood spot, in order to avoid a potential missampling due to compounds modifying the haematocrit, it is often not practicable to do so with the circular punches. Spots may not fit into the disk shape and cutting out the whole spots would be time consuming, particularly when compared to the time taken to process samples with our current CMS workflow. Current LC-MS technologies should be sufficiently sensitive to detect nanomolar levels of analytes in complex matrices like blood. However, blood volumes used in this CMS method can pose an analytical challenge as smaller samples and dilution with washout will reduce analyte concentrations in comparison to the DBS samples.
A benefit of introducing smaller sampling volumes, while only a reduction from 20 µl to 10 µl, is the shortening of time spent bleeding animals. This reduces stressful handling and is considered a refinement of the technique by providing welfare benefits through reduction in restraint time. This reduction in volume could potentially be applied to the DBS method although manufacturers of the cards that we use recommend 15-20 μl per spot. Perceived limitations in DBS have led to a renewed interest in liquid microsampling using capillaries [11,12].
There is a significant financial advantage to using CMS over DBS methodology.
CMS sampling uses one capillary for sampling and storage, whereas DBS requires both a larger sampling capillary and DBS card. The cards have space to spot four samples but are relatively expensive in comparison to heparinized capillaries.
Replacing the use of DBS cards by using the CMS methodology therefore results in a significant decrease in consumable costs; capillaries for a full CMS PK cost seven-fold less than a set of DBS cards and sampling capillaries. Overall, a seven-fold reduction in costs is achieved by eliminating DBS in favour of CMS. While CMS samples require freezing this does not incur any additional cost as both in vivo and bioanalytical laboratories have freezers and very little space is needed to store the sample plate. Although this study used       a -80°C freezer, we have successfully carried out additional studies with other compounds where a more affordable -20°C freezer was used. Extraction of both CMS and DBS requires plates and solvents and there is little financial difference in these consumables, rather the saving comes in the form of analyst time as CMS extraction takes less time than the manual punching of the DBS cards.
CMS can be readily integrated to workflows using liquid handling systems with little extra expense, unlike DBS where automated extraction is possible but requires extra instrumentation and/or addons to mass spectrometers. Capillary based studies also have the advantage of a choice of matrices; blood can be extracted directly or the capillaries can be centrifuged to yield plasma [13].
Using laboratory automation platforms increases costs compared to manual pipetting although this is balanced by an increase in consistency and precision. There is also the added benefit of speeding up the analytical workflow and reducing the risk of repetitive strain injuries through pipetting and punching cards.
Lowering the sample volume allows flexibility for multiple aliquots per time point offering a second analytical chance (providing Home Office guidelines and project licences are not exceeded). Separate samples taken at the same time point would enable analysis by two different methods, for example in PK/PD studies drug determination and biomarker estimation may require different detection techniques.
In conclusion, this study demonstrates that results from DBS and CMS protocols are comparable for novel and structurally differentiated chemical entities from four oncology drug discovery programmes in the ICR portfolios. CMS has great potential to replace DBS sampling in pre-clinical studies; it is easy to perform and readily integrates with automated liquid handling platforms with a much lower financial outlay compared to an automated DBS system. Replacing the manual DBS workflow with an automated CMS methodology offered a simple solution to streamline workflows between in vivo and bioanalytical laboratories. Refining the sampling technique by reducing sample volumes improved animal welfare by reducing restraint times.