Subscriber access provided by RYERSON UNIV
Article
Semi-empirical rules to determine drug sensitivity and ionization efficiency in SIMS using a model tissue sample. Jean-Luc Vorng, Anna M. Kotowska, Melissa K. Passarelli, Andrew West, Peter S Marshall, Rasmus Havelund, Martin P Seah, Colin T. Dollery, Paulina D. Rakowska, and Ian S Gilmore Anal. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.analchem.6b02894 • Publication Date (Web): 11 Oct 2016 Downloaded from http://pubs.acs.org on October 19, 2016
Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Analytical Chemistry is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.
Page 1 of 10
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Analytical Chemistry
Semi-empirical rules to determine drug sensitivity and ionization efficiency in SIMS using a model tissue sample. Jean-Luc Vorng,† Anna M. Kotowska,† Melissa K. Passarelli,† Andrew West,‡ Peter S. Marshall,‡ Rasmus Havelund,† Martin P. Seah,† Colin T. Dollery,‡ Paulina D. Rakowska†* and Ian S. Gilmore†* †
National Centre of Excellence in Mass Spectrometry Imaging (NiCE-MSI), National Physical Laboratory, Hampton Rd Teddington, Middlesex, TW11 0LW, UK; ‡GlaxoSmithKline, Gunnels Wood Rd, Stevenage, Hertfordshire SG1 2NY, UK. There is an increasing need in the pharmaceutical industry to reduce drug failure at late stage and so reduce the cost of developing a new medicine. Since most drug targets are intracellular this requires a better understanding of the drug disposition within a cell. Secondary ion mass spectrometry has been identified as a potentially important technique to do this as it is label-free and allows imaging in 3D with sub-cellular resolution and recent studies have shown promise for amiodarone. An important analytical parameter is sensitivity and we measure this in a bovine liver homogenate reference sample for twenty drugs representing important class-types relevant to the pharmaceutical industry. We also measure the sensitivity for pure drug and show, for the first time, that the SIMS positive ionization efficiency for small molecules is a simple power-law relationship to the log P value. This discovery will be important for advancing the understanding of the SIMS ionization process in small molecules that has, hitherto, been elusive. This simple relationship is found to hold true for drug doped in the bovine liver homogenate reference sample, except for fluticasone, nicardipine and sorafenib which suffer from severe matrix suppression. This relationship provides a simple semi-empirical method to determine drug-sensitivity for positive secondary ions. Furthermore, we show, on chosen models, how the use of different solvents during sample preparation can affect the ionization of analytes.
INTRODUCTION The pharmaceutical industry is facing a major challenge as the cost of bringing a new drug to market continues to rise 1,2 and the targeting of drugs to smaller patient groups (precision medicine) increases. Drugs that fail at late stage contribute a large burden to the cost of new medicines. Consequently, there is a major drive to identify failure at early stages of drug discovery and development. Recently, Dollery has identified a strategically important need to measure the concentration of a drug at the target, to accurately predict its pharmacological effect.3 Since most drug targets are intracellular, analysis with sub-micron resolution is required. Furthermore, label-free strategies are needed since changes of the drug chemistry, with for example fluorescent labels, can invalidate results. In the early stages of drug discovery the amounts of material are limited and, therefore, isotopic labelling is not generally attractive. Secondary ion mass spectrometry (SIMS) has been identified as a technique with considerable promise as it possesses both of these important attributes of label-free detection and sub-micron resolution.4 SIMS is a powerful chemical imaging tool. The recent technological developments of SIMS instrumentation and the implementation of large cluster ion sources have greatly enhanced the sensitivity of the technique.5,6 In consequence, SIMS offers a label-free analysis with high sensitivity throughout a large mass range combined with 3D imaging capabilities and high lateral resolution. As a result, SIMS has become well-suited to biological and pharmaceutical applications. In fact, the use of SIMS in pharmacology is steadily growing, covering a range of investigations including determination of spatial co-localization of
pharmaceuticals with biomolecules or drug diffusion and distribution in single cells and tissues.7-10 We have shown, for the first time, the ability of SIMS to image the drug, amiodarone, at a therapeutic dose in NR8383 macrophage cells. 11 The protonated molecular ion signal of amiodarone is detected, giving unambiguous identification, with sub-cellular resolution. The use of the characteristic negative iodine signal, which is more intense, allows 3D imaging at a resolution of 550 nm. This clearly demonstrates the utility of the method for amiodarone. For industry, important analytical questions are: how sensitive is the method for other drug compounds and can general rules be identified to predict whether a particular drug is detectable? In SIMS, the measurement sensitivity also dictates the achievable lateral resolution.12 The sensitivity is not the simple case of determining a sensitivity factor for the pure drug compound, because the ionisation efficiency is strongly affected by the local chemical environment, the so-called matrix effect.13 This can be a limiting factor, as the local environment can enhance or suppress the ionisation efficiency.14,15 There is, therefore, a need to understand these underlying processes to eventually overcome the issues arising from sample complexity. To do this, a reference sample is needed of a relevant biological material that is uniformly doped with a drug at a known concentration. Methods that dispense a solution of drug in solvent onto the surface of a tissue section suffer from several artefacts including a non-uniform distribution of drug vertically through the tissue (preferentially at the surface) and the cellular heterogeneity of the tissue resulting in high variability of matrix effects across the sample.16 In this study, we develop a reference sample using a BCR 185-R bovine liver standard and a proce-
ACS Paragon Plus Environment
Analytical Chemistry
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
dure to dose the tissue with drug that results in very good homogeneity across the sample and a reasonably good constancy with depth. We use this reference sample, dosed separately with 20 drug compounds. The compounds span many important classes of small molecule drugs for the pharmaceutical industry. We show that there is a good correlation between the SIMS sensitivity and the partition coefficient (log P) value. This simple empirical relationship is helpful to analysts to estimate the sensitivity of SIMS and the suitability for particular applications in the pharmaceutical industry.
EXPERIMENTAL Materials. Drug compounds: acetazolamide, amiodarone hydrochloride, budesonide, carbamazepine, cefoperazone, chloroquine diphosphate, chlorpromazine, ciprofloxacin, cisplatin, clozapine, doxorubicin hydrochloride, elacridar, fluticasone propionate, haloperidol, indomethacin, ketoprofen, nicardipine hydrochloride, sorafenib, tamoxifen and warfarin were purchased from Sigma Aldrich or Alfa Aesar and used without further purification. Reagents: Deionised (DI) water, methanol, anhydrous dimethyl sulfoxide (DMSO), ammonium formate, and ethanol. All reagents were of HPLC or MS grade purity. Standard tissue preparation. Lyophilized bovine liver tissue standard, BCR 185-R,17 was purchased from LGC Standards. The powder was first suspended in DI water and next washed three times by centrifugation (5 min at 7500 rpm) with 0.1 M ammonium formate pH 7.4, and re-suspended in DI water to the final concentration of 0.1 g/mL. The final pH of the tissue was measured to be around 7.5. The washing with buffered ammonium formate minimizes the amount of salts in the tissue sample and enhances the detection of biomolecules or other analytes (See Figure S-1, Figure S-2). In addition, it allows samples to be at pH close to the physiological values. The BCR 185-R bovine liver tissue standard was chosen as a model sample to serve as a relevant and well defined biological matrix. This is a Certified Reference Material with carefully defined properties. It should be remembered that the BCR 185-R standard, due to the preparation and handling procedures,17 may not retain the original biochemical properties of the tissue e.g. factors such as drug transporters or uptake into cellular lysosomes can be affected. Nevertheless, this standard presents the molecular characteristics of a regular tissue. SIMS depth profiling through the tissue samples did not reveal any notable variations in the distribution of tissue compounds (data not shown), proving the sample homogeneity, which is important for our experiments. A selection of 20 compounds were used in this study, ranging from antipsychotic to chemotherapy drugs (Table 1). The drugs were spiked to the tissue samples to the final concentration of 167 µg/mL. To ensure the accuracy of the results, all the samples were prepared the same way and all the parameters in the sample preparation process were rigorously controlled. Drug solutions preparation. Stock solutions of the drug compounds were prepared by dissolving the drugs in appropriate solvents, i.e. DI water, methanol, or a mixture of the two, to a concentration of 1 mg/mL. The solutions were then used to dope the tissue samples at the desired final concentrations. Tissue sample preparation for SIMS. Prepared tissue samples were drop-deposited onto cleaned silicon substrates and allowed to dry at room temperature overnight. Pure drug compound sample preparation for SIMS. Drug compound solutions were drop-deposited on a cleaned silicon wafer and allowed to dry.
Page 2 of 10
SIMS spectra of pure drug compounds. All experiments were carried out on an ION TOF IV instrument (ION TOF GmbH, Germany) using 25 keV Bi3+ ions. Spectra were acquired from a 100 µm x 100 µm area with a square raster of 256 pixels by 256 pixels and 1 shots per pixels for all the drugs. The Bi3+ beam size was ~ 2 µm. The ion beam currents were measured separately before each measurement using a Faraday cup. 18 The pulsed Bi3+ current was 0.1 pA, ion dose of 9.00x1012 ion/cm2 and a pulse width of 13 ns. Spectra were acquired in both positive and negative polarity for 65 seconds (10 scans). SIMS depth profiling of tissue. All experiments were carried out using an ION TOF IV. All spectra and depth profiling experiments were performed in dual beam mode, using Bi3+ analysis gun and an argon cluster gun as the sputter gun (10 keV, Ar5000). To resolve any issues of sample charging, a 20 eV electron flood gun with a beam current of approximately 10 μA was used for charge compensation. The damage caused by electron flood gun during depth profiling is considered to be negligible.19 Depth profiles were performed over a 200 µm x 200 µm area centered in a 500 µm x 500 µm crater with a square raster of 256 pixels by 256 pixels and 1 shot per pixel. The Bi3+ beam size was ~ 2 µm. The ion beam currents were measured separately before each measurement using a Faraday cup. The pulsed Bi3+ current was 0.1 pA and the GCIB current for 10 keV Ar5000 was 1.15 nA. Depth profile acquisitions were performed in positive and negative secondary ion mode for 520 seconds. Spectra and depth profiles were calibrated using H+, C+, CH3+,C2H5+, C3H7+ for positive secondary ions and H-, C-, C2and C3- for negative secondary ions. The total primary ion dose was kept less than 1% of the sputtering dose, as previously recommended.20 Data extraction and processing. The analysis of pure drug compounds was performed on two different locations on each sample. To mitigate the possibility of increased molecular density on the edges of the drying sample, the locations in the centre of the sample, away from the sample ridges were chosen for analysis. The coverage of the substrate surface by the compounds was determined by the Surface Lab 6 software (ION TOF GmbH, Germany) (see supplementary information (Figure S-3)). The secondary ion intensities obtained from the two locations on each sample were averaged. Depth profiling of tissue samples was also carried out at two different locations on each sample. Mass spectra were calibrated and the data reconstructed using Surface Lab 6. For each acquisition, secondary ion depth profiles were reconstructed from the total analytical area and on three selected regions of interest (ROI) each of 14% of the total. This allowed the lateral homogeneity of drug disposition to be measured. The secondary ion intensity for each drug compound was dead-time corrected21 and the background-subtracted and then summed from scan 20 to 33, as discussed later in the text. This was repeated for each region of interest giving 6 measurements used to calculate the average intensity.
RESULTS AND DISCUSSION Relative sensitivity of SIMS for pharmaceutical compounds A representative depth profile obtained from a drug doped tissue model sample is shown in Figure 1.
ACS Paragon Plus Environment
Page 3 of 10
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Analytical Chemistry
Table 1. Chemical structures of drug compounds used in the study along with their corresponding molecular weights and computed by ChemAxon log P and log D (at pH 7) values.*
*Data
were obtained from DrugBank (www.drugbank.ca)
Figure 1. Lateral and vertical homogeneity of the amiodarone-doped tissue sample: a) depth profile showing the intensity of molecular ion of amiodarone [M+H]+ at m/z 646.0 (gray Y scale) and a phosphocholine fragment at m/z 184 (orange Y scale); the ion dose corresponds to the dose of 10 keV Ar5000+ GCIB on the total area of analysis; b) total ion image, from the same depth profile experiment, showing the chosen ROIs for the relative sensitivity calculations.
In this example, the secondary ion signal intensity for the intact protonated molecular ion of the drug amiodarone ([M+H] + at m/z 646) and phosphocholine fragment ([C5H15NO4P]+ at m/z 184) are compared as a function of depth. Here, the phosphocholine signal is most intense at the surface. In the first few sputter cycles, its signal intensity decreases and eventually plateaus, reaching a steady state at about
3.54x1013 ions/cm2. This dose will sputter a thickness of 10 nm in 10 to 100 kDa materials.22,23 Similarly, the depth profile for the protonated molecular ion of amiodarone, the maximum intensity is at the surface and the signal rapidly drops in the near-surface region, before hitting a steady state at 3.54x1013 ions/cm2. Similar depth profiles were observed for all the drugs spiked in to tissue. For the amiodarone sample, the
ACS Paragon Plus Environment
Analytical Chemistry steady-state signal intensity exhibited the most fluctuation compared to the other compounds. The variability is believed to be an artifact of crystallization in the model tissue mixture and for amiodarone the relative variation integrated over the defined limits was only 10%. In order to compare the signal intensity for various compounds, we need uniform signal intensity in both the lateral and vertical dimensions. In the vertical dimension, the secondary ion signal intensities were summed across the steady-state region from 3.54x1013 ions/cm2 (~10 nm) to 6.01x1013 ions/cm2 (~17 nm). In the lateral dimension, the secondary ion signal intensities were summed within regions of interest (ROI), each corresponding to 14 % of the total analytical area (see Figure 1b). The ROI were specifically selected to avoid large topographic defects. Large cracks in the tissue material, seen in the total ion image are artifacts of the drying step in the sample preparation method. We define the ion yield (relative sensitivity) as the sum of the detected ion counts for selected secondary ions, as listed in Table S-1, divided by the total number of Bi3+ ions used to generate that signal. For each drug the average relative sensitivity, Yi, is calculated from the mean of the 6 repeats (2 separate areas of analysis and 3 ROIs from each measurement), as is the associated relative standard deviation. For amiodarone (Figure 1) the relative standard deviation is 5.45 % and the average relative standard deviation for all drug samples is 11.5 %. This repeatability illustrates sample homogeneity, uniform drug distribution and a good lateral and vertical measurement procedure for the reference sample. For most drugs, the characteristic molecular secondary ion is detected either as M+ and [M+H]+ in positive ion mode and as M- and [M-H]- for negative ion mode. In each case, we summed the ions intensity in the calculation of Yi. Furthermore, for drugs containing chlorine or platinum, there is a significant isotope distribution and these signals along with 13C isotopes are integrated into the calculation of Yi. Examples for chloroquine, chlorpromazine clozapine, haloperidol and cisplatin are shown in Figures S-4 and S-5 of the supporting information. For elacridar, the most intense signal appears at m/z 562 which corresponds to [M-H] + ion. This signal is observed in both the pure compound and the doped tissue spectra (see supplementary information Figure S-6). The loss of hydrogen atom could be caused by the rearrangement of the molecule during the ionization process. All peaks for the drug compounds used for the calculations are summarised in Table S-1 in the supporting information. Drug-related peaks with signal-to-noise ratios less than 3 or unresolved mass peaks were not included in the Yi calculations. In order to identify potential spectral interference between the drug-related peaks and the reference tissue material, we compared the doped sample spectrum to a control spectrum of the reference tissue. Figure 2 presents the ion yield, Yi, for each detected drug ordered from high to low. No obvious correlation was observed between the molecular weight of the compounds and their signal intensity. However, since the samples were prepared using the same weight concentration for all the compounds (167 µg/mL), the molecular weight (MW) of the drugs had to be accounted for. Therefore, Yi values were normalized to the same molar fraction of the drugs in tissue relative to the amiodarone sample (YMi). The normalization was done assuming linearity at this dilute level and by multiplying the Yi by the
ratio between amiodarone molecular weight and the molecular weight of the drugs. Figure 2 includes the data acquired in both the positive and the negative ion mode. 0.001
+ +
0.0001
+
-
Yi
YMi
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Page 4 of 10
0.00001
+
+
+
+
+
+
-
0.000001
Figure 2. The ion yield, YMi, of the 13 drugs, detected in the tissue samples, ordered from high to low. ‘+’ and ‘-‘ indicate the polarity, in which the drugs were observed. Error bars show the standard deviation for the 6 measurements done on each sample.
From the 20 compounds, 9 compounds (amiodarone, tamoxifen, haloperidol, elacridar, chlorpromazine, clozapine, chloroquine, carbamazepine and ciprofloxacin) were detected as positive secondary ions, 4 drugs (warfarin, acetazolamide, indomethacin and cisplatin) were detectable in the negative mode and 7 compounds were not detected in either polarity (budesonide, cefoperazone, doxorubicin, fluticasone, ketoprofen, nicardipine and sorafenib). It has to be noted that peaks of cisplatin and acetazolamide are detected but are not completely resolved from the peaks from the tissue material, as shown in the supplementary information Figures S-5 and S-7, respectively. Nevertheless, their peaks are rather evident and included in Figure 2 to show their relative position between other drugs. The lack of the detection of some of the drug molecular ions, or their fragments, could be caused by the weak ionization potential of the compounds (we discuss this later) or by the strong matrix effect and the drug signal suppression by other molecules in this complex biological environment.
An empirical prediction of relative sensitivity In drug discovery, physicochemical parameters are effectively used to predict and classify the behavior of compounds a priori. These include, for example, lipophilicity e.g. log P,24,25,26 acidity constant dissociation (pKa),27 coefficient of diffusion (log D)28,29 or the chromatographic hydrophobic index (CHI).30 The log P, or a partition coefficient, is a parameter often used in early drug discovery to assess the drug-likeness31,32 of a given molecule. It is associated with the ratio of the concentration of a compound in two immiscible solvents at the equilibrium. It is constant and represents the intrinsic lipophilicity or hydrophobicity of the compound for a given solvent system.25,33,34 The higher the log P value, the higher the lipophilicity / hydrophobicity of the drug. Therefore, this parameter can be linked to the bioactivity of molecules.35,36 Log P can be determined experimentally by the shaking flask method.37 This experimental method is the most accurate way, however, it is time consuming. Additionally, taking into account the large number of available molecules and potential pharmaceuticals, various software packages have been constructed, which are able to compute simulated log P values.
ACS Paragon Plus Environment
Page 5 of 10
Cheminformatics programs such as ALOGPS38 and ChemAxon39 provide physical and chemical predictions as well as analysis tools for R&D.40 Table S-2 summarizes the experimental and the computed log P values using both algorithms for the twenty drugs in this study. All the values were obtained from DrugBank.41 In Figure 3, we plot the ion yield YMi for the 9 drugs detected in the model tissue, in the positive mode, versus the log P values computed by ChemAxon39 (see Tables 1, S-1 and S-2). We use computed log P values as they are generated the same way for each drug, as opposed to the experimental values coming from different reports in the literature.
the pH of the environment of the molecule and takes into account all ionisable groups that the drug contains. The log D values in pH 7 for the studied drugs (Tables 1 and S-2), were also simulated in ChemAxon online predictor. pH 7 was set, as it corresponds to the pH of the tissue samples, ruled by the buffer used for the tissue washing. Figure 4 shows the relation of Log D to the YMi for the compounds detected in the positive ion mode. 0.001
YMi
0.0001
0.001
0.00001
0.0001
YMi
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Analytical Chemistry
0.000001 -2
0
2
4
6
8
Log D
0.00001 Amiodarone Chlorpomazine Elacridar
Carbamazepine Ciprofloxacin Haloperidol
Chloroquine Clozapine Tamoxifen
0.000001
-2
0
2
4
6
8
Log P Amiodarone Chlorpomazine Elacridar
Carbamazepine Ciprofloxacin Haloperidol
Chloroquine Clozapine Tamoxifen
Figure 3. YMi versus log P (from ChemAxon39) for positive secondary ions of the drugs in the tissue samples. The trend line indicates the observed linear correlation. YMi is expressed in counts per ion and scaled for the same amount of drug molecules in the sampled volume
Among positive ion mode detected compounds in the tissue model there is a reasonably good linear correlation with log P: Log YMi = 0.2164 log P – 5.4406 (1) 2 with a least squares fit R =0.83 and standard error of ± 0.26 (or a scatter factor of ×÷ 1.82 in linear space). The trend highlighted in Figure 3 suggests a positive correlation between the drugs lipophilicity and SIMS sensitivity as defined by YMi. Perhaps, the relative solubility of the drug in the lipid membrane influences its ionization efficiency. It is possible that the reduced salt concentration, associated with the cellular environment, mitigates ion signals suppression. However, experimental data discussed later in this report show that this is not the case. The relative sensitivity for the drug compounds detected as negative ions appears to be much less dependent on the log P value and the number of data points is rather low (n = 4) to draw a firm conclusion (see Figure S-10). For comparison, the same type of plots were created for log P generated by ALOGPS38 (Figure S-8, Table S-2,) and for the experimental log P values (Figure S-9, Table S-2,). Since the log P values obtained from the different sources are relatively similar, the linear correlation trend is still observed. This allows an estimate of the SIMS relative sensitivity, Yi, of drug in tissue directly using equation (1). We also investigated the correlation of the obtained drug intensities with another pharmacological parameter: log D. This parameter represents the effective lipophilicity of a molecule at a given pH.42,43 Its value is not constant and it depends on
Figure 4. YMi versus log D for the drug compounds detected in tissue in the positive ion mode. Computed by ChemAxon Log D values are used in the plot. The trend line marks the observed linear correlation. YMi is expressed in counts per ion and scaled for the same amount of drug molecules in the sampled volume.
Again, a linear correlation is observed, similar to the one with Log P and is given by: Log YMi = 0.2565 log D – 5.3303 (2) 2 with a least squares fit R = 0.80 and standard error of ± 0.28 (or a scatter factor of ×÷ 1.9 in linear space). The fit is close but not as good as the one to log P. In this case, log D at the given pH might not be a relevant parameter to consider for the purpose of this study, mainly due to the impossibility to determine the pH of the tissue after a stay in high vacuum. As before, no correlation is seen between the relative sensitivity and log D for drug compounds detected as negative ions (see Figure S-11). We now investigate if the log P relationship is caused by matrix effects in the sample, or whether this is an intrinsic property of the ionisation efficiency for these molecules. To do this, drugs were drop-deposited on cleaned silicon substrates. Ideally, a flat uniform film is required. However several drug compounds crystallized, forming non-homogenous surfaces. To mitigate the non-homogenous spreading of some of the drugs on the surface, we estimated the effective coverage of the analytical surface area. This task was completed using Surface Lab 6 image processing tools (see Figure S-3). The coverage was estimated as the fraction of the analytical area (the number of pixels relative to the total number of pixels) in which the molecular ion intensity was above 10% of the maximum molecular ion intensity. The coverage varied from 67 % (acetazolamide) to 100% (e.g. amiodarone), with the mean of 95% across all the studied drugs. As before, the isotopic peaks are summed together. In Figure 5, we plot the ion yield Yi for all 20 studied drugs versus their corresponding log P values (calculated in ChemAxon39). The plot summarises and compares results for
ACS Paragon Plus Environment
Analytical Chemistry drugs detected in both polarities and as both the pure compounds and in the model tissue.
where NA is Avogadro's number, ρ is the density in g/cm3 and MW is the molecular mass. 0.1
1 Pure drugs + Drugs in tissue + 0.1
Pure drugs -
0.01
Drugs in tissue -
Yia3, nm3
0.01
Yi,YMi
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Page 6 of 10
0.001
0.001
0.0001
0.00001
0.0001
-2
0.000001 -2
0
2
4
6
Log P
Figure 5. Yi, YMi versus log P (from ChemAxon39) for positive (filled symbols) and negative (empty symbols) secondary ions of drugs in the tissue (red) and for pure compounds (blue). Yi, YMi is expressed in counts per ion. YMi is scaled for the same amount of drug molecules in the sampled volume.
What we see in Figure 5 is that the results for positive ions of the pure drugs form a well ordered array, similar to the drugs detected in the model tissue. Here, the ratio of the tissue and pure material intensities for the same primary ion dose and for the 9 drug materials in common gives a value of 0.004 ± 0.004, whereas the 4 negative secondary ions have a higher ratio of 0.017 ± 0.021. Clearly, in terms of predictability for an unknown material, the positive secondary ions are more effective, whereas in terms of signal level the negative ions in tissue can be more intense. It has to be remembered that the tissue data derive from an average of many spectra, as shown in Figure 1, whereas the pure material data are for the outer surface. The decay of signal exhibited in Figure 1 means that the pure drug data may be 1.5 times too high, so the ratio of the drugs in tissue to the pure material is, in fact, ~ 0.006 for a composition of 0.1%, i.e. 6 times higher than expected on a linear basis. For negative ions it is closer to 26. These are significant enhancements and will give very non-linear responses at high compositions13 but at the levels of interest here, for concentrations below 1%, good linearity is retained. If we wish to consider the relationship shown in Figure 5 more closely, it should be noted that the smaller molecules will be sputtered in greater numbers, assuming a constant sputtering yield. In terms of the drug contained in a tissue sample, the sputtering rate is controlled by the tissue rather than the dilute drug molecule, therefore any dilute drug in any given tissue will experience the same overall sputtering yield. Thus, in the tissue mixture Yi is proportional to V/a3 where V is the volume of sample sputtered by one Bi3+ ion and a3 is the molecular volume of the drug molecule. So, in Figure 6 we plot, for the positive ions from the pure material Yia3 versus log P. Here, V will be of the order 100 to 500 nm3 and is very much larger than a3 and is omitted. Here, a3 is measured in units of nm3.22 The molecular volume, a3, in nm3 is simply given by:
M a 3 10 21 W NA
0
(3)
2
4
6
8
Log P
8
Amiodarone Chlorpomazine Elacridar Nicardipine
Carbamazepine Ciprofloxacin Fluticasone Sorafenib
Chloroquine Clozapine Haloperidol Tamoxifen
Figure 6. Yia3, nm3, versus log P (from ChemAxon39) for the pure drugs acquired in the positive ion mode. The trendline marks the observed linear correlation. Yia3 is expressed in in counts per ion nm3.
In this case, there is also a correlation with log P and the positive ions. This is remarkable and shows, for the first time, that the ionisation efficiency in SIMS for complex molecules may be predicted for quite a wide range of small molecules. Nicardipine, sorafenib and fluticasone, which were before obscured in the tissue, are now detected as positive ions and exhibit the same relationship given by equation (4): Log Yi a3= 0.1574 log P – 2.9797 (4) 2 with a least squares fit R = 0.67 and a standard error of ± 0.25 (or a scatter factor of ×÷ 1.78 in linear space). This proves that a matrix suppression effect is responsible for the earlier lack of detection of nicardipine, sorafenib and fluticasone in the reference tissue sample. At this stage, we are not able to explain what causes the signal suppression of these three compounds. This requires further investigation, including a larger number of chemical compounds. Nevertheless, this provides a good evidence that the reference sample may be used to measure the matrix effects. The results for negative ions are shown in Figure S-12 in supplementary information. The value of Yi is approximately constant and consequently no relationship with log P is observed. This is in good agreement with Figure S-10, where the negative ion data do not follow equation (1).
Solvent influence on the spectra We studied the influence of the solvent, used to dissolve the drugs, on their ion signal intensities in SIMS. Here, we compared two polar solvents routinely used in biological sample preparations: aprotic dimethylsulfoxyde (DMSO)44,45 and protic methanol (MeOH).46 Four compounds were selected for the analysis: clozapine, haloperidol, tamoxifen and warfarin. The drugs were dissolved in either DMSO or MeOH before dosing, at fixed concentration, the BCR-185-R standard. ToF SIMS depth profiling analyses were performed in positive mode for clozapine, haloperidol, and tamoxifen and negative for warfarin.
ACS Paragon Plus Environment
Page 7 of 10 a)
b)
0.00006
0.00003
0.00005 0.00004
Ion yield Yi
Ion yield Yi
0.000025 0.00002
0.00003
0.000015 0.00001
0.00002
0.000005
0.00001
0
DMSO
c)
0
MeOH
d)
0.0001
DMSO
MeOH
DMSO
MeOH
0.00008
0.00009
0.00007
0.00008
0.00006
Ion Yield Yi
0.00007
Ion yield Yi
0.00006 0.00005 0.00004 0.00003
0.00005 0.00004 0.00003 0.00002
0.00002
0.00001
0.00001
0
0
DMSO
MeOH
Figure 7. Example mass spectra for four drugs prepared either with the use DMSO (blue) or MeOH (green): a) clozapine b) haloperidol c) tamoxifen and d) warfarin.
The obtained data were processed as previously described and related isotopes on protonated ([M+H]+) ions for haloperidol, tamoxifen and clozapine or de-protonated ([M-H]-) molecular ion of Warfarin were taken into account. A clear difference in the drug ion intensities was observed between the samples prepared with the two solvents. The use of DMSO noticeably reduced the signal intensity of three of the drugs: haloperidol, tamoxifen and warfarin (Figure 7b, c, d), in comparison to the signal obtained from the samples where methanol was used A slight rise of clozapine signal was observed after treatment with DMSO (Figure 7a), however this is not significant. Overall, these results show the major influence the applied sample preparation procedures can have on the sensitivity of measurements.
SIMS intensity linearity with concentration Finally, we studied the changes in the molecular ion intensities as a function of concentration of the drugs within the tissue samples. Four pharmaceuticals were investigated, of which protonated molecular ions were clearly observable before in the positive ion mode: amiodarone, chloroquine, clozapine and haloperidol. Each drug was dosed in to the tissue homogenate at the range of concentrations (final amounts varying from 1.7 µg/mL to 167 µg/mL in the sample). To establish the reliability of the experiment and the homogenous distribution of the drug in the sample, four SIMS depth profiling experiments were carried out for each concentration of each drug. The average drug signal intensities (n = 4) were normalized to the primary ion current, as previously described. The intensities were calculated by integrating the main drugs peaks including, as before, related isotopes. In Figure 8 we show reasonable linear relationships with concentration for all four compounds. This is consistent with the matrix suppression and enhancement effects described by Shard et al.11
0.00014
Amiodarone
y = 7E-07x R² = 0.9247
Chloroquine 0.00012
Haloperidol Clozapine
0.0001
y = 5E-07x R² = 0.862
0.00008
Yi
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Analytical Chemistry
y = 3E-07x R² = 0.887
0.00006
0.00004 y = 2E-07x R² = 0.9226
0.00002
0
0
20
40
60
80
100
120
140
160
180
Drug concentration in the tissue (µg/mL)
Figure 8. Drug signal intensity for amiodarone, chloroquine, clozapine and haloperidol normalized to the number of primary ions for concentrations up to 167 µg/mL.
CONCLUSIONS We have developed a biologically relevant reference sample to study the sensitivity of SIMS to pharmaceutical compounds in tissue. The samples consists of BCR 185-R bovine liver standard dosed with one of twenty drugs. A carefully designed procedure results in good lateral and vertical homogeneity of the drug disposition so that repeat SIMS measurements have an average relative standard deviation of 11.5%. We use these samples with a drug concentration of 167 µg/mL to measure the relative sensitivity. Of the 20 drugs evaluated, 9 were observed as protonated molecular positive ions and 4 as deprotonated negative molecular ions. The other 7 pharmaceuticals did not provide any signal in either polarity because of strong matrix suppression effects. We find that most of the drugs observable in the positive polarity contain a tertiary amine, which corresponds to the protonation site of the drug molecule. We show that there is a linear correlation between the relative sensitivity (ion yield, Yi) and the drug log P value for drugs detected as positive ions. This provides a convenient empirical relationship for practical analysis to determine if a drug is likely to be detected in tissue at a given concentration. The log P value can be conveniently calculated using algorithms such as ALOGPS and ChemAxon, both used routinely in the pharmaceutical industry. The effective lipophilicity (log D at
ACS Paragon Plus Environment
Analytical Chemistry
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
pH 7) was also correlated to the relative sensitivity of pharmaceuticals observed in positive polarity. We also show, for the first time, that the positive secondary ion yield for a wide range of small molecules studied as pure compounds is directly proportional to the log P values. This is quite remarkable and will provide important insight into the elusive theory of molecule ionisation in SIMS. When comparing ion yields of drugs in tissue to the yields of pure compounds there is an enhancement of the signals above simple linearity by about 6 for positive and around 26 for negative ions. At the studied levels (concentrations of drugs in tissue below 1%) a good linearity is retained. However, at higher compositions, such enhancements could result in very non-linear responses. We observed the influence of solvent used for the sample preparation on the signal intensity. The results show clearly, based on the comparison of two different solvent systems, that the applied preparation procedures can have a significant effect on the results and ionisation efficiency. The reference samples were also used to investigate the linearity of the signal intensity with drug concentration up to ~160 µg/mL. A linear behaviour was observed. We believe that this reference sample may have utility in the characterization and optimization of SIMS instrument performance as well as other techniques including MALDI and Raman. In addition, it serves as a reliable test material for evaluating signal enhancement strategies (e.g. post-ionisation) for drug dosed tissue investigations. The ability to have a multi-drug dosed sample (“cassette dose”) is being investigated.
ASSOCIATED CONTENT SUPPORTING INFORMATION BCR-185-R tissue washing with ammonium formate; Summary of the studied compounds and their corresponding ions; Example SIMS spectra of drug compounds and related ions; The calculation of surface coverage of pure drug compounds; Summary of log D and Log P values from different sources; Relative sensitivity correlation to Log P from different sources; Relative sensitivity vs Log P for the negative secondary ion mode.
AUTHOR INFORMATION Corresponding Authors *E-mail:
[email protected],
[email protected] Notes The authors declare no competing financial interest.
ACKNOWLEDGMENTS This work forms part of the 3D NanoSIMS project in the Chemical and Biological programme of the National Measurement System of the UK Department of Business, Innovation and Skills.
REFERENCES (1) Dickson, M.; Gagnon, J. P. Nat. Rev. Drug. Discov. 2004, 3, 417-429. (2) Kaitin, K. I. Clin. Pharmacol. Ther. 2010, 87, 356-361. (3) Dollery, C. T. Clin. Pharmacol. Ther. 2013, 93, 263266. (4) Gilmore, I. S. J. Vac. Sci. Technol. A 2013, 31, 050819. (5) Matsuo, J.; Ninomiya, S.; Nakata, Y.; Honda, Y.; Ichiki, K.; Seki, T.; Aoki, T. Appl. Surf. Sci. 2008, 255, 1235-1238.
Page 8 of 10
(6) Seah, M. P. Surf. Interface. Anal. 2007, 39, 890-897. (7) Greer, T.; Sturm, R.; Li, L. J. Proteomics. 2011, 74, 2617-2631. (8) Sugiura, Y.; Setou, M. J. Neuroimmune Pharmacol. 2010, 5, 31-43. (9) Lietz, C. B.; Gemperline, E.; Li, L. Adv. Drug Deliv. Rev. 2013, 65, 1074-1085. (10) Rubakhin, S. S.; Jurchen, J. C.; Monroe, E. B.; Sweedler, J. V. Drug Discov. Today 2005, 10, 823. (11) Passarelli, M. K.; Newman, C. F.; Marshall, P. S.; West, A.; Gilmore, I. S.; Bunch, J.; Alexander, M. R.; Dollery, C. T. Anal. Chem. 2015, 87, 6696-6702. (12) Passarelli, M. K.; Wang, J.; Mohammadi, A. S.; Trouillon, R.; Gilmore, I.; Ewing, A. G. Anal. Chem. 2014, 86, 9473-9480. (13) Shard, A. G.; Spencer, S. J.; Smith, S. A.; Havelund, R.; Gilmore, I. S. Int. J. Mass Spectrom. 2015, 377, 599609. (14) Jones, E. A.; Lockyer, N. P.; Vickerman, J. C. Int. J. Mass Spectrom. 2007, 260, 146-157. (15) Lanekoff, I.; Stevens, S. L.; Stenzel-Poore, M. P.; Laskin, J. Analyst 2014, 139, 3528-3532. (16) Norris, J. L.; Caprioli, R. M. Chem. Rev. 2013, 113, 2309-2342. (17) Lamberty, A.; Kramer, G. N. BCR information; Reference Materials 1998. (18) Gilmore, I. S.; Seah, M. P. Surf. Interface Anal. 1995, 23, 248-258. (19) Havelund, R.; Seah, M. P.; Shard, A. G.; Gilmore, I. S. J. Am. Soc. Mass Spectrom. 2014, 25, 1565-1571. (20) Muramoto, S.; Brison, J.; Castner, D. Surf. Interface Anal. 2011, 43, 58-61. (21) Seah, M. P. Surf. Interface Anal. 1995, 23, 729-732. (22) Seah, M. P. J. Phys. Chem. C 2013, 117, 12622-12632. (23) Seah, M. P. Surf. Interface Anal. 2015, 47, 169-172. (24) Waring, M. J. Expert Opin.Drug Discov. 2010, 5, 235248. (25) Takács-Novák, K.; Avdeel, A. J. Pharm. Biomed. Anal. 1996, 14, 1405-1413. (26) Lipinski, C. A.; Lombardo, F.; Dominy, B. W.; Feeney, P. J. Adv. Drug Deliv. Rev. 2001, 46, 3-26. (27) Manallack, D. T. Perspect. Medicin. Chem. 2007, 1, 25-38. (28) Hollósy, F.; Valkó, K.; Hersey, A.; Nunhuck, S.; Kéri, G.; Bevan, C. J. Medicin. Chem. 2006, 49, 6958-6971. (29) Kah, M.; Brown, C. D. Chemosphere 2008, 72, 14011408. (30) Valkó, K.; Bevan, C.; Reynolds, D. Anal. Chem. 1997, 69, 2022-2029. (31) Leeson, P. Nature 2012, 481, 455-456. (32) Meyer, E. E.; Rosenberg, K. J.; Israelachvili, J. Proc. Natl. Acad. Sci. U. S. A. 2006, 103, 15739-15746. (33) Wermuth, C. G.; Ganellin, C. R.; Lindberg, P.; Mitscher, L. A. Pure Appl. Chem. 1998, 70, 1129. (34) Leo, A.; Hansch, C.; Elkins, D. Chem. Rev. 1971, 71, 525-616. (35) Geldenhuys, W. J.; Malan, S. F.; Bloomquist, J. R.; Marchand, A. P.; Van der Schyf, C. J. Med. Res. Rev. 2005, 25, 21-48. (36) Hansch, C.; Maloney, P. P.; Fujita, T.; Muir, R. M. Nature 1962, 194, 178-180.
ACS Paragon Plus Environment
Page 9 of 10
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Analytical Chemistry
(37) Sarkar, A.; Kellogg, G. E. Curr. Top. Med. Chem. 2010, 10, 67-83. (38) Tetko, I. V.; Bruneau, P. J. Pharm. Sci. 2004, 93, 31033110. (39) ChemAxon. https://wwwchemaxoncom/. (40) Tetko, I. V.; Poda, G. I. J. Med. Chem. 2004, 47, 56015604. (41) Law, V.; Knox, C.; Djoumbou, Y.; Jewison, T.; Guo, A. C.; Liu, Y.; Maciejewski, A.; Arndt, D.; Wilson, M.; Neveu, V.; Tang, A.; Gabriel, G.; Ly, C.; Adamjee, S.; Dame, Z. T.; Han, B.; Zhou, Y.; Wishart, D. S. Nucleic Acids Res. 2014, 42, D1091-D1097. (42) Hill, A. P.; Young, R. J. Drug Discov. Today 2010, 15, 648-655.
(43) Young, R. J.; Green, D. V. S.; Luscombe, C. N.; Hill, A. P. Drug Discov. Today 2011, 16, 822-830. (44) Amodeo, P.; Motta, A.; Picone, D.; Saviano, G.; Tancredi, T.; Temussi, P. A. J. Magn. Reson. 1991, 95, 201207. (45) Jacob, S. W.; Wood, D. C. Am. J. Surg. 1967, 114, 414426. (46) Pollice, A. A.; Philip Mccoy, J.; Shackney, S. E.; Smith, C. A.; Agarwal, J.; Burnolt, D. R.; Janocko, L. E.; Hornicek, F. J.; Singh, S. G.; Hartsock, R. J. Cytometry 1992, 13, 432-444.
ACS Paragon Plus Environment
Analytical Chemistry
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Table of Contents artwork:
ACS Paragon Plus Environment
Page 10 of 10