Exemplifying the Screening Power of Mass Spectrometry Imaging over Label-Based Technologies for Simultaneous Monitoring of Drug and Metabolite Distributions in Tissue Sections

Abstract

Mass spectrometry imaging (MSI) provides pharmaceutical researchers with a suite of technologies to screen and assess compound distributions and relative abundances directly from tissue sections and offer insight into drug discovery–applicable queries such as blood-brain barrier access, tumor penetration/retention, and compound toxicity related to drug retention in specific organs/cell types. Label-free MSI offers advantages over label-based assays, such as quantitative whole-body autoradiography (QWBA), in the ability to simultaneously differentiate and monitor both drug and drug metabolites. Such discrimination is not possible by label-based assays if a drug metabolite still contains the radiolabel. Here, we present data exemplifying the advantages of MSI analysis. Data of the distribution of AZD2820, a therapeutic cyclic peptide, are related to corresponding QWBA data. Distribution of AZD2820 and two metabolites is achieved by MSI, which [¹⁴C]AZD2820 QWBA fails to differentiate. Furthermore, the high mass-resolving power of Fourier transform ion cyclotron resonance MS is used to separate closely associated ions.

Introduction

During the discovery and development of new therapeutics, an understanding of compound biodistribution, accumulation, efficacy, and toxicity is of fundamental importance. Differentiating between alternative compounds in a series, or between chemical classes, requires the screening of many analogues to determine which would be most advantageous to progress. For later-stage compounds, this can be time-consuming and laborious, as indirect monitoring is typically performed using labeled compounds. ¹ The use of labeled compound, typically ³H- or ¹⁴C labeled in quantitative whole-body autoradiography (QWBA), risks misrepresenting the abundance of a compound, as all measurements are based on the abundance of the label. Should a compound be metabolized, leaving the label with any subsequent metabolite, it cannot be distinguished from the parent compound. Similarly, pharmacodynamics understanding requires additional labels or probes to measure the distribution of endogenous compounds.

Overcoming the concern of tracking only the label often requires extensive bioanalytical investigation using sample homogenization, which results in losing all the spatial distribution information. ² Other drawbacks in label-based studies include the time required to synthesize the probe and the associated direct and indirect costs. ³ This limits the number of compounds that can be evaluated. Nevertheless, QWBA remains a crucial component of studies in the later stages of compound development prior to a compound becoming a clinically approved drug. This is due to the sensitivity of the methodology and applicability of the technique to address specific questions about exposure and accumulation in different tissues and to relate to human mass balance studies. In addition, data can feed into investigational new drug applications, and there is also regulatory acceptance of the methodology. ⁴

New analytical techniques that allow higher-throughput screening of compounds and a direct measurement of compound biodistribution in tissues can significantly affect drug discovery. Recently mass spectrometry imaging (MSI) technologies are offering drug discovery researchers the ability to screen compounds without the requirement of labeled compound or indirect probe-based measurements. By direct sampling from the surface of a tissue section, using one of several, distinct ionization methods, MSI technologies are able to delineate between parent compound and metabolites. Images representing the relative distribution of compound can be generated by relating the spatial position at which the mass spectra were acquired with the relative abundance of the selected target ion.^5,6 Further, relative abundance images can be generated by selection of other target ions within the acquired spectra that can either be compound metabolites or endogenous small molecules, lipids, or peptides. Currently, there are three commonly employed MSI ionization methods that have particular utility for research in drug discovery. ⁷ The most common is matrix-assisted laser desorption ionization (MALDI), ⁸ with desorption ionization electrospray^9,10 and secondary ionization mass spectrometry^11,12 providing additional modes of sample ionization applicable to drug distribution analysis. These techniques have been extensively reviewed, and each has advantages and limitations that determine their applicability to wide-ranging applications.^13–16

Presented here is an example of the utility of MSI to measure the distribution of compounds and their metabolites directly from tissue sections. The distribution of a therapeutic cyclic peptide, AZD2820, within a mouse kidney tissue section is made by MALDI-MSI and compared directly with corresponding QWBA data for a ¹⁴C radiolabel analogue of the same peptide. AZD2820 is a cyclic heptapeptide with partial agonist properties at the melanocortin-4 receptor mediating appetite suppression, intended to be a parenterally administered antiobesity drug. This compound was chosen for this study because of the availability of QWBA study data and additional tissue for MSI analysis. No inference regarding the efficacy or safety of AZD2820 can be made from the data presented. The data presented exemplify the potential of direct label-free analysis and the utility of applying such technologies to screening compound biodistribution in early drug discovery.

Materials and Methods

Chemicals and Reagents

AZD2820 (C₄₉H₆₉FN₁₆O₉) and [¹⁴C]AZD2820 were supplied by AstraZeneca compound management, Sweden. Solvents were high-performance liquid chromatography (HPLC) grade (Fisher Scientific, Loughborough, UK). All other chemicals were purchased from Sigma-Aldrich (Dorset, UK) unless otherwise stated.

Animals

Tissues from AZD2820-treated animals were obtained from AstraZeneca (DMPK R&D, Södertälje, Sweden, ethical approval No. 50/09). Male C57Bl:6J pigmented mice weighing 19 to 22 g were purchased from Charles River Laboratories (Germany). The mice were acclimatized for 1 wk prior to dosing. Animal identification and conditions of housing, acclimatization, environment, diet, and water were in accordance with facility standard operating procedures.

Quantitative Whole-Body Autoradiography

The QWBA experiment was performed following industry standard protocols. In brief, the distribution of radioactivity in mouse tissue was measured after subcutaneous administration of [¹⁴C]AZD2820 using phosphor-imaging. A single dose of 4 mg/kg (3.8 µmol/kg, 7 MBq/kg) of [¹⁴C]AZD2820 was given to male mice. The mice were sacrificed between 0.5 h and 7 d after administration. Compound was formulated for dosing as solution in 5% mannitol (v/w) for subcutaneous administration. The animals were sacrificed with enflurane (Efrane, Abbott Laborotories), frozen in acetone, and cooled to −70 °C with solid CO₂. After removal of limbs and tail, the carcass was embedded, with left lateral side uppermost, in a 2.5% aqueous solution of carboxymethyl cellulose and frozen for at least 10 min in acetone at −70 °C. The animal blocks were stored at −20 °C until sectioning. At sectioning, each block was mounted in a Leica CM3600 Cryomacrotome (Leica Microsystems GmbH, Germany) maintained at approximately −20 °C. Sagittal whole-body sections (30 µm) were obtained to include organs and tissues of interest. The sections were mounted on invisible tape (type 810; 3M) and numbered consecutively with radioactive ink. All sections were dried at −20 °C for at least 1 d prior to exposure on phosphor-imaging plates. Sections were chosen for phosphor-imaging to best represent the tissues and organs of interest. Together with two sets of calibration standards, the sections were placed on phosphor-imaging plates, which had been precovered with a thin plastic film. The imaging plates were exposed and enclosed in light-tight cassettes in a lead-shielding box to protect from environmental radiation. Following exposure, the imaging plates were scanned at a pixel size of 50 µm using FLA-7000 (FujiFilm Sverige AB, Sweden).

MSI Analysis

AZD2820 mouse kidney tissues were prepared for MALDI analysis following similar methods as previously described. ¹⁷ In brief, following a single subcutaneous administration of nonlabeled AZD2820 (25 mg/kg), mice were sacrificed. Direct matching of time points was not possible as nonradiolabel experiments were performed primarily for reasons beyond the scope of this experiment and kidney tissue collection was opportunistic. Animal welfare prohibits rerunning animal experiments for publication purposes alone. Kidneys were dissected 30 and 120 min postdose and snap frozen on dry ice–chilled isopropanol and stored at −80 °C. Tissue sectioning was performed on Leica CM3050S cryostat (Leica, Microsystems GmbH, Germany) at −20 °C. Kidney sections (12 µm thick) were thaw mounted onto indium tin oxide–coated mass spectrometry–compatible glass slides (cat. No. 237001; Bruker Daltonics, Germany). Subsequent transfer of samples was performed on dry ice, and storage was at −80 °C. Tissue washing was performed prior to matrix coating and consisted of 30 s in 70% ethanol 30% water followed by 30 s in 100% ethanol before being dried under a stream of nitrogen. Wet matrix (4 mg/mL α-cyano-4-hydroxycinnamic acid [CHCA] in 50% acetonitrile/50% 0.1% trifluoroacetic acid in HPLC-grade water) was applied manually using a pneumatic TLC sprayer. The MALDI target was repeatedly rotated during the coating procedure to prevent any bias in matrix thickness. MALDI MSI was performed using an Ultraflex II TOF/TOF and 12 T SolariX Fourier transform ion cyclotron resonance (FTICR) MS (Bruker Daltonics) in positive ion mode using Smartbeam lasers. The laser spot size was set at medium focus (approximately 50 µm laser spot diameter), and laser power intensity was optimized at the start of each run and then fixed for the MSI experiment. Data were analyzed and normalized using FlexImaging version 2.0, and extracted masses were selected with a 0.2 Da window. An estimation of the sensitivity of the analysis was performed by spotting different concentrations of AZD2820 on control tissue and then performing the analysis in exactly the same way as the dosed tissues. The following amounts were put on tissue in 0.2 µL (50:50 ethanol:water): 0.32 pmol, 0.064 pmol, 12.8 fmol, 2.6 fmol, 0.5 fmol, and 0 fmol. The ion distribution of AZD2820 is visualized on a rainbow scale from 0% to 10% of maximum intensity. The experiment was repeated four times, and the average signal from 0.5 fmol on tissue was approximately three times the background signal (summarized in Suppl. Fig. S5). For FTICR analysis, instrument control used SolariX control v1.5.0 (build 42.8) and Hystar 3.4 (build 8). Prior to analysis, the instrument was calibrated using Agilent tune mix and a quadratic best fit curve. Conformation of the accurate masses were performed using individual spectra in Data Analysis 4.1, and a single-point calibration against the [2M+Na] 1+ CHCA cluster ion was used. Each analysis was the result of 200 laser shots, using a medium focus laser spot (approximately 60 µm laser spot diameter). Ions were detected between m/z 200 and 3000, yielding a 1 Mword time-domain transient. Extracted masses were selected with a 0.002 Da window. Use of a 0.002 Da mass selection window was more stringent than required by resolution of collected spectra; however, the mass selection window was centered on the apex of the peak, and the resulting image was not negatively affected by the narrow range. It should be reiterated that the efficacy, safety, metabolism, and pharmacokinetic and pharmacodynamic properties of the drugs used during the experiments described here are beyond the scope of this research. They have been selected as tool compounds to exemplify the applicability of MSI for obtaining more information than was achievable by use of labeled compound or tissue homogenization. No further inference should be drawn about the compounds other than the biodistributions presented.

Results and Discussion

A primary function of QWBA analysis is to obtain a quantitative overview of the distribution of the radiolabel and then infer the distribution of the administered compound. This can be achieved from whole-animal tissue sections, as shown in Figure 1 . The image is representative of data obtained for the distribution of [¹⁴C]-labeled AZD2820 cyclic peptide (structure shown in Suppl. Fig. S1). It is worth noting that the production of the radiolabel took longer than 6 months because of the complexity of the synthesis. The image shows that a systemic distribution is achieved over the mouse, excluding intestine. The compound appears to be at highest abundance at the site of the subcutaneous injection, the kidney, and in the bladder. However, as for all label-based assays, it does not differentiate between parent compound, any metabolites containing the labels, or if the compounds are free or covalently bound. To determine whether the QWBA image represents the true distribution of the compound, MALDI-MSI analysis was performed. Kidney tissue sections (12 µm thick) were cut from snap-frozen samples and analyzed at 50 µm spatial resolution using a MALDI–time-of-flight (TOF) mass spectrometer. Multiple tissue sections were thaw mounted to the glass slide, but only a single section from each group was analyzed (vehicle control, 30 and 120 min postdose), as summarized in Supplementary Figure S2.

Figure 1.

[¹⁴C]AZD2820 male pigmented mouse, 1 h after subcutaneous administration. Autoradiogram showing the distribution of radioactivity (dark areas) over a whole-body sagittal tissue section from a male C57Bl:6J mouse 1 h after single subcutaneous administration of [¹⁴C]AZD2820.

As [¹⁴C]AZD2820 has two major metabolites that retained the radiolabel (Suppl. Fig. S1), the autoradiograph images in Figure 2 will be an amalgamation of the summed total of the abundance of the three labeled molecules at an unknown ratio. To determine if MSI can differentiate between metabolites, where autoradiography could not, the relative abundance of the corresponding m/z were mapped for AZD2820 [M+H]⁺ at m/z 1045 and for M1 metabolite [M+H]+ at m/z 878. Ion distribution images were generated and are summarized in Figure 2 . The distribution of AZD2820, displayed on a heat map scale image, confirmed that the compound was detected in the 30 and 120 min samples and was not detected in the vehicle control. The compound was predominantly detected in the outer medulla and cortex at 30 min postdose, but by 120 min, it was detected only in the medulla and toward the kidney pelvis. The M1 metabolite was predominantly detected in the cortex at 30 and 120 min but with relatively higher abundance at 120 min compared with the earlier time point. In the QWBA images, from animals most closely matching time postdose for the dissected kidneys, a uniform distribution over the kidney tissue section was detected with a dark aggregate toward the center of each kidney. This central accumulation (pelvis) corresponds to high abundance of the compound in urine. For normal MALDI-MSI analysis of small-molecule drugs and endogenous compounds, tissue samples are typically not washed prior to matrix application and analysis. However, for larger molecular weight targets or proteomic analysis, tissue sections are washed briefly to remove low-molecular-weight salts and cell debris that suppress ionization of larger molecular weight targets.^18–20 Washing of sections was performed in this study and improved detection of the cyclic peptide and its metabolites (as well as removed the urine from the pelvis region of the sample). The effect of washing on the mass spectra is demonstrated in the supplementary information (Suppl. Fig. S3), where relative abundance and signal to noise were improved for the target masses for the same number of summed laser shots. Data not presented compared target distributions before and after washing and showed limited delocalization but improved detection sensitivity. It is worth noting that we are assuming that we have no major delocalization of drug or drug metabolites during tissue washing. This is because detected distributions were similar but with improved signal to noise from washed tissues. More extensive experiments, including analysis of wash solution after use for abundance of metabolites, would be needed to obtained exhaustive distribution characterization. The MSI analysis performed here displays only the detected relative abundance and distribution of the target molecule and does not estimate absolute quantitation. Data presented in Figure 2 illustrate the extracted molecular images on a heat map scale; however, they can also be displayed on monochromatic scales. If these are displayed on the same monochromatic scale, molecular images representing combined distribution that replicate QWBA distribution can be produced (Suppl. Fig. S4). It should be noted that although specificity for different molecular targets is greater by MS analysis, there is typically greater sensitivity by QWBA.

Figure 2.

Matrix-assisted laser desorption ionization mass spectrometry and autoradiogram images of distribution of AZD2820 in mouse kidney. (A) Autoradiogram showing the distribution of radioactivity (dark areas) within the sagittal kidney section from male C57Bl:6J mice 30 min and 1 h after single subcutaneous administration of [¹⁴C]AZD2820. (B) Relative abundance of AZD2820 [M+H]⁺ at m/z 1045.7 and M1 metabolite [M+H]⁺ at m/z 878.6 at 30 min and 2 h postdose within the sagittal kidney section from male C57Bl:6J mice. Mass spectrometry imaging data collected at 50 µm spatial resolution. Presented on heat map, scale inserted. Scale bar 2 mm.

Analysis of the distribution of the M2 metabolite was more complex as it has a molecular weight similar to that of the parent compound. Figure 3 shows the theoretical isotope distribution of AZD2820 (m/z 1045.55) and the M2 metabolite (m/z 1046.53). With the mass-resolving power of the MALDI-TOF mass spectrometer, the second isotope peak of the AZD2820 will mask the monoisotopic peak of the M2 metabolite. This is shown in the overlaid theoretical isotope distributions. This simulated theoretical isotope distribution was similar to the actual detected spectra collected from the kidney tissue sections, as seen in Figure 3A . Therefore, both parent compound and M2 metabolite will contribute to extracted ion images for m/z 1046.53 (using a 0.2 Da mass window typical for MALDI-TOF MS). MSI typically overcomes multiple compounds contributing to the relative abundance of a target mass by performing fragmentation of the target and mapping the unique fragment masses. However, the similarity of the composition of these compounds could generate similar fragments. To overcome this issue, the analysis was repeated on a mass spectrometer with higher mass-resolving power. FTICR MS determines the m/z of ions using the cyclotron frequency of the ions in a magnetic field. ²¹ As a result, the mass-resolving power can be in excess of 1,000,000 on instruments with a higher magnetic field rather than the 15,000 achieved on the TOF mass analyzers. Figure 3B shows that with increased mass-resolving power, it was possible to differentiate between the AZD2820 (C₄₉H₆₉FN₁₆O₉) second isotope peak at m/z 1046.5524 and the M2 metabolite (C₄₉H₆₈FN₁₅O₁₀) m/z 1046.5336. Figure 3C compares lower spatial resolution MALDI FTICR MS image data to that collected at high spatial resolution on the MALDI-TOF MS. The parent compound distributions are comparable at both 30 and 120 min postdose. The significant difference between FTICR and TOF acquired data is seen for M2 metabolite at 120 min postdose. The MALDI-TOF is unable to distinguish between the second parent isotope and the first M2 metabolite isotope peak. Therefore, the MALDI-TOF extracted image for the M2 metabolite is the same as the parent distribution. The MALDI FTICR can differentiate between the second parent isotope peak and the M2 metabolite. The M2 metabolite is detected at 30 min postdose but has substantially decreased in abundance by 120 min. We attempted to map the distribution of the M2 metabolite using the second (m/z 1047.55), third, and fourth isotope peaks in the MALDI-TOF MSI data. However, the parent isotope peak signal still contributed to the abundance for the first two isotope peaks. For the last isotope peak, there was an endogenous interfering signal, which makes the measurements less specific. Therefore, FTICR-MS analysis was performed. This exemplifies the power of label-free distribution analysis and highlights that validation by higher mass-resolving power FTICR MS is often required. Although not performed during this study, it is worth reiterating that high spectral resolution MSI analysis of drugs can also simultaneously detect and monitor endogenous metabolites. These can then be used as pharmacodynamic or toxicity markers. ²² Although other groups have previously published data on small-molecule QWBA versus MSI, ²³ we have presented on large-molecule therapeutics, which to our knowledge have not previously been compared by MSI and QWBA. We also provide all parent and metabolite structures. Furthermore, we present distribution data not at a systemic level but focus on a suborgan level and show how different the MSI data are compared with QWBA. What is also highlighted here is the complexity of closely matched metabolite and parent compounds, requiring multiple MSI platforms to truly elucidate the relative distributions. The experiment outlined here demonstrates the ability to simultaneously monitor the distribution of parent compound and drug metabolites and exemplifies the usefulness of MSI as a screening tool for drug discovery.

Figure 3.

Comparison of mass-resolving power of Fourier transform ion cyclotron resonance (FTICR) and time of flight (TOF) to distinguish between AZD2820 and M2 metabolite. (A) Theoretical isotope distributions of [M+H]⁺ for AZD2820 (C₄₉H₆₉FN₁₆O₉) and M2 metabolite (C₄₉H₆₈FN₁₅O₁₀) and examples of spectra detected from different regions of mouse kidney 30 min postdose of AZD2820. (B) Example spectra acquired by matrix-assisted laser desorption ionization (MALDI) FTICR MS from mouse kidney 30 min postdose of AZD2820. The larger image is the zoomed region of the spectra shown in the insert. Marked are the AZD2820 second isotope peak and the monoisotopic peak for M2 metabolite. (C) Relative abundance of AZD2820 [M+H]⁺ at m/z 1045.7 and M2 metabolite [M+H]⁺ at m/z 1046.7 at 30 min and 2 h postdose within the sagittal kidney section from male C57Bl:6J mice for data acquired by MALDI FTICR and MALDI-TOF. MALDI-TOF image for M2 metabolite misrepresenting true distribution because of the inability to distinguish closely matching mass-to-charge peaks. MALDI FTICR data collected at 150 µm spatial resolution and MALDI-TOF at 50 µm spatial resolution. Presented on heat map, scale inserted. Scale bar 2 mm.

We believe the data presented highlight that MSI data can complement that generated by QWBA. The data show that for certain targets, higher spatial resolution and quicker MALDI-TOF MSI are sufficient (detection and mapping parent and M1 metabolite). However, the utility of FTICR MSI needs to be employed for accurate mass validation and M2 metabolite analysis. No one analytical technique is universally advantageous; however, MSI may offer researchers in drug discovery a relatively rapid screening tool for assessing compound and metabolite distribution at a tissue level.