Investigating Nephrotoxicity of Polymyxin ... - ACS Publications

Aug 20, 2015 - Log In Register .... Biomolecular Imaging and Proteomics, National Center for Mass ... the utility of MSI by measuring the distribution...
6 downloads 0 Views 9MB Size
Article pubs.acs.org/crt

Investigating Nephrotoxicity of Polymyxin Derivatives by Mapping Renal Distribution Using Mass Spectrometry Imaging Anna Nilsson,†,# Richard J. A. Goodwin,‡,# John G. Swales,‡ Richard Gallagher,§ Harish Shankaran,∥ Abhishek Sathe,⊥,▽ Selvi Pradeepan,⊥,▽ Aixiang Xue,∥ Natalie Keirstead,∥,▽ Jennifer C. Sasaki,∥ Per E. Andren,† and Anshul Gupta*,⊥,▽ †

Biomolecular Imaging and Proteomics, National Center for Mass Spectrometry Imaging, Department of Pharmaceutical Biosciences, Uppsala University, Uppsala SE-75237, Sweden ‡ Drug Safety & Metabolism, Innovative Medicines, AstraZeneca R&D, Cambridge CB4 OWG, U.K. § Oncology DMPK, Innovative Medicines, AstraZeneca R&D, Alderley Park, Macclesfield SK10 4TF, U.K. ∥ Drug Safety & Metabolism, Innovative Medicines, AstraZeneca R&D, Waltham, Massachusetts 02451, United States ⊥ Infection DMPK, Innovative Medicines, AstraZeneca R&D, Waltham, Massachusetts 02451, United States S Supporting Information *

ABSTRACT: Colistin and polymyxin B are effective treatment options for Gram-negative resistant bacteria but are used as last-line therapy due to their dose-limiting nephrotoxicity. A critical factor in developing safer polymyxin analogues is understanding accumulation of the drugs and their metabolites, which is currently limited due to the lack of effective techniques for analysis of these challenging molecules. Mass spectrometry imaging (MSI) allows direct detection of targets (drugs, metabolites, and endogenous compounds) from tissue sections. The presented study exemplifies the utility of MSI by measuring the distribution of polymyxin B1, colistin, and polymyxin B nonapeptide (PMBN) within dosed rat kidney tissue sections. The label-free MSI analysis revealed that the nephrotoxic compounds (polymyxin B1 and colistin) preferentially accumulated in the renal cortical region. The less nephrotoxic analogue, polymyxin B nonapeptide, was more uniformly distributed throughout the kidney. In addition, metabolites of the dosed compounds were detected by MSI. Kidney homogenates were analyzed using LC/MS/MS to determine total drug exposure and for metabolite identification. To our knowledge, this is the first time such techniques have been utilized to measure the distribution of polymyxin drugs and their metabolites. By simultaneously detecting the distribution of drug and drug metabolites, MSI offers a powerful alternative to tissue homogenization analysis and label or antibody-based imaging.



INTRODUCTION Multidrug resistant bacteria continue to be a growing challenge among infectious disease causing pathogens. Whereas many antibacterial agents are no longer effective against these “superbugs,” polymyxins still remain an efficacious treatment option.1,2 However, clinicians use polymyxins (colistin methanesulfonate or polymyxin B) as the last-line therapy due to dose limiting renal toxicity.3,4 Although the mechanism of toxicity has not been definitively established, new insights into the fate and distribution of these agents may enable development of safer polymyxin analogues. Growing literature evidence suggests that these small peptides are reabsorbed in the proximal convoluted tubules by active transport processes resulting in accumulation in the kidney tissue.5,6 Given that the toxicity is primarily observed in the proximal convoluted tubules,7,8 more extensive accumulation is suspected in the renal cortical region. Investigators working in this field have utilized traditional bioanalytical approaches such as tissue homogenate analysis to determine the exposure of polymyxins in kidney.5,9 Whereas these techniques are useful in providing © 2015 American Chemical Society

information on the overall exposure levels, they fail to provide the complete picture of compound exposure due to loss of spatial information. The ability to measure the distribution of a compound at suborgan level may allow us to investigate if nephrotoxicity is related to the reduced exposure of polymyxin analogs in specific regions of the kidney. Traditionally, assessment of such distribution is by probe and label based imaging strategies such as immunohistochemistry or whole body autoradiography. For example, a method for detecting the biodistribution of polymyxin B in a mouse nephrotoxicity model using in situ immunostaining technique was recently published.10 Although such methods do provide detection at cellular resolution, they are of limited effectiveness in screening compound series during early drug discovery due to the associated cost/time required to produce robust probes or labeled compounds, which subsequently only measure the biodistribution of a single target. Furthermore, it is ultimately Received: June 21, 2015 Published: August 20, 2015 1823

DOI: 10.1021/acs.chemrestox.5b00262 Chem. Res. Toxicol. 2015, 28, 1823−1830

Article

Chemical Research in Toxicology

to obtain a therapeutically relevant or supratherapeutic exposure level that would induce kidney injury representative of colistin-induced nephrotoxicity in patients. Control animals were dosed with the vehicle, sterile saline, by injection. All groups were sacrificed and subjected to necropsy 3 h following the final dose of their respective vehicle or test article. Serial plasma and urine samples were obtained at various time points throughout the duration of dosing. Terminal plasma and kidney samples were collected at necropsy following CO2 euthanasia. All of the samples were stored at −80 °C until further analysis. Kidneys were dissected and trimmed of fat before being bisected (either transversely or longitudinally) using a scalpel. Half of each kidney was processed for MSI analysis, and the other half was homogenized for LC/MS analysis. MSI kidneys were snap-frozen free floating in dry ice chilled isopropanol and stored at −80 °C until further analysis. All subsequent transfers were performed on dry ice. Pharmacokinetic Analysis (Quantitation). Plasma and kidney homogenate were analyzed following methods previously described.8 Briefly, plasma samples (0.05 mL) were precipitated by using a crash solution (acetonitrile(ACN)/formic acid (95:5 v/v) (0.25 mL) containing 0.1 μg/mL fibrinopeptide B). Samples were mixed and centrifuged, and the supernatant was dried under N2 followed by reconstitution in a solution containing water/ACN/formic acid (95:5:0.1 v/v/v). Kidney samples were homogenized using a tissuemiser homogenizer (ThermoFisher Scientific, Waltham, MA) in 3-times volume of water. The kidney homogenate was diluted in plasma to allow direct comparison of plasma and tissue homogenates against the same calibration standard curve. This reduces the impact of matrix interference on absolute quantitation. The samples were then processed as described above. LC/MS/MS analysis was performed: colistin was monitored by fragmentation at m/z 578.6 > 227.1, PMB B1 at m/z 602.6 > 100.8, PMBN at m/z 482.6 > 120.3, and internal standard fibrinopeptide B (IS) at m/z 786.6 > 333.3. Mass Spectrometry Imaging. Tissue sections were cut using a cryostat microtome following mounting using distilled water. Sections were cut at 14 μm thickness and thaw-mounted onto indium tin oxide coated MALDI target slides (Bruker Daltonics, Bremen, Germany). Sections were taken at approximately equal depth to allow visualization of similar morphology between both transversely and longitudinally cut samples. Kidneys from day 1 dosed animals were sectioned longitudinally to allow sufficient area for LESA analysis. Subsequent kidneys were sectioned transversely to reduce the area of analysis and hence reduce acquisition times. Sections from multiple animals were mounted on each slide with a vehicle control section to minimize variability caused by variations in matrix application or sample handling. Tissue sections were analyzed randomly and nonsequentially to limit the risk of any observed variation in relative abundance being as a result in loss of analyzer sensitivity during the course of the analysis. Samples were analyzed by liquid extraction surface analysis (LESA) and MALDI MSI as previously described.17 Briefly, LESA analysis was performed using a Triversa Nanomate chip based electrospray ionization system (Advion, Ithaca, NY, USA) coupled to a QTRAP 5500 (AB Sciex, Framingham, MA, USA) mass spectrometer operated in positive ion MRM mode. LESA experiments are a direct infusion experiment and should not be confused with LC-MS/MS experiments where chromatographic separation occurs prior to MS analysis. Relative abundance was determined between sampling positions by comparison of MRM transition following optimization of compound standards. LESA data was processed using a purpose built software package capable of extracting relative abundance values from Analyst 6.1 (ABSciex, Framingham, MA, USA), and distribution images were created using in-house developed software capable of color grading ion intensities acquired from each individual LESA spot in a heat map configuration. Transitions for endogenous compounds were also monitored to confirm successful liquid microjunction extraction and ionization. Transitions and collision energies are summarized in Supporting Information (Table S4). For MALDI MSI, matrix solution (35 mg/mL DHB in 50:50:0.2 ACN/water/trifluoroacetic acid v/v/v) was applied to the tissue sections as previously described.17 For low spatial resolution, MALDI

the distribution of the label that is measured, and hence, differentiating between target and modified metabolites is not possible. Additionally, attaching such probes to the molecule raises concerns regarding effects on the biodistribution.11 We have previously demonstrated that the rat can serve as an excellent model to differentiate nephrotoxic agents such as colistin from non-nephrotoxic agents such as polymyxin B nonapeptide, PMBN.8 The intent of this study was to determine if the emerging mass spectrometry imaging (MSI) based technologies can be utilized to differentiate polymyxin analogues with varying degrees of toxicity based on suborgan distribution pattern in the rat nephrotoxicity model. MSI approaches encompass a range of direct ionization methods that sample from the surface of a tissue section and when coupled to mass spectrometry enable simultaneous, multiplexed, and label-free analysis of endogenous and exogenous compounds directly from tissue sections.12−14 Significantly, as the various MSI technologies do not rely on a label or probe but detect molecular mass-to-charge values, they are able to differentiate between drugs and their metabolites with higher specificity. Given the advantages that the MSI techniques offer over methodologies such as immunohistochemistry or whole body autoradiography, they are now being widely applied to pharmaceutical research and development.15,16 We have previously published on the application of the two MSI techniques utilized within these experiments for analysis of both exogenous and endogenous targets17,18 along with a detailed review on appropriate sample processing.19 Here, we have used lower spatial resolution liquid extraction surface analysis (LESA) MSI data to complement the distribution data obtained by matrix assisted laser desorption ionization (MALDI) MSI performed at 100 and 20 μm spatial resolution. Results obtained were further validated by corresponding whole organ homogenization and LC/MS/MS analysis. To our knowledge, this is the first time, data on polymyxin drugs, their metabolites, and simultaneous distribution of endogenous compounds is presented using MSI based label-free methods.



EXPERIMENTAL PROCEDURES

Compounds. Colistin sulfate (USP grade) was purchased from Sigma-Aldrich (St Louis, Missouri). Polymyxin B (PMB) was purchased from Apex Pharmaceuticals (South Plainsfield, NJ). Given that PMB is a multicomponent mixture of natural peptides, we purified the component present in highest abundance, polymyxin B1, in order to reduce lot to lot variability and used it for studies described here. Polymyxin B nonapeptide (PMBN) was prepared as described previously.20 Analytical grade 2,5-dihydroxybenzoic acid (DHB) were purchased from Sigma-Aldrich (Poole, Dorset, UK). Animals. Male Han Wistar rats weighing 275−375 g were purchased from Charles River Laboratories (Raleigh, NC, USA). The rats were given a minimum of 72 h to acclimate before the start of the experiments. Animal identification and conditions of housing, acclimatization, environment, diet, and water were in accordance with facility Standard Operating Procedures. All animal procedures were conducted in an Association for Assessment and Accreditation of Laboratory Animal Care accredited facility under an Institutional Animal Care and Use Committee approved protocol. Study Design. Animals were dosed with either colistin, PMB B1, or PMBN for 1, 2, or 7 days according to a previously described protocol.8 Briefly, each group was dosed 4 times daily (QID) for the duration of dosing by sc injection in the intrascapular region, approximately every 6 h. The colistin- or the PMB B1-treated groups were dosed with 6.25 mg/kg QID (25 mg/kg/day), and the PMBN group was dosed with 10 mg/kg QID (40 mg/kg/day). Doses of each compound were based on the results of maximum tolerated dose and 1824

DOI: 10.1021/acs.chemrestox.5b00262 Chem. Res. Toxicol. 2015, 28, 1823−1830

Article

Chemical Research in Toxicology MSI matrix was applied by a manual pneumatic sprayer. For higher spatial resolution analysis, the matrix was applied by an automated pneumatic sprayer (TM-sprayer, HTX Imaging). For standard resolution (100 μm) MALDI-MSI analysis was performed using a MALDI q-TOF MS (MALDI SYNAPT G2 HDMS, Waters Corporation, U.K.). Positive ion data were acquired over the range of m/z 100−1200 with 300 laser shots per raster position using a 1 kHz laser. Optimization of the mass spectrometer was achieved by tuning acquisition settings while collecting data from a manually deposited control spot of the target compounds (0.5 μL of 2 mM drug standard solution mixed 1:1 with matrix solution v/v). High resolution (20 μm) MALDI MSI experiments were carried out in positive reflectron mode over a mass range of m/z 100 to 1500 using the ultrafleXtreme MALDI-TOF/TOF MS (Bruker Daltonics) equipped with a 2 kHz, smartbeam-II Nd:YAG laser. Data were collected by summing up 500 laser shots/raster position. FlexImaging 4.1 (Bruker Daltonics) was used for data analysis and molecular image extraction. Both standard and high resolution analysis data were normalized by total ion count (TIC) and mass filter windows selected with a precision of ±0.04 Da. Images generated are displayed on heat-map relative abundance scales for a selected m/z or as monochrome scales for multiple m/z displayed simultaneously. Metabolite Identification Analysis. For tissue homogenate analysis, separation of the drug standards and their metabolites was carried out using a UHPLC system (Waters Acquity System, Manchester, UK) fitted with a Waters BEH C18 column with the following dimensions: 100 mm × 2.1 mm i.d., 1.7 μm particle size (Waters, Manchester, UK). The column was heated to 60 °C, and the mobile phase flow rate was 450 μL/min. For all separations, eluent A was 0.1% (v/v) aqueous formic acid, and eluent B was 0.1% formic acid in methanol. Ten microliters of each sample was injected and subject to the following mobile phase gradient: linear gradient used (T = minutes), at T = 0.0, 95%A/5%B; T = 9.0, 90%A/10%B; T = 9.1, 2% A/98%B; T = 11.0, 2%A/98%B; T = 11.01, 95%A/5%B; T = 13.00, 95%A/5%B. Mass spectrometry analysis was performed using a Thermo Fusion mass spectrometer fitted with a heated electrospray ionization source (ThermoFisher Scientific, Hemel Hempstead, UK). The mass spectrometer was operated in positive ionization mode in a data dependent mode using the following scan cycle; scan 1, full scan m/z 200 to 1400; scan 2, MS/MS product ion scan of the most intense ion detected in scan 1 at 80 eV collision energy. All data were collected at 15k resolution (50% valley definition). Accurate mass measurement was maintained to within 2 ppm by using two solvent/gas background impurities as lock masses: butylbenzenesulfonamide (m/z 214.0896) and di-iso octylphthalte (m/z 391.2842). For urine analysis, 400 μL of urine was added to equal volume of ACN, vortex mixed for 5 s, and centrifuged at 14k rpm for 10 min. Six hundred microliters of supernatant was added to 1 mL of water, and 20 μL was injected onto the LC/MS system and analyzed as for tissue homogenates.

Figure 1. MSI images for polymyxin B1 displaying relative abundance distribution in rat kidney tissue sections. (A) LESA MSI at 1000 μm spatial resolution for a 14 μm thick kidney section from 1 day dosed animal showing [M + H]2+ MS/MS transition 602.6 > 100.8 (scale bar = 5 mm) adjacent to the optical scanned image. (B) MALDI MSI of two animals from each 1, 2, and 7 day groups for polymyxin B1 dosing analyzed at 100 μm spatial resolutions: (i) example of optical image of sections on slide prior to MALDI or LESA analysis; (ii) example of endogenous compound marker (m/z 853.70) used to map tissue morphology across drug dosed and vehicle control sections. (iii) Relative abundance distribution of polymyxin B1 [M + H)+ at m/z 1203.8 (scale bar = 5 mm). Heat-map scale bars are representative of 0−100% A.U. (arbitrary units) with no thresholding.

predominantly localized to the cortical region of the kidney. LESA MSI distribution data were rapidly obtained and did not require any optimization such as matrix screening, which is often required for MALDI analysis. The liquid microjunction extraction method and solvent composition were the same as previously used for small molecule exogenous compound analysis by LESA MSI.17 Maximum abundance for polymyxin B1 was 4502 arbitrary units (a.u.), mean tissue abundance 880 a.u., while vehicle control tissue (data not shown) generated maximum abundance for the 602.6 > 100.8 transition of 19 a.u. Following LESA profiling, MALDI MSI was performed at 100 μm (Figure 1B). Tissues were analyzed from duplicate animals in all dose groups (1, 2, and 7 day dosing) by MALDI MSI. Tissue sections were thaw-mounted to the same slide prior to DHB matrix application. Following analysis, the relative abundance for arbitrarily selected endogenous compounds showed similar distributions across all samples indicating sections representing similar gross morphology of the kidney from both transversely and longitudinally processed tissues. A representative sample showed the relative abundance of m/z 853, an endogenous marker localized to the cortical region (Figure 1Bii). The molecular image for m/z 1203 corresponding to [M + H]+ of polymyxin B1 shows greatest abundance in



RESULTS AND DISCUSSION Polymyxins are cationic polypeptide drugs that despite excellent efficacy suffer from their potential to cause nephrotoxicity, and as previously alluded to, the mechanisms of toxicity are not clearly understood. However, with recent research in this area, it is becoming clear that understanding regional tissue exposure in the kidney is critical for refinement in dosing of standard-ofcare polymyxins but also to aid developing newer/safer polymyxin analogues. Therefore, the utility of MSI was evaluated by assessing the distribution of polymyxin B1, colistin, and polymyxin nonapeptide directly from rat kidney tissue sections. Systemic Tissue Imaging. Preliminary tissue screening experiments to determine a systemic overview of polymyxin B1 relative abundance distribution was achieved by analyzing kidney sections using LESA MSI at 1000 μm spatial resolution (Figure 1A). Analysis of a section taken from rat kidney following 1 day of dosing showed that polymyxin B1 1825

DOI: 10.1021/acs.chemrestox.5b00262 Chem. Res. Toxicol. 2015, 28, 1823−1830

Article

Chemical Research in Toxicology

Figure 2. High spatial resolution (20 μm) MALDI MSI of colistin in rat kidney following 1 to 7 days of repeat dosing. (A) Optical images of kidney sections with regions of analysis (ROA) marked by white dotted lines. (B) MSI relative abundance images for each ROA of extracted masses for colistin [M + H]+ at m/z 1169.9, with endogenous compounds localizing to the heme (m/z 616.2), cortex (m/z 724.3), and medulla (m/z 257.2), each individually displayed on a heat-map scale. Note: Heme abundance for day one appears to be off tissue contamination. (C) Expanded image of the ROA for a 7 day dosed animal with each extracted mass displayed simultaneously on monochromatic color scales. Heat-map scale bars are representative of 0−100% A.U. (arbitrary units) with no thresholding.

the cortex region with increasing duration of dosing. To enable visualization of the compound distribution relative to endogenous molecular markers, the expanded image of the region of day 7 displays simultaneously each marker on monochromatic color scales (Figure 2C). Using the mass spectra data sets collected, it was possible to compare the different tissues and identify molecular masses that were only present or have significantly different abundances between certain groups or tissue areas. Comparing summed spectra from different regions of interest (ROI) in gel view, as previously described, allowed rapid visualization of these differences enabling detection of compound adducts and possible drug metabolites and is summarized in Figure 3.21 This approach enabled the colistin sodium and potassium adducts to be detected. When these are codisplayed, it highlighted that there was no difference in adduct distribution (Figure 3B). Figure 3 also summarizes a number of molecular m/z values that were found to be present only after treatment with colistin. These m/z values were also observed to localize to the cortical region, in similar distribution as colistin. Furthermore, they were detected with increased abundance following increased exposure to the compound from repeated dosing. These m/z values were therefore likely metabolites of

the cortex for all dose groups, matching the LESA distribution. However, the MALDI analysis allows multiple samples to be coanalyzed and shows there was also increased cortical accumulation with increased number of days dosed (Figure 1B, iii). High Resolution Imaging. With both LESA and lower spatial resolution MALDI generating similar global distributions over entire tissue sections, it was rational to perform subsequent higher resolution MALDI analysis on only smaller regions of the tissue sections. Therefore, 20 μm spatial resolution imaging was performed for all three benchmark compounds over a 5 mm wide area encompassing cortical and medullary regions. As MALDI analysis could be performed in full spectrum mode, it was again possible to extract the distribution images for endogenous compounds as well as the drug and metabolites. High spatial resolution (20 μm) MALDI MSI analysis of colistin dosed tissues is summarized in Figure 2. The distribution of colistin [M + H]+ at m/z 1170 is shown in relation to the heme at m/z 616 and two endogenous compounds that localize to the cortex or medulla, respectively (Figure 2B). The endogenous markers had a similar abundance and distribution within the control and 3 dose groups; however, there was a greater abundance and accumulation of colistin in 1826

DOI: 10.1021/acs.chemrestox.5b00262 Chem. Res. Toxicol. 2015, 28, 1823−1830

Article

Chemical Research in Toxicology

Figure 3. High spatial resolution (20 μm) MALDI MSI of colistin and colistin metabolites in rat kidney following 1 to 7 days of repeat dosing. (A) Optical images of kidney sections with regions of analysis (ROA) marked by white dotted lines. (B) MSI relative abundance images for each ROA of extracted masses for colistin [M + H]+ at m/z 1169, colistin distribution combined with adducts each on monochromatic green scale ([M + H]+, [M + Na]+, and [M+K]+), and masses detected in tissues following drug treatment that are absent from the vehicle control section. Each one is individually displayed on heat-map scale. Heat-map scale bars are representative of 0−100% A.U. (arbitrary units) with no thresholding.

colistin and hence targets for subsequent metabolite identification using homogenized tissues extracts. Similar high resolution analysis was performed on the polymyxin B1 and PMBN tissues, and the results are summarized in Supporting Information. As previously observed by lower resolution MALDI and LESA MSI, polymyxin B1 was detected in the cortical region with increasing abundance with repeated dosing. The polymyxin B1 adducts, as well as a number of other molecular ions, were also detected predominantly in the cortical region; increasing in abundance with repeated dosing (Figure S1). For PMBN a similar, though less pronounced, profile was detected (Figure S2). There was less marked localization to the cortical region and less prominent accumulation after 7 days dosing compared to the increased relative accumulation seen in polymyxin B1 or colistin after 7 days of dosing. There were also fewer putative metabolite m/z values detected. Cortex to Medulla Ratios. MSI data displayed as images on a relative abundance heat-map color scale are valuable in providing an overview of compound distribution, as well as how that distribution relates to other target molecules. However, such images can over- or under-represent changes in compound distribution or relative abundance.22 Therefore, the ratio of the relative abundance of each compound within the cortex and medulla region was calculated (Table 1). This simple numeric ratio provides insight into the changing relative abundance of the different compounds. PMB B1 showed an initial cortex to medulla ratio of 5:1 after 1 day of dosing; increasing to 7.4 after

Table 1. Ratio of Relative Abundance for Each Compound, [M + H]+ for the Cortical Region Compared to the Medullary Regiona compound colistin

polymyxin B1

PMBN

day

cortex/medulla ratio

1 2 7 1 2 7 1 2 7

3.3 4.5 5.8 5.0 7.4 25.4 2.1 2.1 2.9

a

Summed spectra are baseline subtracted and normalized to TIC. The relative abundance of the protonated, sodium, and potassium adducts ratios ([M + H]+, [M + Na]+, and [M+K]+) were measured and shown to be consistent both in the cortex and in the medulla, and therefore, only protonated ratios are presented here. Regions of interest were randomly selected to cover a large area of the cortex or medulla for each sample and the abundance of target mass summed and averaged by MSI software (FlexImaging, Bruker).

2 days. However, following 7 days of dosing of the ratio had reached 25:1. Colistin also exhibited an increase in cortex accumulation with almost double the abundance after 7 days dosing compared to a single day’s dosing but certainly less pronounced compared to that of PMB B1. PMBN, the least nephrotoxic of the compounds, not only had the lowest cortex 1827

DOI: 10.1021/acs.chemrestox.5b00262 Chem. Res. Toxicol. 2015, 28, 1823−1830

Article

Chemical Research in Toxicology

vastly different between colistin and PMBN. The most significant change in absolute quantitation was for polymyxin B1, with absolute concentration increasing to over 350 μg/g after 7 days of dosing compared to approximately 100 μg/g following 2 days dosing. The MSI data indicate that the increase in abundance of polymyxin B1 is not uniform within the tissue but localized to the cortex region (Figure 1 and Table 1) as the cortex to medulla ratio increased from 5:1 to 25:1 over the 7 days. This suggests that polymyxin B continues to accumulate over the dosing period in the kidney compared to that of colistin and PMBN. Since the kidney exposure would be a function of plasma concentrations, we further calculated the terminal kidney/ plasma ratio. Results from such analyses are summarized in Table 3 along with typical PK parameters. Kidney/plasma ratios were high ranging from 9.8 to 58 suggesting preferential partitioning in the kidney of all the polymyxins tested. These ratios increased for PMBN (