maldi imaging ms reveals candidate lipid markers of ... were acquired in esi nega- tive ion mode...

This article is available online at http://www.jlr.org Journal of Lipid Research Volume 54, 2013 2785

Copyright © 2013 by the American Society for Biochemistry and Molecular Biology, Inc.

Autosomal recessive polycystic kidney disease (ARPKD) is a severe childhood-onset, hepatorenal fi brocystic disease (HRFCD), affecting approximately 1 in 20,000 children ( 1 ). It is characterized by the progressive formation of massive renal cysts and by biliary dysgenesis resulting in congenital hepatic fi brosis ( 2, 3 ). Human ARPKD is caused by mutations in polycystic kidney and hepatic disease 1 , one of the largest known human genes, which is expressed in adult and fetal kidney, liver, pancreas, and lung ( 4 ). Clinical outcome of ARPKD is highly variable with different grades of renal and/or hepatic symptoms, but ultimately fatal no later than in early adulthood ( 1, 5 ). While hepatocellular function remains relatively stable in ARPKD, decreased urinary concentration capability is commonly the first symptom, followed by successive renal insuffi ciency. As a consequence, the majority of ARPKD patients undergo kidney transplantation with or without liver transplanta-tion ( 6, 7 ). No disease-specifi c treatment is currently avail-able for ARPKD patients. Therapy is directed at long-term complications, such as hypertension, portal hypertension, renal infections, and recurrent cholangitis ( 7, 8 ). Up to now, no simple diagnostic test for ARPKD has been avail-able. Clinical diagnosis and disease monitoring is currently based on imaging techniques, such as ultrasound com-bined with radiography, magnetic resonance, or computed

Abstract Autosomal recessive polycystic kidney disease (ARPKD) is a severe, monogenetically inherited kidney and liver disease. PCK rats carrying the orthologous mutant gene serve as a model of human disease, and alterations in lipid profi les in PCK rats suggest that defi ned subsets of lipids may be useful as molecular disease markers. Whereas MALDI protein imaging mass spectrometry (IMS) has be-come a promising tool for disease classifi cation, widely applicable workfl ows that link MALDI lipid imaging and identifi cation as well as structural characterization of candi-date disease-classifying marker lipids are lacking. Here, we combine selective MALDI imaging of sulfated kidney lipids and Fisher discriminant analysis (FDA) of imaging data sets for identifi cation of candidate markers of progressive dis-ease in PCK rats. Our study highlights strong increases in lower mass lipids as main classifi ers of cystic disease. Struc-ture determination by high-resolution mass spectrometry identifi es these altered lipids as taurine-conjugated bile ac-ids. These sulfated lipids are selectively elevated in the PCK rat model but not in models of related hepatorenal fi -brocystic diseases, suggesting that they be molecular mark-ers of the disease and that a combination of MALDI imaging with high-resolution MS methods and Fisher discriminant data analysis may be applicable for lipid marker discovery. —Ruh, H., T. Salonikios, J. Fuchser, M. Schwartz, C. Sticht, C. Hochheim, B. Wirnitzer, N. Gretz, and C. Hopf. MALDI im-aging MS reveals candidate lipid markers of polycystic kidney disease. J. Lipid Res. 2013. 54: 2785–2794.

Supplementary key words autosomal recessive polycystic kidney dis-ease • imaging mass spectrometry • Fisher discriminant analysis • tau-rocholic acid • Fourier transform ion cyclotron resonance mass spectrometry • Matrix-assisted laser desorption/ionization

This work was supported by the Baden-Württemberg Ministry of Science and Culture (INST 874/2-1 LAGG to C.H.) and by a joined grant (“ZAFH ABI-MAS”) from ZO IV by the Landesstiftung Baden-Württemberg and the Europäis-cher Fonds für regionale Entwicklung (EFRE; to N.G., B.W., and C.H.).

Manuscript received 16 May 2013 and in revised form 28 June 2013.

Published, JLR Papers in Press, July 12, 2013 DOI 10.1194/jlr.M040014

MALDI imaging MS reveals candidate lipid markers of polycystic kidney disease

Hermelindis Ruh , * ,†,§, ** Theresia Salonikios , †,†† Jens Fuchser , §§ Matthias Schwartz , †,†† Carsten Sticht , †,§, ** Christina Hochheim , * ,†,§ Bernhard Wirnitzer , †,†† Norbert Gretz , †,§, ** and Carsten Hopf 1, * ,†,§

Institute of Instrumental Analytics and Bioanalytics,* Center for Applied Research in Biomedical Mass Spectrometry (ABIMAS), † and Institute of Digital Signal Processing, †† Mannheim University of Applied Sciences , 68163 Mannheim, Germany ; Institute of Medical Technology, § University of Heidelberg and Mannheim University of Applied Sciences , 68167 Mannheim, Germany ; Medical Research Center, ** University of Heidelberg, 68167 Mannheim, Germany; and Bruker Daltonik GmbH , §§ D-28359 Bremen, Germany

Abbreviations: 9-AA, 9-aminoacridine; ACN, acetonitrile; ADPKD, autosomal dominant polycystic kidney disease; A/I, peak area/inten-sity average of individual sample; ALAT, alanine aminotransferase; ALP, alkaline phosphatase; ARPKD, autosomal recessive polycystic kid-ney disease; BA, bile acid; ChSulf, cholesterol sulfate; FDA, Fisher dis-criminant analysis; FTICR, Fourier transform ion cyclotron resonance; FTMS , Fourier transform mass spectrometry; GGT, gamma-glutamyl transferase; GLDH, glutamate dehydrogenase; GSL, glycosphingolipid; H&E, hematoxylin and eosin; HRFCD, hepatorenal fi brocystic disease; IMS, imaging mass spectrometry; PI, phosphatidylinositol; SDR, Sprague Dawley rat; S/N, signal to noise; TCA, taurocholic acid.

1 To whom correspondence should be addressed. e-mail: [email protected]

The online version of this article (available at http://www.jlr.org) contains supplementary data in the form of fi ve fi gures and four tables.

by guest, on May 26, 2018

ww

w.jlr.org

Dow

nloaded from

.html http://www.jlr.org/content/suppl/2013/07/12/jlr.M040014.DC1Supplemental Material can be found at:

http://www.jlr.org/

http://www.jlr.org/content/suppl/2013/07/12/jlr.M040014.DC1.html


2786 Journal of Lipid Research Volume 54, 2013

discovery of candidate molecular lipid markers in rat ARPKD kidneys.

Statistical analysis of IMS data sets remains a major chal-lenge in marker discovery utilizing MALDI-IMS ( 28 ). Most available computing processes and algorithms, such as principal component analysis or hierarchical clustering, enable the classifi cation of different tissues, e.g., healthy versus diseased, but fail to pinpoint discriminating m/z values that are needed for identifi cation of the correspond-ing biomolecules. Recently, nonimaging MALDI-TOF MS has been combined with linear discriminant analysis to achieve that goal ( 29 ). In microarray data sets even larger than MALDI images, Fisher discriminant analysis (FDA), a related discriminant analysis method, has successfully been applied for gene selection ( 30 ); it has previously been used for discrimination of magnetic resonance or positron emission tomography images ( 31–33 ). FDA does not make the assumptions of a t -test, such as normally dis-tributed classes, and may therefore be suitable for analysis of MALDI-IMS data sets for which no normal distribution can be assumed.

Here, we present a new workfl ow for identifi cation of candidate lipid biomarkers of ARPKD, comprising MALDI-IMS with a selective matrix and FDA for feature selection in IMS data sets. This approach revealed a small set of top-scoring FDA m/z values of unexpectedly low mass ( m/z 512.3 and 514.3). Detailed investigation of these m/z values in kidney and liver extracts by MALDI-TOF MS indicated strong increases in diseased animals that corre-lated with disease progression. Furthermore, to enable noninvasive detection and diagnosis, we established rela-tive quantifi cation of these marker candidates in urine. Using chemical reasoning and Fourier transform ion cy-clotron resonance (FTICR) MS, taurine-conjugated bile acids were identifi ed as differentiating metabolites that were able to distinguish ARPKD and related rodent HRFCDs.

MATERIALS AND METHODS

Materials Analytical grade solvents [acetonitrile (ACN), acetone, chloro-

form, methanol, acetic acid], were obtained from Merck (Darm-stadt, Germany). DEAE A-25, potassium acetate, potassium chloride, sodium carbonate, potassium hydroxide, and tauro-cholic acid (TCA) sodium salt were purchased from Sigma-Aldrich (Taufkirchen, Germany), and SDS, from Fluka. MALDI matrix 9-AA was obtained from Merck, and acidic semisynthetic (3 ′ -Sulfo)Gal- � -Cer(d18:1,C17:0) lipid standard, from Avanti Polar Lipids (Alabaster, AL).

Animals All experiments were conducted in accordance with the German

Animal Protection Law and were approved by the local authority (Regierungspräsidium Karlsruhe, Germany). The following inbred rat strains were used: PCK as model for ARPKD, with Sprague Dawley rats (SDR) as healthy controls. PKD2 mut and PKD/Mhm rats were used as models for ADPKD. For the latter, heterozygous (cy/+) animals (referred to as PKD) develop renal cysts. Unaffected (+/+) rats of the same strain that do not carry the mutant allele were used as controls. Rats received standard rat diet (containing 19%

tomography imaging ( 1, 2 ). As clinical manifestations in ARPKD vary in renal and hepatic involvement and as mor-phological changes might not be indicative for a single type of HRFCD ( 9, 10 ), the identifi cation of body fl uid-derived markers that could distinguish ARPKD from other HRFCDs is urgently needed.

Recently, glycosphingolipids (GSL) and other sphingo-lipids have been associated with a variety of diseases, in-cluding kidney cancer and some forms of HRFCD ( 11 ). For instance, neutral GSL and the acidic ganglioside GM3 are markedly increased in autosomal dominant polycystic kidney disease (ADPKD) and nephronophthisis with mu-tations in Nphp3 or Nphp9 genes ( 12–14 ). However, little is known about GSL levels in ARPKD: In contrast to eleva-tions of some acidic GSLs in ADPKD models, sulfatide levels, for instance, are decreased in cpk/cpk mice, an ARPKD model ( 15 ). In contrast to these mice, the disease gene in PCK rats is orthologous to the polycystic kidney and hepatic disease 1 gene affected in human ARPKD ( 16 ). Several characteristics of human disease are manifest in homozygous PCK rats ( 17 ), allowing for investigation of the molecular pathogenesis and testing of experimental therapeutics ( 18 ). We therefore set out to establish a workfl ow for identifi cation of candidate lipid makers that link molecular alterations in relevant tissues with analytes that could possibly be measured in body fl uids.

Until now, use of MALDI imaging mass spectrometry (IMS) for identifi cation of candidate biomarkers has pri-marily focused on the examination of protein patterns ( 19–21 ). However, the identifi cation of proteins corre-sponding to disease-classifying m/z values, although occa-sionally accomplished ( 22 ), remains a challenging analytical task, as MALDI-IMS protein signals are typically weak and obtained in low-resolution linear detection mode ( 20 ). Protein IMS biomarker discovery workfl ows therefore re-quire follow-up with complementary proteomics methods ( 22, 23 ) and immunohistochemical validation ( 19, 21 ). On the other hand, mounting experimental evidence sug-gests that GSLs could be useful molecular markers of dis-ease ( 24 ), while at the same time being present in higher concentrations than most proteins and being more readily extracted and analyzed in organic solvents typically used for MALDI-IMS ( 25 ). Due to a mass usually below 1,500 Da, lipids are also amenable to MALDI-IMS in refl ectron mode, resulting in higher resolution and enabling frag-mentation of selected ions for identifi cation by tandem mass spectrometry ( 25 ). Therefore, MALDI-IMS of tissue slices may be a useful tool for unbiased preclinical and clinical lipidomic studies ( 26 ). In contrast to the analysis of lipid extracts that requires time-consuming handling steps and where distinct classes of lipids might be depleted during certain purifi cation steps (for example, alkaline hy-drolysis of phospholipids), MALDI-IMS facilitates explor-atory data analysis of numerous lipid classes within their biological environment. We have recently introduced 9-aminoacridine (9-AA) as an IMS matrix for selective and label-free imaging of sulfated lipids in kidney cryosec-tions ( 27 ). Hence, we sought to apply MALDI-IMS for the


ww

w.jlr.org

Dow

nloaded from


http://www.jlr.org/



Identifi cation of lipid markers by MALDI-imaging MS Q2 2787

2 21 1 2 2

1 2

1 1

n 2

n nn

Eq. 3

and where n 1 and n 2 denote the number of observations of class 1 and class 2.

FTICR mass spectrometry FTICR MS was performed on a solariX 12T Fourier transform

mass spectrometer (FTMS) (Bruker Daltonics) equipped with a dual ESI/MALDI ion source. Spectra were acquired in ESI nega-tive ion mode with 50 scans for each spectrum; 1 MW acquisition size and 2 MW processing size. Measurement was done in broad-band mode for overview and SORI-CID for MS/MS, isolating the respective precursor mono-isotopically, including single shots for clean precursor isolation. Liver extract was dissolved in 20 vol and urine extract in 1 vol ACN/H 2 O (1/1, v/v) and compared with 1 pmol of TCA in ACN/H 2 O (1/1, v/v). Mass spectra were calibrated in DataAnalysis, where the spectrum of TCA was inter-nally calibrated (single point correction for [M-H] � ). This cali-bration was used to externally calibrate the urine- and liver-extract spectra.

Preparation of GSL extracts Kidneys and livers were weighed and homogenized for 2 min

on ice in 10 vol of sodium carbonate buffer (200 mM, pH 9.3) using an Polytron PT 1200E (Kinematica, Luzern, Switzerland). Urine was diluted in 2 vol of sodium carbonate buffer. The homo-genate and diluted urine was freeze-dried at � 49°C and 0.2 mbar, followed by lipid extraction with 20 vol chloroform/methanol/water (twice with 10:10:1, v/v/v, once with 30:60:8). The remain-ing pellet was dried over night at 70°C and weighted for normal-ization of the GSL extract to the kidney dry weight. Pooled supernatants were dried with a rotary evaporator at 45°C and re-dissolved in 5 vol of 0.1 M potassium hydroxide in methanol for mild alkaline hydrolysis of phospholipids (37°C). The suspen-sion was sonicated every 20 min for 1 min. Six µl acetic acid per ml of 0.1 M potassium hydroxide was added after 2 h to neutral-ize the samples. The extract was dried again and resuspended in 10 vol of aqueous 0.1 M potassium chloride with sonication. For solid phase extraction, 500 mg Bond Elute C18 cartridges (Agilent Technologies) per 500 mg kidney/liver wet weight and up to 5 ml urine were used. After activating the column stepwise with (a) chloroform, (b) chloroform/methanol (1:1, v/v), (c) methanol, (d) methanol/water (1:1, v/v), and (e) 0.1 M potassium chlo-ride, the extract was loaded. The sample tube was rinsed with 0.1 M potassium chloride and applied onto the column. The fl ow through was loaded again, before desalting the column with 10 column volumes of distilled water. The lipid extract was eluted with four column volumes of methanol and four column volumes of chloroform/methanol (1:1, v/v). Pooled eluates were loaded on DEAE Sephadex A-25 preconditioned with 1M methanolic potassium acetate. The column was rinsed with 9 ml methanol containing neutral GSLs. Acidic GSL were eluted with 6 ml of 1M methanolic potassium acetate followed by evaporation of the organic solvent. The dried acidic lipid fraction was redissolved in 6 ml distilled water and was applied again on a 500 mg Bond Elut C18 cartridge for desalting as described above. Finally, the eluate was dried under air stream (45°C), and stored at � 20°C.

MALDI-TOF mass spectrometry on ground steel target MALDI-TOF MS was performed on an Autofl ex III MALDI-TOF/

TOF instrument (Bruker Daltonics) equipped with a smartbeam laser (200 Hz) and controlled by fl exControl 3.0 software (Bruker Daltonics). The instrument was set to an acceleration voltage of

protein) and tap water ad libitum. Male rats were anesthetized with isofl urane prior to sacrifi cing by neck fracture at 0 (2–3 days), 4, or 10 weeks of age. Kidneys and livers were snap-frozen in liquid nitrogen and stored at � 80°C until lipid and RNA extraction. For cryosectioning, organs were slowly frozen on ice-cold isopentane, wrapped in aluminum foil, further frozen in liquid nitrogen and stored at –80°C. For urine collection, 10-week-old male rats were kept in metabolic cages for 24 h and pcy mice (16 weeks of age ) for 16 h.

MALDI IMS MALDI imaging was performed as described previously ( 27 ).

In brief, frozen rat kidneys were sliced into 10 � m sections using a Leica CM 1900 cryostat (Leica Biosystems, Nussloch, Germany) at � 20°C and thaw-mounted onto indium tin oxide-coated con-ductive glass slides (Bruker Daltonics, Bremen, Germany). Kidney tissue sections of 0- and 4-week-old PCKs and SDRs were measured on the same indium tin oxide glass slide. Likewise, 10-week-old PCKs and SDRs were placed and analyzed together. A following section was transferred onto a normal glass slide for hematoxylin and eosin (H&E) staining. 9-AA [20 mM in ACN/water (80:20, v/v)] was deposited on cryosections using a matrix sprayer (SunChrom, Friedrichsdorf, Germany), where air pressure was set to 2.5 bar. Matrix was deposited in seven layers, where fl ow rate for the fi rst layer was set to 10 µl/min; second layer, 15 µl/min; third, 20 µl/min, and fourth to seventh, 25 µl/min. Mass spectrometric measurements were performed using Autofl ex III MALDI-TOF/TOF instruments (Bruker Daltonics) in nega-tive ion-refl ector mode with baseline subtraction in the range from 450 to 1000 Da. Images were acquired at a spatial resolution of 100 � m with 200 laser shots per position by fl exImaging 2.1 software (Bruker Daltonics).

MALDI-IMS data classifi cation by Fisher linear discriminant analysis

MALDI imaging raw spectra were converted to a MATLAB (Version 7.11.0584, R2010b) readable format. The conversion software is available from the authors. Un-binned m/z -ratios were averaged over class 1 (SDR) and class 2 (PCK). For noise suppres-sion and data reduction, all m/z with intensity threshold < 1 [a.u.] in both averaged spectra were determined and subsequently excluded from the entire set of spectra. The resulting fi ltered spectra sets of class 1 and class 2 were normalized to standard deviation = 1 and mean = 0 (z-score normalization). Subsequently, for each remaining m/z i , the Fisher discriminant value ( FD i ) was determined:

2

1 22 21 2

i ii

i i

FD

Eq. 1

where � i1 and � i2 denote the mean intensities of m/z i of class 1 and class 2, and � i1

2 and � i2 2 denote the corresponding variances

(supplementary Fig. I). Resulting FD i were ranked in descending order, where m/z values with the most discriminating FD i were chosen for further evaluation. For comparison, a two-sample t -test with pooled variance was performed on the same MALDI-imaging data sets (supplementary Fig. II). The null hypothesis was that the data are independent random samples from normal distributions with equal means, against the alternative that the means are not equal. The formula is

1 2 1 2

1 2

t i i n nn n

Eq. 2

where


ww

w.jlr.org

Dow

nloaded from


http://www.jlr.org/




RESULTS AND DISCUSSION

MALDI IMS and FDA of imaging data sets reveal changes in acidic lipids in the PCK rat model of ARPKD

Pursuing our goal of using lipid MALDI-IMS for the dis-covery of candidate markers in a rat model of ARPKD, we chose 9-AA as matrix, which preferentially facilitates ion-ization of sulfated lipids ( 27, 35 ). Zero, 4, and 10-week-old cystic PCK kidneys were compared with SDR kidneys as healthy controls ( Fig. 1A and supplementary Fig. III). Ad-jacent sections were processed for MALDI-IMS analysis. Signal intensity and distribution of rodent kidney house-keeping lipids, such as phosphatidyl inositol PI(18:0;20:4) ( m/z 885.6) did not differ between healthy and cystic kid-neys ( Fig. 1B ), confi rming that matrix coverage of the tis-sue slices was homogeneous and that MALDI-IMS data acquisition was robust. The most prominent sulfatide in

19 kV; spectra were acquired in refl ector negative ion mode. Acidic kidney extracts were dissolved in 25 vol ACN/H 2 O (1:1, v/v) of dry weight, supplemented with fi nal concentrations of 0.25 µM (3 ′ -Sulfo)Gal- � -Cer(d18:1,C17:0) and 0.56 µM SDS; liver extracts were dissolved in 60 vol ACN/H 2 O (1:1, v/v) containing 4.1 µM (3 ′ -Sulfo)Gal- � -Cer(d18:1,C17:0) and 115 µM SDS; and urine extracts were dissolved in 2 vol ACN/H2O (1:1, v/v) sup-plemented with 0.31 µM (3 ′ -Sulfo)Gal- � -Cer(d18:1,C17:0) and 0.88 µM SDS for relative quantifi cation. This solutions were mixed 1:1 (v/v) with 4 mg/ml 9-AA in ACN/H 2 O (80:20, v/v) as matrix. An AutoXecute method was defi ned to sum up 4,000 la-ser shots at 20 different positions of a spot. Only spectra with minimum half width of less than 0.2 Da were accumulated, and acquisition was quit after 50 consecutive failed judgments. For relative quantitation of defi ned lipids, nine spots were prepared for each sample on a ground steel target. MALDI mass spectra were calibrated using the above-mentioned internal standards. Baseline subtraction, peak signal to noise (S/N) ratios, intensity, resolution, and area were determined with fl exAnalysis 3.0 soft-ware (Bruker Daltonics). Mass spectrometric data were evaluated using ClinProTools 2.2 software (Bruker Daltonics), where peak picking was done with S/N threshold of >5 with a maximal peak number of 80 per single spectrum. Minimal occurrence for peak aggregation was set to 20% with aggregation width of 500 ppm. Peak calculation was based on intensities.

Biochemical analysis of blood and urine Blood for determination of kidney and liver function pa-

rameters was taken retrobulbary under isofl urane anesthesia in Li-Heparin-containing microfuge tubes. Blood was centrifuged at 3,000 g for 15 min at 4°C, and plasma was stored at � 20°C until further processing. Biochemical parameters, including choles-terol, triglycerides, total bilirubin, urea, creatinine, Ca 2+ , Na + , K + , PO 4

3- , glucose, total protein, and enzyme activity of glutamate de-hydrogenase (GLDH), � -glutamyl transferase (GGT), alanine aminotransferase (ALAT), and alkaline phosphatase (ALP), were determined by standard laboratory methods (cobas c311 ana-lyzer, Roche Diagnostics, Mannheim, Germany).

Gene expression profi ling Total RNA was isolated from frozen liver tissue of 10-week-old

rats with Trizol reagent (Invitrogen) as described by the manufac-turer and subjected to genome-scale gene expression profi ling us-ing GeneChip® Rat Gene 1.0 ST Array and GeneChip® Scanner 3000 7G (Affymetrix). The raw fl uorescence intensity values were imported with a Custom CDF Version 15 with Entrez-based gene defi nition. The values were normalized applying quantile nor-malization. Differential gene expression was analyzed with one-way ANOVA, using a commercial software package (SAS JMP7 Genomics, version 4; SAS Institute, Cary, NC). A false-positive rate of a = 0.05 with false discovery rate correction was taken as the level of signifi cance. Gene set enrichment analysis was used to determine whether defi ned sets of genes exhibit a statistically signifi cant bias in their distribution within a ranked gene list ( 34 ). Pathways belonging to various cell functions, such as cell cycle or apoptosis, were obtained from public external databases (KEGG, http://www.genome.jp/kegg/).

Statistical analysis Averaged peak area/intensity (A/I) ratios of the individual

samples were calculated using ClinProTools software. Normaliza-tion to either (3 ′ -Sulfo)Gal- � -Cer(d18:1,C17:0) or SDS standard, calculation of mean, standard deviation for diseased and control samples, and two-tailed t -tests were performed with Excel. ANOVA was calculated by GraphPad Prism 5.04 software. The level of sig-nifi cance was set to P < 0.05 unless stated otherwise.

Fig. 1. MALDI IMS reveals only subtle differences in renal sul-fatide levels in the PCK rats. Imaging experiments were performed with kidney tissue sections of 0- and 4-week-old PCK and SDR placed on the same glass slide, and 10-week-old PCK and SDR measured together on a second glass slide. (A) Micrographs of H&E-stained adjacent tissue sections; anatomical structures are visible, such as papillae, medulla, and cortex, as well as cystic regions affecting pre-dominantly the PCK rat medulla. (B) Uniform ion intensity distribu-tion of the membrane component PI(18:0;20:4) with m/z 885.6 in healthy and cystic kidneys. (C) Prominent (3 ′ -Sulfo)Gal- � -Cer (d18:1;h24:0) with m/z 906.7 (red) is primarily localized to the inner medulla and to a lesser extent to the outer medulla, whereas (3 ′ -Sulfo)Gal- � -Cer(d18:1,26:0) with m/z 918.6 (green) is only found in the papilla. (D) Ion intensity distribution of (3 ′ -Sulfo)Gal- � -Cer(t18:0,h24:0) with m/z 924. 6 (red) in diseased and control kid-neys. For (3 ′ -Sulfo)Gal- � -Cer isoforms, a tendency toward lower ion intensities in PCK compared with SDR kidneys is observed.


ww

w.jlr.org

Dow

nloaded from


http://www.jlr.org/




of m/z 514.3 and 512.3 was similar, with the exception of low intensity m/z 514.3 signals that were already detected in neonate kidneys ( Fig. 3C ). This ion was found in both healthy and cystic kidneys, perhaps refl ecting differences in metabolism between newborn rat organs and later devel-opmental stages or an unresolved unrelated ion (see below). In contrast, no differences in signal intensity and no local-ization to fl uid-fi lled cysts were observed for m/z 465.3, the mass of which is consistent with cholesterol sulfate (ChSulf) ( Fig. 3D ).

To validate the fi ndings obtained by MALDI-IMS and sta-tistical analysis and to take biological variability into account, we extracted lipids from six PCK and six SDR kidneys (both 10 weeks of age) and subjected them to MALDI-TOF MS analysis with 9-AA matrix in a dried droplet prepara-tion on a ground steel target. A/I ratios for individual m/z were calculated by ClinProTools software, and A/I values of peaks corresponding to high FD values were normalized to SDS. Interestingly, A/I values of m/z 512.3 and 514.3 in kidney extracts confi rmed the dramatic increase of these biomolecules in PCK kidneys. Compared with unaffected SDR normalized A/I ratios for m/z 512.3 were � 58-fold and for m/z 514.3, � 46-fold increased, whereas m/z 465.3 (ChSulf) remained unchanged ( Fig. 4A ). Encouraged by these unexpected findings, we wondered whether m/z 512.3 and m/z 514.3 could be detected in urine of PCK rats as well. Aiming for an easily accessible marker for ARPKD, we established extraction and relative quantifi cation of TCA in urine by MALDI-TOF MS using acidic extracts sup-plemented with SDS as standard. Surprisingly, A/I ratios of the two molecular species were � 12-fold ( m/z 512.3) and � 21-fold ( m/z 514.3) higher in PCK compared with SDR urine ( Fig. 4B ). As hepatic involvement is a common

rat kidney, (3 ′ -Sulfo)Gal- � -Cer(d18:1;h24:0) ( m/z 906.7), comprising a singly unsaturated sphingosine base (d18:1) and a hydroxylated fatty acid (h24:0), was mainly localized to the inner and to a lesser extent to the outer medulla in 4- and 10-week-old animals ( Fig. 1C ). For more informa-tion on nomenclature of sulfatides, see Ref. 27 . (3 ′ -Sulfo)Gal- � -Cer(d18:1,26:0) ( m/z 918.6 ) served as a reference point in images ( Fig. 1C, D ). In contrast to sphingosine-based (3 ′ -Sulfo)Gal- � -Cer(d18:1;h24:0), the phytosphin-goid-based (3 ′ -Sulfo)Gal- � -Cer(t18:0;h24:0) ( m/z 924.6) with identical sugar head group and fatty acid was found in the papilla as well as in the inner medulla of 4- and 10-week-old PCK and SDR kidneys ( Fig. 1D ), confi rming our earlier fi ndings of distinct localizations of (3 ′ -Sulfo)Gal- � -Cer iso-forms containing phytosphingoid bases ( 27 ). Interestingly, intensities of all detected sulfatides in healthy and diseased kidneys were very low in neonate organs, suggesting that lipid metabolism might be different. In contrast to expec-tations based on studies in cpk/cpk mice ( 15 ), however, sul-fatides appeared to be at best marginally reduced in PCK rats ( Fig. 1C, D ). For verifi cation of whether sulfatides were reduced in the ARPKD model and possibly for iden-tifi cation of other acidic lipids or unknown, low-mass me-tabolites that might serve as marker for ARPKD, we sought to develop a statistical method for the analysis of complex MALDI-IMS data sets and for selection of discriminant m/z values. Linear FDA generally compares two or more classes of data, for which (in contrast to t -tests or ANOVA analy-sis) no normal distribution is assumed. Here, all spectra recorded from the PCK kidney (PCK class) and those re-corded from the SDR kidney (SDR class) were compared and underwent discriminant analysis. A high FD value was obtained for those m/z values that are characterized by a large difference between the mean ion intensities of the PCK and SDR classes and low variances within each class. Two independent MALDI-IMS data sets, biological repli-cates of different 10-week-old PCK and SDR kidneys, were analyzed. Reproducibly, a small number of high FD values were found in the m/z range between 512.2 and 515.6 ( Fig. 2 and supplementary Table I). Data analysis by Student t -test would be a less preferred data analysis option: On the one hand, it has the fundamental malus of assuming a normal distribution of data that may not be given for MALDI-IMS data sets. Nevertheless, for comparison, a Student t -test was performed on the same MALDI-IMS data (supple-mentary Fig. II). Indeed, the highest t -values were obtained for m/z 512.3 and 514.3. However, the plot of m/z versus t -value was noisier than that against FD values ( Fig. 2 ver-sus supplementary Fig. II). We concluded that m/z 512.3 and 514.3 were the highest scoring candidates with both data analysis methods and therefore best suited for experi-mental follow-up.

Signal intensity in MALDI MS images of both m/z 512.3 and 514.3 correlated with cyst progression in PCK kidneys ( Fig. 3 ). Whereas m/z 512.3 was detected in neither SDR kidneys nor neonate PCK rat kidneys, it was localized to papilla and medulla of 4-week-old PCK kidneys, and it was detected in virtually everywhere in 10-week-old organs, in-cluding most of the fl uid-fi lled cysts ( Fig. 3B ). Localization

Fig. 2. FDA identifi es m/z values that distinguish between cystic PCK rat and healthy Sprague-Dawley rat kidneys. (A, B) FDA of MALDI-IMS data sets obtained from two biological replicates, in which kidney sections from two different 10-week-old PCK and SDR were compared. Top scoring disease-discriminating m/z values were obtained in the mass range 512–516 Da.


ww

w.jlr.org

Dow

nloaded from


http://www.jlr.org/




High resolution FTMS identifi ed m/z 512.3 and 514.3 as TCAs

We next sought to identify the molecular species corre-sponding to m/z 512.3 and 514.3. Database search (http://www.lipidmaps.org/ and http://www.genome.jp/kegg/pathway.html) revealed several options for lipids possessing a mass of 514.3 ± 0.2 Da ([M-H + ] � ) ( Fig. 5A ). To discrimi-nate between these options, we performed high-resolution FTICR MS of a urine sample and obtained an accurate mass of 514.28407 amu for the deprotonated molecular ion in urine and the likely elemental sum formula C 26 H 44 NO 7 S, both of which were indicative of TCA ( Fig. 5B ). In con-trast, monoacylglycero-phosphocholine (18:4/0:0) and 2-aminoethylphosphocholate, two molecules consistent with a mass of 514.3 and corresponding to the sum for-mula C 26 H 46 NO 7 P, would have been detected with theo-retical m/z 514.29391 in negative ion mode ( Fig. 5A, B ). The fine structure of m/z 514.28407 in the highly re-solved spectrum unambiguously pointed to the presence of the element sulfur and thus again to the presence of TCA ( Fig. 5C ). In contrast to phosphorus, sulfur has several stable isotopes, the most prominent being 34 S that is indi-cated as m/z 516.280193 in the simulation of the sulfur fi ne structure in C 26 H 44 NO 7 S ( Fig. 5C ). MS/MS experiments

feature in ARPKD, we addressed the question whether m/z 512.3 and 514.3 could be found in cystic livers, too. To this end, we prepared liver acidic lipid extracts and analyzed them by MALDI-TOF MS. We noticed that concentrations of m/z 512.3 and 514.3 in liver were about 200 times higher than in kidney extracts (data not shown), suggesting that the origin of these molecular species may be liver. A/I ratios of m/z 512.3 were already signifi cantly elevated in neonate PCK rats, with a striking increase in 4-week-old livers and a moderate increase thereafter. Much lower but signifi cant levels of m/z 512.3 were observed in SDR livers ( Fig. 4C ). However, the time course of m/z 514.3 in liver was different from that of m/z 512.3: neonate levels of m/z 514.3 in both PCK and SDR livers were higher than in 4-week-old livers (compare Fig. 4D with Fig. 3B, C ). Again, we wondered whether another, unresolved molecular spe-cies was interfering.

Fig. 3. MALDI IMS of m/z with top-scoring FDA values illustrate dramatic increases of m/z 512.3 and 514.3 in PCK kidneys during development. (A) Micrographs of H&E-stained adjacent tissue sec-tions. (B, C) Localization of m/z 512.3 and m/z 514.3, molecular species with highest FDA values. While m/z 512.3 is neither de-tected in PCK nor SDR new born rat kidneys, low intensity signals of m/z 514.3 are localized in cystic as well as healthy 0-week-old (0 w) kidneys. Both candidate markers are clearly visible in the in-ner part of 4-week-old (4 w) cystic kidneys. In 10-week-old (10 w) PCK kidneys, ion intensities of m/z 512.3 and 514.3 are dramatically increased, predominantly in cysts and cortex. (D) Ion distribution of m/z 465.3, consistent with ChSulf, is evenly distributed in 4- and 10 week-old diseased and control kidneys. Images are representa-tive of two independent experiments.

Fig. 4. MALDI-TOF MS analysis reveals time-dependent increase of m/z with top-scoring FDA values in extracts of PCK kidney, urine, and liver. (A) Kidney A/I ratios of m/z 512.3 and 514.3 are strongly enhanced in 10-week-old (10 w) cystic PCK kidneys, whereas ChSulf ( m/z 465.3) remains unchanged. Means ± SD were calculated for n = 6 each of male PCKs and SDRs for four technical replicates. (B) Urine A/I ratios of marker candidates are � 12-fold ( m/z 512.3) and � 21-fold ( m/z 514.3) higher in PCK compared with SDR urine. Means were calculated for PCK n = 3 and SDR n = 3. Results are means of three technical replicates ± SD. (C, D) Liver A/I ratios of m/z 512.3 and m/z 514.3 are signifi cantly increased in 0-, 4-, and 10-week-old PCK rats. Means were determined for 0-week-old PCK (n = 5) and SDR (n = 6), for 4-week-old PCK (n = 3) and SDR (n = 3), and for 10-week-old PCK (n = 6) and SDR (n = 6) male rats. Means ± SD were calculated for four technical replicates. * P < 0.01, ** P < 0.001, and *** P < 0.0001 by two-tailed t -test.


ww

w.jlr.org

Dow

nloaded from


http://www.jlr.org/




atoms indicates an additional double bond in otherwise related molecules, which was confi rmed by FTMS SORI-CID fragmentation of m/z 512.3 (data not shown). The position of the proposed double bond in m/z 512.3 cannot be inferred from these studies. However, it may be at C-5 ( 36 ). For validation of the unusually high m/z 514.3 levels in neonate PCK and SDR rat livers (i.e., to rule out an in-terfering, unresolved analyte), we compared fragments of this precursor ion in 0-week-old livers with fragments of the same m/z in urine of 10-week-old PCK rats by MALDI-TOF/TOF MS (supplementary Fig. IV-A, B). In all cases, we de-tected the same fragments, suggesting that m/z 514.3 in

were performed to further prove this fi nding. FTICR SORI CID after mono-isotopic isolation directly yielded frag-ments of urine and liver m/z 514.3 [ m/z 79.95739 (SO 3 ), m/z 106.98087 (C 2 H 3 O 3 S), and m/z 124.00742 (NC 2 H 5 O 3 S)], which were indicative of a taurine-conjugated bile acid (BA) ( Fig. 5D, E ). To validate our chemical reasoning, we obtained pure TCA substance and demonstrated that its parent mass, derived sum formula, and fragmentation pat-tern were identical with the metabolite in urine and liver referred to as m/z 514.3 ( Fig. 5F ). Interestingly, m/z 512.26845 was measured with exactly 2.01562 Da less than 514.28407 ( Fig. 5B ). The molecular weight of two hydrogen

Fig. 5. High-resolution FTMS identifi es candidate molecular markers as taurine-conjugated bile acids. (A) Structures, sum formulae, and theoretical m/z values of candidate molecules suggested by a lipid database search (515.3 ± 0.5 Da): Monoacylglycero-phosphocholine (18:4/0:0), 2-aminoethylphosphocholate, and TCA (the latter with possible MS/MS fragments of the taurine conjugate). (B) Zoom-in of the high resolu-tion FTMS spectrum of PCK urine extract in the range m/z 511.9–514.7 obtained in ESI negative ion mode. The elemental sum formula for the accurate mass of m/z 514.28407 was calculated to be C 26 H 44 NO 7 S, which is consistent with the sum formula of TCA. A mass difference of 2.01562 between m/z 512.26845 and m/z 514.28407 corresponds to two hydrogen atoms. (C) Zoom-in of the M+2 isotope peak of m/z 514.28407 and (D) simulation of the characteristic 34 S M+2 peak for C 26 H 44 NO 7 S. The presence of the element sulfur (arrow) in the spectrum instead of phosphorus confi rms the identity of TCA in PCK urine extracts. SORI-CID was performed for MS/MS comparison of m/z 514.28 from (D) urine and (E) liver extracts. Parent molecular ions and characteristic fragmentation patterns with m/z 79.95739 (SO 3 ), m/z 106.98087 (C 2 H 3 O 3 S), and m/z 124.00742 (NC 2 H 5 O 3 S) are consistent with the (F) parent mass and fragmentation pattern obtained for pure reference TCA.


ww

w.jlr.org

Dow

nloaded from


http://www.jlr.org/




study, signifi cant hyperlipidemia with an increase in both serum cholesterol and triglycerides had been reported for PCK rats of about 18 weeks of age ( 17 ). Owing to their detergent properties, excessively accumulated liver bile ac-ids are known to be cytotoxic and to cause hepatocyte damage ( 38, 39 ). Therefore, we assessed liver function by plasma clinical chemistry analysis of total bilirubin and enzyme activity of GLDH, GGT, and ALAT. In accordance with Ref. 17 , an increase in bilirubin was observed ( Fig. 6C ). Both GGT and GLDH activity were dramatically increased ( Fig. 6D, E ), suggesting acute liver damage by accumu-lated bile acids, whereas ALAT, another indicator for he-patocellular injury, in particular for viral infections such as hepatitis or mononucleosis, was reduced ( Fig. 6F ). In con-trast to recent disease stage characterization in the PCK rat ( 17 ), ALP, a marker for bile duct obstruction, was un-changed, together with unaffected levels of glucose (sup-plementary Table II). Renal function was not deteriorated in 10-week-old PCK rats, as total protein, urea, sodium, and creatinine levels in plasma were equal to the exam-ined SDR control group (supplementary Table II).

In an attempt to elucidate some of the underlying mech-anisms leading to increases in cholesterol and taurine-con-jugated bile acids in PCK rats, we performed genome-scale gene expression profi ling of PCK and SDR livers (10 weeks of age) using the GeneChip® Rat Gene 1.0 ST. Several pathways were found to be upregulated in cystic livers (supplementary Table III). The functional signifi cance of most upregulated pathways for pathogenesis of cystic dis-ease in PCK rats remains to be established, but enhanced expression of the steroid biosynthesis pathway (RNO00100) may be the cause of the observed increased levels of cho-lesterol ( Fig. 6A ). Surprisingly, the primary bile acid biosynthesis (RNO00120) and cysteine and methionine

neonate PCK (supplementary Fig. IV-B) and SDR livers (data not shown) was the same molecule as that in urine of adult PCK rats (supplementary Fig. IV-A). It is tempting to specu-late that the observed biphasic behavior of m/z 514.3 in liver might be explained by changes in homeostasis of choles-terol-oxidoreductase enzyme activities during rat develop-ment ( 36 ).

Synthesized in hepatocytes from cholesterol, individual BAs and their metabolites play an important role in a num-ber of physiological functions, such as lipid absorption and signaling ( 37 ). Nevertheless, accumulated BAs are cy-totoxic and may cause severe hepatocyte damage ( 38, 39 ). Conjugation of BAs with glycine or taurine increases their water solubility and therefore supports elimination and excretion of cholesterol and BAs into feces or urine ( 37 ). It is now recognized that cilia play an important role in the regulation of bile fl ow ( 40 ). The pathogenesis of several hepatorenal fi brocystic diseases is linked to ciliary sensory dysfunction ( 1 ), as all defective proteins identifi ed so far are localized to these organelles. The 447 kDa membrane-bound ARPKD-gene product of polycystic kidney and hepatic disease 1 , fi brocystin/polyductin localizes at the basal bod-ies to primary cilias of cholangiocytes ( 41, 42 ). In complex with other functional ciliary proteins, fi brocystin/polyductin is thought to regulate intracellular calcium responses ( 43, 44 ). Decreased intracellular calcium levels and increased cAMP concentrations in rodent models of ARPKD were found to be associated with increased proliferation ( 8, 43, 45 ) and secretion of bile acids ( 40 ).

As conjugated BAs are catabolic end products of choles-terol ( 38, 39 ), we further evaluated cholesterol concentra-tions in plasma, revealing 2-fold increased levels ( Fig. 6A ), whereas concentrations of triglyceride were decreased ( Fig. 6B ) in 10-week-old diseased animals. In a previous

Fig. 6. Cholesterol and liver function parameters are changed in PCK rats. (A) Cholesterol is increased � 2-fold in plasma of 10-week-old PCK, whereas (B) triglyceride concentration is decreased. (C) Total biliru-bin levels as well as enzyme activity of (D) GGT and (E) GLDH were found to be markedly enhanced in plasma of cystic animals. In contrast, (F) ALAT activity decreased. Means ± SD were calculated for diseased PCK (n = 3) and control SDR (n = 3) male, 10-week-old rats. * P < 0.05 by two-tailed t -test.


ww

w.jlr.org

Dow

nloaded from


http://www.jlr.org/




with the unaffected rat strains SDR and PKD (+/+) by kidney lipid extraction and MALDI-TOF MS. Interestingly, A/I ra-tios for TCAs in ADPKD kidney extracts were as low as in healthy kidneys. Compared with all the tested rat strains, A/I ratios of TCAs were 34- to 58-fold ( m/z 512.3) and 24- to 46-fold ( m/z 514.3) increased in PCK kidney extract ( Fig. 7A ). In addition, urine of the pcy mouse model of nephronophthisis caused by a mutation in the gene orthologous to human Nphp3 was compared with PCK urine. Here, A/I ratios were � 11-fold ( m/z 512.3) and � 17-fold ( m/z 514.3) higher in PCK compared with pcy urine ( Fig. 7B ).

Metabolomic profi ling is now increasingly used for bio-marker discovery in several pathologies, including hepato-biliary diseases ( 46 ). For example, metabolites involved in bile acid biosynthesis, such as glycochenodeoxycholic acid 3-sulfate, glycocholic acid, glycodeoxycholic acid, TCA, and taurochenodeoxycholate, are downregulated in hu-man hepatocellular carcinoma ( 47 ), whereas urinary bile acid excretion of sulfated chenodeoxycholic acid is mark-edly increased in primary biliary cirrhosis and cholestatic liver disease ( 39, 48 ).

SUMMARY

Taken together, our results suggest that MALDI-IMS in combination with FDA can be useful in the discovery of disease-differentiating metabolites. Aiming for an easily ac-cessible marker for ARPKD, we established relative quantifi -cation of the marker candidates in urine by MALDI-TOF MS. Among other possible biomolecules, TCA was precisely identifi ed due to the pinpoint mass accuracy of FTICR MS. Finally, strong increases of TCA in kidney and urine of dis-eased animals were able to distinguish ARPKD and related rodent HRFCDs.

The authors thank Victoria Skude, Elisabeth Wühl, Isabell Moskal, Cathleen Fichtner, Peter Heinz, and Petra Prochazka for their assistance with the rats and mice. Clinical plasma chemistry is gratefully acknowledged to Petra Schwartz and Elisabeth Seelinger.

REFERENCES

1 . Gunay-Aygun , M. 2009 . Liver and kidney disease in ciliopathies . Am. J. Med. Genet. C. Semin. Med. Genet. 151C : 296 – 306 .

2 . Harris , P. C. , and V. E. Torres . 2009 . Polycystic kidney disease . Annu. Rev. Med. 60 : 321 – 337 .

3 . Wen , J. 2011 . Congenital hepatic fi brosis in autosomal recessive polycystic kidney disease . Clin. Transl. Sci. 4 : 460 – 465 .

4 . Ward , C. J. , M. C. Hogan , S. Rossetti , D. Walker , T. Sneddon , X. Wang , V. Kubly , J. M. Cunningham , R. Bacallao , M. Ishibashi , et al . 2002 . The gene mutated in autosomal recessive polycystic kid-ney disease encodes a large, receptor-like protein . Nat. Genet. 30 : 259 – 269 .

5 . Bergmann , C. , J. Senderek , F. Kupper , F. Schneider , C. Dornia , E. Windelen , T. Eggermann , S. Rudnik-Schoneborn , J. Kirfel , L. Furu , et al . 2004 . PKHD1 mutations in autosomal recessive polycystic kid-ney disease (ARPKD) . Hum. Mutat. 23 : 453 – 463 .

6 . Chapal , M. , A. Debout , A. Dufay , R. Salomon , G. Roussey , S. Burtey , E. A. Launay , C. Vigneau , G. Blancho , C. Loirat , et al . 2012 . Kidney

metabolism (RNO00270) pathways, both of which are in-volved in biosynthesis of TCA, were found to be downregu-lated in diseased livers (supplementary Table IV). Whereas this fi nding may suggest a compensatory mechanism in liver gene expression aiming at reducing levels of toxic BA, it does not explain the presence of very high TCA levels in the fi rst place. Possible functional explanations for TCA accumulation include but are not limited to increased bio-synthesis rate by regulation on the translational level, faster enzyme kinetics, or strongly reduced secretion or excretion. Whereas each of these possible mechanisms warrants experimental follow-up, expression levels of ste-roid and bile acid biosynthesis genes alone comprise a transcriptional signature that distinguishes between dis-eased and healthy livers, as judged by hierarchical cluster-ing analysis (supplementary Fig. V).

TCA is specifi c for the PCK rat model of ARPKD and is not detected in models of related hepatorenal fi brocystic diseases

Given that TCA measured in kidney and urine extract was able to discriminate ARPKD rats from healthy animals, we wondered whether it could further distinguish ARPKD from related HRFCDs, such as ADPKD and nephronophthisis. Two ADPKD models were examined. PKD2 mut transgenic rats and PKD (cy/+) were compared with PCK rats together

Fig. 7. TCA is specifi cally increased in the PCK rat model of AR-PKD but not in models of related HRFCDs. (A) Normal levels of TCA in kidney extracts of ADPKD rat models PKD2 mut and PKD. A/I ratios were 34- to 58-fold ( m/z 512.3) and 24- to 46-fold ( m/z 514.3) increased in PCK compared with ADPKD and healthy con-trol rat strains. MALDI-TOF MS analysis was performed with SDR (n = 6), PKD2 mut (n = 3), unaffected PKD(+/+) (n = 5), cystic PKD(cy/+) (n = 6), and PCK (n = 6), all 10-week-old males. Values represent means ± SD of three independent experiments. **** P < 0.0001 as level of signifi cance by a one-way ANOVA. (B) Normal lev-els of TCA in urine of the nephronophthisis pcy mouse model. A/I ratios were � 11-fold ( m/z 512.3) and � 17-fold ( m/z 514.3) higher in PCK compared with pcy urine. Means were calculated for PCK (n = 3) and pcy (n = 3). Values are expressed as means ± SD of three technical replicates. * P < 0.01, ** P < 0.001 by two-tailed t -test.


ww

w.jlr.org

Dow

nloaded from


http://www.jlr.org/




28 . Alexandrov , T. 2012 . MALDI imaging mass spectrometry: statis-tical data analysis and current computational challenges . BMC Bioinformatics . 13 ( Suppl. 16 ): S11 .

29 . Hettick , J. M. , M. L. Kashon , J. E. Slaven , Y. Ma , J. P. Simpson , P. D. Siegel , G. N. Mazurek , and D. N. Weissman . 2006 . Discrimination of intact mycobacteria at the strain level: a combined MALDI-TOF MS and biostatistical analysis . Proteomics . 6 : 6416 – 6425 .

30 . Fung , E. S. , and M. K. Ng . 2007 . On sparse Fisher discriminant method for microarray data analysis . Bioinformation . 2 : 230 – 234 .

31 . Merhof , D. , P. J. Markiewicz , G. Platsch , J. Declerck , M. Weih , J. Kornhuber , T. Kuwert , J. C. Matthews , and K. Herholz . 2011 . Optimized data preprocessing for multivariate analysis applied to 99mTc-ECD SPECT data sets of Alzheimer’s patients and asymp-tomatic controls . J. Cereb. Blood Flow Metab. 31 : 371 – 383 .

32 . Markiewicz , P. J. , J. C. Matthews , J. Declerck , and K. Herholz . 2011 . Verifi cation of predicted robustness and accuracy of multivariate analysis . Neuroimage . 56 : 1382 – 1385 .

33 . Mayerhoefer , M. E. , W. Schima , S. Trattnig , K. Pinker , V. Berger-Kulemann , and A. Ba-Ssalamah . 2010 . Texture-based classifi cation of focal liver lesions on MRI at 3.0 Tesla: a feasibility study in cysts and hemangiomas . J. Magn. Reson. Imaging . 32 : 352 – 359 .

34 . Subramanian , A. , P. Tamayo , V. K. Mootha , S. Mukherjee , B. L. Ebert , M. A. Gillette , A. Paulovich , S. L. Pomeroy , T. R. Golub , E. S. Lander , et al . 2005 . Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profi les . Proc. Natl. Acad. Sci. USA . 102 : 15545 – 15550 .

35 . Cheng , H. , G. Sun , K. Yang , R. W. Gross , and X. Han . 2010 . Selective desorption/ionization of sulfatides by MALDI-MS facilitated using 9-aminoacridine as matrix . J. Lipid Res. 51 : 1599 – 1609 .

36 . Shea , H. C. , D. D. Head , K. D. Setchell , and D. W. Russell . 2007 . Analysis of HSD3B7 knockout mice reveals that a 3alpha-hydroxyl stereochemistry is required for bile acid function . Proc. Natl. Acad. Sci. USA . 104 : 11526 – 11533 .

37 . Huang , J. , S. P. Bathena , I. L. Csanaky , and Y. Alnouti . 2011 . Simultaneous characterization of bile acids and their sulfate me-tabolites in mouse liver, plasma, bile, and urine using LC-MS/MS . J. Pharm. Biomed. Anal. 55 : 1111 – 1119 .

38 . Alnouti , Y. 2009 . Bile Acid sulfation: a pathway of bile acid elimina-tion and detoxifi cation . Toxicol. Sci. 108 : 225 – 246 .

39 . Hofmann , A. F. , and L. R. Hagey . 2008 . Bile acids: chemistry, pathochemistry, biology, pathobiology, and therapeutics . Cell. Mol. Life Sci. 65 : 2461 – 2483 .

40 . Larusso , N. F. , and T. V. Masyuk . 2011 . The role of cilia in the regu-lation of bile fl ow . Dig. Dis. 29 : 6 – 12 .

41 . Ward , C. J. , D. Yuan , T. V. Masyuk , X. Wang , R. Punyashthiti , S. Whelan , R. Bacallao , R. Torra , N. F. LaRusso , V. E. Torres , et al . 2003 . Cellular and subcellular localization of the ARPKD protein; fi brocys-tin is expressed on primary cilia . Hum. Mol. Genet. 12 : 2703 – 2710 .

42 . Zhang , M. Z. , W. Mai , C. Li , S. Y. Cho , C. Hao , G. Moeckel , R. Zhao , I. Kim , J. Wang , H. Xiong , et al . 2004 . PKHD1 protein encoded by the gene for autosomal recessive polycystic kidney disease associ-ates with basal bodies and primary cilia in renal epithelial cells . Proc. Natl. Acad. Sci. USA . 101 : 2311 – 2316 .

43 . Wang , S. , J. Zhang , S. M. Nauli , X. Li , P. G. Starremans , Y. Luo , K. A. Roberts , and J. Zhou . 2007 . Fibrocystin/polyductin, found in the same protein complex with polycystin-2, regulates calcium responses in kidney epithelia . Mol. Cell. Biol. 27 : 3241 – 3252 .

44 . Sweeney , W. E. , Jr ., and E. D. Avner . 2006 . Molecular and cellular pathophysiology of autosomal recessive polycystic kidney disease (ARPKD) . Cell Tissue Res. 326 : 671 – 685 .

45 . Wang , X. , P. C. Harris , S. Somlo , D. Batlle , and V. E. Torres . 2009 . Effect of calcium-sensing receptor activation in models of auto-somal recessive or dominant polycystic kidney disease . Nephrol. Dial. Transplant. 24 : 526 – 534 .

46 . Griffi ths , W. J. , T. Koal , Y. Wang , M. Kohl , D. P. Enot , and H. P. Deigner . 2010 . Targeted metabolomics for biomarker discovery . Angew. Chem. Int. Ed. Engl. 49 : 5426 – 5445 .

47 . Ressom , H. W. , J. F. Xiao , L. Tuli , R. S. Varghese , B. Zhou , T. H. Tsai , M. R. Ranjbar , Y. Zhao , J. Wang , C. Di Poto , et al . 2012 . Utilization of metabolomics to identify serum biomarkers for he-patocellular carcinoma in patients with liver cirrhosis . Anal. Chim. Acta . 743 : 90 – 100 .

48 . Raedsch , R. , B. H. Lauterburg , and A. F. Hofmann . 1981 . Altered bile acid metabolism in primary biliary cirrhosis . Dig. Dis. Sci. 26 : 394 – 401 .

and liver transplantation in patients with autosomal recessive poly-cystic kidney disease: a multicentric study . Nephrol. Dial. Transplant. 27 : 2083 – 2088 .

7 . Dell , K. M. 2011 . The spectrum of polycystic kidney disease in chil-dren . Adv. Chronic Kidney Dis. 18 : 339 – 347 .

8 . Sato , Y. , X. S. Ren , and Y. Nakanuma . 2012 . Caroli’s disease: cur-rent knowledge of its biliary pathogenesis obtained from an orthol-ogous rat model . Int. J. Hepatol. 2012 : 107945 .

9 . Salomon , R. , S. Saunier , and P. Niaudet . 2009 . Nephronophthisis . Pediatr. Nephrol. 24 : 2333 – 2344 .

10 . Torres , V. E. , and P. C. Harris . 2009 . Autosomal dominant polycys-tic kidney disease: the last 3 years . Kidney Int. 76 : 149 – 168 .

11 . Mather , A. R. , and L. J. Siskind . 2011 . Glycosphingolipids and kid-ney disease . Adv. Exp. Med. Biol. 721 : 121 – 138 .

12 . Natoli , T. A. , L. A. Smith , K. A. Rogers , B. Wang , S. Komarnitsky , Y. Budman , A. Belenky , N. O. Bukanov , W. R. Dackowski , H. Husson , et al . 2010 . Inhibition of glucosylceramide accumulation results in effective blockade of polycystic kidney disease in mouse models . Nat. Med. 16 : 788 – 792 .

13 . Natoli , T. A. , H. Husson , K. A. Rogers , L. A. Smith , B. Wang , Y. Budman , N. O. Bukanov , S. R. Ledbetter , K. W. Klinger , J. P. Leonard , et al . 2012 . Loss of GM3 synthase gene, but not sphin-gosine kinase 1, is protective against murine nephronophthisis-related polycystic kidney disease . Hum. Mol. Genet. 21 : 3397 – 3407 .

14 . Chatterjee , S. , W. Y. Shi , P. Wilson , and A. Mazumdar . 1996 . Role of lactosylceramide and MAP kinase in the proliferation of proximal tubular cells in human polycystic kidney disease . J. Lipid Res. 37 : 1334 – 1344 .

15 . Deshmukh , G. D. , N. S. Radin , V. H. Gattone 2nd , and J. A. Shayman . 1994 . Abnormalities of glycosphingolipid, sulfatide, and ceramide in the polycystic (cpk/cpk) mouse . J. Lipid Res. 35 : 1611 – 1618 .

16 . Nagao , S. , M. Kugita , D. Yoshihara , and T. Yamaguchi . 2012 . Animal models for human polycystic kidney disease . Exp. Anim. 61 : 477 – 488 .

17 . Mason , S. B. , Y. Liang , R. M. Sinders , C. A. Miller , T. Eggleston-Gulyas , R. Crisler-Roberts , P. C. Harris , and V. H. Gattone 2nd . 2010 . Disease stage characterization of hepatorenal fi brocystic pathology in the PCK rat model of ARPKD . Anat. Rec. (Hoboken) . 293 : 1279 – 1288 .

18 . Renken , C. , D. C. Fischer , G. Kundt , N. Gretz , and D. Haffner . 2011 . Inhibition of mTOR with sirolimus does not attenuate progression of liver and kidney disease in PCK rats . Nephrol. Dial. Transplant. 26 : 92 – 100 .

19 . Balluff , B. , C. Schone , H. Hofl er , and A. Walch . 2011 . MALDI im-aging mass spectrometry for direct tissue analysis: technological advancements and recent applications . Histochem. Cell Biol. 136 : 227 – 244 .

20 . Cazares , L. H. , D. A. Troyer , B. Wang , R. R. Drake , and O. J. Semmes . 2011 . MALDI tissue imaging: from biomarker discovery to clinical applications . Anal. Bioanal. Chem. 401 : 17 – 27 .

21 . Schwamborn , K. 2012 . Imaging mass spectrometry in biomarker discovery and validation . J. Proteomics . 75 : 4990 – 4998 .

22 . Rauser , S. , C. Marquardt , B. Balluff , S. O. Deininger , C. Albers , E. Belau , R. Hartmer , D. Suckau , K. Specht , M. P. Ebert , et al . 2010 . Classifi cation of HER2 receptor status in breast cancer tissues by MALDI imaging mass spectrometry . J. Proteome Res. 9 : 1854 – 1863 .

23 . Hanrieder , J. , G. Wicher , J. Bergquist , M. Andersson , and A. Fex-Svenningsen . 2011 . MALDI mass spectrometry based molecular phenotyping of CNS glial cells for prediction in mammalian brain tissue . Anal. Bioanal. Chem. 401 : 135 – 147 .

24 . Battula , V. L. , Y. Shi , K. W. Evans , R. Y. Wang , E. L. Spaeth , R. O. Jacamo , R. Guerra , A. A. Sahin , F. C. Marini , G. Hortobagyi , et al . 2012 . Ganglioside GD2 identifi es breast cancer stem cells and pro-motes tumorigenesis . J. Clin. Invest. 122 : 2066 – 2078 .

25 . Berry , K. A. , J. A. Hankin , R. M. Barkley , J. M. Spraggins , R. M. Caprioli , and R. C. Murphy . 2011 . MALDI imaging of lipid biochemistry in tissues by mass spectrometry . Chem. Rev. 111 : 6491 – 6512 .

26 . Liu , Y. , Y. Chen , A. Momin , R. Shaner , E. Wang , N. J. Bowen , L. V. Matyunina , L. D. Walker , J. F. McDonald , M. C. Sullards , et al . 2010 . Elevation of sulfatides in ovarian cancer: an integrated tran-scriptomic and lipidomic analysis including tissue-imaging mass spectrometry . Mol. Cancer . 9 : 186 .

27 . Marsching , C. , M. Eckhardt , H. J. Grone , R. Sandhoff , and C. Hopf . 2011 . Imaging of complex sulfatides SM3 and SB1a in mouse kid-ney using MALDI-TOF/TOF mass spectrometry . Anal. Bioanal. Chem. 401 : 53 – 64 .


ww

w.jlr.org

Dow

nloaded from


http://www.jlr.org/



maldi imaging ms reveals candidate lipid markers of ... were acquired in esi nega- tive ion mode...

Documents