RESEARCH PAPER

Novel “omics” approach for study of low-abundance,low-molecular-weight components of a complex biological tissue:regional differences between chorionic and basal platesof the human placenta

Komal Kedia1 & Caitlin A. Nichols1 & Craig D. Thulin2& Steven W. Graves1

# Springer-Verlag Berlin Heidelberg 2015

Abstract Tissue proteomics has relied heavily on two-dimensional gel electrophoresis, for protein separation andquantification, then single protein isolation, trypsin digestion,and mass spectrometric protein identification. Such methodsare predominantly used for study of high-abundance, full-length proteins. Tissue peptidomics has recently been devel-oped but is still used to study the most highly abundant spe-cies, often resulting in observation and identification ofdozens of peptides only. Tissue lipidomics is likewise new,and reported studies are limited. We have developed an“omics” approach that enables over 7,000 low-molecular-weight, low-abundance species to be surveyed and have ap-plied this to human placental tissue. Because the placenta isbelieved to be involved in complications of pregnancy, itsproteomic evaluation is of substantial interest. In previousresearch on the placental proteome, abundant, high-molecular-weight proteins have been studied. Application oflarge-scale, global proteomics or peptidomics to the placentahave been limited, and would be challenging owing to theanatomic complexity and broad concentration range of pro-teins in this tissue. In our approach, involving protein deple-tion, capillary liquid chromatography, and tandem mass spec-trometry, we attempted to identify molecular differences

between two regions of the same placenta with only slightlydifferent cellular composition. Our analysis revealed 16 spe-cies with statistically significant differences between the tworegions. Tandem mass spectrometry enabled successful se-quencing, or otherwise enabled chemical characterization, oftwelve of these. The successful discovery and identification ofregional differences between the expression of low-abun-dance, low-molecular weight biomolecules reveals the poten-tial of our approach.

Keywords Tissue proteomics . Capillary liquidchromatography–MS . Low-molecular-weight proteins .

Placenta . Peptidomics . Lipids

Introduction

Tissue proteomics has historically relied on single or two-dimensional gel electrophoresis (2-DGE) for separation andquantification of tissue proteins. This has typically beenfollowed by isolation of single proteins, trypsin digestion,and mass spectrometric (MS) identification of the proteinsby database comparison [1, 2]. These approaches are mostapplicable to highly abundant, large-molecular-weight pro-teins. These methods, without tandemMS fragmentation stud-ies using collisional-induced dissociations or electron-transferdissociation, do not provide direct amino acid sequences.None of these methods, in the absence of fragmentation, en-ables identification of post-translational modifications (PTM)or chemical changes that can occur in response to reactiveoxygen species (ROS) or chemical modifiers. Other “global”proteomic approaches, for example multi-dimensional proteinidentification technology (MudPIT), rely on successive chro-matographic separations, most often after trypsin digestion,rather than electrophoretic methods [3]. They have also been

Electronic supplementary material The online version of this article(doi:10.1007/s00216-015-9009-3) contains supplementary material,which is available to authorized users.

* Steven W. Gravesswgraves@chem.byu.edu

1 Department of Chemistry and Biochemistry, Brigham YoungUniversity, BNSN C-212, Provo, UT 84602, USA

2 Department of Chemistry, Utah Valley University, 800W. UniversityParkway, Orem, UT 84058, USA

DOI 10.1007/s00216-015-9009-3

Received: 24 March 2015 /Revised: 4 August 2015 /Accepted: 26 August 2015 /Published online: 8 September 2015

Anal Bioanal Chem (2015) 407:8543–8556

used for study of tissues [4]. These global approaches ofteninvolve amino acid sequencing of tryptic digest fragments toidentify proteins and associated PTM. Such global methodsenable study of a greater number of full-length proteins than 2-DGE but generally still suffer from ion suppression andmasking of low-abundance components. They are not current-ly designed to study peptides or lipids. More recently, “top-down” methods (i.e. methods which start with the intact, full-length protein) have been developed that enable injection ofproteins into the MS and allow amino acid sequencing andobservation of PTM [5, 6]. However, such approaches oftenrequire more than one separation step before MS analysis, areapplicable to proteins with high charge states and sufficientlylow mass-to-charge ratios, and typically require instrumentswith very high mass-accuracy, for example Fourier-transformion cyclotron resonance mass spectrometers. Although prote-omics studies are valuable, they still focus on high-molecularweight, more highly abundant proteins. The highly expressedcell proteins typically studied are often cyto-skeletal species orare involved in protein folding and trafficking or perform avariety of housekeeping functions and are less frequently in-volved in regulation and disease processes. Hence, monitoringof small, low-abundance components of cells may be as im-portant as probing high-abundance proteins.

As an alternative, tissue peptidomics has recently been de-veloped for cataloguing the peptides present in a particulartissue or for quantitative comparisons between tissues usingboth label-free [7] and isotope-labeled tags (ICAT, iTRAQ) [8,9]. Labelling techniques such as iTRAQ, when applied totissue peptides, seem to enable analysis of dozens of peptides[9]. However, different labeling agents react inconsistentlywith different peptides, suggesting variable quantification[9]. Such techniques are not useful for lipids. Thesepeptidomics methods have potential but seem limited in thenumber and type of small molecular constituents assayed.Tissue lipidomics frequently requires tissue homogenizationfollowed by organic solvent extraction and injection of ex-tracted molecules into to the MS without or with prior chro-matographic separation [10–12]. Fragmentation studies havebeen attempted but there is no well-developed reference data-base enabling lipid identification by comparison of fragmentmass. Identification requires careful assessment of mass dif-ferences between individual fragment peaks to characterizethe neutral loss components and thereby identify molecularfragments present in the pre-fragmentation parent species.

We have developed a unique and complementary methodfor analysis of low-abundance, low-molecular-weight compo-nents of tissue specimens [13].We use a protein depletion step[14]. This desorbs low-molecular-weight (LMW; as appliedhere <5,000 Daltons for peptides and <1,500 Daltons forlipids) species from binding partners and, with the removalof highly abundant proteins, enables study of several thousandadditional molecular species. The method has been applied to

tissue as part of an initial study to evaluate its analytical reli-ability and its breadth of coverage [13]. It was sufficientlyreproducible to enable label-free quantitative comparisons ofdifferent organs and enabled observation of more than 7,000molecular species [13]. The method can be used to study low-abundance, small proteins, peptides, lipids, and potentiallyother biomolecules and enables relatively seamless tandemMS fragmentation studies to provide de-novo amino acid se-quencing for polypeptides and chemical characterization andclassification of lipids [15]. Such fragmentation data can beused to identify post-translational modifications, oxidativedamage, and/or other non-physiologic chemical modifica-tions. In the work discussed in this report this novel approachwas applied to human tissue.

One of the most challenging but important organs for tissueproteomics is the human placenta. The placenta is of clinicalinterest because of its almost certain involvement in pre-eclampsia (PE), a disease complicating 3–5 % of pregnancieswhich results in as many as 75,000 maternal deaths yearly[16]. The placenta is required for PE and evidence suggestsalterations in placental biochemistry occur before, and proba-bly lead, to PE [17].

One rather unique challenge when working with the pla-centa is that it is formed from cells originating from both thefetus and the mother but the cells are interspersed rather thancompartmentalized [18]. Emerging evidence suggests thatspecific early changes that may contribute to PE arise some-times in the fetal tissues and sometimes in the maternallyderived tissues [19]. Although these studies have usually stud-ied single proteins, the results nonetheless reveal the impor-tance of distinguishing between fetal and maternal regions ofthe placenta if PE is to be fully understood [20]. Fixing andstaining tissue may enable identification of a particular celltype, and immuno-staining may enable monitoring of a spe-cific protein’s abundance in that cell type, but this approach isnot amenable to global analysis of molecular mediators andmodified dynamics. Furthermore, quantitative comparativeproteomic approaches would be difficult, if not impossible,to perform after such chemical modifications, even thoughthere are methods that can reverse much of the cross-linkinginitiated by fixatives [21]. Placental proteomics has only beenattempted relatively recently and has mostly been used forstudy of highly abundant proteins [22, 23]. A small numberof more global proteomics studies of the placenta have beenattempted [24, 25]. These studies have focused on a singleclass of biomolecules, for example proteins [24] or metabo-lites [25]. However, low-molecular-weight components of theplacenta do not seem to have been investigated in other“omics” studies. In addition, no effort has been made in pre-vious placental proteomic studies to investigate differencesbetween the regions of the placenta . Consequently, we con-sidered use of our novel tissue “omics” approach for this organas an important demonstration of the method’s utility. The

K. Kedia et al.8544

intent of this study was to assess the robustness of the methodfor finding region-specific, quantitative differences betweenmolecular components, but not to develop a set of cell-specific biomarkers. If this approach were successful, it wouldprovide a method for investigation of thousands of previouslyunstudied, low-molecular-weight, low-abundance compo-nents in complex tissues. It would also provide a way of in-vestigating events potentially mediating diseases, includingPE, in which one would expect to see many more differencesbetween regions of the placenta as part of disease. Hence thisapproach may enable clarification not only of which mole-cules are modified but, more specifically, clarify whether theywere of fetal or maternal origin.

Materials and methods

Tissue collection

The Internal ReviewBoards of Intermountain Health Care andits affiliated hospitals and of Brigham Young University ap-proved of these studies. Placentas were regarded as discardedmaterials and no informed consent was obtained, and no per-sonal information was provided. Placentas were collected im-mediately after uncomplicated vaginal or C-section deliveries(n=12). The fetal amniotic membrane was carefully removedand thin pieces of tissue were dissected from the extreme outersurface of the chorionic plate. Placentas were then invertedand on the opposing surface (basal plate), thin tissue pieceswere sliced from the very outermost surface of the cotyledons.Both tissue types were collected from a location midway be-tween the cord and the peripheral edge of the placenta alongthe same vertical axis. Care was taken to maintain consistencyin this process. After collection, tissues were rapidly frozen ondry ice and maintained frozen until processed further.

Tissue homogenization

Frozen placental tissue from one location was finely mincedby use of a sterile stainless steel blade. These shavings(approx. 500 mg) were then placed in a 5-mL stainless steelball mill cylinder containing 12–15, 3-mm chromium-steelgrinding balls. The vessel was flash frozen in liquid nitrogenfor 4 min and shaken by use of a ball mill dismembrator(Mikro-Dismembrator S; Sartorius, Göttingen, Germany) op-erating at 2000 rpm for 8–10 min to form a paste-like homog-enate. The tissue homogenate was re-suspended in 5 mL coldphysiological phosphate-buffered saline (PBS, pH 7.2). Thismixture was vortex mixed thoroughly and centrifuged (SorvalRT7; Kendro Laboratory Products, Newton, CT, USA) for10 min at 4000 rpm (8046g) at 4 °C. Portions (200 μL) ofthe supernatant were stored at −80 °C for further processing. Abroad-range proteolytic enzyme inhibitor cocktail was added

to the specimen before to processing (Sigma–Aldrich, cata-logue number P9599). In addition, processing at low temper-atures and immediate freezing were used to further preventproteolysis.

Protein depletion

Highly abundant, high-molecular-weight proteins were pre-cipitated by adding 2:1 (v/v) acetonitrile to the homogenate.The mixture was vortex mixed briefly, left at room tempera-ture for 30 min, and centrifuged (IEC Mcromax RF; ThermoFisher Scientific, Waltham, MA, USA) at 14,000 rpm (13,107g) for 10 min at 4 °C. The pellet was discarded and the super-natant (~550 μL) was transferred to a new Eppendorf tubecontaining 300 μL HPLC-grade water. The total volume wasthen reduced to 200 μL by use of a vacuum centrifuge(CentriVap Concentrator; Labconco, Kansas City, MO,USA) at 37 °C. The apparent protein concentration of eachsample was determined (Pierce Microplate BCA Protein As-say Kit; Thermo Scientific). Avolume of the lyophilized sam-ple equivalent to 20 μg protein was then dried in vacuo to avolume of 10 μL. Formic acid (88 %, 10 μL) was then added,each sample was briefly vortex mixed, and 5 μL sample ofconcentration 1 μg μL−1 apparent protein was injected forcapillary liquid chromatography–mass spectrometric (cLC–MS) analysis. This single protein-depleted specimencontained small proteins, peptides, lipids, and potentially oth-er metabolites.

Chromatographic separation

Reversed-phase capillary liquid chromatography (cLC) wasperformed with an LC Packings Ultimate Capillary HPLCpump system, with a Famos autosampler (Dionex, Sunnyvale,CA, USA). The system comprised a 1 mm (16.2 μL) dry-packed MicroBore guard column (IDEX Health and Science,Oak Harbor, WA, USA), coupled to a 15 cm×250 μm i.d.capillary column, slurry-packed in-house with Poros R1reversed-phase medium (Applied Biosystems, Foster City,CA, USA). The mobile phase was a gradient generated byuse of an aqueous phase: 98 % HPLC-grade water, 2 % ace-tonitrile, 0.1 % formic acid, and an organic phase: 98 % ace-tonitrile, 2 % HPLC-grade water, 0.1 % formic acid. Thispacking material had been previously used over a prolongedperiod for experiments involving both serum and tissue pro-teomics and enabled continuity in experiments with goodchromatographic reproducibility [9]. The gradient began with3 min of 95 % aqueous and 5 % organic phase, followed by alinear increase in organic phase to 60 % over the next 24 min.The gradient was then increased linearly to 95 % organicphase–5 % aqueous phase over the next 7 min, held at 95 %organic phase for 7 min and returned to 95 % aqueous phase

Novel “omics” approach surveying low molecular weight molecules 8545

over 5 min. The columnwas allowed to re-equilibrate until theend time of the run (58 min). The flow rate was 5 μL min−1.

Mass spectrometric analysis

Eluate was introduced through an electrospray needle into atandem mass spectrometer (QSTAR Pulsar І quadrupole or-thogonal time-of-flight, QTOF, Applied Biosystems). Theelectrospray source was at 4800 V. All samples were run inthe positive-ion mode. Other MS conditions used during theanalyses were: microchannel plate voltage (MCP)=2380 V,plate voltage=330 V, curtain gas pressure=20 psi, sourcegas pressure=20 psi. A mass spectrum was obtained every1 s from m/z 500–2500 over 5–55 minutes elution. Fragmen-tation studies were conducted with an acquisition rate of 1spectrum per second. MS calibration was performed by injec-tion of a non-physiological peptide, Glu-1-fibrinopeptide(GluFib) with an m/z of 785.85 (z=+2). Resolution of theinstrument during the time when these samples were run var-ied between 9,500 and 10,000. The results were analyzed byuse of Analyst QS 1.1 software (Applied Biosystems). On thebasis of tissue homogenates spiked with this peptide, the ap-proach was estimated to have a limit of detection (LOD) of 5–10 nmol L−1; for the great majority of the species analyzed,estimated concentrations were 7-200 nmol L−1, equivalent toas little as 2 pg material per MS peak.

Time normalization and data analysis

Elution times can vary from day to day and occasionally fromrun to run. To compensate, we defined 2 min windows withinthe useful chromatogram. In each window we selected a cen-tral endogenous peak representing a molecule found in allspecimens.

These central peaks eluted at approximately 2 min intervals.These were used for time alignment among different runs. Tofind regional tissue markers, mass spectra from chorionic plateand basal plate were color coded, overlaid, centered on a timemarker, and peaks were visually inspected to find featuresappearing to differ in abundance. Every peak which seemedindicative of quantitative differences between comparisongroups was further evaluated. This process involved“extracting” the candidate marker, by use of Analyst software,and recording the maximum height of the XIC (extracted ionchromatogram) peak. Time markers selected to correct for var-iation of elution times (in order of increasing elution times) hadmass-to-charge ratios as follows: 695.12 (z=+4), 827.77 (z=+6), 685.9 (z=+ 4), 672.7 (z=+3), 1009.4 (z<+10), 616.19 (z=+1), 526.3 (z=+1), 524.3 (z=+1), 650.4 (z=+1), and 675.56(z=+1). The charge on the m/z 1009.4 peptide was estimatedto be +14. As a consequence the isotopic envelope was notsufficiently resolved for the summed spectra compiled over

the two-minute elution window to enable determination of theexact value of z.

The analytical performance of the method has already beenevaluated [13]; coefficients of variation from 23.0 to 26.8 %were obtained for single analytes measured in different precip-itates from the same tissue. This compares favorably with pro-teomics results obtained by use of 2D GE-MS methods [26,27]. Quantitative reproducibility has not frequently been report-ed for other global or “top-down” proteomics methods, but onestudy reported a CVof 36 % for technical replicates [28].

Statistical analysis

A Student’s t-test was performed to determine whether theabundance of an extracted peakwas different in the two regions.A p-value of less than 0.05 was regarded as statistically signif-icant. In addition, the consistency of greater amounts of thesedifferently expressed molecules in the chorionic or basal plateswas assessed by constructing 2×2 contingency tables and ap-plying a Fisher exact test to the results (Table 1). Finally, as anindication of biologic variability, we calculated relative standarddeviations (%RSD) for each of the 16 differently expressedspecies in each region; the results are included in Table 1.

Normalization of candidate markers to reducenon-biological variation

For all statistically significant candidate markers a referencepeak was selected. This was an endogenous peak that elutedclose to the candidate marker and was present in all samples incomparable abundance across comparison groups. To normal-ize candidate peaks, the intensity of the candidate marker wasdivided by the intensity of the reference peak. Reference peaksselected for normalization of species in elution windows 1, 2,3, 4, 6, 7, and 10 had the mass-to-charge ratios: 1241.17 (z=+4), 993.12 (z=+5), 660.38 (z=+1), 660.38 (z=+1), 913.48(z=+1), 913.48 (z=+1), and 563.57 (z=+1) respectively. Noneof these reference peaks was quantitatively different betweenchorionic and basal plates (p>0.1). This approach has previ-ously been evaluated with an overall coefficient of variation of~25 % in terms of combined pre-analytical and analyticalvariability [13]. A Student’s t-test was then performed a sec-ond time on the normalized values. The peaks that differedstatistically were also required to show the same trend for atleast eight of the 12 placentas.

Identification of peptides of interest by tandem MS

Significantly different molecules were submitted for tandemMS analysis with collisionally-induced dissociation (CID).The fragmentation pattern produced was used in an attemptto determine the amino acid sequence and the parent protein.One or more samples containing high levels of the peptide of

K. Kedia et al.8546

interest were thawed and 5μg of the sample containing 2.5μLof 88 % formic acid was hand-injected into the cLC–MS sys-tem. Fragmentation data were collected for 2 min with argonor nitrogen as collision gas. Different collision energies wereused to achieve more complete production of b-series and y-series fragment ions. Spectra obtained by use of different col-lision energies were summed to obtain more complete cover-age. Peaks in the export data list were checked for anymisassigned charge states. Charge states for all the peaks wereconverted to their +1m/z values by use of the formula +1mass=m/z value×charge−(charge−1H+). This corrected listwas then submitted to the search engine Mascot (Mascot2.3, www.matrixscience.com) which correlated thefragmentation data with archived protein sequence data.

De-novo sequencing

De-novo sequencing was attempted on peptides when Mascotfailed [29]. Individual amino acids were assigned by measur-ing the mass differences between fragment ions. The presence

of immonium ions aided identification of specific aminoacids. This process proved to be much more complicated,because these peptides were not produced by trypsin diges-tion. A BLAST search by use of the National Institutes ofHealth website (http://blast.ncbi.nlm.nih.gov) was performedon the sequence of specific amino acids observed from eitherMascot assignment or de-novo sequencing; the calculatedprobability (E score) of having identified the parent proteinis given in Table 2.

Identification of lipid markers

The processing step, involving acetonitrile treatment, pro-duced a specimen that contained, in addition to polypeptides,polar or positively charged lipids, including neutral lipids thatformed adducts with cations in the elution buffer. During thecLC step, these lipids, mainly phosphatidylcholines (PC) andsphingomyelins (SM), typically eluted at mobile phase organ-ic solvent concentrations of 40 % or greater. These lipids be-come positively charged when sprayed in an acidic mobile

Table 1 List of 16 differently expressed molecular species across all 10 time windows

Marker (m/z) Charge state (z) Student’s t-test Fisher’s exact test %RSD (raw) %RSD (normalized)

Raw p-value Normalized p-value p-value Basal plate Chorionic plate Basal plate Chorionic plate

Window 1 (XIC range: m/z 695–696)

624.32 +2 0.02 0.02 0.0033 31.7 32.5 44.7 45.8

718.36 +2 0.005 0.007 0.0001 54.9 47.6 68.1 58.1

Window 2 (XIC range: m/z 827–828)

614.38 +3 0.05 0.02 0.0033 41.4 41.8 36.7 29.5

808.87 +2 0.05 0.05 0.0033 54.8 49.9 58.4 45.7

Window 3 (XIC range: 685–686)

760.4 +4 0.06 0.01 0.039 33.0 38.2 29.0 34.0

Window 4 (XIC window: 672–673)

857.15 +10 0.01 0.02 0.22 17.3 18.1 28.8 30.0

Window 6 (XIC range: 616–617)

520.33 1 0.02 0.03 0.039 42.3 41.6 41.9 41.7

542.31 1 0.01 0.03 0.039 34.1 38.9 38.9 38.1

544.33 1 0.01 0.01 0.039 46.2 45.1 46.1 44.9

566.32 1 0.02 0.03 0.039 39.0 21.9 38.9 21.5

568.33 1 0.03 0.04 0.039 56.6 44.7 54.1 44.6

Window 7 (XIC range: 526–527)

570.34 1 0.03 0.04 0.22 50.4 36.4 48.9 35.8

1159.50 1 0.02 0.02 0.0001 33.2 20.4 31.5 15.7

Window 10 (XIC range: 675–676)

647.55 1 0.07 0.01 0.039 36.5 39.9 18.1 22.5

673.52 1 0.01 0.001 0.0033 29.8 37.5 20.9 23.4

701.55 1 0.009 0.01 0.039 19.5 27.9 23.6 24.0

XIC, extracted ion count. This was used as a measure of the abundance of a given species. Fisher’s exact test was used to determine the consistency forone region with more of a species than another regions, as described in the Materials and methods section. %RSD is the relative standard deviation as apercentage calculated as SD/mean×100 to provide some measure of biological variability

Novel “omics” approach surveying low molecular weight molecules 8547

phase, because of the presence of a quaternary nitrogen in thecholine group and protonation of the phosphate group, andmost often enter the mass spectrometer as the [M+H]+ spe-cies. Lipids were characterized by the same tandem MS–MSapproach as used for peptides with the same instrument set-tings. The collision gas was usually nitrogen, and 1 or 2 col-lision energies were sufficient to fragment the lipid of interest.The LIPID MAPS (lipidmaps.org) database was used to in-vestigate the structure on the basis of the parent mass andmolecular formula of the lipid. LIPID MS predictor (http://www.lipidmaps.org/tools/index.html) was used to calculatepossible fragment ions.

Results

Time markers to align chromatographic elution times

As described above, the total ion chromatogram (TIC) wasdivided into 2-min windows by using time-alignmentmarkers. These time windows covered most of the usefulchromatogram with its attendant TIC. The many mass spectrarecorded during the 2 min were averaged. The time markersare listed in the Materials and methods section, above.

Low abundance, LMW biomolecules with differentabundance in the chorionic and basal plates

MS data were color coded by region and overlays of MS datafor a given time window were visually inspected to find ob-vious differences between peak abundance in comparisongroups (at least 1.5× or greater). As expected, most peaks werecommon to both sides of the placenta and were present incomparable abundance. Importantly, several peaks differedsignificantly in their intensity between chorionic plate andbasal plate tissue, indicating regional selectivity. Figure 1shows an example. The results revealed 16 peaks with statis-tically significantly different abundances in the two plates.The MS characteristics of these 16 peaks are summarized inTable 1. Box-and-whisker plots for these 16 differently

expressed molecules are provided in the Electronic Supple-mentary Material (ESM).

Identification of peptides

Tandem mass spectrometry (MS–MS) was performed to de-termine the amino acid sequence of significantly different re-gional peptide markers. Tandem MS–MS was also used tofragment candidate lipid markers, enabling the lipid class tobe established.

Of the seven peptide peaks, two represented the same mo-lecular species observed in two different charge states. Of thesix unique peptide biomarkers, Mascot provided sequencesfor two. For example, peptide ofm/z 718.36 (z=+2) was iden-tified as fetal Hgb (γ subunit) (Fig. 2).

Two peptides were identified by de-novo sequencing. Themarker with m/z 808.82 (z=+2) and a neutral mass of1615.64 Dawas identified as fibrinopeptide A phosphorylatedat Ser-3. We were able to determine eight amino acid in seriesfrom the MS–MS spectrum. A BLAST search performed byuse of the NIH protein database identified the parent protein asfibrinopeptide A. This peptide has a predicted neutral mass of1535.68 Da, which differed from the parent mass of the mark-er by ~80 Da, indicating the presence of a phosphate group.The sequence of this peptide of fibrinopeptide Awas enteredinto a fragment ion calculator to predict individual b and yions. Avalue of 80 was added to the serine residue, as the siteof phosphorylation (given the absence of threonine and tyro-sine, ADSGEGDFLAEGGGVR) and resulted in good agree-ment between predicted and observed fragments. This con-firmed the identity of this peptide as a phosphorylated frag-ment of fibrinopeptide A (Fig. 3).

A peptide withm/z 760.4 and a charge state of +4 was alsoidentified by de novo sequencing. We identified four aminoacids in series and performed a BLAST search. The parentprotein for this peptide was assigned as hemoglobin (Hgb)alpha-2 on the basis of sequence homology and the ability toachieve the observed overall mass of the peptide within thesequence of the parent protein.

Table 2 Summary of the chemical identities of peptides differently expressed in chorionic and basal plates of the human placenta

Peptide marker (m/z) z Neutral mass Parent protein Amino acid sequencea E-score

718.36 +2 1434.72 Hgb subunit gamma (Fetal Hgb) GHFTEEDKATITS 5×10−7

614.38 +3 1840.14 Hgb subunit alpha-1 VLSPADKTNVKAAWGKVG 8×10−11

808.82 +2 1615.64 Phosphorylated fibrinopeptide A ADpS*GEGDFLAEGGGVR 0.013

760.68 +4 3038.72 Hgb subunit alpha-2 VLSPADKTNVKAAWGKVGAHAGEYGAEALE 3.3

a Sequence represents the amino acid composition of the peptide identified; *pS in the sequence of 808.82 represents phosphorylation at the amino acidserine.

Hgb, hemoglobin

K. Kedia et al.8548

Two peptides with m/z 857.49 and 624.32 could not besequenced because of their high charge state and low abun-dance, respectively. All peptide identities are listed in Table 3.

Identification of lipids

Ten quantitatively different, region-specific markers werelikely to be lipids, given their elution later in the cLC chro-matogram, consistent with their being non-polar, and giventheir fragmentation patterns (Table 3). Fragmentation spectrafurnished an intense peak at m/z 184.07 characteristic of aphosphocholine group found in both phosphatidylcholines(PC) and sphingomyelins (SM). An accurately determinedmass for each candidate lipid was entered into the searchfunction of the LIPID MAPS database, which provided pos-sible structures and molecular formulas for the lipids, hencedifferentiating PC from SM. This software, however, gives noinformation about likely tandemMS fragment ions. To furthercharacterize the markers, LIPID MS predictor was used; thispredicts, in-silico, possible fragments, with their masses. Ifthese m/z values are then observed in fragmentation spectra,they help establish the marker’s identity.

Three differently expressed lipids were not classified byuse of the LIPID MAPS database. The fragmentation spec-trum of one of these, with anm/z of 544.33 (z=+1), containednot only an abundant peak at m/z 184.07, characteristic of aphosphocholine, but also peaks of several product ions sug-gesting it was a sodium (Na) adduct [30]. A prominent peak atm/z 485.23, indicative of loss of trimethylamine (TMA, loss of59.07 mass units) confirmed it was a Na adduct. This interac-tion of Na with the lipid makes the negatively charged oxygen

available for nucleophilic reactions, and one commonresult is loss of TMA. There was also a peak at m/z339.26, generated by loss of sodium-associatedphosphocholine (neutral loss of 205 mass units) fromthe parent. Further, because of loss of an acyl-sn-glycerol group from the fragment ions generated byneutral loss of 59.07, there was a prominent peak atm/z 146.97. The peak at 104.10 was also evidence thePC was an LPC [31]. These findings confirm the iden-tity of the lipid marker as an LPC with the elementalcomposition of [C26H52NO7P]Na

+. This is same molec-ular formula as the LPC with neutral mass of 521.33(Fig. 4). Because of the inability of mass spectrometryto determine the location of fatty acids and doublebonds within fatty acids, it cannot be confirmed unam-biguously that these two molecules represent the samespecies, but it is, nonetheless, likely. Similarly, the sec-ond unidentified marker with m/z 566.32 was identifiedas a sodiated LPC with elemental composition as[C28H50NO7P]Na

+.The third differently expressed lipid not identified by ref-

erence to LIPID MAPS had an m/z of 570.34. Fragmentationof this molecule resulted in an intense peak at 184.07, indic-ative of a PC or SM. Exact mass studies determined its ele-mental composition as C30H52NO7P. Thus, this lipid moleculeis an LPC with a fatty acid 22:5 (docosapentaenoic acid). Thegeneral classification and constituents of the other lipids aresummarized in Table 3.

The other differently expressed lipids (m/z values 1159.57and 647.52) did not furnish interpretable MS2 fragment spec-tra owing to their low initial abundance.

Fig. 1 Overlay of 16mass spectra in the region containing peptidem/z 718.36 with its isotope envelope (z=+2). Red, tissue collected from the chorionicplate; blue, tissue collected from outer cotyledons (basal plate). This species was more abundant in the chorionic plate of the placenta (p=0.007)

Novel “omics” approach surveying low molecular weight molecules 8549

Discussion

Tissue proteomics is an increasingly accepted tool for inves-tigating biochemical changes accompanying normal physiol-ogy and pathophysiology. Global, unbiased proteomic ap-proaches using multidimensional separations followed byMS are able to identify hundreds to a few thousand proteins

but are not quantitative without a single, predetermined targetand the availability of a heavy isotope analogue of the target, adefined amount of which is added to the specimen. Such ap-proaches do not reveal low-abundance proteins or other low-abundance species. Historically, electrophoretic gel methods,enabling some quantitative assessment, have been coupledwith MS for protein identification [21]. More recent

m/z,amu

Inte

nsity

, cou

nts

a

b

Fig. 2 a Averaged MS2 spectrum of peptide m/z 718.36, z=+2 (massaccuracy of the precursor peak=36 ppm). The assigned MS2 spectrumlabels of observed b and y-fragment ions are shown. bMascot prediction

of amino acid sequence for the peptide. The x-axis represents m/z in amu;the y-axis represents ion counts

K. Kedia et al.8550

comparative and quantitative proteomic approaches includeICAT or iTRAQ, in which heavy or light isotope labeling oftwo different tissue homogenates or cell lysates enable

comparison of those two states or groups [32, 33]. Recentadvances in “top-down” proteomics involving multiple sepa-rations coupled with high-resolution MS have enabled

y13y12

y11

y10

y9y8

y7

y6y5

b5*

b3*

GE

GD

F

L

AE

Fig. 3 Averaged MS2 spectra of a second differently expressed peptidem/z 808.82 (z=+2) (mass accuracy of the precursor peak=27 ppm) and the MSfragment-ion calculator’s prediction of b and y ions. This peptide was identified as a fragment of phosphorylated fibrinopeptide A

Table 3 Summary of the chemical classes and components of the lipids differently expressed in chorionic and basal plates of the human placenta

Lipid marker (m/z) z Neutral mass Lipid class Molecular formula Fatty acid components

520.33 +1 519.37 Glycerophospholipids (LPC) C26H50NO7P (18:2/0:0)a

542.32 +1 541.31 Glycerophospholipids (LPC) C28H48NO7P (20:5/0:0)

544.33 [M+Na]+ 521.33 Glycerophospholipids (LPC) C26H52NO7P (18:1/0:0)

566.32 [M+Na]+ 543.37 Glycerophospholipids (LPC) C28H50NO7P (20:4/0:0) or (0:0/20:4)

568.33 +1 567.33 Glycerophospholipids (LPC) C30H50NO7P (22:6/0:0)

570.34 +1 569.33 Glycerophospholipids (LPC) C30H52NO7P (22:5/0:0) or (0:0/22:5)

673.52 +1 671.52 Phosphosphingolipids C37H73N2O6P (18:2/14:0)

701.55 +1 700.58 Phosphosphingolipids C39H77N2O6P (18:1/16:1) or (16:1/18:1) or (18:2/16:0)

a The annotation (e.g. (18:2/0:0)) describes the number of carbon atoms in the fatty acid chain with the number of double bonds. In this case, the fatty acidis composed of 18 carbons with two double bonds at the sn-1 position of the glycerol backbone, whereas 0:0 represents the absence of a fatty acid chain atthe sn-2 position

LPC, lysophosphocholine

Novel “omics” approach surveying low molecular weight molecules 8551

sequencing of intact proteins and identification of hundreds ofproteins or proteoforms, including PTM [5, 6, 34]. However,these methods, including “top-down” methods, are suitablefor the most abundant, full length proteins, only. Proteins haveobvious importance in cell or tissue physiology, but smallproteins, peptides and lipids are also recognized as mediatingor regulating many intracellular processes [35].

Peptidomics, including tissue peptidomics, use similar ap-proaches. Both isotopic tagging (ICAT or iTRAQ) and non-isotopic quantification techniques have been used [8, 9, 36].Typically, peptides, after their separation from larger proteins,are submitted for analysis by liquid chromatography coupledwith tandem MS without trypsin digestion, given they arepeptides already. Currently, automated sequencing, by usingdata-dependent acquisition to select peptides for fragmenta-tion, is used in conjunction with a tandem mass spectrometer.This results in the top 3 to 20 (depending on the instrument)most abundant peptide species in any initial mass spectrumbeing submitted to fragmentation and the fragment profilesubmitted to search engines to determine amino acid

sequence. Such approaches report dozens of automaticallyidentified sequences [37], but only for high-abundance pep-tides (top three to top 20 peaks) in the initial MS and onlythose with sufficiently high abundance to provide substantial-ly complete b and y-fragment ion series in a single pass, andonly those in low charge states. De-novo sequencing is dra-matically more complicated and involves several runs at dif-ferent fragmentation energies with compilation of data fromthose runs. Typically, these fragments cannot be simply com-pared with those in a database. Yet this approach is requiredfor study of low-abundance species. Hence, low abundancespecies, as monitored by use of our method, cannot currentlybe sequenced by use of automated approaches, and in depth,de-novo analysis has been reserved solely for only those thatwere quantitatively different in these studies. However, theapproach can be applied to any species of any abundance ifit is of interest.

Tissue lipidomic approaches are also relatively new andhave been reported primarily for brain tissues [10–12]. Thesetypically use organic solvent extraction methods and direct

CH

P

O

HO

O

+H

N+

C-

N+

O

P OHO

OH

+Na

P

O

O

O

Phosphocholine headgroup

+H

N+

O

P OHO

O

OH

OH

sn-1 acyl loss

+H

O

HO

O

R1

[M + Na – 205]+

Proposed chemical structure of precursor m/z 544.33

[M + Na– H2O]+

[M + Na – 59.07]+ [M + Na]+

Fig. 4 Summed MS2 spectrum of a differently expressed sodiatedlysophosphatidylcholine (m/z 544.33, z=+1) (mass accuracy of theprecursor peak=51), with its collision product ions and the proposedchemical structures or the component lost on fragmentation. Mass

accuracies of individual product ions having m/z values of 86.09, 104.1,146.97, 184.06, 258.1, 339.26, 485.23 were 64, 59, 68, 42, 37, 63, and53 ppm, respectively

K. Kedia et al.8552

injection, although liquid chromatography coupled with MShas also been used [11]. Such methods have targeted thoselipids partitioned into the non-polar organic extract and assuch they are likely to sample a somewhat different set oflipids than those studied in this work [11]. Hence, the methodsshould be complementary.

In the approach described here a protein-depletion step wasused to remove highly abundant species and markedly reduceion suppression. We studied 7000–8000 unique, novel, low-abundance small proteins, peptides and lipids—thousandsmore than any other current method. Moreover, the great ma-jority of these molecules would not be surveyed by otherproteomic, peptidomic, or lipidomic approaches because ofmethodology differences. Hence, this approach adds substan-tially to the array of tissue molecules that can now be studiedbyMS. The reproducibility was sufficient [13] to enable label-free quantitative comparisons. Post-translational modifica-tions and modifications because of reactive oxygen speciesor alkylating agents can be observed. Although identificationor characterization can be applied to any of the species ob-served, automated sequencing procedures would not be ade-quate for most of the molecules seen in our studies, because ofinstrument limitations and the low concentrations of species.Having to perform intense, multi-stepped de-novo sequencingis not an advantage of this approach, but, with current MSinstrumentation, one would have to do this irrespective ofmethod for any of the low-abundance, smaller species in a cellor tissue.

The test of our approach was on placenta. The placenta isinteresting clinically but challenging anatomically. Earlierproteomic studies on the human placenta have investigatedabundant proteins, identifying many structural and house-keeping proteins, including actin, other cytoskeletal proteins,molecular chaperones, and proteins involved in transport [38].Broadly or narrowly, regional differences within the placentaitself do not seem to have been previously investigated by useof global proteomic methods but may be very important. Oth-er studies have looked selectively at trophoblast cells releasedfrom the placenta [39]. The human placenta is probably in-volved in disorders of pregnancy, including PE [16], but howis incompletely understood [40]. There is early evidence thatmaternal and fetal cells may participate in different ways in PE[19]. A more in-depth knowledge of the peptide and lipidcontent, by region, of the placenta would be likely to yieldadditional details of the pathologic pathways involved, espe-cially if the number of molecules investigated was greater. Our“omics” approach (with measurement of both peptides andmany lipids) studies many thousand additional small proteins,peptides and lipids in an organ which may participate mecha-nistically or be indicative of changes in tissue function.Changes in placental expression by region are likely to bemore dramatic in PE than in the unstimulated or non-pathologic placentae studied here. Having the ability to

regionally locate these molecular changes may contribute sub-stantially to our understanding of the placenta’s involvementin PE.

We studied two regions of the placenta: the chorionic plate,composed primarily of fetal cells, and the basal plate,consisting of a mixture of fetal (trophoblast) andmaternal cells[18]. It was recognized that most biomolecules were found inall cells and that with overlap of cell type in these two regionsnot many differences would be found. Despite this overlap,we observed 16molecular species with statistically significantdifferences between these two placental regions. This suggeststhe method was sufficiently robust and reproducible to over-come this challenge. Of these 16, six were found to be moreabundant in the chorionic plate; the other 10 were more abun-dant in the basal plate. These differences were consistentlyobserved in most of the 12 placentas characterized (Fig. 5).Fragmentation and characterization studies enabled successfulsequencing or chemical classification of 12 markers. Of these,four were peptides and the others were lipids.

One of the peptides was derived from fetal hemoglobin(Hgb γ) [41] and was more abundant in the chorionic plateof the placenta, which is predominantly of fetal origin. Anoth-er peptide, with an m/z of 614.38 and a charge state of +3 wasidentified as a peptide derived fromHgb alpha-1, and was alsomore abundant in the chorionic plate, probably in conjunctionwith the higher amount of Hgb γ present.

Another peptide (m/z=808.87, z=+2) was a fragment offibrinopeptide A, phosphorylated at serine 3 and more abun-dant in the chorionic plate. This agrees with previous research[42]. Fibrinopeptide A (FbA) was found in greater abundancein the placenta at term. It has also been shown that fetal fibrin-ogen contains twice as much phosphate as adult fibrinogen[42]. The biological significance of phosphorylation of fetalfibrinogen and of this particular peptide is not known, but ithas been proposed that phosphorylation of FbA at Ser-3 pro-duces a more efficient enzyme–substrate complex for thehighly regulated blood coagulation system, as is required dur-ing pregnancy [43]. The fourth peptide (m/z=760.4, z=+4)was a fragment of Hgb alpha-2. This peptide was also moreabundant in the chorionic plate. These peptide differencesmake physiologic sense. The two other peptides could notbe sequenced because of low concentration and/or highercharge state.

In addition, regional differences between placental lipids wereobserved. Most of these lipids were glycerophosphocholines,PC, although two were sphingomyelins, SM; all were moreabundant in the basal plate. A possible explanation ofthis might be that implantation of the blastocyst in theuterine wall is accompanied by transformation of theendometrial lining into decidua, becoming maternal cellsin the placenta and more evident in the basal plate [44].This region has a high lipid content [44] and may ex-plain higher concentrations of these in the basal plate.

Novel “omics” approach surveying low molecular weight molecules 8553

One limitation of all MS lipidomics is the inability toidentify the location of fatty acids in the lipids or thelocation of double bonds within the unsaturated fattyacid constituents. Because of the very low abundanceof our lipid regional markers, it was not possible touse other techniques, for example NMR, to identifythe location of double bonds in the fatty acid chains.Characterization of lipids and their fatty acids are aidedonly modestly by the very limited available lipid data-bases and lipidomics-focused software.

In this study we evaluated a novel “omics” approach forinvestigation of the low-molecular-weight species present in acomplex organ, the placenta, and identified subtle regionaldifferences between the chorionic plate, consisting primarilyof trophoblast and fetal cells, and the outer surface of thecotyledons, or basal plate, consisting of a mixture of fetaland decidual cells. This same approach could be used to in-vestigate regional changes that may be part of either normal ordisease processes. It is likely to be possible to determinewhether changes arise and progress within a fixed locationor whether they occur broadly across an organ. By use ofanimal models, it seems possible this approach could be usedto document the location and pathway of progression of spe-cific changes in organs.

Conclusion

Collectively, these results demonstrate the feasibility of usingour low-molecular-mass “omics” method to investigate thou-sands of low-abundance, low-molecular-weight species in tis-sue and to identify regional differences within the same organ,enabling study of temporal and spatial changes as part of bothnormal and pathologic processes. The method also enablesquantitative comparison across many tissues and identifica-tion of differently expressed molecules in a single approach.This method should substantially add to other proteomics orpeptidomics methods and enable more comprehensive analy-sis of tissues.

Acknowledgments This work was supported by the Department ofChemistry and Biochemistry, Brigham Young University. The authorswould like to extend their gratitude to several individuals who participat-ed in parts of this study: Dr Moana Hopoate-Sitake, Bruce Jackson, JodyJones, Dr M. Sean Esplin, and the Mass Spectrometry Facility at BYU.We gratefully acknowledge the support provided by IntermountainHealth Care (IHC) hospitals in making placental tissue samples available.

Conflict of interest The authors declare no competing financial con-siderations or other conflicts of interest.

References

1. Chen J, Kahne T, Rocken C, Gotze T, Yu J, Sung JJY, Chen M, HuP, Malfertheiner P, Ebert MPA (2004) Proteome analysis of gastriccancer metastasis by two-dimensional gel electrophoresis and ma-trix assisted laser desorption/ionization–mass spectrometry foridentification of metastasis-related proteins. J Proteome Res 3(5):1009–1016. doi:10.1021/pr049916l

2. Kim J, Kim SH, Lee SU, Ha GH, Kang DG, Ha NY, Ahn JS, ChoHY, Kang SJ, Lee YJ, Hong SC, Ha WS, Bae JM, Lee CW, KimJW (2002) Proteome analysis of human liver tumor tissue by two-dimensional gel electrophoresis and matrix assisted laser desorp-tion/ionization–mass spectrometry for identification of disease-related proteins. Electrophoresis 23(24):4142–4156

0

50

100

150

200

250

300

350

400

450

500

XIC

inte

nsity

0Basal plate

Line plot for peptide 718.38 (z=+2)

Chorionic plate

a

b

Fig. 5 Different expression of the peptide with m/z of 718.36 (z=+2). aLine plot of the abundance of the peptide in the basal versus chorionicplate of each of the 12 replicates, demonstrating the consistency of itsdifferent expression. Each colored line represents a single specimen. bBox and whisker plot for the same peptide and its mean abundance in thesame two regions of the 12 placentas

K. Kedia et al.8554

3. Washburn MP, Wolters D, Yates JR (2001) Large-scale analysis ofthe yeast proteome by multidimensional protein identification tech-nology. Nat Biotechnol 19(3):242–247

4. Kline KG, Wu CC (2009) MudPIT analysis: application to humanheart tissue. Methods Mol Biol 528:281–293

5. Catherman AD, Skinner OS, Kelleher NL (2014) Top Down pro-teomics: facts and perspectives. Biochem Biophys Res Commun445(4):683–693. doi:10.1016/j.bbrc.2014.02.041

6. Catherman AD, Durbin KR, Ahlf DR, Early BP, Fellers RT, TranJC, Thomas PM, Kelleher NL (2013) Large-scale top-down prote-omics of the human proteome: membrane proteins, mitochondria,and senescence.Mol Cell ProteomicsMCP 12(12):3465–3473. doi:10.1074/mcp.M113.030114

7. Heroux MS, Chesnik MA, Halligan BD, Al-Gizawiy M, ConnellyJM, Mueller WM, Rand SD, Cochran EJ, LaViolette PS, MalkinMG, Schmainda KM, Mirza SP (2014) Comprehensive character-ization of glioblastoma tumor tissues for biomarker identificationusing mass spectrometry-based label-free quantitative proteomics.Physiol Genomics 46(13):467–481

8. Wardman J, Fricker LD (2011) Quantitative peptidomics of micelacking peptide-processing enzymes. Methods Mol Biol 768:307–323

9. Che FY, Fricker LD (2005) Quantitative peptidomics of mousepituitary: comparison of different stable isotopic tags. J MassSpectrom 40(2):238–249

10. Tajima Y, Ishikawa M, Maekawa K, Murayama M, Senoo Y,Nishimaki-Mogami T, Nakanishi H, Ikeda K, Arita M, TaguchiR, Okuno A, Mikawa R, Niida S, Takikawa O, Saito Y (2013)Lipidomic analysis of brain tissues and plasma in a mouse modelexpressing mutated human amyloid precursor protein/tau forAlzheimer's disease. Lipids Health Dis 12:68. doi:10.1186/1476-511x-12-68

11. Chan RB, Oliveira TG, Cortes EP, Honig LS, Duff KE, Small SA,Wenk MR, Shui G, Di Paolo G (2012) Comparative lipidomicanalysis of mouse and human brain with Alzheimer disease. JBiol Chem 287(4):2678–2688. doi:10.1074/jbc.M111.274142

12. Korkes HA, Sass N, Moron AF, Câmara NOS, Bonetti T, CerdeiraAS, Da Silva IDCG, De Oliveira L (2014) Lipidomic assessment ofplasma and placenta of women with early-onset preeclampsia.PLoS ONE 9(10), e110747. doi:10.1371/journal.pone.0110747

13. Alvarez MT, Shah DJ, Thulin CD, Graves SW (2013) Tissue pro-teomics of the low-molecular weight proteome using an integratedcLC-ESI-QTOFMS approach. Proteomics 13(9):1400–1411

14. Esplin MS, Merrell K, Goldenberg R, Lai Y, Iams JD, Mercer B,Spong CY, Miodovnik M, Simhan HN, van Dorsten P,Dombrowski M (2010) Proteomic identification of serum peptidespredicting subsequent spontaneous preterm birth. Am J ObstetGynecol 204(5):391.e391–391.e398. doi:10.1016/j.ajog.2010.09.021

15. Merrell K, Thulin CD, EsplinMS, Graves SW (2009)An integratedserum proteomic approach capable ofmonitoring the lowmolecularweight proteome with sequencing of intermediate to large peptides.Rapid Commun Mass Spectrom 23(17):2685–2696

16. Roberts DJ, Post MD (2008) The placenta in pre-eclampsia andintrauterine growth restriction. J Clin Pathol 61(12):1254–1260.doi:10.1136/jcp.2008.055236

17. Prouillac C, Lecoeur S (2010) The role of the placenta in fetalexposure to xenobiotics: importance of membrane transportersand human models for transfer studies. Drug Metab Dispos38(10):1623–1635. doi:10.1124/dmd.110.033571

18. Benirschke K, Kaufmann P, Baergen RN (2006) Pathology of thehuman placenta, 5th edn. Springer, NY

19. Hwang HS, Park SH, Park YW, Kwon HS, Sohn IS (2010)Expression of cellular prion protein in the placentas of women withnormal and preeclamptic pregnancies. Acta Obstet Gynecol Scand89(9):1155–1161

20. Damsky CH, Fitzgerald ML, Fisher SJ (1992) Distribution patternsof extracellular matrix components and adhesion receptors are in-tricately modulated during first trimester cytotrophoblast differenti-ation along the invasive pathway, in vivo. J Clin Invest 89(1):210–222

21. Reimel BA, Pan S, May DH, Shaffer SA, Goodlett DR, McIntoshMW, Yerian LM, Bronner MP, Chen R, Brentnall TA (2009)Proteomics on fixed tissue specimens—a review. Curr Proteomics6(1):63–69

22. Robinson JM, Vandre DD, Ackerman WE (2009) Placental prote-omics: a shortcut to biological insight. Placenta 30(9):13

23. Centlow M, Hansson SR, Welinder C (2009) Differential proteomeanalysis of the preeclamptic placenta using optimized protein ex-traction. J Biomed Biotechnol 2010 (2010). doi:10.1155/2010/458748

24. Wang F, Shi Z, Wang P, You W, Liang G (2013) Comparativeproteome profile of human placenta from normal and preeclampticpregnancies. PLoS ONE 8(10):2013

25. Dunn WB, Brown M, Worton SA, Crocker IP, Broadhurst D,Horgan R, Kenny LC, Baker PN, Kell DB, Heazell AE (2009)Changes in the metabolic footprint of placental explant-conditioned culture medium identifies metabolic disturbances relat-ed to hypoxia and pre-eclampsia. Placenta 30(11):974–980

26. Andersen HU, Fey SJ, Larsen PM, Nawrocki A, Hejnaes KR,Mandrup-Poulsen T, Nerup J (1997) Interleukin-1beta inducedchanges in the protein expression of rat islets: a computerized da-tabase. Electrophoresis 18(11):2091–2103. doi:10.1002/elps.1150181136

27. Schröder S, Zhang H, Yeung ES, Jänsch L, Zabel C, Wätzig H(2008) Quantitative gel electrophoresis: sources of variation. JProteome Res 7(3):1226–1234. doi:10.1021/pr700589s

28. Mazur MT, Cardasis HL, Spellman DS, Liaw A, Yates NA,Hendrickson RC (2010) Quantitative analysis of intact apolipopro-teins in human HDL by top-down differential mass spectrometry.Proc Natl Acad Sci U S A 107(17):7728–7733

29. Ma B, Zhang K, Hendrie C, Liang C, Li M, Doherty-Kirby A,Lajoie G (2003) PEAKS: powerful software for peptide de novosequencing by tandem mass spectrometry. Rapid Commun MassSpectrom 17(20):2337–2342

30. Xu F, Zou L, Lin Q, Ong CN (2009) Use of liquid chromatography/tandemmass spectrometry and online databases for identification ofphosphocholines and lysophosphatidylcholines in human red bloodcells. Rapid Commun Mass Spectrom 23(19):3243–3254

31. Al-Saad KA, Siems WF, Hill HH, Zabrouskov V, Knowles NR(2003) Structural analysis of phosphatidylcholines by post-sourcedecay matrix-assisted laser desorption/ionization time-of-flightmass spectrometry. J Am Soc Mass Spectrom 14(4):373–382

32. Shi Z, Long W, Zhao C, Guo X, Shen R, Ding H (2013)Comparative proteomics analysis suggests that placental mitochon-dria are involved in the development of pre-eclampsia. PLoS ONE8(5):2013

33. DeSouza L, Diehl G, Rodrigues MJ, Guo J, Romaschin AD,Colgan TJ, Siu KW (2005) Search for cancer markers from endo-metrial tissues using differentially labeled tags iTRAQ and cICATwith multidimensional liquid chromatography and tandem massspectrometry. J Proteome Res 4(2):377–386

34. Kelleher NL (2004) Top-down proteomics. Anal Chem 76(11):197A–203A

35. Gygi SP, Rist B, Griffin TJ, Eng J, Aebersold R (2002) Proteomeanalysis of low-abundance proteins using multidimensional chroma-tography and isotope-coded affinity tags. J Proteome Res 1(1):47–54

36. Xu Z, Wu C, Xie F, Slysz GW, Tolic N, Monroe ME, Petyuk VA,Payne SH, Fujimoto GM, Moore RJ, Fillmore TL, Schepmoes AA,Levine DA, Townsend RR, Davies SR, Li S, Ellis M, Boja E,Rivers R, Rodriguez H, Rodland KD, Liu T, Smith RD (2014)Comprehensive quantitative analysis of ovarian and breast cancer

Novel “omics” approach surveying low molecular weight molecules 8555

tumor peptidomes. J Proteome Res 14(1):422–433. doi:10.1021/pr500840w

37. Michalski A, Cox J, Mann M (2011) More than 100,000 detectablepeptide species elute in single shotgun proteomics runs but themajority is inaccessible to data-dependent LC-MS–MS. JProteome Res 10(4):1785–1793. doi:10.1021/pr101060v

38. Paule S, Li Y, Nie G (2011) Cytoskeletal remodelling proteins iden-tified in fetal–maternal interface in pregnant women and rhesusmonkeys. J Mol Histol 42(2):161–166

39. Orendi K, Kivity V, Sammar M, Grimpel Y, Gonen R, Meiri H,Lubzens E, Huppertz B (2011) Placental and trophoblastic in vitromodels to study preventive and therapeutic agents for preeclampsia.Placenta 2011(32):023

40. Mutter WP, Karumanchi SA (2008) Molecular mechanisms of pre-eclampsia. Microvasc Res 75(1):1–8

41. Bauer DE, Orkin SH (2011) Update on fetal hemoglobin gene reg-ulation in hemoglobinopathies. Curr Opin Pediatr 23(1):1–8. doi:10.1097/MOP.0b013e3283420fd0

42. Seydewitz HH,Witt I (1989) The fraction of high molecular weight(HMW) fibrinogen and phosphorylated fibrinopeptide A in fetalfibrinogen. Thromb Res 55(6):785–790

43. Maurer MC, Peng JL, An SS, Trosset JY, Henschen-Edman A, Scheraga HA (1998) Structural examinationof the influence of phosphorylation on the binding offibrinopeptide A to bovine thrombin. Biochemistry37(17):5888–5902

44. Lu G, Zhu S, Ke Y, Jiang X, Zhang S (2013) Transplantation-potential-related biological properties of decidua basalis mesenchy-mal stem cells frommaternal human term placenta. Cell Tissue Res352(2):301–312

K. Kedia et al.8556