Proteome Mapping of Human Skim Milk Proteins in Term and Preterm Milk † † ‡ § § Claire E

Article

pubs.acs.org/jpr

Proteome Mapping of Human Skim Milk Proteins in Term and Preterm Milk † † ‡ § § Claire E. Molinari,*, Ylenia S. Casadio, Ben T. Hartmann, Andreja Livk, Scott Bringans, † † Peter G. Arthur, and Peter E. Hartmann † School of Chemistry and Biochemistry, The University of Western Australia, Crawley, 6009, Australia ‡ Perron Rotary Express Milk Bank (PREM Bank) Neonatal Paediatrics, King Edward Memorial Hospital, Subiaco, 6008, Australia § Proteomics International, Perth, Western Australia, Australia

*S Supporting Information

ABSTRACT: The abundant proteins in human milk have been well characterized and are known to provide nutritional, protective, and developmental advantages to both term and preterm infants. However, relatively little is known about the expression of the low abundance proteins that are present in human milk because of the technical difficulties associated with their detection. We used a combination of electrophoretic techniques, ProteoMiner treatment, and two-dimensional liquid chromatography to examine the proteome of human skim milk expressed between 7 and 28 days postpartum by healthy term mothers and identified 415 in a pooled milk sample. Of these, 261 were found in human skim milk for the first time, greatly expanding our understanding of the human skim milk proteome. The majority of the proteins identified were involved in either the immune response (24%) or in cellular (28%) or protein (16%) metabolism. We also used iTRAQ analysis to examine the effects of premature delivery on milk protein composition. Differences in protein expression between pooled milk from mothers delivering at term (38−41 weeks gestation) and preterm (28−32 weeks gestation) were investigated, with 55 proteins found to be differentially expressed with at least 90% confidence. Twenty-eight proteins were present at higher levels in preterm milk, and 27 were present at higher levels in term milk. KEYWORDS: human milk, protein, proteomics, ProteoMiner, iTRAQ, 2D LC−MS

■ INTRODUCTION human milk be characterized. There are two main reasons for The importance of human milk proteins to the growth and this. First, it is possible that these proteins play significant roles development of breastfed infants is well established. They not in infant growth and development. Second, knowledge of how only provide a digestible source of amino acids to infants, but the expression of low abundance proteins differs between term also confer immunological protection and perform developmen- and preterm milk may be useful diagnostically, as a reflection of tal and regulatory functions, exerting both long and short-term the developmental changes occurring in the mammary gland benefits compared to formula feeding.1,2 during pregnancy and lactation. Human milk proteins are particularly important for infants Historically, there have been a number of technical who are born prematurely. Recent studies stress the importance challenges associated with characterizing the low abundance of both the total amount of protein and the ratio of protein/ proteins in human milk. Initial studies using gel electrophoretic energy that preterm infants receive for their growth and devel- methods coupled with mass spectrometry were unable to detect opment.3 Significantly, the protein composition of milk from more than 10 different gene products in either human or bovine milk, despite observing hundreds of distinct protein preterm mothers is known to differ from that of term mothers. 11−13 The concentration of total protein is typically higher in preterm spots. This difficulty results from the fact that six proteins, α β milk;4 however, while some individual proteins are expressed at -lactalbumin, -casein, secretory immunoglobulin A, lysozyme, higher levels in preterm milk, others are present at lower con- lactoferrin, and secretory component, constitute over 90% of − 14 centrations.5 7 the total protein content in mature human milk, obscuring Hitherto, most studies investigating milk protein composi- the detection of less abundant proteins of potential biological tion have focused upon the most abundant proteins present, interest. resulting in their relative concentrations in term and preterm − milk being well-defined.8 10 However, it is also important that Received: September 2, 2011 the identity and behavior of the lower abundance proteins in Published: February 7, 2012

More recently, proteomic studies employing strategies to (7−14 days lactation) and 16 term mothers (7−14 days lactation) deplete the high abundance proteins or to extensively frac- were used. All milk samples were thawed, pooled, and centri- tionate the sample prior to analysis have been more successful fuged at 10000g for 10 min to remove the cream layer. A at identifying low abundance proteins than the earlier gel-based − mammalian protease inhibitor cocktail containing 4-(2- methods.14 17 Two studies in particular have identified a large aminoethyl)benzenesulfonyl fluoride, E-64, bestatin, leupeptin, number of low abundance human milk proteins. Palmer et al.18 aprotinin, and sodium EDTA was added to each pooled sample, used immunodepletion columns to deplete the five most which was then depleted of casein using a previously described − 22 abundant proteins in colostrum prior to 2D LC MS/MS an- method. Briefly, to deplete the samples of casein, CaCl2 was alysis and were able to identify 151 proteins. More recently, added to a concentration of 60 mM, and the pH was adjusted 19 Liao et al. employed combinatorial hexapeptide ligand to pH 4.3. Samples were then centrifuged at 189000g at 4 °C libraries (ProteoMiner) to enrich the low abundance milk for 60 min, and the supernatant was collected. proteins before analysis by LC−MS/MS and identified 115 For the ProteoMiner-treated samples (Figure 1), casein- 19 proteins. Liao et al. also showed that many of these proteins depleted skim milk was dialyzed against 10 volumes of 10 mM change in expression over the course of 12 months of lactation, Tris, pH 7 at 4 °C, with three buffer changes at two-hour inter- highlighting the dynamic nature of milk composition. One of vals using dialysis tubing with a MW cutoff of 3500 Da the advantages of using a ProteoMiner bead approach is that it (Spectrapor Membrane Tubing, Spectrum Medical Industries, does not require tailored antibodies or optimization. When a Rancho Domingues, CA). The samples were then lyophilized sample is applied to the ligand library, the high abundance pro- and reconstituted in water. The samples were analyzed for their teins saturate their ligands and the excess remains unbound, protein content and diluted such that 1 mL of 50 g/L protein whereas the lower abundance proteins bind completely, result- solution was loaded onto the hexaligand library beads ing in an overall compression of the dynamic range. Recent (ProteoMiner Large Capacity Protein Enrichment Kit, BioRad, studies have also shown the ProteoMiner treatment to be com- Gladesville, NSW, Australia), according to the manufacturer’s patible with downstream quantitative analyses of low abundance 20,21 instructions. Briefly, after swelling the beads using the buffer proteins. provided, the samples were loaded onto individual ProteoMiner The aim of the present study was two-fold. First, we aimed to columns and rocked for 6 h at room temperature. For the further characterize the proteome of mature human skim milk pooled term milk samples (collected 15−28 days postpartum), from established lactation (7−28 days postpartum), using a the unbound proteins were then washed through the column, combination of ultracentrifugation and ProteoMiner enrich- and the bound proteins were eluted using the acidic elution ment to compress the dynamic range of the proteome prior to buffer provided in the kit. For the pooled term and preterm analysis by 2D LC−MS/MS. Second, we sought to quanti- milk samples to be subsequently labeled with iTRAQ reagents tatively examine whether there are differences in protein − expression between term and preterm milk that reflect the (collected 7 14 days postpartum), the bound proteins were physiological and metabolic effects of preterm delivery upon eluted sequentially using four different buffers: 1 M sodium the mammary gland. chloride in 20 mM HEPES (pH 7.0), 0.2 M glycine (pH 2.4), 60% (w/v) ethylene glycol in water, and 33.3% (v/v) 2-propanol, ■ EXPERIMENTAL PROCEDURES 16.7% (v/v) acetonitrile, 0.1% (v/v) trifluoroacetic acid. Protein Concentration Determination Materials Protein concentrations of human milk samples were deter- Unless otherwise stated, all chemicals and reagents were ob- mined using a Bicinchoninic acid kit (Sigma-Aldrich, NSW, tained from Sigma-Aldrich (NSW, Australia). Australia), as described in a previous study.23 Sample Collection Sodium Dodecyl Sulfate Polyacrylamide Gel Term and preterm milk samples were obtained from healthy Electrophoresis (SDS-PAGE) lactating mothers at King Edward Memorial Hospital, Subiaco, Western Australia. Participating term mothers had delivered SDS-PAGE analysis was conducted using the HOEFER gel between 38 and 41 weeks of gestation, and their infants had a apparatus (HOEFER Scientific Instruments, San Francisco, chronological age of between 7 and 28 days at the time of CA) and the Laemmli gel system using 12.5% polyacrylamide gels.24 Gels were run using a constant current of 15 mA for sample collection. Participating preterm mothers had delivered ° between 28 and 32 weeks of gestation, and their infants had a 16 h at 4 C, fixed for 2 h in 50% methanol/10% trichloroacetic chronological age of between 7 and 14 days at the time of acid, destained using double deionized water, stained using sample collection. All donors gave written informed consent for Coomassie Brilliant Blue R-250 overnight, and scanned using their donations to be used in this research, and this study was an Epson Perfection V700 photographic flatbed color image approved by the University of Western Australia, Human scanner (Epson, Nagano, Japan). The intensity of protein bands were measured using the open access software package Research Ethics Committee and King Edward Memorial Hospital, 25 Human Ethics Research Committee. All samples were frozen at Image J 1.410. −20 °C within an hour of expression and transferred to −80 °C Differential in-Gel Electrophoresis (DIGE) storage within 3 days. Cy5 and Cy3 activated ester dyes were purchased from Sample Treatment Lumiprobe (Lumiprobe Corp, FL). Skim milk samples before The sample analysis workflow is described in Figure 1. For and after casein depletion and ProteoMiner treatment were the purposes of protein identification in mature term milk, each labeled with one of the dyes. The DIGE experiment was milk samples collected from 8 mothers (15−28 days lactation) conducted in duplicate, and the order of the dyes were swapped were pooled. For the quantitative comparison between for the duplicate experiment. Fifty micrograms of protein sample term and preterm milk, milk samples from 16 preterm mothers was labeled according to the CyDye DIGE Flours (minimal dyes)

1697 dx.doi.org/10.1021/pr2008797 | J. Proteome Res. 2012, 11, 1696−1714 Journal of Proteome Research Article

Figure 1. Experimental workflow. for Ettan DIGE system protocol (GE Healthcare, Pittsburgh, PA), LC−MS/MS and the samples were mixed together. Sample Preparation. Protein samples were precipitated by Isoelectric focusing (IEF) and strip equilibration was performed adding five volumes of cold acetone to the treated samples according to the Ettan DIGE system protocol (GE Healthcare), (Figure 1), incubating for 1 h at −20 °C, and pulse centrifuging using ready-to-use Immobiline DryStrip gel strips, linear pH for 5−10 s. The protein pellets were resuspended in 0.5 M gradient 3−10, length 18 cm (GE Health Care) and an triethylammonium bicarbonate (TEAB) (pH 8.5) by shaking IPGphor isoelectric focusing unit (IPGPhor, GE Health Care). before reduction and alkylation according to the iTRAQ protocol (Applied Biosystems, Foster City, CA). A total of The strips were actively rehydrated for 10 h at a constant volt- μ μ age of 200 V and a constant temperature of 20 °C. One hundred 55 g of each sample was digested overnight with 5.5 g trypsin at 37 °C in 0.5 M TEAB. micrograms of protein was loaded. Following rehydration, the 1D-LC. Peptides were separated on a C18 PepMap100, strips were exposed to a linear increase to 1000 V over 4 h 3 μm column (LC Packings, Sunnyvale, CA) with a gradient of followed by 8000 V until a total of 45000 V hr was reached. 10−45% acetonitrile, 0.1% trifluoroacetic acid over 165 min, The second dimension separation was carried out in the dark using the Ultimate 3000 nano HPLC system (LC Packings- according to the method of Lui, Lipscombe, and Arthur.26 Gels Dionex). Every 30 s, the eluent was mixed with matrix solution were imaged using a Typhoon Trio scanner (GE Health Care), (5 mg/mL CHCA) and spotted onto a 384 well Opti-TOF with the Cy5 and Cy3 labeled samples visible using 633/670 nm plate (Applied Biosystems) using a Probot Micro Fraction and 532/580 nm excitation/emission filters, respectively. Gels Collector (LC Packings-Dionex). μ were post stained for total protein content using Coomassie 2D-LC. Peptides were desalted on a Strata-X 33 m poly- Brilliant Blue. DIGE gels were analyzed using the Progenesis meric reversed phase column (Phenomenex) and dissolved in a SameSpots software package (Nonlinear Dynamics Ltd., New- buffer containing 10 mM potassium hydrogen phosphate, pH 3 in 10% acetonitrile, before separation by strong cation exchange castle upon Tyne, U. K.). Spots with a normalized spot volume chromatography on an Agilent 1100 HPLC system (Agilent of less than 500 were excluded from the analysis. All data is Technologies, Palo Alto, CA) using a PolySulfethyl column ± presented as mean SEM. (4.6 × 100 mm, 5 μm, 300 Å, Nest Group, Southborough, For the purposes of downstream mass spectrometry analysis, MA). Peptides were eluted with a linear gradient of 0−400 mM a preparative 2D gel analysis was also conducted. Five hundred KCl. Eight fractions containing the peptides were collected and micrograms of the ProteoMiner-treated skim milk sample was desalted on Strata-X columns. Each peptide fraction was then loaded, and the first and second dimensions were carried out as separated and spotted onto a 384-well Opti-TOF plate above. Gels were fixed and stained as described above. Co- according to the 1D-LC protocol described above, excepting − omassie stained bands and spots of interest were cut from the that a 10 40% acetonitrile (0.1% trifluoracetic acid) gradient gel, destained, and digested as described by Shevchenko et al.27 was used. iTRAQ. The tryptic digests were dried in a SpeedVac, re- Mass spectrometry was conducted as described in a pre- suspended in 30 μL of 0.5 M TEAB, and labeled by adding vious study using an UltraFlex MALDI-TOF/TOF instrument 23 iTRAQ reagents to preterm and term milk samples, respectively, (Bruker Daltonics, Bremen, Germany). MS/MS data was im- according to the iTRAQ protocol (Applied Biosystems). For ported into the database search engine Mascot (Version 2.3.01, the iTRAQ experiment comparing term and preterm milk www.matrixscience.com) and searched against the Swiss-Prot samples without ProteoMiner treatment (Figure 1), duplicate Mammalia database (49 887 sequences). pooled preterm milk samples were labeled with iTRAQ reagents

1698 dx.doi.org/10.1021/pr2008797 | J. Proteome Res. 2012, 11, 1696−1714 Journal of Proteome Research Article

114 and 116. Duplicate pooled term milk samples were labeled of the parameter, iTRAQ 4plex (peptide labeled) modification, with iTRAQ reagents 115 and 117. For the iTRAQ experiment and the Quantitate tab checked. MS/MS spectra were searched comparing term and preterm milk samples after ProteoMiner against the Swiss-Prot human genomic database (2011_5). For treatment (Figure 1), duplicate pooled preterm milk samples quantitation analysis, the duplicates were analyzed separately. were labeled with iTRAQ reagents 116 and 117. Duplicate Average protein ratios and p-values to indicate significant differ- pooled term milk samples were labeled with iTRAQ reagents ential expression were calculated by the software. To be con- 114 and 115. Excess iTRAQ reagent was quenched by adding sidered as being differentially expressed, proteins were required 1 mL of water; the samples were then combined, desalted on a to have an unused protein score greater than 1.3, corresponding Strata-X 33 μm polymeric reverse phase column (Phenomenex, to a confidence interval of 95%, and have significantly different Torrance, CA), and analyzed using the 2D-LC protocol de- protein ratios in both replicates, also at a confidence level of scribed above. 95% (p < 0.05). Identified proteins for which a difference was MALDI-MS/MS. Peptides were analyzed on a 5800 MALDI- found at a confidence level of 90% (p < 0.1) in both replicates TOF/TOF mass spectrometer (Applied Biosystems) operated were also reported. The p values represent the variation in the in reflector positive mode. MS data were acquired over a mass reported iTRAQ ratios for all the peptides of the associated range of 800−4000 m/z, and for each spectrum, a total of 400 protein and do not relate to either biological variation or shots were accumulated. A job-wide interpretation method technical reproducibility. selected the 20 most intense precursor ions above a signal/ The false discovery rate was less than 1%, calculated using a noise ratio of 20 from each spectrum for MS/MS acquisition database containing reversed sequences. In order to categorize but only in the spot where their intensity was at its peak. MS/ the identified proteins, the results were analyzed using the MS spectra were acquired with 4000 laser shots per selected ion software program IPA (Ingenuity Databases) and the UniProt with a mass range of 60 to the precursor ion −20. Database release 2011_6 (http://www.uniprot.org/). Results were compared to a recent comprehensive review publication28 Data Analysis. Protein identification was performed using 19 ProteinPilot 4.0.8085 Software (Applied Biosystems). MS/MS and a subsequent research paper in order to determine which proteins had not been previously identified. spectra were searched against the Swiss-Prot human genomic database (2011_3 for the 2D-LC analysis, 2011_5 for the 1D- ■ RESULTS LC analysis). Search parameters were as follows: Cys alkylation, MMTS; Digestion, trypsin; Instrument, 5800; Special factors, Casein Depletion none; Species, none; Quantitate tab, unchecked; Detected The SDS-PAGE gel analysis showed that casein depletion re- protein threshold (unused ProtScore), 1.3, which corresponds sulted in the removal of 67 ± 3% (n = 5) of the 30 kDa β-casein to proteins identified with <95% confidence. band present in the pooled skim milk sample (Figure 2). The κ- For the iTRAQ experiment, MS/MS spectra were analyzed casein protein band at 38 kDa and a number of other β-casein as above with ProteinPilot 4.0.8085 software, with the addition bands in the 20−30 kDa region present in the skim milk sample

Figure 2. SDS-PAGE electrophoretograms and mass spectrometry identifications of milk protein fractions. Proteins (15 μg) from (A) skim milk, (B) skim milk after ProteoMiner treatment, (C) skim milk after casein depletion, and (D) skim milk after casein depletion and ProteoMiner treatment were analyzed (n = 4, 2 depicted). Significant protein identifications obtained using MALDI MS/MS are displayed adjacent to the corresponding band of the casein-depleted ProteoMiner-treated skim milk.

1699 dx.doi.org/10.1021/pr2008797 | J. Proteome Res. 2012, 11, 1696−1714 Journal of Proteome Research Article also appear to be present at considerably lower levels in the casein-depleted sample (Figure 2). ProteoMiner Treatment: SDS-PAGE The ProteoMiner treatment compressed the dynamic range of the proteins present in human milk. However, ProteoMiner treatment resulted in a far greater enrichment of lower abundance proteins when used after casein depletion, compared to when it was performed on the delipidated skim milk alone (Figure 2). As a result, all samples were depleted of casein prior to further treatment and analysis (Figure 1). In the SDS-PAGE analysis of the casein-depleted protein sample, the five most intense protein bands constituted 52% before ProteoMiner treatment and 24% after treatment (p < 0.001) (Figure 2). The major protein bands of β-casein (30 kDa), α-lactalbumin (14 kDa), serum albumin (65 kDa), and sIgA α-chain C region (60 kDa) were all depleted after the combinatorial ligand library treatment (p < 0.002) (Figure 2). It is also possible to see the resulting enrichment of lower abundance proteins. The xanthine dehydrogenase band at ∼170 kDa was enriched by 5.4 fold (p < 0.001), and there were a number of additional bands seen in the 30−50 kDa range and between 20 and 25 kDa (Figure 2). ProteoMiner Treatment: 2D-DIGE The compression of the protein dynamic range caused by the casein depletion and ProteoMiner treatment was visible to a greater extent in the 2D-DIGE analysis. There were 308 and 320 spots present on the duplicate skim milk gels, compared to 359 and 364 spots on the duplicate gels of the casein-depleted, ProteoMiner-treated samples (Figure 3). Over two-thirds (68 ± 5%) of the spots were present at a relatively higher intensity after casein depletion and ProteoMiner treatment compared to in the skim milk. Similarly, 28 ± 3% of the spots in the treated sample occupied a 2-fold greater percentage intensity, and 20 ± 0.5% of spots were present at a 3-fold higher percentage intensity relative to the skim milk. In the skim milk gels, 60 ± 4% of the total gel intensity was occupied by the β-casein and α-lactalbumin spot clusters at 25− 30 kDa and 14.4 kDa, respectively. Casein depletion and ProteoMiner treatment reduced the dominance of the abundant proteins, with only 40 ± 3% of the total gel intensity being occupied by the same β-casein and α-lactalbumin spots after Figure 3. Two-dimensional fluorescent differential in-gel expression treatment (Figure 3). In the skim milk gels, 54 ± 0.1% of spots (DIGE) experiment. (A) Skim milk protein fraction, (B) skim milk were present at very low levels, each occupying less than 0.04% protein fraction after casein depletion and ProteoMiner treatment. of the total gel intensity. In the treated sample, however, only The gels depicted are one pair out of a set of duplicate experiments. 29 ± 0.5% of the spots on the gel were present at an intensity less than this 0.04% threshold. analysis, and 29 proteins being identified in the iTRAQ experi- For the preparative 2DE analysis, 110 spots were processed ment without ProteoMiner treatment (Figure 4A,B). With for MALDI-MS/MS analysis, and 61 of these spots were regard to the stage of lactation, 174 proteins were identified in both the pooled term milk collected 7−14 days postpartum and positively identified. Many of the spots that were not identified − were of extremely low abundance and were only faintly visible in the pooled term milk collected 15 28 days postpartum. There using Coomassie Brilliant Blue staining. There was a great deal were 141 proteins unique to the pool collected earlier in lactation, and 100 proteins were identified only in the pooled term milk of redundancy, with the 61 spots corresponding to only − 21 gene products (Supporting Information Data Files 2 and 3). collected 15 28 days postpartum (Figure 4C). The 415 proteins were categorized by both their subcellular location and their Protein Identifications functions, according to their annotations in the UniProt Database A total of 415 proteins were identified at a confidence level of release 2011_6 (http://www.uniprot.org/) (Figure 5). The 95% in human milk collected between 7 and 28 days post- majority of the proteins found in the present study were cyto- partum from term mothers, 261 of which had not previously plasmic (46%) or extracellular space proteins (38%). Functionally, been identified in human skim milk (Table 1). With regard to the majority of the proteins identified were involved in either in methodology, the majority of these proteins were identified in the immune response (24%) or in cellular metabolism and the 2D LC−MS/MS analyses after ProteoMiner treatment, cellular growth (28%). See Supporting Information Data File 4 with 15 proteins being identified in only the 1D LC−MS/MS for a description of how each protein was categorized.

1700 dx.doi.org/10.1021/pr2008797 | J. Proteome Res. 2012, 11, 1696−1714 Journal of Proteome Research Article Table 1

protein name accession no. unuseda % covb peptides methodc 1 14-3-3 protein epsilond P62258 2 20 3 3, 4 2 14-3-3 protein gammad P61981 2 20.2 2 4 3 14-3-3 protein zeta/delta P63104 11.01 40.4 6 3, 4, 6 4 40S ribosomal protein S10d P46783 6 27.9 4 6 5 40S ribosomal protein S12d P25398 1.68 17.4 1 4 6 40S ribosomal protein S14d P62263 2.88 22.5 2 4, 6 7 40S ribosomal protein S19d P39019 19.09 55.9 9 4, 6 8 40S ribosomal protein S20d P60866 6.04 37.8 4 6 9 40S ribosomal protein S25d P62851 4.63 27.2 3 4, 6 10 40S ribosomal protein S28d P62857 1.42 17.4 1 5 11 40S ribosomal protein S3d P23396 3.24 14.4 3 4, 6 12 40S ribosomal protein S3ad P61247 4.25 16.3 3 6 13 40S ribosomal protein S5d P46782 1.44 4.4 1 6 14 40S ribosomal protein SAd P08865 1.63 8.5 3 6 15 45 kDa calcium-binding proteind Q9BRK5 6 9.7 4 4, 6 16 4F2 cell surface antigen heavy chaind P08195 1.55 1.6 1 6 17 6-phosphogluconate dehydrogenase, decarboxylatingd P52209 5.67 21.7 7 3, 4, 6 18 6-phosphoglucolactonased O95336 4.54 22.1 2 6 19 60S acidic ribosomal protein P0-liked Q8NHW5 2 15.1 3 4 20 60S acidic ribosomal protein P1 P05386 1.82 14 1 6 21 60S acidic ribosomal protein P2d P05387 2 28.7 1 4 22 60S ribosomal protein L12d P30050 2 22.4 2 6 23 60S ribosomal protein L23d P62829 1.82 20.7 2 6 24 60S ribosomal protein L24d P83731 3.5 22.9 2 6 25 60S ribosomal protein L31d P62899 2.02 22.4 2 6 26 78 kDa glucose-regulated protein P11021 23.53 41.3 19 2, 4, 6 27 Acetyl-CoA acetyltransferase, cytosolicd Q9BWD1 4 13.9 2 4, 6 28 Acid sphingomyelinase-like phosphodiesterase 3b Q92485 5.09 12.1 4 4 29 Actin-related protein 2/3 complex subunit 1Bd O15143 2 8.3 2 4, 6 30 Actin, cytoplasmic 2 P63261 25.01 57.3 36 1, 2, 3, 4, 5, 6 31 Acyl-CoA binding proteind P07108 3.07 20.7 2 5 32 ADP-ribosylation factor 1d P84077 2 17.7 1 4 33 ADP ribosylation factor 5d P84085 1.66 9.4 2 6 34 Adenine phosphoribosyltransferased P07741 1.44 18.9 3 6 35 Alanine aminotransferase 1d P24298 2.02 3.4 2 6 36 Alcohol dehydrogenase [NADP+] P14550 12.62 35.7 8 4, 6 37 Alpha-(1,3)-fucosyltransferased P51993 2.24 12 2 6 38 Alpha-1-acid glycoprotein P02763 7.07 28.9 6 5 39 Alpha-1-antichymotrypsin P01011 22.54 39.2 21 4, 5, 6 40 Alpha-1-antitrypsin P01009 44.3 61.7 70 3, 4, 5, 6 41 Alpha-1B-glycoprotein P04217 4.47 11.5 2 6 42 Alpha-2-antiplasmind P08697 5.32 10 4 4, 6 43 Alpha-2-HS-glycoprotein P02765 7.07 20.4 5 4 44 Alpha-2-macroglobulind P01023 7.93 7.3 6 4, 6 45 Alpha-actinin-4 O43707 2 8 1 3 46 Alpha-aminoadipic semialdehyde dehydrogenased P49419 2 5.2 1 4 47 Alpha-amylase 2Bd P19961 2.01 5.9 2 6 48 Alpha-enolase P06733 18.01 45.2 14 2, 3, 4, 5, 6 49 Alpha-lactalbumin P00709 31.46 82.4 71 1, 2, 3, 4, 5, 6 50 Alpha-S1-casein P47710 38.86 81.1 71 1, 2, 3, 4, 5, 6 51 Amyloid beta A4 protein P05067 4.02 16.2 6 4, 6 52 Angiopoietin-related protein 4d Q9BY76 7.87 20.4 6 4, 6 53 Antileukoproteinased P02647 3.88 37.9 3 4, 6 54 Antithrombin-III P01008 9.92 39.4 9 4, 6 55 Apolipoprotein A-I P02647 29.14 67.8 30 1, 2, 3, 4, 5, 6 56 Apolipoprotein A-II P02652 9.75 41 9 3, 4, 6 57 Apolipoprotein A-IV P06727 17.91 54.3 14 3, 4, 6 58 Apolipoprotein B-100d P04114 63.34 13.2 32 3, 6 59 Apolipoprotein D P05090 8.43 30.7 7 4, 6 60 Apolipoprotein E P02649 15.73 46.1 13 3, 4 61 Arylsulfatase Ad P15289 3.42 9.9 3 4, 6

1701 dx.doi.org/10.1021/pr2008797 | J. Proteome Res. 2012, 11, 1696−1714 Journal of Proteome Research Article Table 1. continued

protein name accession no. unuseda % covb peptides methodc 62 Beta-1,4-galactosyltransferase 1 P15291 15.28 53.8 24 3, 4, 6 63 Beta-2-microglobulin P61769 7.41 32.8 5 4, 5 64 Beta-casein P05814 140.3 82.3 195 1, 2, 3, 4, 5, 6 65 Bifunctional purine biosynthesis protein PURHd P31939 2 2.7 2 6 66 Bile salt-activated lipase P19835 64.64 52.5 96 1, 3, 4, 5, 6 67 Biotinidase P43251 2.99 12.3 6 4, 5 68 Bone morphogenetic protein 1d P13497 2.01 2.5 2 6 69 Butyrophilin subfamily 1 member A1 Q13410 39.81 45.8 48 1, 3, 4, 5, 6 70 C-1-tetrahydrofolate synthase, cytoplasmicd P11586 3.83 6.2 2 6 71 C-type lectin domain family 11 member A Q9Y240 4 16.1 2 6 72 C−C motif chemokine 28d Q9NRJ3 2 11.8 1 4 73 C-X-C motif chemokine 2 P19875 4.28 38.3 3 6 74 C4b-binding protein alpha chain P04003 9.6 17.3 10 4, 6 75 Cadherin-1d P12830 3.18 7.8 5 4, 6 76 Calmodulin P62158 7.74 38.3 4 4, 6 77 Calmodulin-like protein 5d Q9NZT1 4.57 37 4 3, 4, 6 78 Calreticulin P27797 7.74 17.3 5 4, 6 79 Calumenind O43852 2 7 2 4 80 Carbonic anhydrase 2d P00918 2.08 16.9 2 6 81 Carbonic anhydrase 6 P23280 19.66 46.8 19 3, 4, 6 82 Carbonyl reductase [NADPH] 3d O75828 1.66 9.7 3 4, 6 83 Carboxypeptidase B2d Q96IY4 2 3.8 1 4 84 Cathepsin Bd P07858 15.51 42.8 14 4, 6 85 Cathepsin S P25774 4.12 13.3 4 4 86 Cation-independent mannose-6-phosphate receptord P11717 1.8 2.1 2 4, 6 87 CD9 antigen P21926 2.23 16.7 13 5 88 CD81 antigend P60033 2 8.5 1 5 89 Cell division control protein 42 homologued P60953 5.92 29.3 4 4, 6 90 Ceruloplasmin P00450 13.06 20.2 16 3, 4, 6 91 Chitinase-3-like protein 1 P36222 1.32 2.9 1 4 92 Chloride intracellular channel protein 1d O00299 4.01 27.4 4 4, 5, 6 93 Chordin-like protein 2 Q6WN34 34.41 47.6 33 3, 4, 5, 6 94 Clusterin P10909 46.61 64.6 87 1, 2, 3, 4, 5, 6 95 Cofilin-1 P23528 14.3 55.4 8 3, 4, 6 96 Coiled-coil domain containing protein 38d Q502W7 1.89 3.4 3 5 97 Complement C1q tumor necrosis factor-related protein 1d Q9BXJ1 2.33 17.8 3 4 98 Complement C2 P06681 2.39 7.5 3 6 99 Complement C3 P01024 113.44 49.3 126 3, 4, 5, 6 100 Complement C4-A P0C0L4 2 54.8 92 4 101 Complement C4-B P0C0L5 130.07 54.8 146 3, 4, 5, 6 102 Complement C5 P01031 1.51 2.1 1 4 103 Complement component C9 P02748 6.73 13.8 5 4 104 Complement factor B P00751 25.91 36.1 26 4, 6 105 Corticosteroid-binding globulin P08185 2.91 9.6 2 4 106 Cyclic AMP-dependent transcription factor ATF-6 betad Q99941 1.51 1.7 1 6 107 Cystatin-Bd P04080 2 33.7 1 4 108 Cystatin-C P01034 4.75 40.4 3 4, 5, 6 109 Cysteine desulfurase, mitochondriald Q9Y697 1.62 2.8 2 5 110 Cysteine-rich motor neuron 1 proteind Q9NZV1 4 6.4 2 3 111 Cytoplasmic aconitate hydratased P21399 2.09 16.7 3 4 112 Cytoplasmic dynein 1 light intermediate chain 2d O43237 2 6.1 1 6 113 Cytosol aminopeptidased P28838 4.59 11 3 4 114 Cytosolic nonspecific dipeptidased Q96KP4 16.12 44.6 20 3, 4, 6 115 D-dopachrome decarboxylased P30046 2 21.2 2 4 116 Destrind P60981 3.01 27.9 3 3, 4 117 Dihydropyrimidinase-related protein 2d Q16555 4.18 20.6 7 3, 4, 6 118 Dihydropyrimidinase-related protein 3d Q14195 11.99 36.8 13 3, 4, 6 119 Dipeptidyl peptidase 1 P53634 4.56 11.7 4 3, 4 120 DNA-binding protein Ad P16989 8.15 30.4 8 4, 6 121 Double stranded RNA-binding protein Staufen homologue 1d O95793 2 1.7 1 6 122 Dyslexia-associated protein KIAA0319-like proteind Q8IZA0 1.52 4.2 2 4, 6

1702 dx.doi.org/10.1021/pr2008797 | J. Proteome Res. 2012, 11, 1696−1714 Journal of Proteome Research Article Table 1. continued

protein name accession no. unuseda % covb peptides methodc 123 Dystroglycan Q14118 1.34 5.8 2 4 124 EGF-containing fibulin-like extracellular matrix protein 1 Q12805 6.28 19.9 10 4, 6 125 Elongation factor 1-betad P24534 2 16.4 2 4 126 Elongation factor 1-gammad P26641 1.64 7.3 2 6 127 Elongation factor 2 P13639 18.49 21.7 18 3, 4, 6 128 Enoyl-CoA hydratase domain-containing protein 1d Q9NTX5 2 3.6 1 6 129 Eosinophil cationic proteind P12724 2 7.5 1 6 130 Epithelial discoidin domain-containing receptor 1 Q08345 3.74 11.2 5 4, 6 131 Eukaryotic peptide chain release factor subunit 1d P62495 2 7.6 2 6 132 Eukaryotic translation initiation factor 5A-1d P63241 2.03 25.3 2 6 133 Ezrin P15311 6.75 21.5 8 3, 5 134 F-actin-capping protein subunit alpha-1d P52907 2 8 2 4 135 Farnesyl pyrophosphate synthased P14324 4 14.6 4 4 136 Fatty acid-binding protein P05413 12.62 53.4 9 5, 6 137 Fatty acid synthase P49327 16.71 13.3 15 1, 3, 4, 6 138 FERM domain-containing protein 4Bd Q9Y2L6 1.72 7 1 3 139 Ferritin heavy chaind P02794 6.01 24 9 3, 4, 6 140 Fibrinogen alpha chaind P02671 6.02 18.7 5 4, 6 141 Fibrinogen beta chain P02675 11.25 30.6 14 4, 6 142 Fibrinogen gamma chain P02679 4.95 23.4 5 4, 6 143 Fibroblast growth factor-binding protein 1 Q14512 3.25 15 2 6 144 Fibronectind P02751 48.84 31.5 55 3, 4, 6 145 Fibulin-1d P23142 3.51 5.4 3 4 146 Ficolin-2d Q15485 2 8.9 2 6 147 Filamin-Bd O75369 2.62 4.7 5 4, 6 148 Flavin reductased P30043 6.44 36.4 5 4, 6 149 Follistatin-related protein 1 Q12841 5.72 11.4 3 4, 6 150 Fructose-1,6-bisphosphatase 1d P09467 4.32 13.3 4 4, 6 151 Fructose-bisphosphate aldolase A P04075 32.69 62.1 21 3, 4, 6 152 Fructose-bisphosphate aldolase Cd P09972 2.02 25.3 6 4 153 G-protein coupled receptor family C group 5 member B Q9NZH0 2 3 2 5 154 Galactose-1-phosphate uridylyltransferased P07902 1.68 6.9 2 6 155 Galectin-3-binding protein Q08380 32.14 36.9 46 1, 3, 4, 5, 6 156 Gamma-glutamyltranspeptidase 1 P19440 6.09 14.4 3 4, 6 157 Gamma-glutamyltranspeptidase 2 P36268 1.73 13.5 7 5 158 Ganglioside GM2 activatord P17900 1.47 16.1 2 4 159 Gelsolin P06396 6.35 20.5 11 3, 4, 6 160 Glucose-6-phosphate 1-dehydrogenased P11413 4.29 16.8 2 6 161 Glutathione peroxidase 3d P22352 5.2 28.3 6 3, 4, 6 162 Glutathione S-transferase omega-1d P78417 2 10 2 6 163 Glyceraldehyde-3-phosphate dehydrogenase P04406 17.65 59.4 15 3, 4, 6 164 Glycerol-3-phosphate dehydrogenase [NAD+], cytoplasmicd P21695 2.02 9.2 2 6 165 Glyoxalase domain-containing protein 4d Q9HC38 1.6 7.3 2 4 166 Golgi-associated plant pathogenesis-related protein 1 Q9H4G4 2 16.9 2 4, 6 167 Granulinsd P28799 15.9 31.9 19 4, 6 168 Gremlin-2d Q9H772 2 10.7 2 4, 6 169 Group 3 secretory phospholipase A2d Q9NZ20 4 8.1 2 6 170 Haptoglobin P00738 34.87 55.9 30 4, 5, 6 171 Heat shock 70 kDa protein 4d P34932 2.02 6.4 2 6 172 Heat shock 70 kDa protein 6d P17066 6 12 8 6 173 Heat shock 70 kDa protein 13d P48723 2 5.5 1 4 174 Heat shock 70 kDa protein 1A/1Bd P08107 12.58 27.3 13 4 175 Heat shock cognate 71 kDa protein P11142 21.49 35.1 19 4, 6 176 Heat shock protein beta-1d P04792 4.02 36.6 2 3, 4 177 Heat shock protein HSP 90-alphad P07900 4.07 14.9 5 4, 6 178 Heat shock protein HSP 90-betad P08238 2 20.2 6 3, 4, 6 179 Heat shock-related 70 kDa protein 2d P54652 6.14 23.6 4 3 180 Heme-binding protein 1 Q9NRV9 5.29 28 6 4, 6 181 Hemoglobin subunit betad P68871 2.83 19.6 2 6 182 Hemopexin P02790 2 4.1 2 5 183 Heterogeneous nuclear ribonucleoprotein Qd O60506 2 6.3 2 4, 6

1703 dx.doi.org/10.1021/pr2008797 | J. Proteome Res. 2012, 11, 1696−1714 Journal of Proteome Research Article Table 1. continued

protein name accession no. unuseda % covb peptides methodc 184 Histamine N-methyltransferased P50135 1.77 8.2 1 4 185 Histone H12d P16403 6 13.6 3 6 186 Histone H2A type 1-Jd Q99878 2.65 27.3 2 4 187 Histone H2B type 1-Ad Q96A08 2.02 15.8 2 6 188 HLA class I histocompatibility antigen, A-69 alpha chaind P10316 3.03 30.1 3 4, 6 189 HLA class II histocompatibility antigen, DR alpha chaind P01903 2.2 13.8 3 4 190 Hornerind Q86YZ3 1.38 4.1 2 4 191 Hsp90 cochaperone Cdc37d Q16543 1.66 5.6 2 6 192 Hyaluronan and proteoglycan link protein 3 Q96S86 21.07 49.4 14 3, 4, 6 193 Hypoxia up-regulated protein 1d Q9Y4L1 15.42 23.9 15 3, 4, 6 194 Ig alpha-1 chain C region P01876 33.39 58.6 42 1, 2, 3, 4, 5, 6 195 Ig alpha-2 chain C region P01877 6.03 58.2 28 2, 3, 4, 5, 6 196 Ig delta chain C regiond P01880 4 19.8 4 4 197 Ig gamma-1 chain C region P01857 3.1 21.5 3 5 198 Ig gamma-2 chain C regiond P01859 2.05 16.9 2 4, 5 199 Ig heavy chain V−I region HG3 P01743 2.35 21.4 3 5 200 Ig heavy chain V−I region V35d P23083 3.14 20.5 2 6 201 Ig heavy chain V−III region BRO P01766 3.96 30.8 3 5, 6 202 Ig heavy chain V−III region TEId P01777 2 16 2 3, 4 203 Ig heavy chain V−III region TILd P01765 2.83 26.1 3 5 204 Ig heavy chain V−III region TURd P01779 2 30.2 2 6 205 Ig kappa chain C region P01834 11.82 67 24 1, 2, 3, 4, 5, 6 206 Ig kappa chain V−I region BANd P04430 2.03 26.9 3 6 207 Ig kappa chain V−I region EUd P01598 2.5 38.9 2 6 208 Ig kappa chain V−I region Haud P01600 2 26.9 3 6 209 Ig kappa chain V−I region Mevd P01612 2 16.5 2 6 210 Ig kappa chain V−I region WEAd P01610 2 30.6 1 4 211 Ig kappa chain V−I region Wesd P01611 2 16.7 1 4 212 Ig kappa chain V−II region GM607 (Fragment)d P06309 4 37.6 3 4, 6 213 Ig kappa chain V−II region RPMI 6410d P06310 1.77 9.8 2 5 214 Ig kappa chain V−III region CLLd P04207 2.63 22.5 2 5, 6 215 Ig kappa chain V−III region HICd P18136 3.09 50.4 4 2, 4, 5, 6 216 Ig kappa chain V−III region VG P04433 4 33 3 5, 6 217 Ig kappa chain V−IV region B17d P06314 2.01 23.9 2 6 218 Ig lambda chain V−I region HA P01700 2.04 16.1 2 5, 6 219 Ig lambda chain V−I region NEWMd P01703 2 16.5 1 3 220 Ig lambda chain V−I region WAHd P04208 2 11.9 1 4 221 Ig lambda chain V−III region LOI P80748 4 41.4 3 4, 5, 6 222 Ig lambda chain V−III region SH P01714 2 25 2 4, 6 223 Ig lambda chain V−IV region Hil P01717 2.07 39.3 3 4, 5, 6 224 Ig lambda chain C regions P01842 8 60 4 3 225 Ig lambda-1 chain C regions P0CG04 2 46.2 9 4 226 Ig lambda-2 chain C regions P0CG05 8.79 67 12 4 227 Ig lambda-3 chain C regionsd P0CG06 8.13 67.9 11 5, 6 228 Ig mu chain C region P01871 25.61 35.4 21 4, 5, 6 229 Immunity-related GTPase family Q proteind Q8WZA9 3.07 12.8 2 6 230 Immunoglobulin J chain P01591 10.14 56 12 2, 3, 4, 5, 6 231 Immunoglobulin lambda-like polypeptide 5d B9A064 4.07 43.5 12 6 232 Insulin-like growth factor-binding protein 2 P18065 6.83 23.4 7 4, 5, 6 233 Interalpha-trypsin inhibitor heavy chain H2d P19823 2.29 6.3 2 4, 6 234 Interalpha-trypsin inhibitor heavy chain H4d Q14624 9.17 15.5 12 4, 6 235 Isocitrate dehydrogenase [NADP] cytoplasmicd O75874 6.39 23 4 3, 6 236 Isopentenyl-diphosphate Delta-isomerase 1d Q13907 3.85 13.2 2 4 237 Kallistatind P29622 3.6 11.7 2 6 238 Kappa-casein P07498 66.42 57.7 135 1, 2, 3, 4, 5, 6 239 Keratin, type I cytoskeletal 10d P13645 18.69 23.6 12 3, 4, 6 240 Keratin, type I cytoskeletal 14d P02533 2.23 13.8 3 6 241 Keratin, type I cytoskeletal 20d P35900 1.8 6.4 1 4 242 Keratin, type I cytoskeletal 9d P35527 4.91 21.2 10 3, 4, 6 243 Keratin, type II cytoskeletal 1d P04264 27.12 48.9 21 3, 4, 6 244 Keratin, type II cytoskeletal 2 epidermald P35908 6.57 25.2 8 3, 4, 6

1704 dx.doi.org/10.1021/pr2008797 | J. Proteome Res. 2012, 11, 1696−1714 Journal of Proteome Research Article Table 1. continued

protein name accession no. unuseda % covb peptides methodc 245 Keratin, type II cytoskeletal 5 P13647 2.61 15.3 7 4, 6 246 Kininogen-1d P01042 16.2 17.6 12 4, 6 d 247 L-lactate dehydrogenase A chain P00338 2.33 10.2 2 6 248 L-lactate dehydrogenase B chain P07195 4.76 26.1 5 3, 4, 6 d 249 L-xylulose reductase Q7Z4W1 2.75 23 4 3, 4, 6 250 Lactadherin Q08431 33.43 63.6 42 1, 3, 4, 5, 6 251 Lactotransferrin P02788 162.54 89.2 325 1, 2, 3, 4, 5, 6 252 Left-right determination factor 2d O00292 5.88 23.2 3 6 253 Legumaind Q99538 3.05 20.3 2 3 254 Leucine-rich alpha-2-glycoprotein P02750 9.82 30.6 13 4, 5, 6 255 Lipolysis-stimulated lipoprotein receptord Q86 × 29 2 6.6 2 6 256 Lipoprotein lipase P06858 12.74 28.2 16 3, 4, 5, 6 257 Lysozyme C P61626 19.74 80.4 50 2, 3, 4, 5, 6 258 Lysyl oxidase homologue 4d Q96JB6 2.6 6.2 2 6 259 Macrophage mannose receptor 1 P22897 2 4.4 2 6 260 Macrophage mannose receptor 1-like protein 1 Q5VSK2 26.97 14.9 22 3, 4, 5 261 Macrophage migration inhibitory factor P14174 4.02 17.4 4 4, 6 262 Malate dehydrogenase, cytoplasmicd P40925 10.48 21 6 3, 4, 6 263 Mannose-6-phosphate receptor-binding protein 1 O60664 10.01 27.9 8 3 264 Mannosyl-oligosaccharide 1,2-alpha-mannosidase 1Ad P33908 2.03 11.8 4 6 265 Matrilin-3d O15232 3.56 19.8 7 4 266 Matrix-GIa proteind P08493 2 15.5 2 6 267 Midkined P21741 3.05 27.3 3 4, 6 268 Moesin P26038 1.92 9.9 3 6 269 Monocyte differentiation antigen CD14 P08571 32.65 60.8 41 2, 3, 4, 5, 6 270 Mucin-1 P15941 9.28 6.1 9 4, 5, 6 271 Mucin-4 Q99102 7.33 7.9 9 4, 5, 6 272 Multiple inositol polyphosphate phosphatase 1d Q9UNW1 2 8.4 2 4 273 Myosin-IXad B2RTY4 2.28 7.8 3 3 d 274 N-acetyl-D-glucosamine kinase Q9UJ70 2 6.7 2 6 275 N-acetylglucosamine-1-phosphotransferase subunit gammad Q9UJJ9 3.6 11.2 2 4, 6 d 276 N-acetylmuramoyl-L-alanine amidase Q96PD5 4.02 14.1 4 4 277 N(G),N(G)-dimethylarginine dimethylaminohydrolase 1d O94760 11.43 26 8 3, 4, 6 278 Na(+)/H(+) exchange regulatory cofactor NHE-RF1d O14745 2.12 14.8 2 4, 6 279 Nephronectind Q6UXI9 12.89 22.5 8 4, 6 280 Nicotinamide phosphoribosyltransferased P43490 2.28 5.7 2 4, 6 281 Nidogen-1d P14543 4.36 6.6 3 6 282 Nuclease-sensitive element-binding protein 1d P67809 4 31.8 6 4 283 Nucleobindin-1 Q02818 34.55 58.8 34 2, 3, 4, 6 284 Nucleobindin-2 P80303 38.52 62.6 49 1, 2, 3, 4, 6 285 Nucleolind P19338 2.02 3.1 4 6 286 Nucleoside diphosphate kinase Bd P22392 2.53 25 4 4, 6 287 Nucleotide exchange factor SIL1 Q9H173 4.85 20.2 8 4, 6 288 Obg-like ATPase 1d Q9NTK5 2.01 7.6 2 4 289 Osteopontin P10451 14.15 36.9 32 3, 4, 5, 6 290 Parathyroid hormone-related protein P12272 2 17.5 4 4, 6 291 Peptidyl-prolyl cis−trans isomerase A P62937 16.36 63 11 3, 4, 6 292 Peptidyl-prolyl cis−trans isomerase B P23284 9 38.4 10 4, 5, 6 293 Peptidyl-prolyl cis−trans isomerase Cd P45877 3.92 42.9 3 4 294 Peptidyl-prolyl cis−trans isomerase FKBP1Ad P62942 3.8 25 3 4, 6 295 Perilipin-2 Q99541 10 18.5 5 6 296 Perilipin-3 O60664 6.63 22.6 8 4, 6 297 Peroxiredoxin-1d Q06830 3.34 9.5 3 6 298 Peroxiredoxin-2d P32119 5.49 18.7 3 4 299 Peroxiredoxin-6d P30041 6.76 25.9 5 4, 6 300 Phosphatidylethanolamine-binding protein 1 P30086 4.38 27.8 5 3, 4, 6 301 Phosphoglucomutase-1d P36871 15.09 32.2 11 3, 4, 6 302 Phosphoglycerate kinase 1d P00558 15.64 34.1 10 3, 4, 6 303 Phosphoglycolate phosphatased A6NDG6 2 15.3 2 6 304 Phospholipid transfer protein P55058 6.69 19.3 4 3, 4 305 Pigment epithelium-derived factord P36955 4.22 16.8 2 6

1705 dx.doi.org/10.1021/pr2008797 | J. Proteome Res. 2012, 11, 1696−1714 Journal of Proteome Research Article Table 1. continued

protein name accession no. unuseda % covb peptides methodc 306 Plasma protease C1 inhibitor P05155 13.26 18.2 8 4, 5, 6 307 Plasminogen P00747 9.29 19.6 6 4, 6 308 Plastin-2 P13796 2.23 7.3 2 4 309 Platelet-derived growth factor Cd Q9NRA1 2 3.5 1 6 310 Platelet glycoprotein 4 P16671 3.11 19.1 9 4, 5 311 Poly(rC)-binding protein 1d Q15365 8 23 4 6 312 Polyadenylate-binding protein 1d P11940 5.3 13.8 4 6 313 Polyadenylate-binding protein 3d Q9H361 1.57 12.8 4 4 314 Polymeric immunoglobulin receptor P01833 54.01 48 68 2, 3, 4, 5, 6 315 Polypeptide N-acetylgalactosaminyltransferase 2d Q10471 5.4 12.3 3 6 316 Polyubiquitind P0CG48 5.5 74 6 5 317 Prefoldin subunit 4d Q9NQP4 2 11.2 1 4 318 Pro-epidermal growth factor P01133 22.61 21.6 20 3, 4, 6 319 Proactivator polypeptide P07602 44.5 58.6 62 3, 4, 6 320 Procollagen-lysine,2-oxoglutarate 5-dioxygenase 1d Q02809 3.51 9.5 5 4, 6 321 Programmed cell death 6-interacting proteind Q8WUM4 2.72 6.5 5 4, 6 322 Prolactin-inducible protein P12273 3.4 33.6 4 4, 5, 6 323 Proline synthase cotranscribed bacterial homologue proteind O94903 2 2.9 1 6 324 Proliferation-associated protein 2G4d Q9UQ80 2 9.9 1 4 325 Prostaglandin reductase 1d Q14914 3.6 10 2 4 326 Proteasome activator complex subunit 1d Q06323 4.86 17.3 4 3, 4 327 Proteasome subunit alpha type-7-liked Q8TAA3 2.65 17.6 2 4 328 Proteasome subunit alpha type-5d P28066 1.72 6.2 2 6 329 Proteasome subunit beta type-1d P20618 8.88 44 9 4, 6 330 Proteasome subunit beta type-2d P49721 2 7.5 1 4 331 Protein CutAd O60888 2 13.4 2 6 332 Protein disulfide-isomerase A3 P30101 3.17 15.3 5 4, 6 333 Protein disulfide-isomerase A6 Q15084 4.01 11.8 3 4 334 Protein disulfide-isomerase P07237 33.62 56.5 27 4, 6 335 Protein DJ-1d Q99497 2 21.2 2 4 336 Protein ERGIC-53d P49257 1.57 8.4 2 4 337 Protein FAM150Ad Q6UXT8 2.7 31.8 2 6 338 Protein NDRG2d Q9UN36 2.51 15.6 2 4 339 Protein OS-9d Q13438 4.35 5.7 4 4 340 Protein S100-A1 P23297 2 38.3 2 3, 4 341 Protein S100-A9 P06702 2 18.2 1 3 342 Protein TFGd Q92734 2 6.5 1 4 343 Prothrombin P00734 4.2 7.6 3 4, 6 344 Purine nucleoside phosphorylased P00491 3.85 20.8 3 4 345 Putative 40S ribosomal protein S26-like 1d Q5JNZ5 2 7.8 1 6 346 Putative elongation factor 1-alpha-like 3d Q5VTE0 10.38 22.7 11 3, 4, 6 347 Putative tropomyosin alpha-3 chain-like proteind A6NL28 3.56 17 2 6 348 Putative trypsin-6d Q8NHM4 2 4 1 4 349 Pyruvate kinase isozymes M1/M2d P14618 4.7 15.8 4 3, 4, 6 350 Quinone oxidoreductased Q08257 2 12.2 2 4 351 Rab GDP dissociation inhibitor beta P50395 10.81 39.1 13 4, 6 352 Radixind P35241 6.73 16 7 4 353 Ras-related C3 botulinum toxin substrate 1d P63000 4.23 29.2 3 4 354 Ras-related protein Rap-1Ad P62834 2 6.5 1 3 355 Ras-related protein Rab-11Bd Q15907 1.56 8.3 2 6 356 Ras-related protein Rab-1Ad P62820 7.25 27.3 5 6 357 Ras-related protein Rab-1Bd Q9H0U4 2.5 18.9 3 4 358 Ras-related protein Rab-21d Q9UL25 4 23.1 2 4 359 Ras-related protein Rab-5Cd P51148 3.68 17.6 3 4 360 Ras-related protein Rab-7ad P51149 8 29.5 4 4, 6 361 Ras-related protein Rab-8Ad P61006 1.83 31.4 3 6 362 Receptor-type tyrosine-protein phosphatase Fd P10586 4.43 5.6 2 6 363 Retinoid-inducible serine carboxypeptidased Q9HB40 4.01 7.1 2 4 364 Rho-related GTP-binding protein RhoBd P62745 2 30.1 2 3 365 Ribonuclease 4d P34096 2.13 23.1 3 4 366 Ribonuclease inhibitord P13489 1.77 3.3 1 4

1706 dx.doi.org/10.1021/pr2008797 | J. Proteome Res. 2012, 11, 1696−1714 Journal of Proteome Research Article Table 1. continued

protein name accession no. unuseda % covb peptides methodc 367 Ribose-phosphate pyrophosphokinase 2d P11908 3.72 9.7 2 3, 4, 6 368 S-acyl fatty acid synthase thioesterase, medium chaind Q9NV23 4.75 24.2 3 4, 6 369 Sclerostin domain-containing protein 1 Q6 × 4U4 12.78 39.8 15 3, 4, 6 370 Secreted frizzled-related protein 1d Q8N474 6.14 18.8 4 4, 6 371 Selenium-binding protein 1 Q13228 27.49 51.3 25 3, 4, 5, 6 372 Semaphorin-4Bd Q9NPR2 2.21 7.6 3 4 373 Serum albumin P02768 41.02 44 51 1, 2, 3, 4, 5, 6 374 Serum amyloid A-4 proteind P35542 3.5 38.5 2 6 375 Sialic acid synthased Q9NR45 2 5 1 4 376 Sortilin Q99523 2.74 3.6 3 4 377 SPARC-like protein 1 Q14515 1.77 6.9 2 5 378 StAR-related lipid transfer proteind Q9P2P6 2.01 2.5 8 5 379 Sulfhydryl oxidase 1 O00391 43.94 51.3 44 3, 4, 5, 6 380 Synaptic vesicle membrane protein VAT-1 homologued Q99536 11.88 35.9 11 4, 6 381 Syntaxin-binding protein 2d Q15833 2.16 11.5 2 6 382 Syntenin-1 O00560 6.91 26.2 5 4 383 T-cell immunomodulatory proteind Q8TB96 2.47 5.4 2 4 384 T-complex protein 1 subunit alphad P17987 3.65 6.6 3 4, 6 385 Tenascin P24821 49.29 27.7 48 4, 6 386 Thrombospondin-1 P07996 19.15 13.6 11 4, 6 387 Thyroxine-binding globulind P05543 2 2.9 2 6 d 388 Tissue alpha-L-fucosidase P04066 4.26 14.8 6 4 389 Transforming protein RhoAd P61586 6.07 42.5 5 4, 6 390 Transketolased P29401 15.81 27.3 16 3, 4, 6 391 Transmembrane protease serine 13 Q9BYE2 2 18.6 4 5 392 Transthyretin P02766 6.01 41.5 6 2, 3, 4, 5, 6 393 Triosephosphate isomerase P60174 9.46 56.2 12 3, 4, 6 394 Tropomyosin beta chaind P07951 1.7 24.7 4 6 395 Trypsin-1d P07477 4.78 19.4 7 6 396 Tryptophanyl-tRNA synthetase, cytoplasmicd P23381 8 31.2 9 4 397 Tubulointerstitial nephritis antigen-liked Q9GZM7 1.47 2.8 1 6 398 Tumor necrosis factor ligand superfamily member 13d O75888 4.01 16.8 4 4, 6 399 Tumor necrosis factor receptor superfamily member 11B O00300 2 9.5 1 4, 6 400 Ubiquitin carboxyl-terminal hydrolase isozyme L3d P15374 2 14.4 2 4, 6 401 Ubiquitin-60S ribosomal protein L40d P62987 2 18.8 2 4 402 UDP-GlcNAc:betaGal beta-1,3-N-acetylglucosaminyltransferase 3d Q9Y2A9 2.17 11 2 6 403 UDP-GlcNAc:betaGal beta-1,3-N-acetylglucosaminyltransferase 7d Q8NFL0 5.6 16.7 4 4, 6 404 UPF0764 protein C16orf89d Q6UX73 1.52 20.4 5 4 405 UMP-CMP kinased P30085 2.15 15.8 2 6 406 UTP-glucose-1-phosphate uridylyltransferase Q16851 13.22 32.7 8 3, 4, 6 407 UV excision repair protein RAD23d P54727 1.62 3.2 2 6 408 V-type proton ATPase subunit S1d Q15904 2 6.2 1 4 409 Vinculind P18206 1.37 6.7 2 6 410 Vitamin D-binding protein P02774 17.3 40.1 12 3, 4, 6 411 Vitronectin P04004 24.59 46.7 28 1, 3, 4, 5, 6 412 Von Willebrand factor A domain containing protein 1 Q6PCB0 1.33 9.2 2 5, 6 413 Xanthine dehydrogenase/oxidase P47989 110.65 59.4 154 1, 3, 4, 5, 6 414 Zinc-alpha-2-glycoprotein P25311 23.66 50.7 16 5 415 Zymogen granule protein 16 homologue B Q96DA0 4 25.5 3 3, 4, 6 aUnused ProtScore is a measure of the protein confidence for a detected protein. An Unused ProtScore of 1.3 corresponds to 95% confidence, with a higher score representing a higher level of confidence. bThe percentage of the total protein sequence covered by the identified peptides. cThe experiment in which the protein was identified. (1) SDS-PAGE, (2) 2D DIGE, (3) 1D LC−MALDI, (4) 2D LC−MALDI, (5) iTRAQ without ProteoMiner treatment, (6) iTRAQ after ProteoMiner treatment. When the protein was identified using more than one experiment, the reported results are from the experiment that gave the highest Unused ProtScore. dProteins that have not previously been identified in human skim milk.

Compatibility of ProteoMiner Treatment with Downstream between proteins of high and low abundance: one in which iTRAQ Quantitation casein-depleted skim milk samples were subjected to Proteo- Low abundance proteins retain their relative abundance after Miner treatment prior to iTRAQ analysis and one in which ProteoMiner treatment, whereas high abundance proteins do casein-depleted skim milk samples were left untreated (Figure 1). not.20,21,29 As described in previous studies,20,30 we conducted The 80 proteins identified in the untreated milk samples were two parallel iTRAQ experiments in order to distinguish classified as proteins of high-abundance. The iTRAQ analysis of

1707 dx.doi.org/10.1021/pr2008797 | J. Proteome Res. 2012, 11, 1696−1714 Journal of Proteome Research Article

Figure 5. Protein classifications. (A) Subcellular location of all 415 proteins identified in the 1D LC−MALDI and 2D LC−MALDI analyses. (B) Functional categorization of 415 proteins identified in the 1D LC−MALDI and 2D LC−MALDI analyses. A number of proteins are included in multiple categories. The number of proteins in each category is indicated in parentheses.

the quantitative information from the 51 proteins identified in both iTRAQ experiments (Figure 4) was compared. Of the 11 proteins found to be differentially expressed in the iTRAQ experiment without ProteoMiner treatment, only four were also differentially expressed after ProteoMiner treatment. Of the 40 proteins found not to be differentially expressed in the iTRAQ Figure 4. Protein identifications in human term milk using different experiment without ProteoMiner treatment, 14 were found to analytical techniques. (A) Human term milk (38−41 weeks gestation) be differentially expressed after ProteoMiner treatment. Of the was collected from 8 mothers between 15 and 28 days postpartum, total 51 proteins, only 28 displayed homodirectional changes in pooled together, treated with ProteoMiner beads to deplete the most the two iTRAQ experiments. These results indicate that the abundant proteins, and then analyzed using different techniques. 274 ProteoMiner treatment does introduce significant error into the proteins were identified in total. (B) Sixteen term (38−41 weeks − relative abundance ratios of these higher abundance proteins. gestation) and 16 preterm (28 32 weeks gestation) mothers donated Therefore, in accordance with previous studies,20 only the milk samples between 7 and 14 days postpartum. Samples in each group were pooled. Two iTRAQ experiments were conducted: one comparing relative abundance ratios of the low abundance proteins (those the protein expression in term and preterm milk without ProteoMiner not also identified in the iTRAQ analysis of untreated samples) treatment and one comparing the protein expression after ProteoMiner were considered to be accurate in the iTRAQ analysis of treatment. 315 proteins were identified in total in the iTRAQ experi- ProteoMiner-treated samples. ments. (C) Combining the results from the experiments described in (A) Quantitative Comparison of Term and Preterm Milk and (B), 415 proteins were identified in total from human term milk (38−41 weeks gestation), expressed between 7 and 28 days postpartum. The protein concentration of the pooled term and preterm milk samples were similar both before casein depletion (15.9 mg/mL the untreated samples enabled accurate quantitation of these and 16.0 mg/mL, respectively) and afterward (7.1 mg/mL and 7.0 mg/mL, respectively). There were 80 abundant proteins high abundance proteins, whereas the iTRAQ analysis of the identified in the iTRAQ experiment of the non-ProteoMiner- ProteoMiner-treated samples enabled quantitation of the low treated samples (Table 1). All of the 80 proteins were found in abundance proteins. both the term and preterm milk samples and in each duplicate. To assess whether ProteoMiner treatment does in fact affect Five proteins constituted a significantly greater proportion of the the relative abundance ratios of the high abundance proteins, protein content in the pooled preterm milk, and 10 proteins

1708 dx.doi.org/10.1021/pr2008797 | J. Proteome Res. 2012, 11, 1696−1714 Journal of Proteome Research Article were present at a significantly higher level in the pooled term ProteoMiner beads, and others present at levels too low to be milk (Table 2). The remaining 65 proteins did not differ in detected. abundance between the two samples. The protein identification study yielded no direct informa- In the iTRAQ analysis of the ProteoMiner-treated pooled tion about the relative concentration of the proteins in human term and preterm milk samples, 286 proteins were identified. milk. However, comparing the results to those of previous Fifty-one of these proteins were also identified in the iTRAQ studies in which fewer proteins were identified suggests that the experiment without ProteoMiner treatment and were thus cytoplasmic proteins are among those of least abundance in classed as high abundance proteins. Of the 235 low abundance human milk. Liao et al.19 found that 23.5% of the 115 proteins proteins identified, 40 differed in abundance between the pooled they identified were cytoplasmic. Similarly, of the 80 proteins term and preterm milk samples (Table 2). Twenty-three of these identified in the iTRAQ experiment without ProteoMiner low abundance proteins were present at higher levels in preterm treatment in the present study, only 15% were cytoplasmic. By milk, and 17 were present at higher levels in term milk. The re- contrast, nearly half (46%) of the total 415 proteins we identi- maining 195 low abundance proteins were present at similar fied were of cytoplasmic origin. As these additional cytoplasmic levels in both term and preterm milk. proteins were only identified after an extensive enrichment and SDS-PAGE analysis of the pooled preterm and term samples separation process, it is likely that they represent the proteins of was conducted to confirm the iTRAQ results of the highly least abundance in milk. This is consistent with the notion that abundant proteins (Figure 6). The identity of protein bands some cytoplasmic proteins are present in human milk as an was confirmed using MALDI-MS/MS in a previous study.23 incidental consequence of the secretory process rather than Bile salt-stimulated lipase, lactoferrin, and serum albumin were being under regulatory control and are thus found in only present at significantly higher levels in preterm milk (p < 0.05). trace amounts. Indeed, Patton and Huston32 reported the Levels of β-casein and α-lactalbumin were similar in the term presence of cytoplasmic crescents within milk fat globules, and preterm milk. having being entrained within the fat globule membrane during the secretory process. Alternatively, the cytoplasmic ■ DISCUSSION proteins found in human milk may be derived from the In the present study, 415 proteins were identified using a breakdown of its cellular content. Histological studies have number of separation and identification techniques, 261 of found significant levels of cellular debris in human milk, 33 which had not been previously found in human skim milk. The mostly derived from secretory epithelial cells. It is likely that majority of the identified proteins (371 of 415) were found cytoplasmic proteins enter human milk through both pro- using a combination of casein depletion, ProteoMiner treat- cesses, and thus it may be possible to view them as a reflection ment, and a 2D LC separation (Figure 4). The efficacy of the of the protein content of the cytosol of mammary secretory ProteoMiner treatment was highlighted by the iTRAQ experi- epithelial cells. ment, in that when using a 2D LC separation alone without One of the challenges associated with proteomic mapping prior ProteoMiner treatment, only 80 proteins were identified experiments is processing and applying the large amounts of in human skim milk, much fewer than the 286 proteins identi- information generated within a particular biological context. fied in the corresponding analysis after ProteoMiner treatment. One way in which the expanded proteome of human milk Similarly, the 2D LC separation provided a far greater level of may be useful is if the proteins identified reveal any infor- proteome coverage compared to a 1D LC separation or gel- mation about either the metabolism or function of the mam- based approach (Figure 4A). Indeed, we identified over three mary gland. Selected metabolites in human milk are currently timesasmanyproteinsasarecentstudythatcoupledProteoMiner used in this manner as markers of mammary gland develop- treatment to a 1D LC separation.19 ment and health. For example, levels of milk citrate, sodium, ProteoMiner treatment was found to be reproducible, with a and lactose are used to indicate successful secretory differentia- high degree of consistency observed in both the SDS-PAGE tion and activation of the mammary gland.34 Similarly, milk and 2D-DIGE analysis of the samples treated in duplicate lactose and glucose can be used as markers of mastitis35 and (Figures 2, 3). However, it was apparent in the SDS-PAGE metabolites in the lactose synthesis pathway used to identify analysis that not all proteins of equivalent initial concentration diabetic mothers.36 In the present study, 140 of the proteins were depleted or enriched to the same extent. For example, identified (28% of total) are constitutively involved in cellular α-lactalbumin was nearly completely removed, whereas lacto- metabolism and normal cellular function. This includes 44 ferrin remained at a relatively high concentration after proteins of the glycolysis, pentose phosphate, citrate, and carbo- ProteoMiner treatment, despite both being present at similar hydrate metabolism pathways, many of which such as pyruvate levels in milk samples (Figure 2). Furthermore, 29 of the 80 kinase (PKM2), transketolase (TKT), malate dehydrogenase proteins that were identified in the casein-depleted skim milk (MDH1), and fructose-1,6-bisphosphatase (FBP) were identi- by 2D LC−MS/MS in the iTRAQ experiment were not found fied for the first time in human milk. While further research in the corresponding ProteoMiner-treated sample (Figure 4C). is required to investigate the expression of these proteins This indicates that some proteins of medium abundance may throughout lactation, it is possible that they may prove useful have a very low binding affinity for the hexapeptide ligand as markers of events that involve alterations to the meta- library and are not retained on the ProteoMiner beads. Indeed, bolic activity of the mammary gland, such as lactogenesis and the different binding affinities of proteins for the ProteoMiner involution. peptide library have been described as one of the major limita- Proteins in human milk may also be used as markers of in- tions of the technique, with an estimated 5−15% of the proteins fection and inflammation of the mammary gland. Several studies in any given mixture not binding at all.31 Therefore, while our re- have reported changes to the levels of immunological and sults represent a considerable expansion of the known proteome inflammatory proteins during mastitis. For example, lactoferrin, of human milk, it is likely that many other milk proteins remain sIgA, secretory leukocyte protease inhibitor, and serum albumin to be identified, with some proteins escaping capture by the have all been found to be present at higher levels in human milk

1709 dx.doi.org/10.1021/pr2008797 | J. Proteome Res. 2012, 11, 1696−1714 Journal of Proteome Research Article

Table 2

name peptidesa unusedb fold differencec Up-regulated in preterm milk Abundant proteins Serum albumin 57 57.57 2.35 Bile-salt activated lipase 26 29.91 5.19 Cysteine desulfurase 1 1.64 1.62 Lactotransferrind 204 127.89 2.83 Lysozyme-Ce 12 8.28 2.30 Low abundance proteins T-complex protein 1 subunit thetad 3 3.65 1.21 Isocitrate dehydrogenase [NADP] cytoplasmic 4 6.39 2.11 40S ribosomal protein S19 9 19.09 1.85 Phosphoglycerate kinase 1 10 15.64 2.52 Alanine aminotransferase 1 2 2.02 1.84 Glyceraldehyde-3-phosphate dehydrogenase 9 10.31 2.11 40S ribosomal protein S20 4 6.04 1.76 Antithrombin-III 2 4 1.67 UDP-GlcNAc:betaGalbeta-1,3-N-acetylglucosaminyltransferase 3 2 2.17 2.10 Nucleolin 3 2.02 2.00 Perilipin-3 7 11.4 2.48 Putative elongation factor 1-alpha-like 3 5 11.9 4.91 Apolipoprotein B-100 32 63.34 4.20 Cytosolic nonspecific dipeptidase 16 27.42 5.27 Apolipoprotein A-II 3 5.82 3.91 S-acyl fatty acid synthase thioesterase, medium chain 3 4.74 3.45 Ras-related protein Rab-1A 5 7.25 6.36 Apolipoprotein E 13 23.25 12.01 DNA-binding protein A 2 4 53.47 Cytoplasmic dynein 1 light intermediate chain 2 2 2 58.65 Syntaxin-binding protein 2 2 2.16 87.90 Ras-related protein Rab-8A 3 1.83 87.90 Elongation factor 1-gamma 2 1.64 56.77 Down-regulated in preterm milk Abundant proteins Polymeric immunoglobulin receptor 59 58.91 −6.67 Clusterin 12 21.97 −2.38 Prolactin-inducible protein 2 3.38 −50.0 Lactadherin 9 14.71 −1.50 Alpha-lactalbumin 22 11.82 −33.3 Mucin-4 3 4.31 −1.02 Ig heavy chain V−I region HG3d 2 2.35 −6.25 Biotinidase 1 1.82 −33.3 Gamma-glutamyl transpeptidase 1 1.73 −1.50 Complement C3e 12 21.1 −2.44 Low abundance proteins Glutathione S-transferase omega-1 2 2 −79.05 C-type lectin domain family 11 member A 2 4 −74.35 Tenascin 45 69.8 −26.18 Lipoprotein lipase 10 13.6 −7.16 Protein disulfide-isomerase 12 23.79 −2.37 Proactivator polypeptide 18 22.11 −2.15 Beta-1,4-galactosyltransferase 1 7 13.1 −2.93 Ceruloplasmin 7 9.79 −2.43 Cathepsin B 6 9.69 −2.16 Vitronectin 16 25.3 −2.20 Lysyl oxidase homologue 4 2 2.6 −1.76 L-xylulose reductase 2 2.29 −1.04 EGF-containing fibulin-like extracellular matrix protein 1 4 7.78 −1.19 Gelsolin 4 2.16 −1.27 Fibrinogen gamma chain 2 3.91 −1.33 Triosephosphate isomerase 10 19.35 −1.22 6-phosphogluconate dehydrogenase, decarboxylating 4 2.97 −1.05 aThe number of peptides identified. bUnused ProtScore is a measure of the protein confidence for a detected protein. An Unused ProtScore of 1.3 corresponds to 95% confidence, with a higher score representing a higher level of confidence. cThe ratio between the abundance of each protein in the pooled term and preterm samples, expressed as an average of the duplicate experiments. The value is positive when the protein is more abundant in the preterm milk, and negative when the reverse is true. dProteins that were found to be differentially expressed in one replicate at a confidence level >95% and in the other replicate at a confidence level >90%. eProteins that were found to be differentially expressed in both replicates at a confidence level >90%.

1710 dx.doi.org/10.1021/pr2008797 | J. Proteome Res. 2012, 11, 1696−1714 Journal of Proteome Research Article

Figure 6. SDS-PAGE analysis of differences in protein composition of pooled term (T) and preterm (PT) milk samples. (A) Representative SDS- PAGE electrophoretograms of PT and T milk. Each sample was run in triplicate (n = 3). (B) The intensity of the protein bands corresponding to bile-salt stimulated lipase (BSSL), lactoferrin (LF), serum albumin (SA), sIgA, β-casein, and α-lactalbumin (ALA) were expressed as percentages of the total intensity of each lane. Differences between the two samples were analyzed using a Student’s t-test. All significant differences between groups are indicated (p < 0.05). during episodes of mastitis.35,37 A recent study in bovine milk role that human milk plays in protecting infants from infection. also identified a number of immunological proteins as being up- Indeed, Vorbach et al.48 argued that it was immune protection regulated during Escherichia coli infection.38 Of the proteins we rather than the provision of nutrition that was the original func- identified, 120 (24% of total) were associated with immune and tion of the mammary gland in premammals. Although there is inflammatory pathways and may have the potential to act as an abundance of literature describing the protective advantages sensitive biomarkers of mammary gland inflammation. Promising conferred upon infants through breastfeeding, the mechanisms candidates include eight heat shock proteins that we identified in involved are far from delineated.49 Individual proteins have human milk for the first time (Table 1), which are known to been shown to exert protective effects through multiple path- respond to cellular insults.39 ways, even after proteolytic digestion.50,51 Furthermore, there is a In addition to yielding information regarding mammary great deal of interaction between different proteins involved in gland function, the expansion of the human milk proteome in the immune response, as well as with the infant’s own defenses.49 the present study is also of potential interest from an infant The identification of additional potential immune proteins pre- centered perspective, in that many of the proteins may be sent in human milk (Table 1) may be of use in further elucidat- involved in providing nutritional, protective, and developmental ing these complex protective mechanisms. Immune response advantages to breastfed infants. With respect to infant growth proteins that we identified in human milk for the first time and development, we identified 33 proteins (7%) as being in- include C1q and tumor necrosis factor related protein-1 volved in tissue development, including a number of proteins (CTRP1), peroxiredoxins-1, -2, and -6 (PRDX1, PRDX2, identified for the first time, such as granulin (GRN), cysteine- PRDX6), glutathione peroxidase 3 (GPX3), and HLA class I rich motor neuron protein (CRIM1), gremlin-2 (GREM2), and II histocompatibility antigens (HLA-A, HLA-DRA). CTRP1 nephronectin (NPNT), and bone morphogenetic protein 1 is a cytokine that possesses both immunomodulatory and meta- (BMP1). GRN is a growth factor with marked functional bolic functionality, inhibiting common pro-inflammatory path- similarities to the epidermal growth factor family and has been ways, as well as being involved in glucose and insulin shown to regulate cellular proliferation, particularly in the regulation.52,53 PRDX1, PRDX2, PRDX6, and GPX3 are all hematopoietic and reproductive systems.40 Similarly, GREM2 involved in systems of oxidative stress regulation, protecting cells, also regulates cellular differentiation and is involved in the Wnt enzymes, and other proteins from oxidative damage,54 whereas signaling pathways.41 NPNT, BMP1, and CRIM1 have been HLA-A and HLA-DRA are both membrane proteins, involved in shown to interact with several growth factors and to play im- presenting foreign antigens to the immune system.55 While the portant roles in the differentiation of osteoclasts, chondrocytes, presence of these proteins in human milk does not indicate that and motor neurons, respectively, although more general roles in they are functionally active in the infant gut, they nonetheless − development have also been proposed.42 44 While further provide useful targets for further investigation. research is required to assess whether these proteins retain their To investigate whether the human milk proteome under- activity upon digestion, it is known that human-milk-fed infants goes detectable changes at different stages of mammary gland experience developmental advantages over formula-fed in- development, a pilot study involving iTRAQ analysis was used − fants,45 47 and it is possible that proteins within this group to compare the protein composition of term and preterm milk. may be partly responsible. Although iTRAQ experiments are designed to detect differ- Proteins involved in immunity and inflammation constitute ences in the fractional composition of protein samples, in the 24% of the proteins identifed (Figure 5), signifying the critical present study the original protein concentration of each pooled

1711 dx.doi.org/10.1021/pr2008797 | J. Proteome Res. 2012, 11, 1696−1714 Journal of Proteome Research Article sample was equivalent. This means that the differences in were differentially expressed in term and preterm milk, with fractional composition also directly correspond to differences in BSSL present at a higher level in preterm milk and both LPL the concentration of individual proteins between term and pre- and BTD found at a higher level in term milk. The presence of term milk. these enzymes in human milk is thought to compensate for a In order to detect differentially expressed proteins of both lack of endogenous enzymes in the immature pancreatic juice high and low abundance in term and preterm milk we adopted of newborns, enabling the efficient digestion of triacylglycerols − a workflow including parallel iTRAQ experiments. In a similar and biotin, respectively.59 61 Given the importance of the 20 manner to a recent publication, iTRAQ comparisons of pool- absorption of fat and biotin to an infant’s growth and develop- ed term and preterm milk samples were conducted both with ment, it is likely that the delivery of these enzymes is of and without prior ProteoMiner treatment. This enabled us to particular importance to preterm infants. Again, their pattern of distinguish between high abundance proteins, those identified expression defies teleological explanation, in that while the without ProteoMiner treatment, and low abundance proteins, higher levels of BSSL in preterm milk may promote additional those identified only when ProteoMiner treatment was used growth in preterm infants, the lower levels of BTD and LPL prior to iTRAQ analysis. We found that ProteoMiner treat- may be of detriment. ment did significantly alter the relative quantitation of the high Identifying proteins that are differentially expressed in term abundance proteins, and therefore in the iTRAQ analysis of and preterm milk may also be useful diagnostically. There have ProteoMiner-treated samples, only the relative abundance ratios been a number of studies investigating the differences in mam- of the low abundance proteins were considered to be accurate. SDS-PAGE analysis was performed to verify the iTRAQ mary gland physiology and metabolism after preterm birth; however, the link between these physiological differences and quantitative analysis of high abundance proteins. The direction 4,34,62 of the differences in expression of five proteins, bile-salt-stimu- preterm milk composition largely remain unknown. The lated lipase, lactoferrin, β-casein, sIgA, and serum albumin, in differential expression of proteins in preterm milk may highlight term and preterm milk were consistent between the two potential regulatory and metabolic pathways that are disrupted analytical methods; however, the iTRAQ experiment over- after preterm delivery. For example, the higher levels of serum estimated the magnitude of the difference. Alpha-lactalbumin albumin that we found in preterm milk may be due to the per- was found to be differentially expressed in the iTRAQ experi- sistence of an open paracellular pathway after delivery in pre- ment but not by SDS-PAGE. It is likely that these differences term mothers, allowing the flow of serum proteins into the between the two methods are due to an overestimation of the mammary alveoli. Indeed, levels of serum albumin in milk have level of background contamination of the iTRAQ spectra by been used previously as a marker of an open paracellular the analysis software.56 Although previous studies report the pathway.63,64 Similarly, the low level of prolactin-inducible pro- quantitative accuracy of iTRAQ analysis of low abundance tein in preterm milk is interesting from a diagnostic perspective, proteins after ProteoMiner-treated samples,20 these results were in that it may reflect low levels of circulating maternal prolactin. not verified in the present study. Despite these limitations, Indeed, it has been shown that preterm mothers are more likely these results provide support for the use of an iTRAQ approach to have lower levels of serum prolactin, which may be re- to identify differentially expressed proteins in human milk and sponsible in part for lower levels of milk production in preterm illustrate how ProteoMiner treatment can be incorporated mothers.65 Tenascin is another protein that was found to be within a quantitative analysis. present at much lower levels in preterm milk compared to term We found 28 proteins that had significantly higher expression milk (Table 2). Having been previously implicated in mammary levels in preterm milk compared to term milk and 27 proteins gland development and differentiation,66,67 its concentration in that had significantly lower levels of expression (Table 2). A milk may also potentially be useful as a marker of mammary number of these differentially expressed proteins such as lacto- gland development after preterm delivery. ferrin, lysozyme, polymeric immunoglobulin receptor, lactadher- It is also of interest that distinct sets of proteins were identi- in, prolactin inducible protein, Ig heavy chain, mucin-4, vitro- fied in the iTRAQ analysis of pooled term milk collected 7−14 nectin, and complement C3 are associated with the immune 57 days postpartum and in the 2D-LC analysis of pooled term milk response, protecting infants against infection. A number of collected 15−28 days postpartum, with over 100 proteins targeted studies have found higher levels of specific immunologic unique to each sample (Figure 4C). Although no quantitative factors in preterm milk and have accounted for these differences comparison was conducted comparing these samples, it is likely on a teleological basis, arguing that the composition of preterm that many of these proteins change in relative abundance in milk differs from term milk to render it more suited to the protection of vulnerable preterm infants.10,58 Our data contra- term milk over this time period and therefore may also re- dicts this assertion, in that we found no concerted difference in present proteins of interest with regard to mammary gland the expression of immunological proteins between term and pre- development throughout lactation. term milk, with some present at higher levels in preterm milk In summary, the present study represents the most com- and vice versa (Table 2). This suggests that differences in protein prehensive study to date of the human milk proteome, iden- composition between term and preterm milk are due to phys- tifying 261 novel proteins, as well as documenting changes in iological differences in the mammary gland affecting protein the relative abundance of proteins in term and preterm milk. synthetic and transport pathways, rather than being the result of Knowledge obtained from this characterization of the proteins differing infant requirements, and supports the idea that as in human milk will provide insights into the regulatory mech- preterm infants have only been able to survive in recent years, anisms involved in the synthesis and secretion of human milk, there has not been any selection pressure to encourage an in- particularly after preterm delivery, as well as identifying poten- crease in protective proteins in preterm milk.4 tial proteins in human milk responsible for the nutritional, Similarly, three digestive enzymes, biotinidase (BTD), lipo- immunological, and developmental advantages conferred onto protein lipase (LPL), and bile-salt-stimulated lipase (BSSL) the breastfed infant.

1712 dx.doi.org/10.1021/pr2008797 | J. Proteome Res. 2012, 11, 1696−1714 Journal of Proteome Research Article ■ CONCLUSIONS (4) Atkinson, S. A. Effects of gestational stage at delivery on human milk components. In Handbook of Milk Composition; Jensen, R. G., Ed.; We investigated the effectiveness of a number of different an- Academic Press: San Diego, 1995; pp 222−237. alytical techniques to analyze the human milk proteome. Dynamic (5) Montagne, P.; Cuilliere, M. L.; Mole, C.; Bene, M. C.; et al. range compression of human skim milk by the depletion of casein Immunological and nutritional composition of human milk in relation and ProteoMiner treatment followed by 2D LC−MS/MS was the to prematurity and mothers’ parity during the first 2 weeks of lactation. most successful approach. In total, 415 proteins were identified, J. Pediatr. Gastroenterol. Nutr. 1999, 29,75−80. over half of which had not been found before in human milk. In (6) Bielicki, J.; Huch, R.; von Mandach, U. Time-course of leptin addition, iTRAQ analysis was used to identify differentially levels in term and preterm human milk. Eur. J. Endocrinol. 2004, 151, expressed proteins between term and preterm milk, providing 271−276. insights into metabolic differences in the mammary gland after (7) Dvorak, B.; Fituch, C. C.; Williams, C. S.; Hurst, A. N. M.; et al. preterm birth in comparison to term birth. Increased epidermal growth factor levels in human milk of mothers with extremely premature infants. Pediatr. Res. 2003, 54,15−19. (8) Montagne, P.; Cuillere, M.; Mole, C.; Bene, M.;et al., Changes in ■ ASSOCIATED CONTENT lactoferrin and lysozyme levels in human milk during the first twelve *S Supporting Information weeks of lactation. In Bioactive Components of Human Milk; Newburg, D., Ed.; Springer: New York, 2001; Vol. 1. Supporting Information Data File 1 contains individual (9) Kunz, C.; Lonnerdal, B. Re-evaluation of the whey protein/casein annotated MS/MS spectra of each of the proteins for which ratio of human milk. Acta Paediatr. 1992, 81, 107−112. only one peptide was identified. Supporting Information Data (10) Grosse, S. J.; Buckley, R. H.; Wakil, S. S.; McAllister, D. C.; et al. File 2 displays the preparative 2DE gel of the skim milk after Elevated IgA concentration in milk produced by mothers delivered of casein depletion and ProteoMiner treatment and each of the preterm infants. J. Pediatr. 1981, 99, 389−393. spots that was analyzed by MS. Supporting Information Data (11) Galvani, M.; Hamdan, M.; Righetti, P. G. Two-dimensional gel File 3 contains tables displaying all of the peptides used for electrophoresis/matrix-assisted laser desorption/ionisation mass spec- identifying proteins in the 1D-LC, 2D-LC, iTRAQ, 1D SDS- trometry of a milk powder. Rapid Commun. Mass Spectrom. 2000, 14, 1889−1897. PAGE, and 2DE experiments and their corresponding (12) Galvani, M.; Hamdan, M.; Righetti, P. G. Two-dimensional gel identification strengths. Supporting Information File 4 is a electrophoresis/matrix-assisted laser desorption/ionization mass spec- table displaying the classification of the proteins according to trometry of commercial bovine milk. Rapid Commun. Mass Spectrom. their function and location. This material is available free of 2001, 15, 258−264. charge via the Internet at http://pubs.acs.org. (13) Holland, J. W.; Deeth, H. C.; Alewood, P. F. Proteomic analysis of kappa-casein micro-heterogeneity. Proteomics 2004, 4, 743−752. ■ AUTHOR INFORMATION (14) Mange, A.; Bellet, V.; Tuaillon, E.; Van de Perre, P.; et al. Comprehensive proteomic analysis of the human milk proteome: Corresponding Author Contribution of protein fractionation. J. Chromatogr., B 2008, 876, − *Tel.: +61 6488 4428. Fax: +61 8 6488 7086. E-mail: 10224872@ 252 256. student.uwa.edu.au. (15) Yamada, M.; Murakami, K.; Wallingford, J. C.; Yuki, Y. Identi- fication of low-abundance proteins of bovine colostral and mature milk using two-dimensional electrophoresis followed by microsequencing ■ ACKNOWLEDGMENTS and mass spectrometry. Electrophoresis 2002, 23, 1153−1160. The MS analyses were performed in facilities provided by the (16) Picariello, G.; Ferranti, P.; Mamone, G.; Roepstorff, P.; et al. Lotterywest State Biomedical Facility-Proteomics node at the Identification of N-linked glycoproteins in human milk by hydrophilic Western Australian Institute for Medical Resesarch. This study was interaction liquid chromatography and mass spectrometry. Proteomics 2008, 8, 3833−3847. funded by an unrestricted grant from Medela AG (Switzerland) to (17) Panchaud, A.; Kussmann, M.; Affolter, M. Rapid enrichment of the University of Western Australia. C.E. Molinari was supported ’ bioactive milk proteins and iterative, consolidated protein identi- by a scholarship from the Western Australian Women s Service fication by multidimensional protein identification technology. Guild (2009−2011), and an Australian Postgraduate Award Proteomics 2005, 5, 3836−3846. (2009−2011). (18) Palmer, D. J.; Kelly, V. C.; Smit, A.-M.; Kuy, S.; et al. Human colostrum: Identification of minor proteins in the aqueous phase by ■ ABBREVIATIONS proteomics. Proteomics 2006, 6, 2208−2216. (19) Liao, Y.; Alvarado, R.; Phinney, B.; Lonnerdal, B. Proteomic 1D-LC, one-dimensional liquid chromatography; 2DE, two- characterization of human milk whey proteins during a twelve-month dimensional gel electrophoresis; 2D-LC, two-dimensional lactation period. J. Proteome Res. 2011, 10, 1746−1754. liquid chromatography; iTRAQ, isobaric tags for relative and (20) Bandhakavi, S.; Van Riper, S. K.; Tawfik, P. N.; Stone, M. D.; absolute quantitation; MALDI, matrix-assisted laser desorp- et al. Hexapeptide libraries for enhanced protein PTM identification tion/ionization; MS, mass spectrometry; PAGE, polyacrylamide and relative abundance profiling in whole human saliva. J. Proteome gel electrophoresis; sIgA, secretory IgA; SDS, sodium dodecyl Res. 2011, 10, 1052−1061. sulfate; TOF, time-of-flight (21) Hartwig, S.; Czibere, A.; Kotzka, J.; Pablack, W.; et al. Combinatorial hexapeptide ligand libraries (ProteoMiner): An innovative fractionation tool for differential quantitative clinical ■ REFERENCES proteomics. Arch. Physiol. Biochem. 2009, 115, 155−160. (1) Lonnerdal, B. Nutritional and physiologic significance of human (22) Kunz, C.; Lonnerdal, B. Human-milk proteins: Analysis of milk proteins. Am. J. Clin. Nutr. 2003, 77, 1537S−1543. casein and casein subunits by anion exchange chromatography, gel (2) Srinivasan, M.; Patel, M. S. Metabolic programming in the electrophoresis, and specific staining methods. Am. J. Clin. Nutr. 1990, immediate postnatal period. Trends Endocrinol. Metab. 2007, 19, 146− 51,37−46. 152. (23) Molinari, C. E.; Casadio, Y. S.; Arthur, P. G.; Hartmann, P. E. (3) Rigo, J.; Senterre, J. Nutritional needs of premature infants: The effect of storage at 25 °C on proteins in human milk. Int. Dairy J. Current issues. J. Pediatr. 2006, 149, S80−S88. 2011, 21, 286−293.

1713 dx.doi.org/10.1021/pr2008797 | J. Proteome Res. 2012, 11, 1696−1714 Journal of Proteome Research Article

(24) Laemmli, U. Cleavage of structural proteins during the assembly (45) Dvorak, B. Milk epidermal growth factor and gut protection. of the head of bacteriophage T4. Nature 1970, 227, 680−685. J. Pediatr. 2010, 156, S31−S35. (25) Abramoff, M. D.; Magelhaes, P. J.; Ram, S. J. Image processing (46) Lucas, A.; Morley, R.; Cole, T. J.; Lister, G.; et al. Breast milk with ImageJ. Biophoton Int. 2004, 11,36−42. and subsequent intelligence quotient in children born preterm. Lancet (26) Lui, J. K. C.; Lipscombe, R.; Arthur, P. G. Detecting changes in 1992, 339, 261−264. the thiol redox state of proteins following a decrease in oxygen (47) Quinn, P. J.; O’Callaghan, M.; Williams, G. M.; Najman, J. M.; concentration using a dual labeling technique. J. Proteome Res. 2009, 9, et al. The effect of breastfeeding on child development at 5 years: 383−392. A cohort study. J. Pediatr. Child Health 2001, 37, 465−469. (27) Shevchenko, A.; Tomas, H.; Havlis, J.; Olsen, J.; et al. In-gel (48) Vorbach, C.; Capecchi, M.; Penninger, J. Evolution of the digestion for mass spectrometric characterisation of proteins and mammary gland from the innate immune system? Bioessays 2006, 28, − proteomes. Nat. Protoc. 2007, 1, 2856−2860. 606 616. (28) D’Alessandro, A.; Scaloni, A.; Zolla, L. Human milk proteins: An (49) Xanthou, M. Immune protection of human milk. Biol. Neonate − interactomics and updated functional overview. J. Proteome Res. 2010, 1998, 74, 121 133. 9, 3339−3373. (50) Lonnerdal, B.; Iyer, S. Lactoferrin: Molecular structure and − (29) Mouton-Barbosa, E.; Roux-Dalvai, F.; Bouyssie, D.; Berger, F.; biological function. Annu. Rev. Nutr. 1995, 15,93 110. (51) Strate, B. v. d.; Belijaars, L.; Molema, G.; Harmsen, M.; et al. et al. In-depth exploration of cerebrospinal fluid by combining peptide − ligand library treatment and label-free protein quantification. Mol. Cell. Antiviral activities of lactoferrin. Antiviral Res. 2001, 52, 225 239. − (52) Davis, K.; Scherer, P. Adiponectin: No longer the lone soul in Proteomics 2010, 9, 1006 1021. − (30) Roux-Dalvai, F.; Gonzalez de Peredo, A.; Sime, C.; Guerrier, L.; the fight against insulin resistance? Biochem. J. 2008, 416,e7 e9. et al. Extensive analysis of the cytoplasmic proteome of human (53) Nayak, A.; Pedenekar, L.; Reid, K. B. M.; Kishore, U. Complement and non-complement activating functions of C1q: A erythrocytes using the peptide ligand library technology and advanced prototypical innate immune molecule. Innate Immun. 2011. mass spectrometry. Mol. Cell. Proteomics 2008, 7, 2254−2269. (54) Rhee, S. G.; Yang, K. S.; Kang, S. W.; Woo, H. A.; et al. (31) Guerrier, L.; Pier Giorgio, R.; Egisto, B. Reduction of dynamic Controlled elimination of intracellular H O : Regulation of peroxir- protein concentration range of biological extracts for the discovery of 2 2 edoxin, catalase, and glutathione peroxidase via post-translational low-abundance proteins by means of hexapeptide ligand library. Nat. − − modification. Antioxid. Redox Signaling 2005, 7, 619 26. Protoc. 2008, 3, 883 890. (55) Garcia-Lora, A.; Algarra, I.; Garrido, F. MHC class I antigens, (32) Patton, S.; Huston, G. Incidence and characteristics of cell immune surveillance, and tumor immune escape. J. Cell. Physiol. 2003, pieces on human milk fat globules. Biochim. Biophys. Acta 1988, 965, − − 195, 346 355. 146 153. (56) Karp, N. A.; Huber, W.; Sadowski, P. G.; Charles, P. D.; et al. (33) Brooker, B. E. The epithelial cells and cell fragments in human Addressing accuracy and precision issues in iTRAQ quantitation. Mol. − milk. Cell Tissue Res. 1980, 210, 321 332. Cell. Proteomics 2010, 9, 1885−1897. (34) Cregan, M. D.; De Mello, T. R.; Kershaw, D.; McDougall, K.; (57) Field, C. J. The immunological components of human milk and et al. Initiation of lactation in women after preterm delivery. Acta their effect on immune development in Infants. J. Nutr. 2005, 135, − Obstet. Gynecol. Scand. 2002, 81, 870 877. 1−4. (35) Fetherston, C. M.; Lai, C. T.; Mitoulas, L. R.; Hartmann, P. E. (58) Chandra, R. K. Immunoglobulin and protein levels in breast Excretion of lactose in urine as a measure of increased permeability of milk produced by mothers of preterm infants. Nutr. Res. (N. Y., NY, U. the lactating breast during inflammation. Acta Obstet. Gynecol. Scand. S.) 1982, 2,27−30. 2006, 85,20−25. (59) Bernback, S.; Blackberg, L.; Hernell, O. The complete digestion (36) Arthur, P. G.; Kent, J.; Hartmann, P. E. Metabolites of lactose of human milk triacylglycerol in vitro requires gastric lipase, pancreatic synthesis in milk from diabetic and non-diabetic women during colipase-dependent lipase, and bile salt-stimulated lipase. J. Clin. Invest. lactogenesis II. J. Pediatr. Gastroenterol. Nutr. 1994, 19, 100−108. 1990, 85, 1221−1226. (37) Semba, R. D.; Kumwenda, N.; Taha, T. E.; Hoover, D. R.; et al. (60) Oizumi, J.; Hirano, M.; Hayakawa, K.; Daimatsu, T.;et al. The Mastitis and immunological factors in breast milk of lactating women significance of breast milk biotinidase. Pediatr. Int. 1989, 31. in Malawi. Clin. Diagn. Lab. Immunol. 1999, 6, 671−674. (61) Suormala, T.; Wick, H.; Baumgartner, E. Low biotinidase (38) Boehmer, J. L.; Bannerman, D. D.; Shefcheck, K.; Ward, J. L. activity in plasma of some preterm infants: Possible source of false- Proteomic analysis of differentially expressed proteins in bovine milk positive screening results. Eur. J. Pediatr. 1988, 147, 478−480. during experimentally induced escherichia coli mastitis. J. Dairy Sci. (62) Anderson, G. The effect of prematurity on milk composition 2008, 91, 4206−4218. and its physiological basis. Fed. Proc. 1984, 43, 2438−2442. (39) Kaufmann, S. H. E. Heat shock proteins and the immune (63) Hartmann, P. E.; Kulski, J. K. Changes in the composition of the response. Immunol. Today 1990, 11, 129−136. mammary secretion of women after abrupt termination of breast (40) Bateman, A.; Bennett, H. P. Granulins: The structure and feeding. J. Physiol. 1978, 275,1−11. function of an emerging family of growth factors. J. Endocrinol. 1998, (64) Stelwagen, K. Mammary epithelial cell tight junction integrity 158, 145−51. and mammary blood flow during an extended milking interval in goats. − (41) Klapholz-Brown, Z.; Walmsley, G.; Nusse, Y.; Brown, P. J. Dairy Sci. 1994, 77, 426 432. Transcriptional program induced by Wnt protein in human fibroblasts (65) Hill, P. D.; Aldag, J. C.; Demirtas, H.; Naeem, V.; et al. suggest mechanisms for cell cooperativity in defining tissue micro- Association of serum prolactin and oxytocin with milk production in − environments. PLoS One 2007, 2, e945. mothers of preterm and term infants. Biol. Res. Nurs. 2009, 10, 340 (42) Kahai, S.; Lee, S.-C.; Seth, A.; Yang, B. B. Nephronectin 349. promotes osteoblast differentiation via the epidermal growth factor- (66) Jones, F. S.; Jones, P. L. The tenascin family of ECM − glycoproteins: Structure, function, and regulation during embryonic like repeats. FEBS Lett. 2010, 584, 233 238. − (43) Kolle, G.; Georgas, K.; Holmes, G. P.; Little, M. H.; et al. development and tissue remodeling. Dev. Dyn. 2000, 218, 235 259. CRIM1, a novel gene encoding a cysteine-rich repeat protein, is (67) Jones, P. L.; Boudreau, N.; Myers, C. A.; Erickson, H. P.; et al. Tenascin-C inhibits extracellular matrix-dependent gene expression in developmentally regulated and implicated in vertebrate CNS develop- mammary epithelial cells. Localization of active regions using ment and organogenesis. Mech. Dev. 2000, 90, 181−193. recombinant tenascin fragments. J. Cell Sci. 1995, 108, 519−527. (44) Anderson, H.; Hodges, P.; Aguilera, X.; Missana, L.; et al. Bone morphogenetic protein (BMP) localization in developing human and rat growth plate, metaphysis, epiphysis, and articular cartilage. J. Histochem. Cytochem. 2000, 48, 1493−1502.

1714 dx.doi.org/10.1021/pr2008797 | J. Proteome Res. 2012, 11, 1696−1714