Serum Proteomes Distinguish Children Developing Type 1 Diabetes in a Cohort with HLA-Conferred Susceptibility

Diabetes Volume 64, June 2015 2265

Robert Moulder,1 Santosh D. Bhosale,1 Timo Erkkilä,2 Essi Laajala,1 Jussi Salmi,1 Elizabeth V. Nguyen,1 Henna Kallionpää,1 Juha Mykkänen,3,4 Mari Vähä-Mäkilä,3,4 Heikki Hyöty,5,6 Riitta Veijola,7 Jorma Ilonen,8,9 Tuula Simell,3,4 Jorma Toppari,3,4,10 Mikael Knip,11–14 David R. Goodlett,1,15 Harri Lähdesmäki,1,2 Olli Simell,3,4 and Riitta Lahesmaa1

Serum Proteomes Distinguish Children Developing Type 1 Diabetes in a Cohort With HLA-Conferred Susceptibility

Diabetes 2015;64:2265–2278 | DOI: 10.2337/db14-0983 GENETICS/GENOMES/PROTEOMICS/METABOLOMICS

We determined longitudinal serum proteomics profiles the serum proteome in healthy children and children from children with HLA-conferred diabetes susceptibil- progressing to type 1 diabetes, including new protein ity to identify changes that could be detected before candidates, the levels of which change before clinical seroconversion and positivity for disease-associated diagnosis. autoantibodies. Comparisons were made between children who seroconverted and progressed to type 1 diabetes (progressors) and those who remained auto- The measurement of islet cell autoantibodies is currently antibody negative, matched by age, sex, sample peri- the principle means of identifying an emerging threat of odicity, and risk group. The samples represented the developing type 1 diabetes (1). The risks associated with prediabetic period and ranged from the age of 3 months the appearance of islet antibodies have been evaluated in to 12 years. After immunoaffinity depletion of the most depth, and overall, the appearance of multiple biochemi- abundant serum proteins, isobaric tags for relative and fi absolute quantification were used for sample labeling. cally de ned autoantibodies correlates with progression Quantitative proteomic profiles were then measured to disease irrespective of family history, genetic risk for 13 case-control pairs by high-performance liquid group, or autoantibody combination (1). Nevertheless, it fi chromatography-tandem mass spectrometry (LC-MS/MS). still remains open whether nding even earlier indications Additionally, a label-free LC-MS/MS approach was used of future disease development is possible. Such markers to analyze depleted sera from six case-control pairs. could shed further light on disease etiology and poten- Importantly, differences in abundance of a set of tially be used in the evaluation of risks and preventive proteins were consistently detected before the appear- treatments. ance of autoantibodies in the progressors. Based on Proteomic analyses in the study of type 1 diabetes has top-scoring pairs analysis, classification of such pro- been previously reviewed (2) and applied in studies gressors was observed with a high success rate. Over- addressing differences in the sera of patients with diabe- all, the data provide a reference of temporal changes in tes and subjects without diabetes (3–5). Zhang et al. (5)

1Turku Centre for Biotechnology, University of Turku, Turku, Finland 12Research Program, Diabetes and Obesity, University of Helsinki, Helsinki, 2Department of Information and Computer Science, Aalto University School of Finland Science, Espoo, Finland 13Department of Pediatrics, Tampere University Hospital, Tampere, Finland 3Department of Pediatrics, University of Turku, Turku, Finland 14Folkhälsan Research Institute, Helsinki, Finland 4Department of Pediatrics, Turku University Hospital, Turku, Finland 15Department of Pharmaceutical Sciences, University of Maryland, Baltimore, MD 5 School of Medicine, University of Tampere, Tampere, Finland Corresponding author: Riitta Lahesmaa, riitta.lahesmaa@btk.fi. 6Fimlab Laboratories, Pirkanmaa Hospital District, Tampere, Finland Received 25 June 2014 and accepted 8 January 2015. 7University of Oulu and Oulu University Hospital, Department of Pediatrics, Oulu, Finland This article contains Supplementary Data online at http://diabetes 8Department of Clinical Microbiology, University of Eastern Finland, Kuopio, .diabetesjournals.org/lookup/suppl/doi:10.2337/db14-0983/-/DC1. Finland © 2015 by the American Diabetes Association. Readers may use this article as 9Immunogenetics Laboratory, University of Turku, Turku, Finland long as the work is properly cited, the use is educational and not for profit, and 10Departments of Physiology and Pediatrics, University of Turku, Turku, Finland the work is not altered. 11Children’s Hospital, University of Helsinki and Helsinki University Central Hos- pital, Helsinki, Finland 2266 Serum Proteomes En Route to Type 1 Diabetes Diabetes Volume 64, June 2015 compared protein levels in plasma from patients with to report longitudinal proteomics profiles in children who type 1 diabetes and healthy subjects, observing significant develop type 1 diabetes as well as such profiles in healthy differences in the abundance of 24 proteins. Similarly, Zhi children. et al. (4) detected differences in the levels of 21 serum RESEARCH DESIGN AND METHODS proteins between patients with type 1 diabetes and healthy subjects; six of the proteins were validated by A schematic of the experimental design is illustrated in immunoassay. Fig. 1. Detailed description of the proteomics measure- Although in-depth comparisons of proteins in samples ments, samples comparisons, and availability of the raw from healthy subjects and patients with type 1 diabetes data are provided as supplementary information. fi have distinguished the diseased state, the identi cation of Subjects and Sample Collection changes preceding this aggressive autoimmune disease is All children studied were participants in the Finnish DIPP important for disease prediction and prevention. McGuire study (9), where children identified as at risk for type 1 et al. (6) used a proteomic approach to identify predictive diabetes based on HLA genotype were followed prospec- markers in the cord blood of children in whom type 1 tively from birth. Venous nonfasting blood samples were diabetes developed later. Although their measurements collected at each study visit; sera were separated and with surface-enhanced laser desorption/ionization mass spectrometry revealed different patterns, the discriminat- ing peaks were not identified. To establish the origin and changes associated with the development of type 1 diabetes, careful selection of appropriate study groups is essential, such as have been established by prospective sampling from at-risk individuals (7,8). The Finnish Type 1 Diabetes Prediction and Prevention (DIPP) project collected samples from Finnish children with HLA-defined predisposition to type 1 diabetes (7,9), thus creating an extensive prospective sample collection from birth to diagnosis or otherwise healthy until 15 years of age. This resource has allowed investiga- tion of the longitudinal profiles of a wide range of factors in children who developed type 1 diabetes, using samples ranging from early infancy to diagnosis, as well as sample measurements from carefully matched control subjects (10–14). In the current study, we determined the longitudinal serum proteomics profiles of a group of children who aretype1diabetessusceptibleenrolledintheDIPP study.Themeasurementsweremadeinserafrom38 children comprising 19 type 1 diabetes case-control pairsmatchedbydateandlocationofbirth,sex,and HLA-conferred genetic risk. The samples selected for analysis represent the time course from autoantibody negativity to seroconversion to diagnosis. The analyses were made using two mass spectrometry–based quantitative proteomics techniques. First, we used isobaric tags for relative and absolute quantification (iTRAQ) reagents, which have previously been extensively used in serum proteomics applications, including the development of robust analytical protocols and applied studies up to the scale of hundreds of subjects (15,16). Second, we used a label-free method, which — has also been applied in serum proteomics analyses Figure 1 Schematic presentation of the study design. Using a prospective longitudinal serum sample collection from children with an (17,18). The present results reveal a spectrum of HLA-conferred risk for type 1 diabetes. Samples were selected changesanddifferencesintheserumproteinprofiles based on clinical outcome and the titers of diabetes-associated between children progressing to type 1 diabetes and autoantibodies. The samples were prepared for proteomics analysis matched control subjects. Some of these changes by mass spectrometry. Comparisons were made between children who developed type 1 diabetes and age-, HLA risk–,andsex- were consistently detected before the appearance of matched control subjects. Two quantitative approaches were ap- autoantibodies. To our knowledge, this study is the first plied: first, iTRAQ reagents and second, a label-free approach. diabetes.diabetesjournals.org Moulder and Associates 2267 stored at 270°C within 3 h from collection. Serum islet LC-MS/MS Data Processing cell autoantibody (ICA) measurements were made as pre- The iTRAQ data were analyzed with ProteinPilot soft- viously described (19). For ICA-positive children, levels of ware using the Paragon identification algorithm (22) with a GAD antibody (GADA), tyrosine phosphatase-related pro- Human Swiss-Prot database (18 August 2011; 20,245 tein antibody (IA-2A), and insulin antibodies (IAA) were entries). The database searches were made in thorough also analyzed. mode, specifying 8plex iTRAQ quantification, trypsin The proteomics measurements were performed on sera digestion, and MMTS (S-methyl methanethiosulfonate) from 19 case children who developed type 1 diabetes modification of cysteine. The QSTAR data were analyzed during the DIPP follow-up. Prospective serum samples directly, and the Orbitrap data were converted to mascot (5–11 per child) were selected to represent phases of generic format using Proteome Discoverer version 1.3 disease progression from autoantibody negativity to sero- (Thermo Scientific) (23). False discovery rates (FDRs) for conversion to overt disease. A persistently autoantibody- protein identification were estimated using the Protein- negative control child was matched with each case child Pilot PSEP functionality (24,25). A confidence threshold (typically in the order of seven samples per child) based of 95% for protein identification was applied. iTRAQ on date and place of birth, sex, and HLA-DQB1 genotype. ratios were calculated using ProteinPilot. The prospective control serum samples were matched LFQ data were analyzed with Proteome Discoverer with the case samples by age at sample draw. Altogether, together with Mascot 2.1 (Matrix Science). The search 266 serum samples were analyzed of which sera from 26 criteria were trypsin digestion, MMTS modification of children (13 case-control pairs) were processed for iTRAQ cysteine, deamidation of N/Q, and methionine oxidation, analysis and sera from 12 children (6 case-control pairs) using the aforementioned database. For the quantitative for analysis using a label-free approach (Table 1, Supple- analysis, Progenesis version 4.0 software was used for mentary Table 1, and Fig. 2). feature detection, alignment, and calculation of intensity- based abundance measurements for each protein (26). To Sample Preparation facilitate comparison of the label-free data with the Serum samples were depleted of the most abundant iTRAQ results, the intensity values of each protein were proteins using immunoaffinity columns from Beckman scaled relative to the median intensity of each protein Coulter (ProteomeLab IgY-12) and Agilent (Hu14) col- across the paired case-control sample series. umns. The same depletion method was always applied to the follow-up samples of each case-control pair. Data Analysis For iTRAQ labeling, the samples were processed in Serum Proteomics Differences Between Healthy accordance with the manufacturer’s protocol for 8plex Children and Type 1 Diabetes Progressors reagents (AB Sciex, Framingham, MA) and then fraction- Case-control abundance ratios were calculated for the ated using strong cation exchange chromatography as pre- paired samples. The ratios were log2 transformed and viously described (20). Samples from 26 children were used in rank product analyses (27) to identify differences compared using the iTRAQ method, applying 27 paired/ throughout the time series (n = 19), before the detection cross-referenced 8plex iTRAQ labeling schemes of the of autoantibody seroconversion (n = 14), and before di- samples. Samples from 12 additional children were ana- agnosis (n = 19). The rank product analyses were made lyzed in quadruplicate using a label-free quantitative with 10,000 times permutations, and an FDR #5% was (LFQ) approach (depletion with the Hu14 system), applied (Benjamini-Hochberg correction). The averaged with concentration and digestion performed in a similar log2 case-control ratios were used to compare time inter- manner as for the iTRAQ samples (21), with the digests vals selected on the basis of the similarity of the sample otherwise unfractionated beforehigh-performanceliquid series as indicated in Table 2. chromatography-tandem mass spectrometry (LC-MS/MS). Because the two data sets (iTRAQ vs. label-free) demonstrated a close overlap of the proteins repeatedly LC-MS/MS Analysis detected and quantified, the ranked results from these LC-MS/MS analyses were performed with a QSTAR Elite fl analyses were combined to investigate the longitudinal time-of- ight instrument and an Orbitrap Velos Pro paired differences and trends. Previous studies have Fourier transform instrument. For the analysis of indicated that these two methods are complementary iTRAQ-labeled samples, the collision-induced dissociation (28,29). If the protein was absent in either the case or the and higher-energy collisional dissociation modes were control subject, the paired measurement was not used or used to record positive ion tandem mass spectra for the imputed to minimize the influence of missing values. QSTAR Elite and Orbitrap Velos, respectively. The LFQ data were acquired with the Orbitrap Velos using Longitudinal Serum Proteomics Profiles in Children collision-induced dissociation. Chromatographic separa- Progressing to Type 1 Diabetes tions were made with 150 mm 3 75 mm internal diam- Spearman rank correlation analyses were used to assess eter columns packed with magic C18-bonded silica (200 Å) whether any of the protein profiles were related to using binary gradients of water and acetonitrile with 0.2% progression to diabetes. The analyses were made for the formic acid. collected case-control abundance ratios of the paired samples. 2268 eu rtoe nRuet ye1Diabetes 1 Type to Route En Proteomes Serum

Table 1—Summary of the children progressing to type 1 diabetes whose samples were studied with proteomics Number of samples analyzed ID Sex HLA-DQB1 risk alleles Age at seroconversion (years) Age at diagnosis (years) Autoantibodies detected Case Control Analysis method D1 M *02, *03:02 1.3 2.2 ICA, IAA, GADA, IA-2A 7 7 iTRAQ D2 M *02, *03:02 1.4 4.0 ICA, IAA, GADA, IA-2A 7 7 iTRAQ D3 M *02, *03:02 1.3 3.9 ICA, IAA, GADA, IA-2A 7 7 iTRAQ D4 F *02, *03:02 1.5 3.3 ICA, IAA, GADA, IA-2A 7 7 iTRAQ D5 F *02, *03:02 3.4 7.0 ICA, IAA, GADA, IA-2A 7 7 iTRAQ D6 F *02, *03:02 0.5 4.0 ICA, IAA, GADA, IA-2A 7 7 iTRAQ D7 F *02, *03:02 0.6 4.1 ICA, IAA, GADA, IA-2A 7 7 iTRAQ D8 M *02, *03:02 1.0 4.4 ICA, IAA, GADA, IA-2A 7 7 iTRAQ D9 M *02, *03:02 2.5 3.6 ICA, IAA, GADA 7 7 iTRAQ D10 M *03:02, x 1.5 2.5 ICA, IAA, GADA, IA-2A 7 7 iTRAQ D11 M *03:02, x 1.3 4.0 ICA, IAA, IA-2A 11 9 iTRAQ D12 M *03:02, x 2.0 2.2 ICA, IAA, GADA, IA-2A 5 7 iTRAQ D13 M *03:02, x 3.5 5.5 ICA, GADA, IA-2A 7 6 iTRAQ D14 M *02, *03:02 6.1 8.8 ICA, IAA, GADA, IA-2A 7 8 Label free D15 M *02, *03:02 2.6 8.3 ICA, IAA 6 7 Label free D16 F *02, *03:02 1.0 10.0 ICA, IAA, GADA, IA-2A 6 6 Label free D17 F *02, *03:02 5.0 7.7 ICA, IAA, GADA, IA-2A 7 7 Label free

D18 M *03:02, x 1.3 12.1 ICA, IAA, GADA, IA-2A 6 8 Label free Diabetes D19 F *03:02, x 1.3 8.6 ICA, IAA, IA-2A 7 8 Label free See also Supplementary Table 1 for more information on the case subjects and matched control subjects. x *02, *03:01, *06:02/3. oue6,Jn 2015 June 64, Volume diabetes.diabetesjournals.org Moulder and Associates 2269

Figure 2—Timing of the serum sample collection (years) relative to the ﬁrst detection of diabetes-associated autoantibodies (A) and relative to the diagnosis of type 1 diabetes (B). ◆, samples proﬁled for the children who progressed to type 1 diabetes. For the comparison between children (healthy vs. progressors), the analyses considered protein abundance throughout the series (119 vs. 119 age-matched samples) and samples before (45 vs. 45 age-matched samples) and after (74 vs. 74 age-matched samples) detection of seroconversion. Other details of the comparisons are indicated in Table 2. An additional 13 case and 13 control samples were included in the measurements. See Supplementary Tables 2 and 3 for further information.

For this comparison, there were 11 well-matched pairs Gene Ontology Annotation and Pathway Analysis with samples before and after seroconversion (Table 3). DAVID (33) was used to perform functional annotation The analysis was repeated separately for the case- and the and pathway analysis of the proteins correlated with time control-to-reference ratios. To unify this analysis, the age/ to diagnosis and to further analyze the protein clusters time axis was scaled between birth and diagnosis (0–1). where the correlation coefficient was $0.6. To reduce bias An absolute Spearman correlation coefficient $0.4 was in these enrichment analyses, the protein background was considered a valid weak correlation (two-tailed P value based scaled to the identified proteins or used in the comparison on 10,000 permutations of the time axis, FDR #5%). (34).

Subject and Status Classification Comparisons With Published Data The top-scoring pairs (TSP) method was applied to We compared the current study results with those from identify whether combinations of the quantified proteins two studies of the serum proteomes of patients with type could classify the samples and subjects (30,31). The leave- 1 diabetes (4,5). Collectively, these reports present 38 one-out method was used for cross-validation. The proteins (in the UniProt database) putatively associated method was applied with the subject-averaged log2 with type 1 diabetes status (Supplementary Table 9). case-to-control abundance ratios for the time periods RESULTS compared and similarly for the log2 subject-to-reference ratios. Although highlighted by the rank product analyses, The iTRAQ measurements detailed, on average, the apolipoprotein C-IV (APOC4) was detected in only 16 of quantitative comparison of 220 proteins, and in total, 19 subject pairs; these 16 were analyzed separately with 658 proteins were identified and quantified with two or the TSP method. The failure to quantify APOC4 in all more unique peptides. In comparing with reference children was attributed to differences in instrument per- concentrations (35) and after excluding depletion targets, formance rather than to its absence. these spanned a range of estimated concentrations of six orders of magnitude. With the analyses using a label-free Hierarchical Clustering and Correlation Analysis approach, 261 proteins were consistently detected and To identify proteins with similar longitudinal profiles and quantified with more than one unique peptide and highlight intersubject differences, k-medians clustering of spanned a similar dynamic range of detection. The com- the diabetes case-control paired subject data was done parison of the proteins identified by the iTRAQ and label- using the Pearson correlation coefficient (k = 15). The free methods revealed an overlap of a core 248 proteins Multiexperiment Viewer was used for these analysis (32). detected with two or more peptides. For the comparative analysis of changes in the complement proteins, the Pearson correlation coefficients Differences Between the Serum Proteomes of Children from each subject of complement component 5 (CO5) Who Developed Type 1 Diabetes and Age-Matched with the other proteins were used together in rank Control Subjects product analyses. CO5 was selected because of its central The children who developed type 1 diabetes had lower role in the formation of the membrane attack complex. levels of APOC4 and apolipoprotein C-II (APOC2) than 2270 eu rtoe nRuet ye1Diabetes 1 Type to Route En Proteomes Serum

Table 2—Serum proteins detected at different levels in children progressing to type 1 diabetes and matched control subjects Average unique Average % Sample Average Entry Average unique peptides label sequence comparisons case-control Protein name Entry name peptides iTRAQ free coverage where detected ratio % FDR Mannose-binding protein C P11226 MBL2 24 9 39 a 0.92 2.0 Complement factor H–related protein 5 Q9BXR6 FHR5 10 9 10 a 1.24 4.0 Complement component C9 P02748 CO9 49 35 36 a,e,f 1.20 3.3 Apolipoprotein C-IV P55056 APOC4 7 6 22 a,b,c,d 0.62 0.6 Apolipoprotein C-II P02655 APOC2 8 5 63 a,b 0.76 0.0 Profilin-1 P07737 PFN1 5 5 30 d 1.42 3.0 e 0.81 0.4 Coagulation factor IX P00740 FA9 11 3 19 g 1.46 4.5 Dopamine beta-hydroxylase P09172 DOPO 10 5 15 e 1.39 0.0 C4b-binding protein beta chain P20851 C4BPB 5 2 21 e 1.32 5.0 Adiponectin Q15848 ADIPO 13 3 33 e 0.68 0.01 Sex hormone–binding globulin P04278 SHBG 30 13 50 e 0.79 0.9 Periostin Q15063 POSTN 13 7 17 e 0.76 0.9 Transforming growth factor-beta–induced protein ig-h3 Q15582 BGH3 17 10 24 e 0.83 1.6 Peptidase inhibitor 16 Q6UXB8 PI16 15 16 19 e 0.83 1.0 Protein S100-A9 P06702 S10A9 6 3 40 f 1.41 4.4 The analyses were made by rank product analysis of the case-control ratios, with the samples considered in relation to time intervals relative to diagnosis and seroconversion (as indicated). a, throughout (n = 19); b, preseroconversion (n = 14); c, 9–12 months preseroconversion; d, 3–6 months preseroconversion; e, postseroconversion and ,1.5 year prediagnosis (n = 19); f, 15–18 Diabetes months postseroconversion (n = 14); g, 3–6 months postseroconversion (n = 14). oue6,Jn 2015 June 64, Volume 2271 Table 3—Longitudinal correlation of serum proteins specific to the children who progressed to type 1 diabetes Average % Correlation Entry Average unique Average unique sequence coefficient Protein name name Entry peptides iTRAQ peptides label free coverage* (Spearman) Fetuin-B FETUB Q9UGM5 18 7 30 0.63 Serum amyloid P-component SAMP P02743 32 9 37 0.51 Clusterin CLUS P10909 70 35 43 0.50

Moulder and Associates C4b-binding protein alpha chain C4BPA P04003 21 11 24 0.49 C4b-binding protein beta chain C4BPB P20851 5 2 20 0.48 Complement factor I CFAI P05156 50 44 48 0.45 Inter-alpha-trypsin inhibitor heavy chain H4 ITIH4 Q14624 251 92 63 0.44 Apolipoprotein C-IV APOC4 P55056 4 6 22 0.44 Insulin-like growth factor–binding protein 3 IBP3 P17936 12 13 27 0.43 Serum amyloid A-4 protein SAA4 P35542 5 10 22 0.43 Complement component C8 alpha chain CO8A P07357 51 35 38 0.42 Complement C1q subcomponent subunit B C1QB P02746 27 17 29 0.42 Hyaluronan-binding protein 2 HABP2 Q14520 20 12 26 0.40 Complement component C8 gamma chain CO8G P07360 28 7 58 0.40 Transforming growth factor-beta–induced protein ig-h3 BGH3 Q15582 17 10 24 20.41 Ectonucleotide pyrophosphatase/phosphodiesterase family member 2 ENPP2 Q13822 11 4 12 20.41 Poliovirus receptor PVR P15151 5 4 8 20.42 Vinculin VINC P18206 6 3 6 20.42

N-acetylmuramoyl-L-alanine amidase PGRP2 Q96PD5 61 31 50 20.42 Contactin-1 CNTN1 Q12860 8 1 9 20.43 L-lactate dehydrogenase B chain LDHB P07195 9 2 24 20.46 Extracellular superoxide dismutase (Cu-Zn) SODE P08294 7 4 28 20.48 Apolipoprotein A-IV APOA4 P06727 189 97 74 20.54 Adiponectin ADIPO Q15848 13 3 33 20.54 Neural cell adhesion molecule 1 NCAM1 P13591 13 5 17 20.60 Insulin-like growth factor–binding protein 2 IBP2 P18065 6 6 19 20.64 A subset of proteins was identified where the absolute Spearman correlation coefficient was .0.4 (permutation-based two-tailed test with FDR #5%) that were not observed at or above these thresholds in the control subjects. The analysis was based on the changes observed in 11 children representing the samples before and after seroconversion (D1, D4, D5, D9, D10, D11, D12, D14, D15, D17, and D19). The functional enrichment for these proteins is shown in Table 4. The equivalent analysis was made for the control subjects, and direct comparison of these lists is shown in Supplementary Table 6A and B. *Coverage based on iTRAQ measurements. diabetes.diabetesjournals.org 2272 Serum Proteomes En Route to Type 1 Diabetes Diabetes Volume 64, June 2015 age-matched healthy control subjects (FDR ,1%). Simi- larly, mannose-binding protein C (MBL2) was also lower than in the matched control subjects (FDR 2%). In contrast, the relative abundance of complement factor H–related protein 5 (FHR-5) and CO9 were higher in the children who developed type 1 diabetes (FDR ,5%) (Table 2). In samples before seroconversion, lower levels of APOC4 and APOC2 were apparent in children who developed type 1 diabetes than in the matched control subjects (FDR ,1% and 4%, respectively) (Supplementary Fig. 9B). Similarly, specific consideration of the samples 3–6 months before seroconversion was consistent with the lower levels of both APOC2 and APOC4 as well as with a larger relative abundance of profilin-1 (PFN1). With a similar analysis of the age-matched data from the period after detection of seroconversion, several proteins were distinguished with a lower relative abundance, including sex hormone–binding globulin, adiponectin (ADIPO), and periostin (FDR ,5%). A higher relative abundance of dopamine b-hydroxylase was observed as well as an apparent peak in protein S100-A9 and a decrease in PFN1 (Table 2). Longitudinal Changes in the Serum Proteomes of Children En Route to Type 1 Diabetes From the analysis of protein abundance ratios, no significant correlations were observed between the case-control ratios and the time to diagnosis, with the strongest being found with Ig mu-chain C region (a depletion target) and tetranectin, which were positively and negatively correlated, respectively. On the contrary, both the case- and the control-to-reference correlations gave a much clearer indication of the longitudinal changesintheserumproteomes.Changesintheabun- dance of 26 proteins (14 increased and 12 decreased, FDR #5%) (Table 3) were distinct from proteins observed to be correlated in both the case and the control children.

Serum Proteomics Classification of the Subjects Progressing to Type 1 Diabetes Figure 3—A: Classification between children who developed type 1 TSP analysis classified the children progressing to type 1 diabetes and age-matched control subjects based on abundance of diabetes at a success rate of 91% (Fig. 3A), the area under APOC4 and AFAM. The TSP method was used, yielding a 91% ▲ □ B the curve being 0.85 (Supplementary Fig. 9A). The classi- success rate. , control subjects; , case subjects. : Relative fi abundance measurements for APOC4 and AFAM for case and con- cation was based on the combination of the relative trol subjects. levels of APOC4 and afamin (AFAM), which were lower and higher than in the control subjects, respectively (P =33 2 10 4, Wilcoxon signed rank test) (Fig. 3B). Similar analysis of the preseroconversion data did not reveal Functional Annotation Enrichment Analysis and any clear classification, whereas for the postseroconver- Hierarchical Clustering sion data, vitronectin (VTN) and CO5 classified the chil- For the proteins observed to be positively correlated dren who progressed to type 1 diabetes with a success rate specifically in children who progressed to type 1 diabetes, of 77% (Supplementary Fig. 10). With the evaluation of a significant enrichment of proteins was associated with longitudinal changes in subjects progressing to type 1 di- inflammation and immune response (Tables 3 and 4). abetes, TSP analysis resulted in classification between the There was no specific functional enrichment in the in- pre- and postseroconversion samples at a success rate of versely correlated proteins. ;80% based on changes in abundance of both apolipopro- From similar enrichment analysis after k-medians tein A-IV and insulin-like growth factor–binding protein clustering of the age-matched subject data, lipid and complex acid labile subunit (Supplementary Fig. 11). cholesterol transport, acute inflammatory response, and 2273

Table 4—GO annotations enriched in proteins increasing in children who progressed to type 1 diabetes Term P value Proteins % FDR

Moulder and Associates GO:0002526 acute inflammatory response 3.8E-04 P07357, Q14624, P04003, P05156, P02746, P20851, P10909, P07360, P02743, P35542 0.3 GO:0019724 B-cell–mediated immunity 8.6E-04 P07357, P04003, P05156, P02746, P20851, P10909, P07360 1.0 GO:0006958 complement activation, classical pathway 8.6E-04 P07357, P04003, P05156, P02746, P20851, P10909, P07360 1.0 GO:0016064 immunoglobulin-mediated immune response 8.6E-04 P07357, P04003, P05156, P02746, P20851, P10909, P07360 1.0 GO:0002455 humoral immune response mediated by circulating immunoglobulin 8.6E-04 P07357, P04003, P05156, P02746, P20851, P10909, P07360 1.0 GO:0006954 inflammatory response 0.0010 P07357, Q14624, P04003, P05156, P02746, P20851, P10909, P07360, P02743, P35542 1.3 Complement pathway 0.0013 P07357, P04003, P05156, P02746, P20851, P10909, P07360 1.3 GO:0002250 adaptive immune response 0.0014 P07357, P04003, P05156, P02746, P20851, P10909, P07360 1.7 GO:0002449 lymphocyte-mediated immunity 0.0014 P07357, P04003, P05156, P02746, P20851, P10909, P07360 1.7 GO:0002460 adaptive immune response based on somatic recombination of immune receptors built from immunoglobulin superfamily domains 0.0014 P07357, P04003, P05156, P02746, P20851, P10909, P07360 1.7 GO:0002443 leukocyte-mediated immunity 0.0018 P07357, P04003, P05156, P02746, P20851, P10909, P07360 2.1 GO:0006952 defense response 0.0032 P07357, Q14624, P04003, P05156, P02746, P20851, P10909, P07360, P02743, P35542 3.8 Innate immunity 0.0039 P07357, P04003, P05156, P02746, P20851, P10909, P07360 3.8 GO:0006959 humoral immune response 0.0041 P07357, P04003, P05156, P02746, P20851, P10909, P07360 4.8 GO:0002541 activation of plasma proteins involved in acute inflammatory response 0.0041 P07357, P04003, P05156, P02746, P20851, P10909, P07360 4.8 GO:0002253 activation of immune response 0.0041 P07357, P04003, P05156, P02746, P20851, P10909, P07360 4.8 GO:0006956 complement activation 0.0041 P07357, P04003, P05156, P02746, P20851, P10909, P07360 4.8 Enrichment was calculated for proteins with an absolute Spearman correlation coefficient $0.4 (FDR #5%) that were not observed at or above these thresholds in the control subjects. A background of the 208 proteins detected for these analyses was used in the enrichment analysis. The protein names and detection details are indicated in Table 3. GO, gene ontology. diabetes.diabetesjournals.org 2274 Serum Proteomes En Route to Type 1 Diabetes Diabetes Volume 64, June 2015 humoral and innate immunity were the most frequently a role of viral infections in type 1 diabetes (13), lower observed enriched functional annotations (FDR #5%) for levels of apolipoproteins have been associated with viral distinct protein profiles between the children progressing infection (36,37). In view of this potential association, we to type 1 diabetes and matched control subjects. These compared the current sample data with enterovirus data clusters frequently included a high representation of the available in the DIPP project. Of the subjects considered complement proteins (data not shown). Because CO5 in the proteomics measurements, neutralizing antibodies plays an important role in the formation of the mem- to coxsackievirus B1 data were available from 12. Six of brane attack complex, we further analyzed the changes eight progressors had antibodies to the virus, whereas in the complement proteins correlated with CO5. This none of the four control subjects was antibody positive. analysis revealed stronger correlations in the children en Due to lack of control data, no further analyses were route to type 1 diabetes than in the age-matched control performed. subjects (Wilcoxon signed rank test P = 0.004 and 0.0002, Analysis with the TSP classifier demonstrated classifi- for positively and inversely correlated proteins, respec- cation within this group of subjects with a 91% success tively). Overall, these analyses supported a strong positive rate based on APOC4 together with AFAM levels. AFAM is correlation of the components of the membrane attack involved in vitamin E transport and has been associated complex with CO5 (i.e., CO8A, CO8B, CO8G, CO6, CO9 with insulin secretion in islet cells (38). However, because [although not CO7]) in both groups, although the inverse these measurements represent a small, yet well-controlled correlations with CO5 in the case subjects were more group of individuals, the global scope of these markers on clearly different from the control subjects (Table 5). a wider scale are not clear and need to be analyzed in From a separate analysis of proteins correlated with other cohorts. CO7, no such clear correlations were observed. In the children who were autoantibody positive, a lower abundance of ADIPO was observed. ADIPO is involved Comparison With Serum Proteomics of Patients With both in the control of fat metabolism and in insulin Type 1 Diabetes sensitivity and has been positively correlated with insulin We compared the present data with observations from sensitivity in patients with type 1 diabetes (4,39). PFN1 two studies of the serum proteomes of patients with type has been associated with inflammation and insulin resis- 1 diabetes (4,5). From the merged list of proteins of the tance (40), and notably, significant differences were fi latter studies, 32 of 38 were clearly de ned in the present detected both before and after seroconversion (decreasing data. In large-scale targeted validations by Zhi et al. (4), in the latter). Further evaluation of this observation is patients with type 1 diabetes were found to have signif- needed. Collectively, these findings appear to reflect met- icantly higher serum levels of ADIPO, insulin-like growth abolic differences and changes preceding the diagnosis of – factor binding protein 2 (IGFBP2), serum amyloid pro- type 1 diabetes. Although we used age-matched control fi tein A, and C-reactive protein and signi cantly lower subjects, these changes should ideally be considered in levels of myeloperoxidase and transforming growth factor- relation to growth and puberty as well as to the broad – beta induced protein (BGH3) (4,5). Comparatively, we found range of ages among the subjects and the size of the fi statistically signi cant lower abundances of both BGH3 cohort. and ADIPO in the children progressing to diabetes within This study revealed correlations in the levels of the the time frame of 18 months from diagnosis (Table 2), complement proteins and in particular, the components whereas a more pronounced time-related decrease was of the membrane attack complex. In general, the compo- observed in these and IGFBP2 (Table 3). Among the bio- nents of this complex circulate independently and interact marker candidates evaluated by Zhang et al. (5), we ob- in a highly specific manner following the cleavage of CO5 b served time-correlated increases of -ala-his dipeptidase (41). We interpreted the observed correlation in the rel- and glutathione peroxidase 3, in both the progressors and ative abundance of circulating concentrations to reflect the control subjects, whereas a more pronounced increase activation of the terminal pathway. The best correlations in clusterin over time was indicated in the progressors for these components were found with CO8G and CO9. (Table 3 and Supplementary Table 9). Because CO8G is less abundant than CO8A and CO8B (35) and CO9 contributes six molecules to the membrane DISCUSSION attack complex, stoichiometry may be the reason that The serum proteomics profiles of a type 1 diabetes risk these proteins provide a better reflection of activation cohort enrolled in the Finnish DIPP study were analyzed. changes. Although the complement system is an integral Comparisons were made between children who developed and functional component of blood, it has been impli- type 1 diabetes–associated autoantibodies and subse- cated as a contributor to the development of various au- quently progressed to overt diabetes and age-matched toimmune diseases, including type 1 diabetes (42). Thus, children who remained persistently autoantibody nega- the present observations of the distinct profiles of the tive. APOC4 and APOC2 were detected at lower levels in complement system in the children progressing to type prospective samples of children progressing to diabetes. 1diabetesmayreflect different challenges to the im- Although recent studies reinforced the hypothesis of mune system or levels of immune responses. Notably, in diabetes.diabetesjournals.org Table 5—Correlation analysis of protein relative abundance profiles with CO5 Protein name Protein abbreviation Accession Case % FDR Type 1 diabetes median Control % FDR Control median Complement component C8 gamma chain CO8G P07360 0.0 0.86 0.0 0.64 Serum amyloid P-component SAMP P02743 0.0 0.82 0.1 0.73 Complement component C9 CO9 P02748 0.0 0.81 0.0 0.73 Complement C1r subcomponent C1R P00736 0.0 0.8 15 0.41 Complement factor H–related protein 5 FHR5 Q9BXR6 0.0 0.78 0.0 0.62 Ceruloplasmin CERU P00450 0.0 0.77 0.0 0.65 Leucine-rich alpha-2-glycoprotein A2GL P02750 0.0 0.76 0.0 0.65 Inter-alpha-trypsin inhibitor heavy chain H4 ITIH4 Q14624 0.0 0.75 0.0 0.68 Complement component C8 beta chain CO8B P07358 0.0 0.74 0.0 0.68 Complement factor B CFAB P00751 0.0 0.72 0.0 0.71 Complement factor I CFAI P05156 0.0 0.72 0.0 0.76 Complement component C6 CO6 P13671 0.1 0.71 0.0 0.69 C4b-binding protein alpha chain C4BPA P04003 0.0 0.71 0.1 0.6 Alpha-1-acid glycoprotein 1 A1AG1 P02763 0.0 0.7 0.0 0.75 Complement C1s subcomponent C1S P09871 0.0 0.7 1.0 0.39 Complement component C8 alpha chain CO8A P07357 0.0 0.66 0.1 0.66 Lipopolysaccharide-binding protein LBP P18428 0.4 0.66 0.2 0.65 Alpha-1-antichymotrypsin AACT P01011 0.0 0.66 0.0 0.77 Protein S100-A9 S100A9 P06702 0.8 0.65 1.0 0.28 Complement C4-B CO4B P0C0L5 0.0 0.65 0.3 0.62 Alpha-1-acid glycoprotein 2 A1AG2 P19652 0.0 0.64 0.0 0.72 Complement factor H–related protein 3 FHR3 Q02985 0.4 0.54 0.2 0.55 Apolipoprotein A-IV APOA4 P06727 0.0 20.51 5.0 20.36 Alpha-2-HS-glycoprotein FETUA P02765 0.5 20.52 9.0 20.29 Complement component C1q receptor C1QR1 Q9NPY3 0.4 20.55 20 20.2 ole n Associates and Moulder Sex hormone–binding globulin SHBG P04278 0.2 20.56 10 20.39 Receptor-type tyrosine-protein phosphatase gamma PTPRG P23470 0.2 20.59 4.0 20.33 Periostin POSTN Q15063 0.0 20.59 10 20.32 Vasorin VASN Q6EMK4 0.0 20.61 0.3 20.42 Peptidase inhibitor 16 PI16 Q6UXB8 0.0 20.63 2.0 20.34 Collectin-11 COL11 Q9BWP8 0.1 20.64 6.0 20.3 72 kDa type IV collagenase MMP2 P08253 0.0 20.65 7.0 20.2

Continued on p. 2276 2275 2276 Serum Proteomes En Route to Type 1 Diabetes Diabetes Volume 64, June 2015

relation to the complement proteins, the TSP analysis of samples postseroconversion relative to control samples provides classification with a success rate of 77% based on CO5 and VTN. VTN is involved in the regulation of the 0.4 0.3 0.48 0.34 0.19 0.48 2 2 2 2 2 2 complement system and minimizes off-target effects during complement-mediated attack (43). In addition, a larger relative abundance of CO9 was observed in the children progressing to type 1 diabetes. From the time-wise correlations, the analysis of data from children progressing to type 1 diabetes emphasizes an increasing immune response before diagnosis (Table 4). Similarly, with the paired subjects, the detection of increasing Ig mu-chain C region could similarly reflect the ongoing immune process. Although MBL2 levels are variable within the population, the progressors displayed predominantly lower levels than matched control subjects. Notably, MBL2 is an important component of immune defense, where it is 0.65 0.9 0.66 3.0 0.66 15 0.67 0.1 0.68 6.0 0.71 5.0 2 2 2 2 2 2 a key protein in lectin activation of the complement pathway, and as a genetic disorder, a deficiency of MBL2 is associated with an increased infection risk (44). How- ever, no strong correlations were identified between the relative abundance of MBL2 and the other proteins quantified. Consistent with a number of other multiple-comparison serum iTRAQ measurements, the data were limited to comparison of ;250 proteins between subject pairs (15,16,45,46). Although the present measurements were mostly restricted to the description of the moderate abundance proteome, an underlying tenet in the proteomics 2 subjects (case and control). To assign consistency in the observations, a rank product analysis was

3 community is that characterization of the lower abundance serum proteins holds the key to finding biomarkers (47). Instrumental improvements (48), alternative frac- tionation strategies, and data-independent approaches have begun to address this problem (49). cients from 19 fi Validation measurements in a larger and independent cohort are an important next step in this research. On the basis of the present findings, studies exploiting selected reaction monitoring (SRM) are currently in progress to validate these putative biomarkers of development of b-cell autoimmunity. As a pertinent example of the application of this methodology, García-Bailo et al. (50) applied SRM protocols on the scale of 1,000 individuals to quantify a similar panel of plasma proteins. Our future goal is to investigate the extent to which we can improve the detection of disease activity and the prediction of type 1 diabetes through the detection of a selected set of se- cients and is represented by the % FDRs for the highest-ranked proteins.

fi rum protein targets by SRM together with other param- eters, such as autoantibody status and other analytes identified from our ongoing omics studies. In summary, using mass spectrometry–based analysis of immunodepleted sera, we demonstrate for the first time serum proteomics profiles of the prediabetic transi- Continued fi

— tion all the way to diagnosis, comparing pro les between children progressing to type 1 diabetes and healthy children. These results demonstrate shared and group- Table 5 Protein nameEndothelial protein C receptor EPCR Protein abbreviation Accession Q9UNN8 Case % FDR Type 1 0.0 diabetes median Control % FDR Control median Collagen alpha-1(I) chain CO1A1 P02452 0.2 Aggrecan core protein PGCA P16112 0.0 Collagen alpha-1(VI) chain CO6A1 P12109 0.0 Contactin-1 CNTN1 Q12860 0.0 Collagen alpha-1(XII) chain COCA1 Q99715 0.0 The table indicates the median value of the top-ranked correlation coef made for the generated coef specific longitudinal changes against a background of wide diabetes.diabetesjournals.org Moulder and Associates 2277 subject heterogeneity, suggesting that components of the 4. Zhi W, Sharma A, Purohit S, et al. Discovery and validation of serum protein moderately abundant serum proteins could indicate the changes in type 1 diabetes patients using high throughput two dimensional liquid emerging threat of type 1 diabetes. Future developments chromatography-mass spectrometry and immunoassays. Mol Cell Proteomics of this nature could include the determination of a panel 2011;10:M111.012203 5. Zhang Q, Fillmore TL, Schepmoes AA, et al. Serum proteomics reveals systemic of thus-related proteins in addition to current markers of dysregulation of innate immunity in type 1 diabetes. J Exp Med 2013;210:191–203 seroconversion as a means to stratify disease risk. 6. McGuire JN, Eising S, Wägner AM, Pociot F. Screening newborns for candidate biomarkers of type 1 diabetes. Arch Physiol Biochem 2010;116:227– Acknowledgments. The authors thank the DIPP families for participation; 232 S. Simell and the staff of the DIPP study for work with the families and obtaining the 7. Salminen K, Sadeharju K, Lönnrot M, et al. Enterovirus infections are as- samples of the study; J. Hakalax, M. Laaksonen, and E. Pakarinen (of the De- sociated with the induction of beta-cell autoimmunity in a prospective birth cohort partment of Pediatrics, Turku University Hospital, Turku, Finland, and the Depart- study. J Med Virol 2003;69:91–98 ment of Pediatrics, University of Turku, Turku, Finland) for handling and managing 8. Johnson SB, Lee HS, Baxter J, Lernmark B, Roth R, Simell T; TEDDY Study study participant data; and V. Simell (of the Department of Pediatrics, Turku Group. The Environmental Determinants of Diabetes in the Young (TEDDY) study: University Hospital, Turku, Finland, and the Department of Pediatrics, University predictors of early study withdrawal among participants with no family history of of Turku, Turku, Finland) for managing the DIPP biobank and samples. The mea- type 1 diabetes. Pediatr Diabetes 2011;12:165–171 surements presented in this work were performed at the Turku Centre for Bio- 9. Kupila A, Muona P, Simell T, et al.; Juvenile Diabetes Research Foundation technology Proteomics core facility from which the excellent technical support of Centre for the Prevention of Type I Diabetes in Finland. Feasibility of genetic and Arttu Heinonen and constructive criticisms of Susumu Imanishi (both of the Turku immunological prediction of type I diabetes in a population-based birth cohort. Centre for Biotechnology, University of Turku, Turku, Finland) are greatly appreci- Diabetologia 2001;44:290–297 ated. Waltteri Hosia is acknowledged for early work in this project and former 10. Oresic M, Simell S, Sysi-Aho M, et al. Dysregulation of lipid and amino acid students Ida Koho and Joona Valtonen (all from Turku Centre for Biotechnology, metabolism precedes islet autoimmunity in children who later progress to type 1 University of Turku, Turku, Finland) are thanked for their assistance. AB Sciex, diabetes. J Exp Med 2008;205:2975–2984 particularly Sean Seymour, is thanked for assistance with ProteinPilot. 11. Oresic M, Gopalacharyulu P, Mykkänen J, et al. Cord serum lipidome in Funding. The work was financially supported by the National Technology prediction of islet autoimmunity and type 1 diabetes. Diabetes 2013;62:3268– Agency of Finland grants (40453/04, 40229/08, and 40398/11), JDRF (17- 3274 2011-586), and Academy of Finland grants (77773, 203725, 207490, 116639, 12. Elo LL, Mykkänen J, Nikula T, et al. Early suppression of immune response 115939, 123864, 126063, 110432 to J.M. and the Centre of Excellence in pathways characterizes children with prediabetes in genome-wide gene ex- Molecular Systems Immunology and Physiology Research, 2012–2017, Decision pression profiling. J Autoimmun 2010;35:70–76 No. 250114 to M.K., H.L., O.S., and R.L.). 13. Laitinen OH, Honkanen H, Pakkanen O, et al. Coxsackievirus B1 is asso- Duality of Interest. No potential conflicts of interest relevant to this article ciated with induction of b-cell autoimmunity that portends type 1 diabetes. Di- were reported. abetes 2014;63:446–455 Author Contributions. R.M. set up the experimental methods, prepared 14. Kallionpää H, Elo LL, Laajala E, et al. Innate immune activity is detected and analyzed samples, analyzed and interpreted the data, and was the key prior to seroconversion in children with HLA-conferred type 1 diabetes suscep- author of the manuscript. S.D.B. prepared and analyzed samples and played tibility. Diabetes 2014;63:2402–2414 an important role in implementing the label-free method. T.E., E.L., and J.S. 15. van der Greef J, Martin S, Juhasz P, et al. The art and practice of systems participated in establishing and testing methods for the analysis of the longitu- biology in medicine: mapping patterns of relationships. J Proteome Res 2007;6: dinal data. E.V.N. contributed to the clustering analysis and preliminary analysis 1540–1559 of the label-free data. H.K. was involved in the interpretation of the observations 16. Cole RN, Ruczinski I, Schulze K, et al. The plasma proteome identifies and discussions throughout the study. J.M. participated in the initiation of the expected and novel proteins correlated with micronutrient status in un- study. M.V.-M. participated in selecting the samples. H.H. was responsible for the dernourished Nepalese children. J Nutr 2013;143:1540–1548 virus analysis within the study. R.V. and M.K. were responsible for the analyses of 17. Adkins JN, Monroe ME, Auberry KJ, et al. A proteomic study of the HUPO diabetes-associated autoantibodies. J.I. was responsible for the DNA isolation and Plasma Proteome Project’s pilot samples using an accurate mass and time tag HLA screening of the study children. T.S., J.T., and O.S. provided the samples and strategy. Proteomics 2005;5:3454–3466 the clinical information of the study children. D.R.G. was involved in selecting the 18. Smith MP, Wood SL, Zougman A, et al. A systematic analysis of the effects analytical methods and data analysis strategies. H.L. supervised T.E. and E.L. and of increasing degrees of serum immunodepletion in terms of depth of coverage participated in the interpretation of the results. O.S. and R.L. initiated the study, and other key aspects in top-down and bottom-up proteomic analyses. Proteo- designed the study setup, supervised the study, and participated in the interpreta- mics 2011;11:2222–2235 fi tion of the results and writing the manuscript. All authors contributed to the nal 19. Kulmala P, Savola K, Petersen JS, et al.; The Childhood Diabetes in Finland version of the manuscript. R.M., H.L., O.S., and R.L. are the guarantors of this work Study Group. Prediction of insulin-dependent diabetes mellitus in siblings of and, as such, had full access to all the data in the study and take responsibility for children with diabetes. A population-based study. J Clin Invest 1998;101:327– the integrity of the data and the accuracy of the data analysis. 336 20. Moulder R, Lönnberg T, Elo LL, et al. Quantitative proteomics analysis of the References nuclear fraction of human CD4+ cells in the early phases of IL-4-induced Th2 1. Ziegler AG, Rewers M, Simell O, et al. Seroconversion to multiple islet differentiation. Mol Cell Proteomics 2010;9:1937–1953 autoantibodies and risk of progression to diabetes in children. JAMA 2013;309: 21. Vähämaa H, Koskinen VR, Hosia W, et al. PolyAlign: a versatile LC-MS data 2473–2479 alignment tool for landmark-selected and -automated use. Int J Proteomics 2. Zhi W, Purohit S, Carey C, Wang M, She JX. Proteomic technologies for the 2011;2011:450290 discovery of type 1 diabetes biomarkers. J Diabetes Sci Tech 2010;4:993–1002 22. Shilov IV, Seymour SL, Patel AA, et al. The Paragon Algorithm, a next 3. Metz TO, Qian W, Jacobs JM, et al. Application of proteomics in the dis- generation search engine that uses sequence temperature values and feature covery of candidate protein biomarkers in a diabetes autoantibody standardiza- probabilities to identify peptides from tandem mass spectra. Mol Cell Proteomics tion program sample subset. J Proteome Res 2008;7:698–707 2007;6:1638–1655 2278 Serum Proteomes En Route to Type 1 Diabetes Diabetes Volume 64, June 2015

23. Rissanen J, Moulder R, Lahesmaa R, Nevalainen OS. Pre-processing of 36. Singh IP, Chopra AK, Coppenhaver DH, Ananatharamaiah GM, Baron S. Orbitrap higher energy collisional dissociation tandem mass spectra to reduce Lipoproteins account for part of the broad non-specific antiviral activity of human erroneous iTRAQ ratios. Rapid Commun Mass Spectrom 2012;26:2099–2104 serum. Antiviral Res 1999;42:211–218 24. Tang WH, Shilov IV, Seymour SL. Nonlinear fitting method for determining 37. Rowell J, Thompson AJ, Guyton JR, et al. Serum apolipoprotein C-III is local false discovery rates from decoy database searches. J Proteome Res 2008; independently associated with chronic hepatitis C infection and advanced fi- 7:3661–3667 brosis. Hepatol Int 7 July 2011 [Epub ahead of print] 25. Tambor V, Hunter CL, Seymour SL, Kacerovsky M, Stulik J, Lenco J. 38. Liu J, Walp ER, May CL. Elevation of transcription factor Islet-1 levels in vivo CysTRAQ - A combination of iTRAQ and enrichment of cysteinyl peptides for increases b-cell function but not b-cell mass. Islets 2012;4:199–206 uncovering and quantifying hidden proteomes. J Proteomics 2012;75:857–867 39. Pereira RI, Snell-Bergeon JK, Erickson C, et al. Adiponectin dysregulation – 26. Fischer R, Trudgian DC, Wright C, et al. Discovery of candidate serum and insulin resistance in type 1 diabetes. J Clin Endocrinol Metab 2012;97:E642 proteomic and metabolomic biomarkers in ankylosing spondylitis. Mol Cell Pro- E647 fi teomics 2012;11:M111.013904 40. Pae M, Romeo GR. The multifaceted role of pro lin-1 in adipose tissue fl – 27. Breitling R, Armengaud P, Amtmann A, Herzyk P. Rank products: a simple, in ammation and glucose homeostasis. Adipocyte 2014;3:69 74 yet powerful, new method to detect differentially regulated genes in replicated 41. Sodetz JM, Plumb ME. Complement: terminal pathway [article online]. eLS 2001. Available from http://onlinelibrary.wiley.com/doi/10.1038/npg microarray experiments. FEBS Lett 2004;573:83–92 .els.0000511/abstract;jsessionid=EDBFEF808E1860D4505A2F29F6536450.f02t01. 28. Wang H, Alvarez S, Hicks LM. Comprehensive comparison of iTRAQ and Accessed 13 June 2013 label-free LC-based quantitative proteomics approaches using two Chlamydo- 42. Chen M, Daha MR, Kallenberg CG. The complement system in systemic monas reinhardtii strains of interest for biofuels engineering. J Proteome Res autoimmune disease. J Autoimmun 2010;34:J276–J286 2012;11:487–501 43. Fink TM, Jenne DE, Lichter P. The human vitronectin (complement S-protein) 29. Li Z, Adams RM, Chourey K, Hurst GB, Hettich RL, Pan C. Systematic com- gene maps to the centromeric region of 17q. Hum Genet 1992;88:569–572 parison of label-free, metabolic labeling, and isobaric chemical labeling for quan- 44. Turner MW. The role of mannose-binding lectin in health and disease. Mol titative proteomics on LTQ Orbitrap Velos. J Proteome Res 2012;11:1582–1590 Immunol 2003;40:423–429 30. Geman D, d’Avignon C, Naiman DQ, Winslow RL. Classifying gene ex- 45. Song X, Bandow J, Sherman J, et al. iTRAQ experimental design for plasma fi pression pro les from pairwise mRNA comparisons. Stat Appl Genet Mol Biol biomarker discovery. J Proteome Res 2008;7:2952–2958 2004;3:Article19 46. Overgaard AJ, Thingholm TE, Larsen MR, et al. Quantitative iTRAQ-based 31. Xu L, Tan AC, Naiman DQ, Geman D, Winslow RL. Robust prostate cancer proteomic identification of candidate biomarkers for diabetic nephropathy in marker genes emerge from direct integration of inter-study microarray data. plasma of type 1 diabetic patients. Clin Proteomics 2010;6:105–114 – Bioinformatics 2005;21:3905 3911 47. Gerszten RE, Asnani A, Carr SA. Status and prospects for discovery and 32. Saeed AI, Sharov V, White J, et al. TM4: a free, open-source system for verification of new biomarkers of cardiovascular disease by proteomics. Circ Res microarray data management and analysis. Biotechniques 2003;34:374–378 2011;109:463–474 33. Huang W, Sherman BT, Lempicki RA. Bioinformatics enrichment tools: 48. Hebert AS, Richards AL, Bailey DJ, et al. The one hour yeast proteome. Mol paths toward the comprehensive functional analysis of large gene lists. Nucleic Cell Proteomics 2014;13:339–347 Acids Res 2009;37:1–13 49. Chapman JD, Goodlett DR, Masselon CD. Multiplexed and data-independent 34. Louie B, Higdon R, Kolker E. The necessity of adjusting tests of protein tandem mass spectrometry for global proteome profiling. Mass Spectrom Rev category enrichment in discovery proteomics. Bioinformatics 2010;26:3007–3011 2014;33:452–470 35. Farrah T, Deutsch EW, Omenn GS, et al. A high-confidence human plasma 50. García-Bailo B, Brenner DR, Nielsen D, et al. Dietary patterns and ethnicity proteome reference set with estimated concentrations in PeptideAtlas. Mol Cell are associated with distinct plasma proteomic groups. Am J Clin Nutr 2012;95: Proteomics 2011;10:M110.006353 352–361