Elevated Serum Levels of Circulating Immunoinflammation-Related

Dec 2, 2013 - ... or thyroid cancer, 141 patients with other types of cancer, 376 patients with benign lung, colorectal, pancreatic, gastric, or thyro...
0 downloads 0 Views 5MB Size
Article pubs.acs.org/jpr

Elevated Serum Levels of Circulating Immunoinflammation-Related Protein Complexes Are Associated with Cancer Yanying Wang,†,∥ Gaoguang Song,†,∥ Yanmin Wang,‡ Ling Qiu,§ Xuzhen Qin,§ Hui Liu,† Fang Li,† Xiaodong Wang,† Fenjie Li,† Shuai Guo,† Yaping Zhang,† and Zhili Li*,† †

Department of Biophysics and Structural Biology, Institute of Basic Medical Sciences, Chinese Academy of Medical Sciences & School of Basic Medicine, Peking Union Medical College, Beijing 100005, P. R. China ‡ Department of Clinical Laboratory, Heze Municipal Hospital, Shandong 274031, P. R. China § Department of Clinical Laboratory, Peking Union Medical College Hospital, Chinese Academy of Medical Sciences & Peking Union Medical College, Beijing 100730, P. R. China S Supporting Information *

ABSTRACT: Disease-specific immune response-related protein complexes in the bloodstream are associated with disease status. We used proteomic technologies to screen novel circulating immunoinflammation-related protein complexes (IIRPCs) and to evaluate their diagnostic accuracy. The discovery study included 96 gastric cancer patients and 83 healthy controls and was designed to isolate and identify the IIRPCs. Then an independent validation study including 1366 patients with lung, colorectal, pancreatic, gastric, or thyroid cancer, 141 patients with other types of cancer, 376 patients with benign lung, colorectal, pancreatic, gastric, or thyroid diseases, and 3707 healthy controls was performed. We observed seven major patterns of the IIRPCs and confirmed the IIRPCs as personalized biomarkers of cancers. The levels of the IIRPCs were significantly increased in cancer patients compared with controls and benign patients (p < 0.0001). Each of the IIRPCs (a2 to a4, a6, a7, and b3 to b5) shows excellent discriminating power for lung, colorectal, pancreatic, and gastric cancer, with the areas under the receiver operating characteristic curves (AUCs) from 0.95 to 0.99 (95% CIs 0.91−1.00), and for thyroid cancer, with the AUCs from 0.87 to 0.96 (95% CIs 0.80−0.98). The IIRPCs can be used as a novel type of broad-spectrum and supramolecular biomarker for personalized cancer diagnosis. KEYWORDS: immunoinflammation-related protein complexes, cancer, supramolecular biomarker, native gel, serum



INTRODUCTION Cancer is a leading cause of death worldwide.1 Given that most cancer cases are diagnosed at advanced stages, the high-risk population should be targeted for prevention and early detection. An early diagnosis of cancer facilitates early decision-making and treatment to improve outcomes. Blood components, including circulating tumor cells, free DNA and RNA, proteins, peptides, and metabolites, are indicators of health status.2 At the present, however, single molecular biomarkers have not provided high diagnostic accuracy.3 Protein complexes assembled by noncovalent bond interactions may be indicators of health status. Trypsin 2−α1 antitrypsin complex has been used to differentiate between acute pancreatitis and extrapancreatic diseases,4,5 to predict severity of acute pancreatitis,5,6 as well as being a biomarker for cholangiocarcinomas,7 with high sensitivity and specificity. Currently, the measurement of these protein complexes is mainly performed by immunoassays, which have the disadvantage of limiting the exploration of biomarkers to known protein complexes. To overcome the above limitations, nativepolyacrylamide gel electrophoresis (native-PAGE),8−10 com© 2013 American Chemical Society

bined with mass spectrometry-based proteomic technologies, has been used to isolate and identify serum protein complexes and cancer-associated protein complexes.11 Both chronic inflammatory processes and humoral immune responses are associated with many human cancers.12−14 Only a few studies to date have tried to correlate humoral immune responses with cancer on a large number of samples.15 Here, we have performed serum protein complex profiling of 96 patients with gastric cancer and 83 healthy controls using an optimized native-PAGE. By analyzing the resulting serum protein complex profiles, we observed seven major patterns of novel serum immunoinflammation-related protein complexes (IIRPCs), and we also found that elevated levels of the IIRPCs are significantly correlated with gastric cancer compared with healthy controls. We then validated these patterns and the diagnostic accuracy of the IIRPCs in an independent study of 5590 individuals, including 1507 cancer cases, 376 benign disease cases, and 3707 healthy controls. Received: August 11, 2013 Published: December 2, 2013 710

dx.doi.org/10.1021/pr4008255 | J. Proteome Res. 2014, 13, 710−719

Journal of Proteome Research

Article

Table 1. Sample Collections in the Studya discovery study health status controls cancer lung colorectum pancreas stomach thyroid other types of cancers (esophageal(n = 24), throat (n = 15), liver (n = 12), prostatic (n = 10), renal (n = 7), uterine (n = 13), and breast (n = 22) cancers, and multiple myeloma (n = 38)) benign diseases lung pneumonia pulmonary fibrosis nodules COPD emphysema tracheitis chronic bronchitis colorectum colitis colonic polyps colonic ulcer Crohn’s disease rectal polyps pancreas pancreatitis chronic pancreatitis acute pancreatis pancreatic cyst benign tumor stomach gastritis chronic gastritis gastric ulcer gastric polyps gastric stromal tumors thyroid thyroid nodules hyperthyroidism goiter chronic thyroiditis a

ind validation study

(n = 179)

(n = 5449)

83 96

3707 1366 410 300 237 212 207

96

all subjects

(n = 141)

(n = 5769)

141

3790 1603

141 376 82 27 4 5 2 3 33 8 61 29 17 4 8 3 79 35 21 8 11 4 79 39 24 12 2 2 75 33 31 8 3

376

COPD denotes chronic obstructive pulmonary diseases.



EXPERIMENTAL SECTION

consent was obtained from all participants or from the parents or guardians of participants aged less than 18 years.

Participants

Study Design

From April 2009 through October 2011, we recruited a total of 5769 participants from two hospitals in China (Table 1). All specimens were the sera that remained from clinical laboratory examination. Approximately 0.3 mL of serum was obtained from each participant. Serum was centrifuged at 3000g for 10 min at 4 °C. The supernatant was harvested, divided into 100 μL aliquots, and immediately stored at −80 °C. All samples underwent only one additional freeze−thaw cycle before electrophoretic analysis. This study was approved by the ethics review committee at the Institute of Basic Medical Sciences, Chinese Academy of Medical Sciences & Peking Union Medical College (Serial No. 002-2009). Written informed

This large-scale case-control study included discovery and independent validation (Figure 1 and Table 1). Sera of all patients were collected at diagnosis. In the discovery study, serum protein complex profiles were first analyzed in 96 patients with gastric cancer and 83 healthy controls by nativePAGE. Next, the components of these protein complexes were isolated and identified. Finally, the correlation of the levels of the IIRPCs with cancer was determined. An independent validation study was used to further assess the association of the IIRPCs with lung, colorectal, pancreatic, gastric, and thyroid cancer, as well as the corresponding benign diseases, in three consecutive sets of samples (Figure 2). In set I, we analyzed 711

dx.doi.org/10.1021/pr4008255 | J. Proteome Res. 2014, 13, 710−719

Journal of Proteome Research

Article

Figure 1. Flow chart of study design: Native-PAGE, nativepolyacrylamide gel electrophoresis; SDS-PAGE, sodium dodecyl sulfate−polyacrylamide gel electrophoresis; MALDI-FTICR MS, matrix-assisted laser desorption/ionization−Fourier transform ion cyclotron resonance mass spectrometer; TRPC, transferrin-related protein complex; IIRPCs, immunoinflammation-related protein complexes; ROC, receiver operating characteristic.

sera from 1366 cancer patients, 376 benign patients, and 3707 controls by native-PAGE and summarized the patterns of the IIRPCs. In set II, to minimize confounding effects on levels of protein complexes, we excluded specimens with macroscopic hemolysis (hemoglobin affected the quantification of the internal reference) or without information on sex or age of the patients. We also excluded controls with incomplete physical examination results. Ultimately, we used 1310 cancer patients aged 10−95 years, 376 benign patients aged 13−91 years, and 528 healthy controls aged 10−96 years to validate an internal reference for quantifying IIRPCs in native gels. In set III, patients without anticancer and/or anti-inflammation treatment and with pattern a or b of IIRPCs were recruited, and controls were matched to cases with lung, colorectal, pancreatic, gastric, and thyroid cancer, as well as each type of benign disease based on the patterns of their IIRPCs, sex, and age. Finally, statistical analyses of the levels of the IIRPCs from 615 cancer patients aged 16−95 years, 251 benign patients aged 13−91 years, and 252 controls aged 18−96 years were performed to evaluate the association of the IIRPCs levels with different cancers. In addition, we also determined the correlation between the levels of interleukin (IL)-6 and IIRPCs and between IL-6 concentration and cancer risk. In addition, we studied 103 patients with seven other epithelial cancers (esophageal, throat, liver, prostatic, renal, uterine, and breast cancer) and 38 patients with multiple myeloma to assess the association between the IIRPC levels and cancer risk.

Figure 2. Steps taken in selecting cancer cases and controls in the independent validation study: IIRPCs, immunoinflammation-related protein complexes; TRPC, transferrin-related protein complex.

concentration between the patients and controls) from the external reference (QC sample), cases, and controls were randomly loaded onto different lanes in native-PAGE gels. The QC sample was a mixed serum from 5 random healthy controls. Electrophoresis was carried out at 10 mA per gel for 2 h, followed by 25 mA per gel for 3.5 h, with 25 mM Tris base and 192 mM glycine as the cathode and anode buffers, respectively. A commercially available native protein mixture (66−669 kDa) (GE Healthcare, Uppsala, Sweden) and thyroglobulin (SigmaAldrich, St. Louis, MO) were used as molecular weight markers. Native gels were stained with Coomassie brilliant blue G-250, the background was destained with water, and then the gels

Separation and Pattern Assignment of Protein Complexes

Double-blind experiments on the separation, pattern assignment, and quantification of the protein complexes were carried out. Serum protein complexes were isolated by native-PAGE on the basis of the approach used by Elsasser et al.10 with some modifications. Briefly, 4% to 10% (w/v) (acrylamide/bisacrylamide =37.5:1) gradient acrylamide gels in TBM (90 mM Tris base/90 mM Boric acid/1.5 mM MgCl2, pH 8.3) were used to separate protein complexes in serum, and a 4% stacking gel16 was overlaid onto the separating gel. Equal volumes of 2 μL of sera (there was no significant difference in total protein 712

dx.doi.org/10.1021/pr4008255 | J. Proteome Res. 2014, 13, 710−719

Journal of Proteome Research

Article

digestion was performed as described previously.16 We identified the components by peptide mass fingerprinting on a matrix-assisted laser desorption/ionization−Fourier transform ion cyclotron resonance mass spectrometer (MALDI-FTICR MS, Bruker Daltonics, Billerica, MA). An aliquot of 1 μL of tryptic digest was spotted onto the MALDI target plate, and dried at room temperature before the addition of 1 μL of 1 μg/ μL α-cyano-4-hydroxycinnamic acid in 30% (v/v) acetonitrile/ 0.1% (v/v) trifluoroacetic acid. The peptide calibration standard II mixture (Bruker Daltonics) was used for external calibration. Mass spectra were acquired over the range of m/z 800−4000 after external calibration. Data mining was performed using BioTools 3.2 software (Bruker Daltonics) and the Swiss-Prot database using the MASCOT search engine with the following parameters: taxonomy, Homo sapiens; enzyme, trypsin; mass tolerance, 10 ppm; the carbamidomethylation of cysteine as a fixed modification, and the oxidation of methionine as a variable modification.

were either scanned with a Umax PowerLook 2100XL scanner (Techville, Inc., USA) for optical densitometry-based quantification, or were immediately followed by SDS-PAGE separation to confirm the components of these complexes and to identify proteins. Based on the positions and the number of the protein complexes in each lane of the gel, the samples were classified into seven major patterns (Figure 3). Briefly, for

Quantification of the TRPC and IIRPCs

Each run included nine samples and one QC sample in a single native-PAGE. The QC sample was a mixture of 5 random healthy control sera. We quantified the gray values of the gel bands from 1 to 18 corresponding to the IIRPCs and band 19 corresponding to transferrin-related protein complex (TRPC) in each gel (Figure 3) with Quantity One software version 4.6.2 (Bio-Rad). The levels of these bands in each gel were exported into Microsoft Excel after the gel background had been subtracted. The level of the TRPC in each sample was also quantified relative to that of the QC sample, and then the standardized levels of the TRPC of participants were statistically analyzed. The levels of the IIRPCs for each sample were calculated relative to the level of the corresponding TRPC, which was normalized to 100. The reproducibility of the method was determined on the basis of three QC samples (mixtures of 5 random healthy control sera, 5 random benign patient sera, and 5 random cancer patient sera). Coefficients of variation for the intraday precision (ten runs) and the interday precision (seven consecutive days) for these three QC samples were 0.92 based on a serial dilution of 5 μL of serum using a 13-cm-long gel. The stability results showed that the IIRPCs (bands 1−18) and TRPC (band 19) remained unchanged even though the sera were stored for seven days at 37 °C, three months at 4 °C, or three years at −80 °C.

Figure 3. Patterns of immunoinflammation-related protein complexes (IIRPCs, bands 1−18) and transferrin-related protein complex (TRPC, band 19) in sera isolated by native-PAGE. Mr and kDa indicate protein marker and kilodalton, respectively. Thyroglobulin is used as a molecular weight marker. a to g are seven major patterns in sera. Shown are the results in cancer cases. The components of bands 1−19 are listed in Table 4.

pattern a, we assigned 9 specific IIRPCs (a1, a2, a3, ..., and a9); for pattern b, 6 specific IIRPCs (b1, b2, b3, ..., and b6); for pattern c, no specific IIRPCs; for pattern d, 7 specific IIRPCs (d1, d2, d3, ..., and d7); for pattern e, 3 specific IIRPCs (e1, e2, and e3); for pattern f, 6 specific IIRPCs (f1, f2, f3, ..., and f6); and for pattern g, 7 specific IIRPCs (g1, g2, g3, ..., and g7). Only 0.07) and are also independent of gastric cancer stage (p > 0.2). The protein complexes (Figure 3) were isolated by SDS-PAGE, and included complement components, immunoglobulin, haptoglobin, apolipoprotein A-I, and transferrin (Table 3 and Supporting Information Figures S1−S4).

Variables with normal and skewed distributions are presented as means ± standard deviation (SD) and medians (with the interquartile range), respectively, and categorical variables that can take on only a few fixed values are presented as numbers and percentages. We compared categorical variables using the Pearson χ2 test and continuous variables between two groups using the Student’s t test or Mann−Whitney U test. To compare the continuous variables of more than two groups, one-way ANOVA or Kruskal−Wallis tests were performed. The Spearman rank-order test was used to assess the correlation between the levels of the IIRPCs and cancer, or between the levels of IL-6 and cancer. Receiver operating characteristic (ROC) curves were constructed to determine the diagnostic performance of the IIRPCs and to compare their accuracies of cancer detection. Two-tailed tests were calculated, and a p value of 60 stage I II III unknown pattern a b c d e f g others

gastric cancer

controls

(n= 96)

(n= 83)

P valueb

70 (73) 57.9 ± 11.7 55 (57) 41 (43)

53 (63) 56.9 ± 11.7 50 (60) 33 (40)

0.141 0.557



0.225 0.559

The major components of these protein complexes are listed in Table 4. Interestingly, there are few IIRPCs in pattern c samples, which may be due to different haptoglobin phenotypes of the serum.11

16 (17) 22 (23) 50 (52) 8 (8)

Number corresponds to the number in Supporting Information Figure S4. *Score corresponds to the band labeled “*” in Supporting Information Figure S4.

Independent Validation of Protein Complexes 28 (29) 23 (24) 16 (17) 15 (16) 6 (6) 2 (2) 5 (5) 1 (1)

23 (28) 24 (29) 17 (21) 12 (15) 2 (2) 2 (2) 2 (2) 1 (1)

0.863

The components of the IIRPCs are not cancer mechanismassociated proteins, but are associated with immune, inflammatory, and complement systems, which play important roles in cancer development, progression, and response to treatment.12,13,17−19 To further confirm the patterns of the IIRPCs and the association of their levels with cancer, we performed an independent validation study with a larger sample size consisting of 1366 patients with lung, colorectal, pancreatic, gastric, or thyroid cancer (LC, CRC, PC, GC, or TC), 141 patients with other types of cancer, 376 patients with benign diseases, and 3707 controls (Figure 2). Patterns of the IIRPCs. In set I, we analyzed the 5449 samples using the native-PAGE, and we confirmed the findings found in the discovery study. Seven major patterns (a, b, c, d, e, f, and g) exist in the 5418 participant sera, and each of the patterns accounts for approximately 36% (n = 1982), 26% (n = 1428), 13% (n = 694), 12% (n = 679), 6% (n = 300), 3% (n = 153), and 3% (n = 182) of the samples, respectively. “Others” (31 subjects, 0.07), indicating that it can be used as an internal reference for quantifying other protein complexes. In participants with patterns a and b (Supporting Information Table S1), the levels of a1 to a9 and b1 to b6 are significantly increased in gastric cancer patients compared with 714

dx.doi.org/10.1021/pr4008255 | J. Proteome Res. 2014, 13, 710−719

Journal of Proteome Research

Article

Table 4. Components of IIRPCs and TRPCa band no.b

protein complex

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

a9 and d7 a8 and d6 a7 and d5 a6 and d4 a1, b1, and d1 e1 b2 a2 b3 and d2 e2 a3 b4 d3 a4 e3 b5 a5 b6 TRPC

participants have no statistical significance (p > 0.06), which is consistent with the results found in the discovery study, and suggests that the TRPC can be as an internal reference for quantifying the IIRPCs. Association between the Levels of IIRPCs and Cancer. In set III, only cancer cases without antitumor treatment before blood collection were considered for analysis because antitumor drugs may alter the circulating levels of the IIRPCs and confound the results. Here, we mainly investigated the association between cancer and the levels of the IIRPCs in patterns a and b, respectively (Table 5 and Figure 2), since both patterns accounted for the majority of the participants, and the sample size with pattern d, e, f, or g was too small for statistical analyses. In agreement with the discovery study, the statistical results show that there is no association of the IIRPCs levels with sex and age in the group of patients or controls (p > 0.05). These data suggest that the association between the levels of the IIRPCs and cancer is independent of the potential confounding factors tested. In addition, the IIRPCs levels also remained statistically significantly associated with cancer even if the sera had been stored for three years at −80 °C. For all participants, the levels of the IIRPCs of cancer patients are significantly increased compared to healthy controls and to patients with benign diseases (p < 0.0001, Supporting Information Tables S5−S9). In addition, the levels of the IIRPCs of the patients with benign diseases are also significantly increased compared with healthy controls (p < 0.02, Supporting Information Tables S5−S9). It was found that the participants with higher levels of the IIRPCs showed an increased risk of cancer compared with those with corresponding benign diseases with low level of the IIRPCs and that the fold changes for LC, CRC, PC, and GC patients are significantly higher than those for TC patients (LC, 4.44−9.34; CRC, 3.47−8.85; PC, 4.38−9.55; GC, 3.78−11.04; TC, 2.01−3.50, Supporting Information Tables S5−S9). AUC values reveal that the IIRPCs are powerful biomarker panels to distinguish patients with four types of cancer (LC, CRC, PC, and GC) from patients with benign diseases and healthy controls (AUCs from 0.91 to 0.96; 95% CIs, 0.88−0.98; Figure 4). For example, if the sensitivity of individual IIRPCs for the

components [(HPT) (CFH) (C3) (IgG1) (IgA1)]

[(HPT) (CFH) (C3) (C4A) (IgG1) (IgA1)]

[(HPT) (C3) (C4A) (C5) (C7) (IgG1) (IgA1)] [(HPT) (C3) (C4A) (IgG1) (IgA1)] [(HPT) (C3) (C4A) (IgG1) (ApoA-I) (IgA1)]

[(HPT) (C3) (ApoA-I) (IgA1)] [(TF) (ApoA-I)]

a

ApoA-I, apolipoprotein A-I; C3, complement component 3; C4A, complement component 4A; C5, complement component 5; C7, complement component 7; CFH, complement factor H; HPT, haptoglobin; IgA1, immunoglobulin A1; IgG1, immunoglobulin G1; IIRPCs, immunoinflammation-related protein complexes; TRPC, transferrin-related protein complex; TF, transferrin. bBand number corresponds to the number in Figure 3.

particular patterns. The combination of patterns a, b, c, d, and e accounts for approximately 93% (n = 5083) of the total (Supporting Information Table S3). TRPC as an Internal Reference. In set II, to further validate whether the TRPC could be used as an internal reference for quantifying the IIRPCs, we statistically analyzed the correlation of its amount with sex, age, patterns of the IIRPCs, or health status, as well as their statistical interactions of 2214 participants (Supporting Information Table S4). The statistical results show that the levels of TRPC in the 2214

Table 5. Characteristics of Healthy Controls and Patients To Evaluate the Association of Serum IIRPC Levels with Cancers in the Independent Validation Studya pattern a (n = 632) health status controls cancer lung colorectum pancreas stomach thyroid benign diseases lung colorectum pancreas stomach thyroid

total subjects (n = 1118)

no. of subjects

male sexno. (%)

252 615 163 133 122 116 81 251 55 45 59 39 53

148 335 97 72 59 62 45 149 31 21 36 25 36

84 (57)

P valueb

pattern b (n = 486)

age, yr

P valuec

60.1 ± 12.6

54 41 32 44 9

(56) (57) (54) (71) (20)

0.87 0.98 0.74 0.05