A Quantitative Proteomic Analysis Uncovers the Relevance of CUL3 in Bladder Cancer Aggressiveness

To identify aggressiveness-associated molecular mechanisms and biomarker candidates in bladder cancer, we performed a SILAC (Stable Isotope Labelling by Amino acids in Cell culture) proteomic analysis comparing an invasive T24 and an aggressive metastatic derived T24T bladder cancer cell line. A total of 289 proteins were identified differentially expressed between these cells with high confidence. Complementary and validation analyses included comparison of protein SILAC data with mRNA expression ratios obtained from oligonucleotide microarrays, and immunoblotting. Cul3, an overexpressed protein in T24T, involved in the ubiquitination and subsequent proteasomal degradation of target proteins, was selected for further investigation. Functional analyses revealed that Cul3 silencing diminished proliferative, migration and invasive rates of T24T cells, and restored the expression of cytoskeleton proteins identified to be underexpressed in T24T cells by SILAC, such as ezrin, moesin, filamin or caveolin. Cul3 immunohistochemical protein patterns performed on bladder tumours spotted onto tissue microarrays (n = 284), were associated with tumor staging, lymph node metastasis and disease-specific survival. Thus, the SILAC approach identified that Cul3 modulated the aggressive phenotype of T24T cells by modifying the expression of cytoskeleton proteins involved in bladder cancer aggressiveness; and played a biomarker role for bladder cancer progression, nodal metastasis and clinical outcome assessment.     

Discovery of Colorectal Cancer Biomarker Candidates by Membrane Proteomic Analysis and Subsequent Verification using Selected Reaction Monitoring (SRM) and Tissue Microarray (TMA) Analysis

Recent advances in quantitative proteomic technology have enabled the large-scale validation of biomarkers. We here performed a quantitative proteomic analysis of membrane fractions from colorectal cancer tissue to discover biomarker candidates, and then extensively validated the candidate proteins identified. A total of 5566 proteins were identified in six tissue samples, each of which was obtained from polyps and cancer with and without metastasis. GO cellular component analysis predicted that 3087 of these proteins were membrane proteins, whereas TMHMM algorithm predicted that 1567 proteins had a transmembrane domain. Differences were observed in the expression of 159 membrane proteins and 55 extracellular proteins between polyps and cancer without metastasis, while the expression of 32 membrane proteins and 17 extracellular proteins differed between cancer with and without metastasis. A total of 105 of these biomarker candidates were quantitated using selected (or multiple) reaction monitoring (SRM/MRM) with stable synthetic isotope-labeled peptides as an internal control. The results obtained revealed differences in the expression of 69 of these proteins, and this was subsequently verified in an independent set of patient samples (polyps (n = 10), cancer without metastasis (n = 10), cancer with metastasis (n = 10)). Significant differences were observed in the expression of 44 of these proteins, including ITGA5, GPRC5A, PDGFRB, and TFRC, which have already been shown to be overexpressed in colorectal cancer, as well as proteins with unknown function, such as C8orf55. The expression of C8orf55 was also shown to be high not only in colorectal cancer, but also in several cancer tissues using a multicancer tissue microarray, which included 1150 cores from 14 cancer tissues. This is the largest verification study of biomarker candidate membrane proteins to date; our methods for biomarker discovery and subsequent validation using SRM/MRM will contribute to the identification of useful biomarker candidates for various cancers. Data are available via ProteomeXchange with identifier PXD000851.Recent advances in proteomic technology have contributed to the identification of biomarkers for various diseases. Improvements in LC-MS technology have led to an increase in the number of proteins that have been identified. In addition, a stable isotopic labeling method using isobaric tag for relative and absolute quantitation (iTRAQ)¹ and stable isotope labeling by amino acids in cell culture has enabled the quantitative analysis of multiple samples (1, 2). Therefore, a large number of proteins have already been identified as biomarker candidates; however, only a few of these have been used in practical applications because most have not yet progressed to the validation stage, in which potential biomarker candidates are quantified on a large scale. The validation of biomarker candidates is generally accomplished using Western blotting and enzyme-linked immunosorbent assays (ELISA) if specific and well-characterized antibodies for these candidates are available. However, highly specific antibodies are not currently available for most novel biomarker candidate proteins, and it takes a significant amount of time and money to obtain these antibodies and optimize ELISA assay systems for many candidates; therefore, another validation assay system needs to be developed. Selected (or multiple) reaction monitoring (SRM or MRM) was previously shown to be a potentially effective method for the validation of biomarker candidates (3–5). The SRM/MRM assay can measure multiple targets at high sensitivity and throughput without antibodies; hence, it is useful for initial quantitative evaluations and the large-scale validation of biomarker candidates, which defines validation of hundreds of biomarker candidate proteins simultaneously.In addition to these technical improvements, the fractionation process also plays an important role in proteome analysis for biomarker discovery. This procedure very effectively analyzes the proteomes of specific cellular compartments or organelles in detail, which reduces sample complexity. The preparation of a membrane fraction was previously shown to be useful for identifying membrane proteins that are generally expressed at relatively low levels. Membrane proteins play critical roles in many biological functions, such as signal transduction, cell-cell interactions, and ion transport, account for ∼38% of all proteins encoded by the mammalian genome and more than one-third of biomarker candidates, and are also potential targets for drug therapy (6, 7). Therefore, membrane proteome analysis is important for biomarker discovery. However, difficulties have been associated with extracting and solubilizing membrane proteins and subsequent protease digestion. Many procedures have consequently been developed to improve the solubilization and digestion of membrane proteins (8–11), and a protocol using phase transfer surfactant (PTS) was shown to be suitable for membrane proteomics using LC-MS/MS (12, 13).The selection of a control group for comparisons is also important for identifying potential biomarkers. Tissue samples from cancer patients have been used in many studies to discover biomarker candidates by proteomic analysis. Previous studies, including our own, attempted to compare cancer tissues with matched normal tissue (14–17). However, marked differences have been reported in the histology, genetics, and proteomics of normal and cancer tissues, and many biomarker candidates have been identified, by making it difficult to narrow down more reliable candidates for further validation. Lazebnik recently emphasized that the features of malignant, but not benign tumors could be used as a hallmark of cancer (18), and also that premalignant lesions were more appropriate controls for cancer tissue than normal tissue for the identification of biomarker candidates involved in cancer progression. Moreover, comparisons of cancer with and without metastasis may also assist in the discovery of biomarker candidates involved in cancer metastasis. Therefore, the identification of biomarker candidates that can be used to diagnose and determine the prognosis of cancer should become more effective by comparing cancer tissues at different stages, including benign tumors.We performed a shotgun proteomic analysis of membrane fractions prepared from colorectal cancer tissue and benign polyps in the present study to identify biomarker candidates for the diagnosis and treatment of cancer. We identified a large number of biomarker candidate proteins associated with the progression of colon cancer by using membrane protein extraction with PTS followed by iTRAQ labeling. SRM/MRM confirmed the altered expression of these biomarker candidates, and these results were further verified using an independent set of tissue samples. A protein with uncharacterized function, C8orf55, was also validated with a tissue microarray that included various types of cancers.     

Depletion of highly abundant proteins from human cerebrospinal fluid: a cautionary note

Affinity-based techniques, both for enrichment or depletion of proteins of interest, suffer from unwanted interactions between the bait or matrix material and molecules different from the original target. This effect was quantitatively studied by applying two common procedures for the depletion of albumin/gamma immunoglobulin to human cerebrospinal fluid. Proteins of the depleted and the column-bound fraction were identified by mass spectrometry, employing ¹⁸O labeling for quantitation of their abundance. To different extents, the depletion procedures caused the loss of proteins previously suggested as biomarker candidates for neurological diseases. This is an important phenomenon to consider when quantifying protein levels in biological fluids.     

Myristoylation of proteins in the yeast secretory pathway.

Protein myristoylation was investigated in the yeast secretory pathway. Conditional secretory mutations were used to accumulate inteRmediaries in the pathway between the endoplasmic reticulum and Golgi (sec 18, 20), within the Golgi (sec 7), and between the Golgi and plasma membrane (sec 1, 3, 4, 5, 6, 8, 9). The accumulation of vesicles was paralleled by the enrichment of a defined subset of proteins modified either via ester or amide linkages to myristic acid: Myristoylated proteins of 21, 32, 49, 56, 75, and 136 kDa were enriched between the endoplasmic reticulum and Golgi; proteins of 21, 32, 45, 56, 75, 136 kDa were enriched by blocks within the Golgi; and proteins of 18, 21, 32, 36, 49, 68, and 136 kDa were trapped in a myristoylated form by blocks between the Golgi and plasma membrane. This enrichment of myristoylated proteins was reversed upon returning the cells to the permissive temperature for secretion. The fatty acid was linked to the 21-kDa protein via a hydroxylamine-resistant amide linkage (N-myristoylation) and to the proteins of 24, 32, 49, 56, 68, 136 kDa via hydroxylamine-labile ester linkage (E-myristoylation). In addition, myristoylated proteins of 21, 56, and 136 kDa were glycosylated via amino linkages to asparagine. This suggests they are exposed to the lumen of the secretory pathway. Three proteins (24, 32, and 56) were E-myristoylated in the presence of protein synthesis inhibitors, indicating this modification can occur posttranslationally. After using cycloheximide to clear protein passengers from the secretory pathway the 21-, 32-, and 56-kDa proteins continued to accumulate in a myristoylated form when vesicular transport was blocked between the Golgi and plasma membrane. These data suggest that myristoylation occurs on a component of the secretory machinery rather than on a passenger protein.     

Proteomics characterization of abundant Golgi membrane proteins

A mass spectrometric analysis of proteins partitioning into Triton X-114 from purified hepatic Golgi apparatus (84% purity by morphometry, 122-fold enrichment over the homogenate for the Golgi marker galactosyl transferase) led to the unambiguous identification of 81 proteins including a novel Golgi-associated protein of 34 kDa (GPP34). The membrane protein complement was resolved by SDS-polyacrylamide gel electrophoresis and subjected to a hierarchical approach using delayed extraction matrix-assisted laser desorption ionization mass spectrometry characterization by peptide mass fingerprinting, tandem mass spectrometry to generate sequence tags, and Edman sequencing of proteins. Major membrane proteins corresponded to known Golgi residents, a Golgi lectin, anterograde cargo, and an abundance of trafficking proteins including KDEL receptors, p24 family members, SNAREs, Rabs, a single ARF-guanine nucleotide exchange factor, and two SCAMPs. Analytical fractionation and gold immunolabeling of proteins in the purified Golgi fraction were used to assess the intra-Golgi and total cellular distribution of GPP34, two SNAREs, SCAMPs, and the trafficking proteins GBF1, BAP31, and alpha(2)P24 identified by the proteomics approach as well as the endoplasmic reticulum contaminant calnexin. Although GPP34 has never previously been identified as a protein, the localization of GPP34 to the Golgi complex, the conservation of GPP34 from yeast to humans, and the cytosolically exposed location of GPP34 predict a role for a novel coat protein in Golgi trafficking.     

An integrated approach for comparative proteomic analysis of human bile reveals overexpressed cancer-associated proteins in malignant biliary stenosis

Proteomics is a key tool in the identification of new bile biomarkers for differentiating malignant and nonmalignant biliary stenoses. Unfortunately, the complexity of bile and the presence of molecules interfering with protein analysis represent an obstacle for quantitative proteomic studies in bile samples. The simultaneous need to introduce purification steps and minimize the use of pre-fractionation methods inevitably leads to protein loss and limited quantifications. This dramatically reduces the chance of identifying new potential biomarkers. In the present study, we included differential centrifugation as a preliminary step in a quantitative proteomic workflow involving iTRAQ labeling, peptide fractionation by OFFGEL electrophoresis and LC-MS/MS, to compare protein expression in bile samples collected from patients with malignant or nonmalignant biliary stenoses. A total of 1267 proteins were identified, including a set of 322 newly described bile proteins, mainly belonging to high-density cellular fractions. The subsequent comparative analysis led to a 5-fold increase in the number of quantified proteins over previously published studies and highlighted 104 proteins overexpressed in malignant samples. Finally, immunoblot verifications performed on a cohort of 8 malignant (pancreatic adenocarcinoma, n = 4; cholangiocarcinoma, n = 4) and 5 nonmalignant samples (chronic pancreatitis, n = 3; biliary stones, n = 2) confirmed the results of proteomic analysis for three proteins: olfactomedin-4, syntenin-2 and Ras-related C3 botulinum toxin substrate 1. This article is part of a Special Issue entitled: Biomarkers: A Proteomic Challenge.     

GS32, a Novel Golgi SNARE of 32 kDa, Interacts Preferentially with Syntaxin 6

Syntaxin 1, synaptobrevins or vesicle-associated membrane proteins, and the synaptosome-associated protein of 25 kDa (SNAP-25) are key molecules involved in the docking and fusion of synaptic vesicles with the presynaptic membrane. We report here the molecular, cell biological, and biochemical characterization of a 32-kDa protein homologous to both SNAP-25 (20% amino acid sequence identity) and the recently identified SNAP-23 (19% amino acid sequence identity). Northern blot analysis shows that the mRNA for this protein is widely expressed. Polyclonal antibodies against this protein detect a 32-kDa protein present in both cytosol and membrane fractions. The membrane-bound form of this protein is revealed to be primarily localized to the Golgi apparatus by indirect immunofluorescence microscopy, a finding that is further established by electron microscopy immunogold labeling showing that this protein is present in tubular-vesicular structures of the Golgi apparatus. Biochemical characterizations establish that this protein behaves like a SNAP receptor and is thus named Golgi SNARE of 32 kDa (GS32). GS32 in the Golgi extract is preferentially retained by the immobilized GST–syntaxin 6 fusion protein. The coimmunoprecipitation of syntaxin 6 but not syntaxin 5 or GS28 from the Golgi extract by antibodies against GS32 further sustains the preferential interaction of GS32 with Golgi syntaxin 6.     

Vacuolar storage proteins are sorted in the cis-cisternae of the pea cotyledon Golgi apparatus

Developing pea cotyledons contain functionally different vacuoles, a protein storage vacuole and a lytic vacuole. Lumenal as well as membrane proteins of the protein storage vacuole exit the Golgi apparatus in dense vesicles rather than in clathrin-coated vesicles (CCVs). Although the sorting receptor for vacuolar hydrolases BP-80 is present in CCVs, it is not detectable in dense vesicles. To localize these different vacuolar sorting events in the Golgi, we have compared the distribution of vacuolar storage proteins and of alpha-TIP, a membrane protein of the protein storage vacuole, with the distribution of the vacuolar sorting receptor BP-80 across the Golgi stack. Analysis of immunogold labeling from cryosections and from high pressure frozen samples has revealed a steep gradient in the distribution of the storage proteins within the Golgi stack. Intense labeling for storage proteins was registered for the cis-cisternae, contrasting with very low labeling for these antigens in the trans-cisternae. The distribution of BP-80 was the reverse, showing a peak in the trans-Golgi network with very low labeling of the cis-cisternae. These results indicate a spatial separation of different vacuolar sorting events in the Golgi apparatus of developing pea cotyledons.     

Localization and targeting of the Saccharomyces cerevisiae Kre2p/Mnt1p alpha 1,2-mannosyltransferase to a medial-Golgi compartment

The yeast Kre2p/Mnt1p alpha 1,2-mannosyltransferase is a type II membrane protein with a short cytoplasmic amino terminus, a membrane- spanning region, and a large catalytic luminal domain containing one N- glycosylation site. Anti-Kre2p/Mnt1p antibodies identify a 60-kD integral membrane protein that is progressively N-glycosylated in an MNN1-dependent manner. Kre2p/Mnt1p is localized in a Golgi compartment that overlaps with that containing the medial-Golgi mannosyltransferase Mnn1p, and distinct from that including the late Golgi protein Kex1p. To determine which regions of Kre2p/Mnt1p are required for Golgi localization, Kre2p/Mnt1p mutant proteins were assembled by substitution of Kre2p domains with equivalent sequences from the vacuolar proteins DPAP B and Pho8p. Chimeric proteins were tested for correct topology, in vitro and in vivo activity, and were localized intracellularly by indirect immunofluorescence. The results demonstrate that the NH2-terminal cytoplasmic domain is necessary for correct Kre2p Golgi localization whereas, the membrane-spanning and stem domains are dispensable. However, in a test of targeting sufficiency, the presence of the entire Kre2p cytoplasmic tail, plus the transmembrane domain and a 36-amino acid residue luminal stem region was required to localize a Pho8p reporter protein to the yeast Golgi.     

Wnt signaling requires retromer-dependent recycling of MIG-14/Wntless in Wnt-producing cells

Wnt proteins are secreted signaling molecules that play a central role in development and adult tissue homeostasis. We have previously shown that Wnt signaling requires retromer function in Wnt-producing cells. The retromer is a multiprotein complex that mediates endosome-to-Golgi transport of specific sorting receptors. MIG-14/Wls is a conserved transmembrane protein that binds Wnt and is required in Wnt-producing cells for Wnt secretion. Here, we demonstrate that in the absence of retromer function, MIG-14/Wls is degraded in lysosomes and becomes limiting for Wnt signaling. We show that retromer-dependent recycling of MIG-14/Wls is part of a trafficking pathway that retrieves MIG-14/Wls from the plasma membrane. We propose that MIG-14/Wls cycles between the Golgi and the plasma membrane to mediate Wnt secretion. Regulation of this transport pathway may enable Wnt-producing cells to control the range of Wnt signaling in the tissue.     

Subcellular localization of UDP-GlcNAc, UDP-Gal and SLC35B4 transporters

The mechanisms of transport and distribution of nucleotide sugars in the cell remain unclear. In an attempt to further characterize nucleotide sugar transporters (NSTs), we determined the subcellular localization of overexpressed epitope-tagged canine UDP-GlcNAc transporter, human UDP-Gal transporter splice variants (UGT1 and UGT2), and human SLC35B4 transporter splice variants (longer and shorter version) by indirect immunofluorescence using an experimental model of MDCK wild-type and MDCK-RCA(r) mutant cells. Our studies confirmed that the UDP-GlcNAc transporter was localized to the Golgi apparatus only and its localization was independent of the presence of endogenous UDP-Gal transporter. After overexpression of UGT1, the protein colocalized with the Golgi marker only. When UGT2 was overexpressed, the protein colocalized with the endoplasmic reticulum (ER) marker only. When UGT1 and UGT2 were overexpressed in parallel, UGT1 colocalized with the ER and Golgi markers and UGT2 with the ER marker only. This suggests that localization of the UDP-Gal transporter may depend on the presence of the partner splice variant. Our data suggest that proteins involved in nucleotide sugar transport may form heterodimeric complexes in the membrane, exhibiting different localization which depends on interacting protein partners. In contrast to previously published data, both splice variants of the SLC35B4 transporter were localized to the ER, independently of the presence of endogenous UDP-Gal transporter.     

Biomarker discovery from pancreatic cancer secretome using a differential proteomic approach

Quantitative proteomics can be used as a screening tool for identification of differentially expressed proteins as potential biomarkers for cancers. Candidate biomarkers from such studies can subsequently be tested using other techniques for use in early detection of cancers. Here we demonstrate the use of stable isotope labeling with amino acids in cell culture (SILAC) method to compare the secreted proteins (secretome) from pancreatic cancer-derived cells with that from non-neoplastic pancreatic ductal cells. We identified 145 differentially secreted proteins (>1.5-fold change), several of which were previously reported as either up-regulated (e.g. cathepsin D, macrophage colony stimulation factor, and fibronectin receptor) or down-regulated (e.g. profilin 1 and IGFBP-7) proteins in pancreatic cancer, confirming the validity of our approach. In addition, we identified several proteins that have not been correlated previously with pancreatic cancer including perlecan (HSPG2), CD9 antigen, fibronectin receptor (integrin beta1), and a novel cytokine designated as predicted osteoblast protein (FAM3C). The differential expression of a subset of these novel proteins was validated by Western blot analysis. In addition, overexpression of several proteins not described previously to be elevated in human pancreatic cancer (CD9, perlecan, SDF4, apoE, and fibronectin receptor) was confirmed by immunohistochemical labeling using pancreatic cancer tissue microarrays suggesting that these could be further pursued as potential biomarkers. Lastly the protein expression data from SILAC were compared with mRNA expression data obtained using gene expression microarrays for the two cell lines (Panc1 and human pancreatic duct epithelial), and a correlation coefficient (r) of 0.28 was obtained, confirming previously reported poor associations between RNA and protein expression studies.     

Immunofluorescent localization of 100K coated vesicle proteins

A family of coated vesicle proteins, with molecular weights of approximately 100,000 and designated 100K, has been implicated in both coat assembly and the attachment of clathrin to the vesicle membrane. These proteins were purified from extracts of bovine brain coated vesicles by gel filtration, hydroxylapatite chromatography, and preparative SDS PAGE. Peptide mapping by limited proteolysis indicated that the polypeptides making up the three major 100K bands have distinct amino acid sequences. When four rats were immunized with total 100K protein, each rat responded differently to the different bands, although all four antisera cross-reacted with the 100K proteins of human placental coated vesicles. After affinity purification, two of the antisera were able to detect a 100K band on blots of whole 3T3 cell protein and were used for immunofluorescence, double labeling the cells with either rabbit anti-clathrin or with wheat germ lectin as a Golgi apparatus marker. Both antisera gave staining that was coincident with anti-clathrin, with punctate labeling of the plasma membrane and perinuclear Golgi apparatus labeling. Thus, the 100K proteins are present on endocytic as well as Golgi-derived coated pits and vesicles. The punctate patterns were nearly identical with anti-100K and anti-clathrin, indicating that when vesicles become uncoated, the 100K proteins are removed as well as clathrin. One of the two antisera gave stronger plasma membrane labeling than Golgi apparatus labeling when compared with the anti-clathrin antiserum. The other antiserum gave stronger Golgi apparatus labeling. Although we have as yet no evidence that these two antisera label different proteins on blots of 3T3 cells, they do show differences on blots of bovine brain 100K proteins. This result, although preliminary, raises the possibility that different 100K proteins may be associated with different pathways of membrane traffic.     

Characteristics of endoplasmic reticulum-derived transport vesicles

We have isolated vesicles that mediate protein transport from the ER to Golgi membranes in perforated yeast. These vesicles, which form de novo during in vitro incubations, carry lumenal and membrane proteins that include core-glycosylated pro-alpha-factor, Bet1, Sec22, and Bos1, but not ER-resident Kar2 or Sec61 proteins. Thus, lumenal and membrane proteins in the ER are sorted prior to transport vesicle scission. Inhibition of Ypt1p-function, which prevents newly formed vesicles from docking to cis-Golgi membranes, was used to block transport. Vesicles that accumulate are competent for fusion with cis-Golgi membranes, but not with ER membranes, and thus are functionally committed to vectorial transport. A 900-fold enrichment was developed using differential centrifugation and a series of velocity and equilibrium density gradients. Electron microscopic analysis shows a uniform population of 60 nm vesicles that lack peripheral protein coats. Quantitative Western blot analysis indicates that protein markers of cytosol and cellular membranes are depleted throughout the purification, whereas the synaptobrevin-like Bet1, Sec22, and Bos1 proteins are highly enriched. Uncoated ER-derived transport vesicles (ERV) contain twelve major proteins that associate tightly with the membrane. The ERV proteins may represent abundant cargo and additional targeting molecules.     

Secretome-derived isotope tags (SDIT) reveal adipocyte-derived apolipoprotein C-I as a predictive marker for cardiovascular disease

We developed a quantitative strategy, named secretome-derived isotopic tag (SDIT), to concurrently identify and quantify the adipocyte-secreted plasma proteins from normal and high-fat-diet (HFD) induced obese mice, based on the application of isotope-labeled secreted proteins from cultured mouse adipocytes as internal standards. We detected 197 proteins with significant changes between normal and obese mice plasma. Importantly, a novel adipocyte-secreted plasma protein, apolipoprotein C-I (apoC-I), significantly increased in the obese mice plasma. The expression and secretion of adipocyte apoC-I was detected in differentiated 3T3-L1 and primary rat adipocytes. Our in vitro experiments proved that functional Golgi apparatus was required for apoC-I secretion. Additionally, obese mice had increased apoC-I production in adipose tissue. Population survey of 367 participants showed that the plasma level of apoC-I was significantly increased in obese individuals compared with healthy individuals. After multiple adjustments for age and sex, the odds ratios for risk factors of cardiovascular disease including high LDL cholesterol, hypercholesterolemia, and hypertriglyceridemia, respectively, were used to compare the highest with the lowest apoC-I quartile. Taken together, our studies provide a novel strategy to concurrently identify and quantify tissue-specific secreted proteins. This strategy can be used to identify the largest global characterization of adipocyte-derived plasma proteome and provides a potential disease-related biomarker for clinical diagnoses. By selectively analyzing adipocyte-secreted proteins in plasma from obese vs lean murine and/or human subjects, we discovered that apoC-I is an adipocyte-secreted plasma protein and a predictive marker for cardiovascular disease.