Proteomic Identification of Insulin-like Growth Factor-binding Protein-6 Induced by Sublethal H2O2 Stress from Human Diploid Fibroblasts*

Fibroblasts are the most ubiquitous cell types within our body. They produce various factors to maintain the texture and structure of a particular organ or tissue. To identify protein factors secreted by fibroblasts and alteration of these protein factors upon oxidative stress, HCA3 human skin diploid fibroblasts were exposed to a sublethal dose of H2O2, which induces a prematurely senescent phenotype. Conditioned media from prematurely senescent cells versus control cells were analyzed for proteins using an LC-MS/MS-based proteomic technique. Collagen α1(VI), collagen α2(I), fibronectin, lumican, and matrix metalloproteinase 2 were among the proteins consistently detected from control and H2O2-treated cells. Insulin-like growth factor-binding protein-6 (IGFBP-6) consistently showed up in the conditioned medium of H2O2-treated cells but not from untreated cells. Increased IGFBP-6 production due to H2O2 treatment was confirmed by RT-PCR and Western blot analyses. While H2O2 induced a dose-dependent elevation of IGFBP-6 mRNA, Western blot analyses detected elevated levels of IGFBP-6 protein in the conditioned medium of H2O2-treated cells. In comparison, fibronectin or matrix metalloproteinase 2 did not show changes at the mRNA level in cell lysates or at the protein level in the conditioned medium by H2O2 treatment. Using several types of toxins at sublethal doses, including cis-platin, hydroxyurea, colchicine, l-mimosine, rhodamine, dithiothreitol, or N-ethylmaleimide, we found that these agents induced increases of IGFBP-6 at mRNA and protein levels. An increased level of IGFBP-6 protein was detected in the plasma of aging mice and of young mice treated with doxorubicin. These data suggest that IGFBP-6 may serve as a sensitive biomarker of cell degeneration or injury in vitro and in vivo.

Fibroblasts reside in the stromal layer of the skin and actively maintain the integrity and architecture of the tissue by secreting matrix proteases and depositing extracellular matrix proteins. During the process of aging, it is thought that such cells show changes in biochemistry and gene expression patterns. Some of these changes, such as alterations in the secretion of degradative enzymes, inflammatory cytokines, and growth factors, are related to the senescent phenotype of normal diploid fibroblasts in culture (1)(2)(3)(4)(5). These changes can alter a microenvironment, disrupt tissue structures, and cause growth of neighboring premalignant cells (1,2,(5)(6)(7)(8)(9)(10). With human diploid fibroblasts (HDFs) 1 in culture, the senescent phenotype is typically achieved by serial passage (11,12). However, recent evidence suggests that early passage HDFs respond to a defined dose range of oxidants by entering a state of arrested growth and altered phenotype resembling replicative senescence (13)(14)(15)(16)(17)(18).
A large volume of literature suggests that oxidative stress contributes to aging and aging-associated diseases (19 -22). Although aging is the highest risk factor for cancer, cardiovascular disease, and neurodegenerative disease, mechanisms underlying the interplay between oxidative stress, aging, and diseases have not been well addressed. Recent experimental evidence supports the hypothesis that induction of the senescent phenotype by oxidants confers a tumor-promoting activity of HDFs (6,23). Proteins secreted by senescent-like fibroblasts appear to exhibit the ability to promote the growth and colony formation of initiated keratinocytes (23). Uncovering the nature of the proteins secreted by prematurely senescent cells becomes important in understanding the interplay between oxidative stress, aging, and aging-associated diseases.
Recent advancement in available genomic sequence information has provided an infrastructure for the emerging field of proteomics (24 -28). Most commonly used proteomic techniques involve separation of a complex mixture of proteins into less complex subgroups, mass spectrometry analysis of peptides derived from the proteins in each subgroup, and data mining using bioinformatic tools. Often two-dimensional gel electrophoresis has been used for protein separation. However, staining two-dimensional gels only detects abun-dant proteins that are visible, and the efficiency of protein recovery from the polyacrylamide gel is often a rate-limiting step that prohibits detection of proteins with low abundance. A "shotgun" approach based on the separation capacity of liquid chromatography instrumentation becomes possible if the number of proteins is not overwhelmingly large, such as from a defined subproteome (29,30). Compared with cell lysates, the subproteome of secreted proteins is less complex and allows meaningful identification of proteins using the shotgun approach. Based on the fact that a mass spectrometer measures a molecule based on its abundance given a mixture of different molecules, methods have been developed not only to identify the nature of proteins in a mixture but also to compare relative levels of a protein between different samples (24,27,28,(31)(32)(33). Protease digestion of the proteins from conditioned media followed by analysis of the resulting peptides using ESI-LC-MS/MS allowed us to measure the alteration of secreted protein factors following oxidative stress.

MATERIALS AND METHODS
Chemicals and Reagents-Chemicals were purchased from Sigma unless otherwise indicated. Stabilized H 2 O 2 (H-1009, Sigma) was used, and the concentration of the stock was verified by absorbency at 240 nm.
Maintenance of Cell Culture-HCA 3 human dermal fibroblasts at the population doubling level (PDL) 20 were obtained from Dr. Olivia Periera-Smith. These cells typically reach replicative senescence after PDL 80 and were used for this study at PDL 26 -40. HCA 3 cells were subcultured weekly in 10 ml of Dulbecco's modified Eagle's medium (DMEM) containing 10% (v/v) fetal bovine serum, 50 units/ml penicillin, and 50 g/ml streptomycin (Invitrogen) at a seeding density of 1 ϫ 10 6 cells/100-mm Falcon dish. Under these conditions, the cells reached confluence 6 -7 days after subculture.
Treatment with H 2 O 2 and Various Toxicants-HCA 3 cells were seeded at a density of 2 ϫ 10 6 /100-mm dish 5 days before treatment. At the time of H 2 O 2 treatment, the cells had reached confluence, and the density of cells was 10.48 Ϯ 0.85 ϫ 10 6 . Confluent cells were treated with 600 M H 2 O 2 in a 100-mm dish containing 10 ml of medium. This dose is equivalent to ϳ0.6 pmol of H 2 O 2 /cell. The dose less than 0.85 pmol/cell has been shown to be non-lethal and induce premature senescence in early passage HDFs (13 Preparation of Conditioned Media-To collect conditioned media of HDFs for proteomic analysis, culture media for HCA 3 cells in 100-mm dishes were removed 3 days after H 2 O 2 or other toxicant treatment. The cells were rinsed two times in DMEM and were placed in 6 ml of fresh DMEM containing 0% FBS for 3 days of culture. The serum-free conditioned media were collected, filtered through a 0.45-m filter to remove cell debris, dialyzed against 0.01 N NH 4 HCO 3 , and concentrated 100 times down in volume using a speed vacuum concentrator. Protein concentration in the concen-trated medium was determined by the Bradford method according to the manufacturer's instructions (Bio-Rad).
LC-MS/MS Analysis-The concentrated media were digested overnight with trypsin at a 50:1 ratio, i.e. 1 g of trypsin/50 g of protein (34). The resulting peptides were analyzed by a ThermoFinnigan (San Jose, CA) LCQ Classic quadrupole ion trap mass spectrometer equipped with a Michrom (Auburn, CA) MAGIC2002 HPLC instrument and a nanospray ion source (University of Washington). A mixture of peptides equivalent to 7 g of proteins was loaded onto a 10-cm-long capillary column with a diameter of 365 m (outer diameter) or 100 m (inner diameter). The capillary column was generated using a P2000 capillary puller (Sutter Instrument Co., Novato, CA) and was packed with 5-6 cm of Vydac C 18 material. Samples were eluted at a flow rate of 200 -300 nl/min into a mass spectrometer using reversed phase solvent conditions. Tandem MS spectra of peptides were analyzed with the Turbo SEQUEST software that assigns peptide sequences to the spectra (28). The software was used to search for known human proteins in the non-redundant database from the National Center for Biotechnology Information (NCBI).
RNA Isolation and Semiquantitative RT-PCR-Total RNA was extracted from cells with TRIzol (Invitrogen) and was used as a template for RT-PCR. Following the RT reaction using 2 g of total RNA from each sample, 3 l of the 35-l RT reaction mixture was used for each PCR. PCR for IGFBP-6 was carried out in 28 cycles with the primer pair 5Ј-GAATCCAGGCACCTCTACCA-3Ј and 5Ј-GGTAGAAGCCTC-GATGGTCA-3Ј at 94°C for 30 s, 62°C for 30 s, and 72°C for 30 s. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as a reference gene and as an internal control. PCR for GAPDH was carried out using the primers 5Ј-CGTCTTCACCATGGAGA-3Ј and 5Ј-CGGCCATACGCCCACAGTTT-3Ј at 94°C for 30 s, 55°C for 30 s, and 72°C for 30 s for 30 cycles. The PCR products were detected by agarose gel electrophoresis and ethidium bromide staining.
Plasma Collection and Administration of Doxorubicin to Mice-Blood was collected from BL6 ϫ 129SF1 J mice at 5-7 weeks (young) or 16 -18 months (old) of age from the abdominal vena cava. Total blood (200 -300 l) from an animal was centrifuged at 2000 rpm at 4°C to remove blood cells and sediments. The remaining plasma (40 -45% of total blood) is clear and transparent. For doxorubicin (Dox) treatment, 5-7-week-old male (18 -22 g) mice were treated with Dox via intraperitoneal injection at the dose of 4 mg/kg (10 ml/kg of body weight) according to the protocol described by Sun et al. (35). The animals were injected twice a week for a total of 10 injections. Control animals were injected with saline at the same volume. The animals were not treated for 2 weeks between the first four injections and the last six injections to allow for recovery of bone marrow depression. The blood from the animals was collected 2 weeks after the last injection. treatment on protein factors secreted, we compared the profiles of proteins in the conditioned medium from control versus H 2 O 2 -treated cells. Serum-free conditioned media were collected for concentration and protease digestion. The resulting peptide mixtures were injected into the ESI-LC-MS/MS instrument. A representative total ion current chromatogram from the conditioned medium of control or H 2 O 2treated cells is shown in Fig. 1. The mass spectrometer was operated in a data-dependent MS to MS/MS switching mode so that precursor peptide ions detected in a MS survey scan trigger an ion fragmentation for obtaining MS/MS spectra for each of the precursor peptide ions. MS/MS spectra indicate primarily fragment ions originating from either the C terminus (y ion series) or N terminus (b ion series) of a peptide and were searched against a human protein sequence database using the Turbo SEQUEST software. This software searches the entries against all peptide sequences in the database and assigns correlation scores for the probability of matches. The judgment of a confident match is largely based on two parameters: "Xcorr" and "Ions." Xcorr represents the cross- The conditioned medium was prepared as described in the legend of Fig. 1 Table I lists proteins identified by the criteria described above in the conditioned media collected from three independent experiments. The reproducibility of the analytical method is 80 -90% between different runs with the same sample. Several proteins consistently showed up in all three experiments in both control and H 2 O 2 -treated groups (Table  I). These proteins include collagen ␣1(IV) chain, collagen ␣2(I) chain, lumican, fibronectin, and MMP-2 (gelatinase A or 72-kDa type IV collagenase). Table II summarizes the scores of Xcorr and Ions and the number of peptides identified for these proteins. IGFBP-6 appeared in the conditioned medium of H 2 O 2 -treated cells from all three experiments (Table II). The MS/MS spectra and SEQUEST Flicka protein information output on IGFBP-6 identified from three experiments are shown in Fig. 2   teins, including fibronectin and MMP-2, from both control and H 2 O 2 -treated cells. To verify these results, we performed Western blot analyses using conditioned media collected from control or H 2 O 2 -treated cells. The data indicate that there is no significant difference in the level of MMP-2 and fibronectin in the conditioned media between control and H 2 O 2 -treated cells (Fig.  3). With cell lysates, a minor elevation of fibronectin was detected with H 2 O 2 treatment. MMP-2 was not detected with cell lysates, suggesting that MMP-2 is a secreted protein.

Identification of Proteins in Conditioned Media of HDFs by LC-MS/MS-Early
IGFBP-6 is known to be O-glycosylated at 5 amino acid residues (Thr 126 , Ser 144 , Thr 145 , Thr 146 , and Ser 152 ) (36,37). Western blot analyses of conditioned media showed two bands of IGFBP-6 ( Fig. 3). Presumably the lower molecular weight band represents the non-glycosylated form, and the higher molecular weight band represents the glycosylated form of IGFBP-6. With either form, IGFBP-6 protein showed an elevation in the conditioned medium of H 2 O 2 -treated cells (Fig. 3). IGFBP-6 protein from cell lysates showed a molecular weight between the two forms present in the conditioned medium (Fig. 3), suggesting that the protein is partially glycosylated. The level of IGFBP-6 protein from cell lysates did not show a dramatic increase with H 2 O 2 treatment. These data demonstrate that we are able to verify the data on IGFBP-6 obtained by LC-MS/MS with Western blot analyses.

FIG. 2-continued
To characterize the induction of IGFBP-6 by H 2 O 2 treatment, we performed dose-response studies to measure levels of IGFBP-6 mRNA in cell lysates and to determine levels of IGFBP-6 protein in the conditioned medium. HCA 3 cells were treated with 0, 150, 300, 450, or 600 M H 2 O 2 . RNA and conditioned media were collected in parallel from the same set of samples. Western blot analyses showed a dose-dependent increase of IGFBP-6 protein, both glycosylated and non-glycosylated forms, in the conditioned medium of H 2 O 2treated cells (Fig. 4A). In comparison, no significant changes of fibronectin and MMP-2 at the protein level were detected in the conditioned media of HCA 3 cells treated with various doses of H 2 O 2 . Consistent with the protein measurement data, semiquantitative RT-PCR showed a dose-dependent increase of IGFBP-6 mRNA with H 2 O 2 treatment (Fig. 4B).
Using IGFBP-6 as an in Vitro and in Vivo Biomarker of Cellular Injury-We extended our study by asking whether or not IGFBP-6 can serve as a biomarker of oxidative injury. HCA 3 cells were treated with a variety of chemicals, including a DNA-damaging agent (cis-platin), DNA polymerase inhibitor (hydroxyurea), microtubule disruptor (colchicine), amino acid analogue (mimosine), mitochondrial uncoupler (rhodamine), and reducing agents (dithiothreitol and N-ethylmaleimide). The nuclear receptor agonist retinoic acid, which has been reported to induce IGFBP-6 expression (38,39), was included as a comparison. Cells were treated with a sublethal dose of toxins and were allowed to recover for 3 days before collecting the conditioned medium and RNA as described under "Materials and Methods." Western blot analyses indicated that the IGFBP-6 protein in the conditioned media increased to various degrees due to the treatment with different chemicals (Fig. 5A). Judging from the molecular weight markers, the increased IGFBP-6 from various chemical treatments is mainly the glycosylated form. In contrast, levels of fibronectin or MMP-2 proteins did not appear to change (Fig. 5A). RT-PCR results also indicated up-regulated IGFBP-6 mRNA levels in cells treated with these chemicals (Fig. 5B). These data suggest that a variety of chemical stresses can induce IGFBP-6.
Based on the fact that IGFBP-6 is a secreted protein that increases its expression when cells encounter damaging agents, we reason that this protein may serve as a biomarker of cell injury if the increase can be detected in the plasma of individuals. Two animal models were used to test this hypothesis: aging and Dox treatment. There is much evidence in support of the fact that oxidative stress contributes to cell degeneration during the process of aging. In comparison, Dox, an antineoplastic drug that is known to produce reactive oxygen species, can induce cardiomyopathy and other types of tissue injury. Plasma was collected from young (5-6 weeks) or old (16 months) mice for Western blot analysis to measure the level of IGFBP-6. The results show that old mice, male or female, exhibited an elevated level of IGFBP-6 in the plasma (Fig. 6A). The protocol of administering Dox, as described under "Materials and Methods," has been shown to induce cardiomyopathy (35). The plasma of Dox-treated mice showed an elevation of IGFBP-6 (Fig. 6B). With serum samples from mice, IGFBP-6 appeared to show in one broad band. It is not known whether this band represents the non-glycosylated, glycosylated, or partially glycosylated form. Regardless our data HCA 3 cells were treated with H 2 O 2 at the dose indicated for 2 h and were allowed to recover for 3 days as described under "Materials and Methods." Cells were harvested for RNA extraction at the time of conditioned medium collection. Concentrated conditioned media containing 20 g of proteins were used for Western blot to detect fibronectin, MMP-2, and IGFBP-6 (A). For RT-PCR (B), 2 g of total RNA from each sample were used for RT, and one-tenth of the RT reaction mixture was used for PCR to amplify IGFBP-6. In a parallel PCR using the same RT reaction mixture, GAPDH was amplified to show equal amount of RNAs between each sample. suggest a potential for using plasma levels of IGFBP-6 as a biomarker of tissue degeneration or injury in vivo.

DISCUSSION
The shotgun approach of ESI-LC-MS/MS-based proteomics has lead to the discovery that H 2 O 2 -induced senescentlike human diploid fibroblasts increase the production of IG-FBP-6 protein. The results obtained by this proteomic analysis have been verified by Western blot and semiquantitative RT-PCR analyses. This study sets an example supporting the importance of proteomic techniques in discovering biomarkers, new targets, and novel pathways associated with a particular cellular condition.
Previous studies from our laboratory and others found that senescent or senescent-like fibroblasts produce protein factors that stimulate the growth of initiated keratinocytes (6,23). In theory, proteomic techniques are most suitable for identifying these protein factors and for elucidating the mystery of how many proteins are actually secreted by fibroblasts. Computational analyses of the human genome predict 5235 secreted proteins based on signal peptide sequence analysis (40). Two-dimensional electrophoresis of secreted proteins from adipocytes and osteocytes found about 1000 spots, indicating about 1000 secreted proteins detectable by this method (41,42). In our experimental system, we have confidently identified 10 -20 proteins in the conditioned medium using the shotgun method. Such a low number reflects the limitation of our current instrumentation. The instrument used in this study, a ThermoFinnigan LCQ Classic quadrupole ion trap mass spectrometer, detects peptides at a sensitivity of 200 fmol. With this conventional ion trap, considerable amounts of potentially valuable information are missed when several peptides co-elute, and the instrument is unable to fragment them all efficiently. Much of the emphasis in the proteomic industry involves improving mass spectrometers in sensitivity, speed, selectivity, dynamic range, and mass accuracy. A linear ion trap is able to collect much more information due to its much faster scanning time. One experiment using a demo model of the ThermoFinnigan linear ion trap LC-MS/MS instrument identified 71 proteins from the conditioned medium of control HDFs. This is a 3-7-fold enhancement in the capacity of protein identification. Two-dimensional linear trap mass spectrometers have recently been developed that offer an increased ion trapping capacity, increased detection sensitivity, and better quality of tandem mass spectra. This type of capillary multidimensional liquid chromatography produces a better separation of complex peptide mixtures with two or more series of orthogonal nano-HPLC columns, allowing the characterization of an entire proteome from cell lysates, which contain 200,000 -300,000 proteins (30,44). The improvement in instrumentation will enhance our capacity to profile secreted proteins toward the level that truly reflects the actual number of secreted proteins from a particular cell type.
Despite the fact that LC-MS/MS analysis can generate data reproducibly for certain proteins, such as fibronectin, MMP-2, and IGFBP-6, we have observed considerable variations in the proteins identified (Table I) detection of ions that are most likely abundant. 2) Classical LC-MS/MS is operated under the assumption of "first come and first serve." Because our current instrument does not have detailed ion separation capacity, the detection is somewhat a process of randomization. 3) The optimal detection range of the mass spectrometer is 700 -3000 Da for a peptide. Depending on the completion of protease cleavage of each protein, a peptide within this mass range may not always appear in a large abundance. 4) The data depicting specific protein identifications are largely dependent on the selection criteria of SEQUEST software. The current setting is quite stringent and therefore filters out the peptide ions that may not meet the level of high confidence. 5) Variations may exist between cells from different passages. Because our three experiments were performed with different passages of cells in culture and HDFs are known to progress toward replicative senescence with each subculture, culture conditions and age of the cells may possibly contribute to variations of proteins in the conditioned medium. Regardless of these caveats, the finding that oxidants and multiple toxicants increase the expression and secretion of IGFBP-6 presents a novel biomarker for diagnosing sublethal cell injury. The data from our animal models of aging and Dox treatment support this biomarker argument. With plasma samples in human studies, a gradual increase in the level of IGFBP-6 up to 2-fold has been documented with increasing age (45). Consistent with this finding, we detected an increased level of IGFBP-6 in the plasma of aged mice. IGFBP-6 belongs to a family that contains six well characterized members. These members share sequence homologies, contain abundant cysteine residues, and have been found in a variety of biological fluids, including the plasma (46,47). IGFBP-6 undergoes glycosylation during the process of secretion (36,37). Our data show two broad bands of IG-FBP-6 in Western blots, suggesting non-glycosylated and glycosylated forms, both of which showed elevation in the conditioned medium of H 2 O 2 -treated cells. However, an in vivo study showed only one band of IGFBP-6. It is not known whether the band represents the glycosylated or non-glycosylated form because the glycosylated band seems to be dominant as evidenced by in vitro studies, but the molecular weight of the band from animal serum is close to what is presumably the non-glycosylated form. Species differences also contribute to the complication. Mouse IGFBP-6 appears to have a slightly lower molecular weight than rat IGFBP-6 (36). Regardless, glycosylation appears to play a role in IG-FBP-6 protein stability (36,48). This feature may contribute in part to the increased level of IGFBP-6 protein in addition to transcriptional activation of the IGFBP-6 gene by H 2 O 2 . Several nuclear receptor ligands, such as retinoids, vitamin D, and 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) (39,49,50), have been shown to induce the expression of IGFBP-6. In contrast, transforming growth factor ␤ and agents that elevate intracellular cAMP concentration cause decreases in IGFBP-6 ex-pression (51). Our finding points to a novel pathway regulating the expression of IGFBP-6 by oxidants and toxicants.
The biological function of IGFBP-6 may vary depending on experimental systems. IGFBP-6 preferentially binds to insulinlike growth factor (IGF)-II rather than to IGF-I. Although IG-FBPs can modulate the activity of IGFs through high affinity binding, IGFBPs may also regulate biological processes, such as cell proliferation or growth arrest in an IGF-independent manner, for example through direct cell association. Glycosylation does not appear to affect IGF-II binding but appears to inhibit direct cell association of IGFBP-6 (36,48). Overexpression of IGFBP-6 arbitrarily in non-small cell lung cancer cells activates programmed cell death (52). With TCDD-induced expression of IGFBP-6, a high degree of IGFBP-6 elevation appeared to enhance apoptosis, whereas a reduction of IGFBP-6 inhibited TCDD from inducing apoptosis in thymoma cells (53). Inactivating the expression of IGFBP-6 in colon cancer cells resulted in a gain of cell proliferation, suggesting that IGFBP-6 may be inhibitory for cell growth (43). Whether IGFBP-6 mediates growth arrest or apoptosis induced by H 2 O 2 treatment remains to be determined.