Development of fragment-specific osteopontin antibodies and ELISA for quantification in human metastatic breast cancer

Recombinant human full-length OPN (fl-OPN), human N-OPN and human C-OPN fragments representing the MMP cleaved products were produced in E. coli, and purified as previously described [22]. OPN trilevel controls were purchased from R&D systems. This recombinant full-length OPN has a C-terminal 6x-His tag and was expressed in NS0 mouse myeloma cells. The ranges given for these controls are 1.66–2.26 ng/ml, 5.03–6.57 ng/ml, 10.13–13.09 ng/ml. Native OPN purified from human milk [23] was a generous gift of Esben Sørensen (University of Aarhus, Denmark). All animal protocols were reviewed and approved by our Institutional Animal Use and Care Committee, and performed under veterinary supervision. The fl-OPN protein was used for immunization of OPN homozygous null mice [24]. Prior to immunization, pre-immune blood was collected from mice by retro-orbital bleed. Antibody generation was accomplished by subcutaneous injection of 100 μg fl-OPN in Complete Freund's adjuvent (Sigma, F5881), followed by subsequent injections of 50 μg fl-OPN in Incomplete Freund's Adjuvent (Sigma F5506) on days 13, 33, 45, and 161 days after the initial injection. The mice were then bled and antibody titer measured by ELISA. Final injections of 30 μg fl-OPN intraperitoneally/20 μg fl-OPN subcutaneously were given 4 days before collection of the spleen for fusion. The hybridomas were generated by fusing splenocytes with FO myeloma cells of Balb/c origin (ATCC, Manassas, VA), according to a standard protocol developed by Kohler and Milstein (Kohler G, Milstein, C, 1975), with minor modifications.

After five rounds of immunization, 0.1 ml blood was collected by retro-orbital bleed to determine the OPN antibody level in the serum. A 96-well microtiter plate (BD Biosciences, Bedford, MA.) was coated with 50 ng/ml (100 μl/well) of purified human fl-OPN and incubated at 4°C overnight. The plate was blocked with 3% nonfat dry milk in PBS-T (PBS pH = 7.3, 0.05% Tween-20) at room temperature for 2 hours. The sample serum was loaded at a 1:100 dilution in 3% nonfat dry milk/PBS-T, followed by serial 5-fold dilutions. After incubation at 37°C for 1 hour, the plate was washed with PBS-T, and incubated at 37°C for 2 hours with a combination of two separate secondary antibodies at 100 μl/well: 1:5000 dilution of HRP-conjugated goat anti-mouse IgG, Fcγ fragment specific, and a 1:5000 dilution of HRP-conjugated AffiniPure goat anti-mouse IgG + IgM (H+L) (both from Jackson ImmunoResearch Laboratories). Antigen was detected by incubation with 100 μl/well of tetramethylbenzidine for 15 minutes, and the reaction was stopped by 100 μl/well of 1 M HCl. The plate was read at 405 nm using an automated microplate reader (Bio-Tek Instruments). Screening of subsequent hybridomas was performed similarly, with HRP-conjugated goat antimouse IgG + IgM used as the secondary antibody. In addition, hybridoma screening included negative selection of clones that recognized an unrelated control his-tagged protein, and positive selection of clones recognizing a second recombinant OPN without a his epitope tag. For crossreactivity studies, wells were coated with either human or mouse fl-OPN, and human N- and Cterminal OPN fragments were used for epitope mapping. To compare epitope specificity, unlabeled anti-OPN antibodies were used to compete with biotinylated 1F11 and 2F10 for binding to specific OPN epitopes. Unlabeled antibodies were incubated first, followed by reaction with the biotinylated antibodies to detect binding.

Ascites fluid was collected using either Hsd athymic nude mice or F1 hybrids (H-2^b/d MHC class) of OPN heterozygous mice (C57BL/6 H-2^b) crossed to Balb/c mice (H-2^d). Mice were primed with an intraperitoneal injection of incomplete Freund's adjuvant, followed by injection of 5 × 10⁵ – 5 × 10⁶ hybridoma cells. Ascites was collected from peritoneal tapping. For antibody purification, ascites were centrifuged at 14,000 g for 15 minutes and diluted with 5 volumes of 20 mM sodium phosphate buffer, pH = 7.0. Antibodies were purified over a protein G column (Sigma), and eluted with 0.1 M glycine-HCl, pH = 2.7 and adjusted to neutral pH with 1 M Tris-HCl, pH = 8. Eluted antibodies were dialyzed against PBS. Antibody concentration was determined with a DC protein assay kit (BioRad). On average, 1–4 mg antibody was purified from 1 ml ascites fluid. Purity of antibodies was assessed by SDS-PAGE and Coomassie blue staining. Antibody isotypes were determined using the IsoStrip assay (Roche) following the manufacturer's instructions.

Sulfo-NHS-LC-Biotin [sulfosuccinimidyl-6-(biotinamido) hexanoate] (MW 556.59, Pierce) was prepared at 10 mg/ml in PBS and added to dialyzed MAbs 1F11, 2F10, and 1E3 at a ratio of 186 μg, 148 μg and 150 μg of biotin per mg of antibody, respectively. IE3, a monoclonal antibody against platelet activation antigen CD62-P was used as a control [25]. The mixtures were incubated at room temperature for 1 hour before being dialyzed extensively against PBS to remove uncoupled biotin. The quantities of the biotinylated proteins were then determined using a DC protein assay kit (Bio-Rad).

A two-antibody sandwich ELISA was developed as a quantitative ELISA for fl-OPN measurement. MAb 2F10, which binds specifically to the N-terminus of OPN, was selected as the capture antibody; while 1F11, specific for the C-terminus, was selected as the detection antibody (1F11 was biotinylated). This assay design aimed to detect only OPN molecules that were intact and not cleaved. Firstly, chessboard titrations were performed with 2F10 as the capture antibody and 1F11b as the detecting antibody and fl-OPN as the antigen. The chessboard showed that a dilution of 1.5625 μg/ml of 2F10 and 0.0975 μg/ml of 1F11b would be the optimal concentrations for the sandwich ELISA.

The capture antibody, 2F10, was then used at a concentration of 1.5625 μg/ml in PBS pH = 8.0 and coated at 100 μl/well overnight at 4°C. PMSF (1 mM) and protease inhibitor cocktail (Sigma, 0.5 μl/ml) were added to the blocking and carrying solution of 1% BSA/PBS (referred to hereafter as 1%BSA/PBS/PIC). Wells were blocked with 1%BSA/PBS/PIC for 2 hours at room temperature on a plate shaker at ~500 rpm, washed, and samples and standards loaded for 1 hour at room temperature. After washing, the biotinylated detecting antibody, 1F11b, was added at 0.0975 μg/ml in 1%BSA/PBS/PIC and incubated for 1 hour. After washing, 1:2000 HRPconjugated streptavidin (Jackson Immunoresearch) was incubated for 1 hour, followed by washing and antigen detection with tetramethylbenzidine (TMB). The chromogenic reaction was stopped with 1 M HCl and the plate read at 405 nm with a correction at 570 nm.

Our Institutional Review Board evaluated and approved the protocols for the use of healthy volunteer and patient samples. Plasma samples were collected from 55 healthy blood donors as well as 40 women with metastatic breast cancer (Dr. Susan Miesfeldt, Maine Center for Cancer Medicine and Blood Disorders, Scarborough, Maine). All samples were collected in EDTA tubes; the plasma was then separated, aliquoted into 200 μl and frozen at -70°C within 4 hours of collection. Data on initial stage, histology, receptor status, and disease free interval are presented in the text.

Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) was performed with a Mini-PROTEAN Gel Electrophoresis Unit (Bio-Rad). Protein samples were diluted appropriately and boiled for 5 minutes in the sample buffer (63 mM Tris, 2% SDS, 0.01% bromophenol blue, 5% β-mercaptoethanol). The resolving gel was 12.5% and the stacking gel was 5%. The running buffer consisted of 25 mM Tris base, 192 mM glycine, 0.1% SDS, pH = 8.8. The separated proteins on the gel were transferred to a polyvinylidene diflouride (PVDF) membrane with the Mini-PROTEAN Gel Electrophoresis unit (Bio-Rad Laboratories) The transfer buffer consisted of 25 mM Tris, 192 mM glycine and 0.1% SDS, pH = 8.3.

Nonspecific binding was blocked with 3% BSA in TBS-T buffer (10 mM Tris, pH = 8.0, 150 mM NaCl, 0.05%Tween-20) at room temperature for 1 hour. The membrane was then cut into strips and incubated in individual tubes with the appropriate dilution of the MAbs in 3% BSA in TBST buffer overnight. After the overnight incubation at 4°C the strips were washed in TBS-T in their own tubes for 1.5 hours. The strips were incubated with 3% BSA for 30 minutes, washed 3X in TBS-T, and incubated with HRP-conjugated goat anti-mouse IgG, Fcγ (1:10,000) in 3% BSA TBS-T for 1 hour, followed by a TBS-T and 3% BSA washes. The bands were visualized using an ECL Western blotting detection reagent (Amersham). Prestained molecular weight standards (Bio-Rad) were used to estimate protein sizes.

Monoclonal antibodies (MAb) recognizing human fl-OPN were purified and characterized, and included MAb 2C5, 2H9, 2F10, 2E11 and 1F11. Antibodies were originally identified by their recognition of fl-OPN by ELISA. All are of the IgG1 subtype, with a κ light chain. During the course of the screening, these antibodies were found not to cross react with a random hiscontaining protein, and also recognize a recombinant OPN lacking a his epitope tag. Using human or mouse recombinant fl-OPN protein, we found that under denaturing, reduced conditions, all MAb recognized the recombinant proteins, although the detection of human OPN was more robust (Fig. 1A). OPN is known to run anomalously by SDS-PAGE (~50–70 kD), not corresponding to its actual molecular mass (~35 kD). Antibody recognition to the OPN molecule was mapped to either an N-terminal or C-terminal epitope by ELISA, using purified recombinant protein (Fig. 1B). Epitopes of MAb 2C5, 2F10, 2H9, and 2E11 map to the human N-terminal fragment (aa1-aa166), while that of 1F11 maps to the C-terminal fragment (aa167-aa314). This was verified by immunoblot analysis using the same recombinant N-terminal or C-terminal OPN fragment (Fig. 1C).

Bild vergrößern

The antibodies were then tested for recognition of native OPN protein by ELISA and immunoblot analysis. We obtained purified native human milk OPN [23] and tested activity of the MAb by ELISA (Fig. 2A). MAb 1F11 and 2F10, which we subsequently used to develop the quantitative sandwich ELISA, both recognized human native milk OPN. We also collected mouse kidneys, which are tissues known to contain endogenous OPN protein [26, 27]. We tested wild kidneys from wild type, OPN heterozygous, and OPN null mice [24], compared to recombinant human OPN (Fig. 2B). OPN was detected in kidneys from wild type and OPN heterozygous mice, but not from kidneys from OPN null animals, showing recognition of mouse native OPN. Finally, since OPN expression in smooth muscle cells is induced by Notch signaling [28], we also tested human aortic smooth muscle cells with activated Notch, and found strong expression of OPN detectable by both 1F11 and 2F10 MAb. These data support the specificity of the antibody in recognition of native mouse and human OPN protein.

Bild vergrößern

MAb 2F10 specific for the N-terminus of OPN and MAb 1F11 specific for the C- terminus were evaluated as a capture-detection antibody pair in a sandwich ELISA for OPN quantification. To aid detection of signals, the detection antibody was biotinylated. 2F10 was designated the capture antibody, and biotinylated 1F11 was the detection antibody. This antibody selection ensured that only relatively intact OPN molecules would be recognized and quantitatively determined. Chessboard reagent titration experiments were carried out to define optimal assay parameters. Based on this analysis (Table 1) using excess antigen, we found that absorbance plateaued from 50 μg/ml to 3.125 μg/ml of capture antibody, with the drop occurring using 0.78 μg/ml. Thus, we determined the optimal capture antibody concentration as 1.56 μg/ml. In the second checkerboard assay, 1.56 μg/ml was used as the coating concentration with a serial dilution of fl-OPN from 0.156 μg/ml to 2.4 pg/ml, with no fl-OPN as a negative control. From columns 1–11 a serial dilution of detecting antibody was used from 1.56 μg/ml to 1.5 ng/ml, with no detecting antibody in column 12. At 9.7 ng/ml of fl-OPN, the absorbance plateaued between 1.56 μg/ml of 1F11b to 0.195 μg/ml, with a drop occurring at 0.098 μg/ml. A titration of the fl-OPN antigen occurs using 0.098 μg/ml 1F11b between 19.5 ng/ml of Fl-OPN to 2.4 pg/ml. Thus, the capture antibody optimum concentration is 1.56 μg/ml while the detecting antibody 1F11b optimum concentration is 0.0975 μg/ml.

A sandwich ELISA was developed using MAb 2F10 (N-terminal OPN specific) as capture antibody and biotinylated 1F11 (C-terminal OPN specific) as detection antibody. This ELISA permits quantitative, rapid and reproducible measurement of OPN levels in blood plasma and other bodily fluids. The fl-OPN was used as antigen in this assay to generate the standard curve for calculating OPN concentration of unknown samples. The resulting standard curve had a linear correlation (R²>0.99, Fig. 3A) between OPN concentration and A_405nm, with a linear doseresponse range of 0.078 ng/ml-10 ng/ml. Thus, the sandwich ELISA generates a standard curve that can be used to provide accurate determinations of fl-OPN concentrations in test samples. The assay protocol was followed subsequently to measure OPN level in test samples. Assay sensitivity was calculated by comparing binding with no fl-OPN compared to 0.156 ng/ml. The detection limit was determined as the concentration of human OPN measured at two standard deviations from 0 pg/ml standard along the curve. The mean OD for the control was 0.049 ± 0.0017, and the mean OD at 0.156 ng/ml was 0.087 ± 0.0025, giving a deltaOD = 0.038. Our calculations yielded a sensitivity of 13.9 pg/ml.

Bild vergrößern

Currently there are three fl-OPN kits on the market (IBL, Assay Designs, R&D Systems) used previously to quantify plasma OPN levels in cancer patients. Thus, we compared our ELISA with the commercially available tests. The detection limits of these assays are: IBL, 5–320 ng/ml; Assay Designs, 2–32 ng/ml; Quantikine R&D, 0.312–20 ng/ml. We obtained fl-OPN control proteins from R&D Systems representing a high (10.1–13.1 ng/ml), middle (5–6.6 ng/ml), and low (1.7–2.3 ng/ml) protein concentration, and compared assay results using these standards (Fig. 3B). The quantification of fl-OPN varied in all assays, and we compared the actual values with the known standards, as well as the relationship between the three standards, which represent a twofold (1:2) and three fold (1:3) decrease from the high and the middle standards, respectively. Our in-house assay and the Quantikine R&D ELISA were the most accurate, while the Assay Designs over-estimated the standard values from 1.2-fold to 3.6 fold. The relationship between the standards from highest to lowest (1:2/1:3) was proportionally consistent with the in-house (1:2/1:3.1) and Quantikine R&D (1:1.8/1:3.1) assays, but not with the Assay Designs kit (1:1.6/1:1.3). The IBL ELISA kit gave extremely high measurements for these same standards, running 13-fold to 18-fold higher than the standard ranges. The relationship between the high and middle standards was measured as expected with the IBL assay (1:1.8), however, was not consistent between the middle and low standard (1:4).

In an initial attempt to evaluate OPN as a biomarker for cancer, the quantitative ELISA was used to measure fl-OPN levels in patients with metastatic breast cancer (Table 2). In addition, because the available fl-OPN assays were so variable, we compared our results to those using the commercially available kits. The patient population with metastatic breast cancer was chosen for testing, since similar populations have previously been studied with the commercially available OPN ELISAs, and provide a basis for comparison. Using our in-house assay, we initially compared fl-OPN levels in fifty-one healthy controls with levels from forty metastatic breast cancer patients (Fig. 4A). The plasma fl-OPN level among control subjects averaged 1.22 ± 0.446 ng/ml, while the average level in breast cancer samples was 4.76 ± 1.233 ng/ml (p = 0.0042).

Bild vergrößern

We then compared our results of fl-OPN quantification in the plasma of normal volunteers or breast cancer patients with the commercially available assays. In all cases, the standard curves were linear, with R² values of 0.99 (Assay Designs), 0.99 (in-house test), 0.98 (R&D), and 0.98 (IBL). As we suspected, the average values obtained for fl-OPN levels were highly variable between assays using the same group of breast cancer patient samples (Fig. 4). Consistent with the prior comparison with the recombinant fl-OPN standards (Fig. 3B), the IBL assay reflected the highest amount of fl-OPN, with an average of 1170 ng/ml, which is 10-fold higher than the results from the other three assays (Assay Designs 44.7 ng/ml, in house 4.3 ng/ml, R&D 129 ng/ml). Surprisingly, however, the in-house assay and the R&D assay, which were concordant in accurately quantifying the recombinant standards (Fig. 3B), were significantly different, with the quantification from the R&D assay yielding values more than 10-fold higher than the in-house ELISA. Also unexpected was the observation that the Assay Design test quantification was intermediate between the in-house and the R&D assay, since it showed consistently higher values than the other two using the recombinant fl-OPN standards. There was heterogeneity in OPN levels in the breast cancer patients (Fig. 4B). Similarly, quantification of plasma OPN in normal healthy volunteers ranged between assays used. In general, the relationship followed the breast cancer plasma quantification, with the in house yielding the lowest mean value, (1.2 ng/ml), followed by Assay Designs (32 ng/ml), R&D (53.9 ng/ml), and IBL (396.3 ng/ml). Using all assays, the amount of circulating plasma OPN was greater in the breast cancer samples compared to those from normal volunteers (Assay Designs p = 0.025, R&D p = 0.0001, IBL p = 0.006). These results show that OPN is difficult to accurately quantify in human blood, and complicating factors need to be considered when evaluating ELISA data quantifying fl-OPN.