Development of a novel ELISA for human insulin using monoclonal antibodies produced in serum-free cell culture medium

Results: The sensitivity of the insulin ELISA was 0.73 uU/mL with a dynamic range of 2–200 uU/mL. No cross-reactivity with either human C-peptide or human proinsulin was observed. The intra- and inter-assay CVs were < 7%. The mean recovery of insulin added to plasma samples was between 102.2% and 105.7%. The mean linearity of dilution was between 93% and 110% of undiluted plasma samples. The animal serum-free (ASF) insulin ELISA showed no marked degradation of any kit component when stored at 37°C for up to 7 days. Significantly higher fasting insulin levels were observed in overweight or obese subjects ( n = 12) compared to lean subjects ( n = 10, p < 0.05). Feeding markedly increased fasting insulin levels in both lean ( p < 0.02) and overweight or obese ( p < 0.005) subjects. Excellent correlation was observed between insulin levels measured by ASF insulin ELISA and another CE marked insulin ELISA ( y = 1.06 x − 0.44, r = 1.00, n = 44). Conclusions: This novel insulin ELISA provides precision and reliability equal to methods currently used in clinical research and serves as a guide for the development of other serum-free immunoassays. Keywords Animal use alternatives Serum-free hybridomas Immunoassay Insulin ELISA Insulin monoclonal antibodies Diabetes clinical research Introduction At present, monoclonal antibody (mAb) production via hybridoma methods is performed using either the in vivo ascites method or in vitro cell culture methods. For clinical applications, current immunological assays used to measure hormones, such as insulin, typically use mAbs derived via the ascites method. However, the use of ascites for mAb production is restricted in several countries [1] . The in vitro production of mAbs can alleviate ethical concerns and allows for relative ease of purification without the complex, postharvest purification steps necessary when using the ascites method. Although in vitro cell culture methods allow for a reduction in the use of animals and significant improvements in technique for mAb production, these methods typically rely on culture media that contains foetal bovine serum (FBS). Concerns have been raised regarding FBS collection methods, including animal welfare issues [2] and the introduction of undesirable variables into research protocols [3] . For example, not only is FBS chemically undefined with a high lot-to-lot variability [4] , it is also a potential source of contaminants such as prions, viruses, and mycoplasms [5] . The high protein content in FBS often requires additional downstream processing steps [6] , and the cost of serum can be high [7] . These obstacles can be overcome by replacing serum-containing media with chemically-defined media that contain no animal sera. The first experiments that eliminated serum from mammalian cell cultures were reported in the laboratory of Ham [8] . The first reported serum-free medium successfully used for hybridoma cultivation was reported by Chang et al. [9] . In that study, productivity of antibody-secreting hybridomas in serum-free medium was comparable to productivity from cells in serum-containing medium. Moreover, the purity of the mAb product was enhanced when hybridomas were cultured in the absence of serum [9–11] . In most cases, preparations of mAb produced in serum-free medium can be used without further purification [12] , an additional time- and cost-saving feature. The use of serum-free and, by extension, animal protein-free, medium may provide better control and more reproducible results during each step in the production of mAbs. In the course of a clinical trial involving patients with type II diabetes, we sought to develop a quantitative assay for human insulin that replaced both the use of the ascites method and the use of FBS for mAb production. This modified immunoassay would also provide equal precision and reliability to conventional immunoassays used in clinical chemistry and research applications. Materials and methods Reagents Murine hybridoma A61130087P cells, secreting IgG1 against human insulin, and murine hybridoma A61140087P cells, secreting IgG1 that cross-reacts with bovine insulin (BiosPacific, Emeryville, CA), were cultured using standard methods in T-flasks with DMEM supplemented with 2% (v/v) FBS and 2 mmol/L l -glutamine (Invitrogen/Gibco, Carlsbad, CA). Foetal bovine serum that had been used in the parent cultures and the control cultures was certified New Zealand sourced material from Invitrogen/Gibco. Cell culture Freshly thawed cells were grown through 3 passages until the cells were stabilized and the viability was > 95%. Cells were adapted to many different protein-free media in parallel. The culture performing best was taken on for mAb production. The adaptation protocol used was a combination of the serial FBS dilution and the constant density methods. In brief, cells from high-density parent cultures were diluted into fresh RPMI 1640/DMEM (1:1, v/v) (Invitrogen/Gibco) supplemented to 4 mmol/L l -glutamine, 4% (v/v) Maxi-MAb Mark II Supplement (Patricell, Nottingham, UK) and 2.5 mL/L Complex Lipids Solution (Patricell) to 2.5–3.0 × 10 5 cells/mL. Maxi-MAb Mark II, an animal component-free, peptide- and protein-free, media supplement was used to replace FBS. Every 2 to 3 days, the cells were counted and divided again to obtain a cell density of 2.5–3.0 × 10 5 cells/mL. Cell cultures were maintained in incubators (NuAire, Plymouth, MN) at 37°C, 5% CO 2 and 95% humidity, and all cultures were maintained free of antibiotic and antifungal agents. Monoclonal antibody production Cells kept in log phase growth were expanded in T-175 culture flasks with vented caps (Corning, Corning, NY) at 50 mL/flask until the spinners could be inoculated at 3 × 10 5 cells/mL in a fresh volume of 500 mL of protein-free medium supplemented to 0.1% (w/v) Pluronic F-68 (Sigma-Aldrich, St. Louis, MO). Hybridoma clone A61130087P was grown and expanded in a 2 L glass stirred spinner (Techne, Cambridge, UK) to a final volume of 1 L. Hybridoma clone A61140087P was grown and expanded in a 3 L glass stirred spinner (Techne) to a final volume of 1.5 L. Glass stirred vessels were placed in the incubator on a 4-gang stir plate (Techne) set at 60 rpm. Cell densities were allowed to increase to the maximum possible by keeping the volume constant at the daily harvests. When the glucose concentration of the cultures dropped to below 12 mmol/L, cells and media were harvested by centrifugation in 250 mL conical polypropylene centrifuge tubes (Corning) for 10 min at 1200 rpm. The harvested cells were resuspended in a final volume of fresh medium and transferred back to spinner flasks. Aliquots were periodically removed for cell counts on a Casy-1 particle counter (Scharfe Systems, Reutlingen, Germany) and for mAb quantification by ELISA. Selected cell culture metabolites were measured on a BioProfile 100 Plus (Nova Biomedical, Waltham, MA). Cell supernatants were collected and stored at 4°C until they were pooled for purification. The mAbs were purified by Protein A chromatography. ELISA method An important modification from conventional ELISA was the use of human serum albumin, instead of bovine serum albumin, in the preparation of all reagents utilized for this assay. Briefly, the ELISA was developed as a two-step sandwich method in which the microtiter plates were coated with 100 uL of 5 ug/mL insulin mAb (Clone A6113008P, Batch A1449). Twenty uL of samples or standards (Recombinant Human Insulin, Eli Lilly, Indianapolis, IN) were incubated for 1.5 h with 50 mmol/L phosphate buffered saline containing 0.1% human serum albumin (Sigma-Aldrich) and 80 ug/mL immunoglobulin inhibiting reagent (IIR, Linco, St. Charles, MO) assay buffer at room temperature in the microtiter plate with shaking. Charcoal treated human serum was used as the assay matrix. This was followed by the addition of 100 uL of 0.25 ug/mL biotinylated detection mAb (Clone A6114008P, Batch A1450) and incubation for an additional 1 h at room temperature with shaking. Next, streptavidin–horseradish peroxidase solution (Linco) was added and samples were incubated for 30 min at room temperature, followed by incubation with tetramethylbenzidine (Linco) as the substrate for 10–12 min at room temperature. The delta optical density at wavelengths 450 nm and 590 nm was determined using Bio-Tek Synergy TM Multi-detection Microplate Reader (Progen Scientific Ltd, South Yorkshire, UK). Comparison with a commercial insulin ELISA Plasma samples were obtained from 22 in-house volunteers (11 males and 11 females) with body mass index (BMI) ranging between 18.0 and 39.0 kg/m 2 and ages between 21 and 62 years. Glycaemic status and medical history were not determined. After an overnight fast, blood samples were collected in tubes containing EDTA. Next, each subject received breakfast and an additional blood sample was drawn 1 h after feeding. Blood samples were kept at 4°C and centrifuged within 2 h of obtaining the blood. Plasma samples were then immediately frozen and kept at − 70°C until analysis. All 44 plasma samples (22 fasting and 22 post-prandial) were analyzed for insulin levels by both the ASF ELISA and a commercial insulin ELISA (Mercodia AB, Uppsala, Sweden). The commercial ELISA is intended for in vitro diagnostic use and also bears CE marking. Statistical significance of the difference between fasting insulin levels in lean and overweight or obese subjects by ASF ELISA was determined using a Student t -test assuming unequal variance. Post-prandial insulin values by ASF ELISA were compared to the corresponding fasting insulin values of both lean and overweight or obese subjects using a Student t -test assuming equal variance. A linear regression analysis was performed for the comparison of insulin levels determined by the two ELISA methods. Results Serum-free monoclonal antibody yield The two murine hybridoma clones, A61130087P and A61140087P, were completely adapted to serum-free media after 7 successive passages (approximately 3 weeks), with culture densities equal to serum-containing control cultures. These serum-free cultures generated 400 mL of A61130087P supernatant and 370 mL of A61140087P supernatant. Antibody yield was estimated by absorbance at 280 nm with final yields of 116 mg of A61130087P mAb and 111 mg of A61140087P mAb from the volumes subjected to purification. Table 1 shows the doubling times and specific mAb production levels for the two hybridoma cell lines in the serum-containing medium compared with the serum-free medium. The doubling times of the cells in serum-free medium were slightly less than that for cells grown in serum-containing medium. Moreover, the specific mAb productivity of serum-free cultures was notably higher than the corresponding cultures containing FBS (5.30–4.66 pg/culture/day versus 4.42–4.30 pg/culture/day, respectively), indicating higher productivity on a per cell basis of serum-free cultures. The purified antibodies were then used to develop a novel insulin ELISA without the use of animal components such as bovine serum albumin and bovine insulin standard. Performance characteristics Sensitivity The limit of detection of the ASF insulin ELISA was 0.73 uU/mL (mean ± 2 S.D., n = 7). The sensitivity of the ASF insulin was determined by StatLIA® v.3.2 (Brendan Technologies, Inc., Carlsbad, CA) using minimum detectable concentration values from 7 different assays and a fixed weighted five parameter (asymmetrical) logistic curve. We report a dynamic range of 2–200 uU/mL using a 20 uL sample size ( Fig. 1 ). Specificity The ASF insulin ELISA showed no cross-reactivity with human C-peptide or human proinsulin at concentrations up to 20 ng/mL and 2 ng/mL, respectively ( Fig. 1 ). It is worthy of note that ELISAs may give falsely elevated insulin values in a small percentage of patient samples due to interference from endogenous heterophilic antibodies. This type of interference can be overcome by addition of IIR. During the development of this assay, we observed markedly higher insulin values in 4 subjects, indicating possible interference from heterophilic antibodies. In these 4 samples, insulin levels were decreased substantially in the presence of IIR (98.0 to 69.6 uU/mL, 82.2 to 35.8 uU/mL, 115.0 to 12.2 uU/mL and 75.4 to 40.3 uU/mL). Precision Low concentration and high concentration quality controls (QCs) were evaluated to determine intra- and inter-assay CVs ( Table 2 ). Quality control samples were prepared by adding insulin to buffer in order to reflect concentrations at the lower and higher ends of the standard curve. The intra-assay variation was calculated from eight replicates with CVs < 2% for both QC samples. The inter-assay variation was calculated from multiple analyses ( n = 5) of two different low- and high-concentration QC samples. The inter-assay CVs were < 7% for both QC samples. Analytical recovery Recovery experiments were carried out with seven plasma samples with varying amounts of endogenous insulin. To these samples, three different amounts of exogenous insulin were added. In each case, the amount recovered was > 87% with mean recovery of added insulin between 102.2 ± 12.5% and 105.7 ± 13.5% ( Table 3 ). Linearity of dilution Six plasma samples, with either exogenously added insulin (Samples 1, 2 and 3) or high endogenous insulin (Samples 4, 5 and 6) were serially diluted with the assay matrix. Mean values of 93–110% of expected insulin concentrations were obtained with 2-, 4-, and 8-fold dilutions ( Table 4 ). Stability of ASF insulin ELISA kit All individual components of the ASF insulin ELISA kit were evaluated for stability by incubating each kit component for 1, 3 and 7 days at room temperature and at 37°C. No substantial loss of activity of any kit component was observed after 7 days at room temperature or at 37°C (data not shown). The stability of the entire kit itself was also determined at 37°C for 1, 3 and 7 days. No appreciable change in the assay performance was observed over the 7-day period at 37°C when compared to a fresh, control kit ( Fig. 2 ). Fasting and post-prandial insulin levels by ASF ELISA For analysis of fasting and post-prandial insulin levels by the ASF ELISA, the 22 subjects were divided into 2 groups based on BMI: lean (BMI < 25 kg/m 2 ) and overweight or obese (BMI > 25 kg/m 2 ). The lean group included 5 males and 5 females with BMI ranging between 18 and 24.9 kg/m 2 and ages between 21 and 61 years. The overweight or obese group included 6 males and 6 females with BMI ranging between 25.8 and 39 kg/m 2 and ages between 22 and 49 years. Fasting insulin levels in the overweight or obese group were 11.0 ± 2.5 uU/mL (mean ± S.E.) and were significantly higher ( p < 0.05) than those in the lean group (5.5 ± 0.9 uU/mL) ( Fig. 3 ). Feeding significantly increased the mean fasting insulin level in both the lean group (32.5 ± 10.3, p < 0.02) and the overweight or obese group (63.6 ± 17.6, p < 0.005). Comparison of insulin values by a commercial ELISA and ASF ELISA Analysis of insulin values in 44 plasma samples (fasting and post-prandial conditions) showed excellent correlation between the two methods over a wide range of insulin levels (1.5–197 uU/mL). Linear regression analysis of insulin values obtained from the two methods yielded the equation: y = 1.06 x − 0.44 ( r = 1.00). Even at lower insulin concentrations (< 50 uU/mL), excellent correlation was seen between the ASF ELISA and the commercial ELISA ( y = 1.07 x − 0.53, r = 1.00) ( Fig. 4 ). Discussion We report the first human insulin ELISA to be developed using nonanimal proteins that meets or exceeds all the technical specifications of a valid immunoassay. While parent cultures of paired anti-human insulin mAbs had been grown in the presence of FBS, daughter cells were weaned off serum and produced a higher yield per day than corresponding serum-containing cultures. The purified mAbs were then incorporated into an ELISA that replaced bovine serum albumin with human serum albumin and animal-derived insulin with recombinant human insulin for use as the standard. The performance characteristics of the new ASF insulin ELISA, as determined by sensitivity, intra- and inter-assay precision, analytical recovery, and linearity of dilution studies, met all required specifications for use in research applications. A comparison of plasma samples using the ASF ELISA and a commercially-available ELISA (for in vitro diagnostic use and bearing CE marking) for the measurement of insulin showed excellent correlation ( r = 1.00) between the two methods over a wide range of insulin values (2–200 uU/mL). Analysis of insulin values, despite the relatively small number of patient samples, still shows hyperinsulinemia in overweight or obese subjects and also demonstrates an increase in circulating insulin levels after feeding. The ASF insulin ELISA, like any ELISA, has the benefit of being free of radioactive materials, resulting in reduced monetary and environmental costs for radioactive isotope disposal and training of lab technicians. Hybridoma cell lines are often grown in mice, rather than in cell culture. When hybridomas are produced in vitro , FBS is often used as a medium supplement. There is increased interest in the use of in vitro derived antibodies for immunoassays [1,13] . Once a hybridoma cell line that secretes the mAb of interest is established, high-density cell cultures allow for exponential growth and a high, constant rate of mAb production. Numerous established hybridoma cell lines have been adapted to serum-free media with no cessation of mAb production [9,10,11,12,14] . In some cases, hybridoma cell lines were maintained for 3–5 months with continuous antibody secretion [9,10] . Furthermore, substantial improvements in the homogeneity of mAb-containing supernatants were described with serum-free and protein-free media. It is also possible that serum use may lead to problems of reproducibility, mainly through the introduction of unknown and uncontrolled substances into cell culture media. In recent years, safety considerations regarding the production of mAbs for human therapeutic use have led to many advances in commercially available serum-free media. One technical challenge of all immunoassays and ELISAs in particular, is the potential for falsely elevated analyte values in a small percentage of patients due to interference from endogenous heterophilic antibodies. Heterophilic antibodies bind to the capture and/or detection antibody in a non-specific manner, leading to erroneously high values. The influence of this type of interference depends on the concentration and combination of antibodies used in a particular assay. Previous reports indicate that 30–40% of patient samples contain heterophilic (anti-IgG) antibodies [15,16,17] . However, a study by Hennig et al. [18] found only 0.3–1% interference in ELISAs using mAbs for both capture and detection. This type of interference can be overcome by addition of heterophilic blocking reagent(s) to remove human anti-mouse antibodies from patient samples [19,20] . The new ASF insulin ELISA uses murine mAbs for both capture and detection and shows a slight interference by human anti-mouse antibodies in approximately 2–4% of patient samples. During development of this assay, we observed markedly higher insulin values in 4 subjects. In an effort to reduce the heterophilic interference, we used F(ab′)2 fragments of the mAbs for capture and/or detection. The use of antibody fragments may, in some cases, be a useful alternative to monoclonal antibodies, or in this case, the use of IIR. However, F(ab′)2 fragments did not reduce heterophilic interference in this study. Therefore, IIR was used in conjunction with this assay to suppress interference from heterophilic antibodies. We are currently investigating alternative sources and methods of production for use as a heterophile blocker. In conclusion, this novel insulin ELISA does not depend on either the use of the ascites method or FBS for the generation of monoclonal insulin antibodies. This ASF insulin ELISA is now commercially available for research purposes ( http://www.lincoresearch.com ) and we believe this methodology could serve as a guide for the development of immunoassays for other hormones and bio-markers. Acknowledgments The authors wish to thank Rick Ryan, Patricia Facchini, and BiosPacific for technical assistance and advice on this project. We are also grateful to Jonathan Balcombe for editorial assistance. References [1] C.F.M. Hendriksen W. de Leeuw Production of monoclonal antibodies by the ascites method in laboratory animals Res. Immunol. 149 1998 535 542 [2] J. van der Valk D. Mellor R. Brands The humane collection of fetal bovine serum and possibilities for serum-free cell and tissue culture Toxicol. In Vitro 18 2004 1 12 [3] C.E. Jochems J.B. van der Valk F.R. Stafleu V. Baumans The use of fetal bovine serum: ethical or scientific problem? Altern. Lab. Anim. 30 2002 219 227 [4] C.B. Morris S. Warburton Serum-screening and selection A. Doyle J.B. Griffiths D.G. Newell Cell and tissue culture: laboratory procedures vol. 2B 1994 Wiley Chichester, UK 1.1 1.5 [5] S.J. Wessman R.L. Levings Benefits and risks due to animal serum used in cell culture production Dev. Biol. Stand. 99 1999 3 8 [6] S.J. Froud The development, benefits and disadvantages of serum-free media Dev. Biol. Stand. 99 1999 157 166 [7] USDA Livestock and Grain Market News. USDA Pharmaceutical Report. http://www.ams.usda.gov/LSMNpubs/pdf_monthly/pharm.pdf (Accessed January 2006). [8] R.G. Ham Clonal growth of mammalian cells in a chemically defined, synthetic medium Proc. Natl. Acad. Sci. U. S. A. 53 2 1965 288 293 [9] T.H. Chang Z. Steplewski H. Koprowski Production of monoclonal antibodies in serum free medium J. Immunol. Methods 39 1980 369 375 [10] W.L. Cleveland I. Wood B.F. Erlanger Routine large-scale production of monoclonal antibodies in a protein-free culture medium J. Immunol. Methods 56 1983 221 234 [11] P. Baker K. Knoblock L. Noll D. Wyatt B. Lydersen A serum independent medium effective in all aspects of hybridoma technology and immunological applications Dev. Biol. Stand. 60 1985 63 72 [12] H. Murakami H. Masui G.H. Sato N. Sueoka T.P. Chow T. Kano-Sueoka Growth of hybridoma cells in a serum-free medium: ethanolamine is an essential component Proc. Natl. Acad. Sci. U. S. A. 79 1982 1158 1162 [13] S. Fazekas de St. Groth Automated production of monoclonal antibodies in a cytostat J. Immunol. Methods 57 1983 121 136 [14] J. Kovar F. Franek Serum-free medium for hybridoma and parental myeloma cell cultivation: a novel composition of growth-supporting substances Immunol. Lett. 7 1984 339 345 [15] L.M. Boscato M.C. Stuart Incidence and specificity of interference in two-site immunoassays Clin. Chem. 32 1986 1491 1495 [16] L.M. Boscato M.C. Stuart Heterophilic antibodies: a problem for all immunoassays Clin. Chem. 34 1988 27 33 [17] G. Ward L. McKinnon T. Badrick P.E. Hickman Heterophilic antibodies remain a problem for the immunoassay laboratory Am. J. Clin. Pathol. 108 1997 417 421 [18] C. Hennig L. Rink U. Fagin W.J. Jabs H. Kirchner The influence of naturally occurring heterophilic anti-immunoglobulin antibodies on direct measurement of serum proteins using sandwich ELISAs J. Immunol. Methods 235 2000 71 80 [19] B.A. Morton M. O'Connor-Tressel B.G. Beatty J.E. Shively J.D. Beatty Artifactual CEA elevation due to human antimouse antibodies Arch. Surg. 123 1988 1242 1246 [20] E.S. Newman L.A. Moskie R.N. Duggas D.M. Goldenberg H.J. Hansen Murine monoclonal antibody adsorbed onto vinylidene fluoride floccules used to eliminate antibody interference in “sandwich”-type immunoassays Clin. Chem. 35 1989 1743 1746