Presented in part as a poster at the 2010 ECVIM Congress, Toulouse, France
Development and analytical validation of an enzyme-linked immunosorbent assay for the measurement of feline tumor necrosis factor α in serum
Article first published online: 27 JUN 2014
© 2014 American Society for Veterinary Clinical Pathology and European Society for Veterinary Clinical Pathology
Veterinary Clinical Pathology
Volume 43, Issue 3, pages 397–404, September 2014
How to Cite
Steiner, J. M., Xenoulis, P. G., Schwierk, V. M. and Suchodolski, J. S. (2014), Development and analytical validation of an enzyme-linked immunosorbent assay for the measurement of feline tumor necrosis factor α in serum. Veterinary Clinical Pathology, 43: 397–404. doi: 10.1111/vcp.12164
- Issue published online: 3 SEP 2014
- Article first published online: 27 JUN 2014
- chronic enteropathy;
- control range;
- ELISA ;
The role of tumor necrosis factor alpha (TNF-α), a cytokine shown to play a crucial role in human Crohn's disease patients, has not been documented in cats with chronic enteropathies. Also, currently, no validated assay for measurement of TNF-α in cats is available.
The objective of this study was to develop and analytically validate an enzyme-linked immunosorbent assay (ELISA) for the quantification of TNF-α in serum from cats.
A sandwich ELISA was developed and analytically validated by assessment of detection limit, linearity, accuracy, precision, and reproducibility. A control range for serum fTNF-α concentration in healthy cats was established. In addition, serum concentrations of fTNF-α in 39 cats with chronic enteropathies were compared with those in 20 healthy cats.
The detection limit of the assay was 38.4 ng/L. Observed-to-expected ratios for serial dilutions of 4 serum samples ranged from 75.1% to 111.9%. Observed-to-expected ratios for spiking recovery for 4 serum samples ranged from 91.3% to 129.7%. Coefficients of variation for intra- and inter-assay variability ranged from 3.9% to 7.6% and from 7.8% to 12.5%, respectively. The control range of the assay was < 38.4–223.5 ng/L. Serum concentrations of feline TNF-α were significantly higher in cats with chronic enteropathies and diarrhea than in cats with chronic enteropathies without diarrhea, or in healthy control cats.
The ELISA described here was suitable for the quantification of fTNF-α in feline serum and should facilitate research into the importance of TNF-α in cats with chronic enteropathies.
Chronic diarrhea is common in cats, and many of these cats are ultimately diagnosed with idiopathic inflammatory bowel disease (IBD). However, the pathogenesis of IBD remains largely unknown. Antigenic stimulation is hypothesized to play an important role, and both food-related and microbial antigens have been implicated to be important. But how antigenic stimulation leads to uncontrolled inflammation and which inflammatory mediators play an important role are largely unknown.
Tumor necrosis factor α (TNF-α) is a cytokine released mainly by macrophages in response to contact with infectious agents and tissue injury, but also in autoimmune conditions. TNF-α has been shown to play a major role in the initiation and enhancement of general inflammation in people. For example, people with IBD have increased serum concentrations of TNF-α, and TNF-α suppression represents an important area of therapeutic intervention in the treatment of IBD in people.[3, 4] However, for cats with IBD, only few data regarding feline TNF-α (fTNF-α) expression are available, and results are conflicting. In one study, healthy cats and cats with IBD showed higher duodenal fTNF-α mRNA expression than cats with other GI diseases, but there was no significant difference in duodenal fTNF-α mRNA expression between healthy cats and cats with IBD. In the same study, many healthy cats had evidence of a mild inflammatory infiltrate of the duodenum, and cats with evidence of an inflammatory infiltrate (healthy or diseased) showed a higher duodenal fTNF-α mRNA expression than cats without evidence of such an infiltrate. Further studies in a larger number of cats are necessary to clarify the significance of these results. Evaluation of fTNF-α concentrations in biologic samples such as serum or feces maybe superior to the evaluation of local mRNA expression in cats with IBD. To date and to the author's knowledge, no assay for the measurement of feline TNF-α in biologic samples from cats is available, although an ELISA optimized for the measurement of fTNF-α in cell culture supernatants (Feline TNF-alpha DuoSet, R&D Systems, Minneapolis, MN, USA) does exist. Cell culture supernatant is a biologic sample with potentially less matrix effect and complexity than serum, and so far, the assay has not been validated for feline serum. To the author's knowledge, crossreactivity between human-specific tests and fTNF-α in biologic samples from cats have also not been evaluated.
Thus, the hypothesis to be tested in the present study was whether fTNF-α concentration is increased in cats with chronic enteropathies. To test this hypothesis, an ELISA for the measurement of fTNF-α was developed using recombinant fTNF-α (rfTNF-α) as an antigen. The aims of this study were to produce rfTNF-α in Escherichia coli, purify the protein, develop and analytically validate an ELISA for the measurement of fTNF-α in serum, and to compare serum concentrations of fTNF-α in cats with chronic enteropathies to fTNF-α levels in serum of healthy control cats.
Materials and Methods
Expression of recombinant feline TNF-α (rfTNF-α) in Escherichia coli
A stock of E coli strain XL1-blue bacteria transfected with a recombinant pGEX-2T plasmid (Pharmacia, Uppsala, Sweden) was obtained from Dr. Espen Rimstad at the Department of Pharmacology, Microbiology and Food Hygiene of the Norwegian School of Veterinary Science, Oslo, Norway. The original mRNA encoding fTNF-α was extracted from feline macrophage cultures, transcribed into complementary DNA, and cloned into the plasmid vector pGex-2T, which was subsequently transfected into E coli strain XL1-blue, as described elsewhere. These transfected bacteria produced a fusion protein composed of rfTNF-α and the affinity tag glutathione-S-transferase (GST). After incubation on an agar-plate (Luria-Bertani medium containing 75 mg/L ampicillin) at 37°C for 12 hours, a single bacterial colony was selected for larger scale expression and purification of rfTNFα, as previously described elsewhere.
Cleavage of GST from rfTNF-α and partial characterization of rfTNF-α
Three columns with a bed volume of 10 mL each and equipped with filter frits were packed with glutathione-agarose beads (Sigma Chemicals, St. Louis, MO, USA), prepared according to the manufacturers' instructions. Two aliquots of 15 mL of fusion protein-containing solution, originating from 1 L of the original bacterial culture solution, were mixed with 140 mg of bead powder and loaded onto the 3 columns. While incubating at 4°C, the columns were slowly inverted for 12 h.
The resin was then washed with 5 column volumes of a 1× PBS-Triton X-100 buffer (Triton x-100: 1% v/v; Sigma Chemicals) followed by PBS buffer using an ÄKTA Purification System (GE Healthcare Bio-Sciences AB, Uppsala, Sweden). Washing with PBS was repeated until the absorbance at 280 nm of the eluent determined by the ÄKTA system approached baseline values. Cleavage of the rfTNF-α portion from the bead-bound GST portion was initiated with 10 μL of bovine thrombin (1 activity unit/μL; Sigma Chemicals) added to each column. After an incubation period of 4 h at 4°C, each column was washed with PBS, and the eluted rfTNF-α fractions were collected until the absorbance at 280 nm returned to baseline values. The purity of the resultant protein was confirmed by SDS-PAGE analysis (data not shown).
The molecular weight of rfTNF-α was estimated by the surface-enhanced laser desorption/ionization time-of-flight mass spectrometry (SELDI-TOF-MS). The amino acid sequence of the first 11 N-terminal amino acids of rfTNF-α was determined using the Edman degradation method at the Protein Chemistry Laboratory, Department of Biochemistry and Biophysics at Texas A&M University.
Production of polyclonal antibodies against rfTNF-α
The production of anti-rfTNF-α antiserum was performed at Lampire Biological Laboratories, Pipersville, PA, USA. Briefly, 2 New Zealand White rabbits (Oryctolagus cuniculus) were injected subcutaneously with 200 μg of rfTNF-α protein in 0.5 mL PBS, mixed with 0.5 mL of complete Freund's adjuvant (Sigma Chemicals). The rabbits were re-inoculated at 2 additional time-points with 150 μg of rfTNF-α in 0.5 mL of PBS spiked with 0.5 mL of incomplete Freund's adjuvant on days 21 and 42 after the initial injection. One of the animals received a fourth injection of 150 μg of rfTNF-α in 0.5 mL of PBS mixed with 0.5 mL of incomplete Freund's adjuvant on day 63 after the initial injection. To assess the production of antibodies against the injected protein, blood samples were collected 10 days after each injection, and a serum antibody titer was estimated by a simplified radioimmunoassay (data not shown). The animals were exsanguinated after a total of 3 and 4 injections, respectively. The separated antiserum was then frozen at −80° until further use.
Purification of anti-rfTNF-α antibodies
Polyclonal anti-rfTNF-α antibodies were purified from the antiserum by affinity chromatography. Briefly, an N-hydroxysuccinimide (NHS)-affinity column (GE Healthcare Bio-Sciences AB) was prepared according to the manufacturer's instructions, permanently binding 1 mg of pure rfTNF-α to the NHS resin. A volume of 20 mL of rabbit anti-serum was thawed and dialyzed against the starting buffer (buffer A; 75 mM Tris-HCl, 150 mM NaCl, pH 8.0). The NHS-affinity column was equilibrated with 5 column volumes of buffer A and 2 mL of the dialyzed antiserum was applied onto the column. The column was washed with buffer A until the absorbance measured in the eluent (at 280 nm) returned to baseline values. Subsequently, the elution buffer (buffer B; 100 mM glycine, 500 mM NaCl, pH 2.7) was applied and eluent fractions were collected. Fractions with a significant absorbance over baseline at 280 nm were pooled and concentrated. The purity of the antibody solution was assessed by SDS-PAGE. Aliquots with a concentration of 1 mg/mL were stored frozen at −80°C until further use.
Biotinylation of anti-rfTNF-α antibodies
For the production of secondary antibodies for the ELISA, 2 mg of the affinity-purified polyclonal antibodies in PBS (1 mg/mL) were mixed with a 20-fold molar excess of EZ-Link Sulfo-NHS-LC tagged biotin (Pierce Chemical CO, Rockford, IL, USA), according to the manufacturer's instruction. The efficacy of the biotinylation process was determined by use of a 2-4′-hydroxyazonbenzene benzoic acid (HABA)/avidin assay kit (Pierce Chemical CO). A biotinylation coefficient between 4.0 and 8.0 mmol biotin/mmol antibody was considered optimal, based on previous experiences in our laboratory. The biotinylated antibodies were stored at −80°C until further use.
Development of ELISA
Calibrator solutions of fTNF-α were prepared using a serial 2-fold dilution of the purified rfTNF-α protein in sample buffer (PBS, 1% BSA, 0.05% Tween) at concentrations of 1250, 625, 312.5, 156.3, 78.1, and 39.1 ng/L rfTNF-α. Three different control samples (low, medium, and high range of the assay) were prepared, which contained purified rfTNF-α in sample buffer at concentrations of 300, 600, and 900 ng/L, respectively.
Each well in a 96-well flat-bottom enhanced binding ELISA plate was coated with 200 ng of the purified anti-rfTNF-α antibodies (primary antibodies) in 100 μL of 0.2 M carbonate-bicarbonate buffer, pH 9.4. Plates were incubated at 37°C for 1 h, while constantly being shaken using an automated plate incubator. The wells were subsequently washed 4 times with 200 μL/well of PBS with Tween (0.05%; pH 7.2). To block the remaining binding sites, the plates were then incubated for 1 h at 37°C with 200 μL/well of a commercially available blocking solution (Superblock, Pierce, Rockford, IL, USA) while shaking continuously. The wells were then washed as described above. Standards, controls, and samples were loaded in duplicates of 100 μL each per well. One hundred microliters of sample buffer served as a negative control. Serum samples were diluted with sample buffer in a 1:2 dilution and were then loaded in duplicates of 100 μL each per well. After loading the samples, the plate was again incubated and washed as described above. Finally, 100 μL containing 100 ng biotinylated anti-rfTNF-α antibodies were added to each well. After 1 h of incubation followed by washing, 8 ng of horseradish peroxidase-labeled streptavidin in 100 μL sample buffer were added to each well and incubated for 1 h, followed by washing. For detection, 100 μL of stabilized 3, 3′, 5, 5′-tetramethylbenzidine (1-StepTM Ultra-TMB-ELISA; Pierce Chemical CO) were added to each well and incubated at room temperature for 20 min in the dark. The 3, 3′, 5, 5′-tetramethylbenzidine-reaction was stopped by adding 100 μL/well of a stopping solution (4 M acetic acid, 0.5 N sulfuric acid). Finally, the absorbance was measured at a wavelength of 450 nm using an automated plate reader (EMax, Molecular Devices, Sunnyvale, CA, USA). A 5-parameter logistic curve fit, using the following mathematical equation, was used to calculate the standard curve:
The value for each study sample was calculated by extrapolating the mean absorbance of each sample duplicate from the calculated standard curve.
Analytical validation of the ELISA
The detection limit of the assay was defined as the concentration of fTNF-α calculated from the absorbance equal to the mean absorbance of 10 duplicates of the negative control plus 3 times its standard deviation (SD). For dilutional parallelism, 4 different feline serum samples were serially diluted 1:2, 1:4, 1:8, and 1:16. Observed-to-expected ratios (O/E) were calculated. Spiking recovery was determined by adding 19.5, 39.1, 78.1, 156.3, 312.5, and 625.0 ng/L of rfTNF-α each to aliquots of 4 different serum samples with a known concentration of fTNF-α and O/E ratios were calculated. Intra-assay variability was determined by the measurement of 4 different serum samples run 10 times in duplicate within the same assay run. Subsequently, the mean and SD were calculated; the coefficient of variation (CV) was calculated as follows: CV = (SD/mean)*100. Inter-assay variability was determined by the measurement of 4 different serum samples in 10 separate assay runs performed on different days. Subsequently, the duplicate means, SD, and CV for each sample were calculated as detailed above.
For validation of the assay, target acceptance criteria for the O/E for dilutional parallelism (linearity) and spiking recovery (accuracy) were set between 80% and 120%. The maximum accepted CV for intra-assay variability (precision) and inter-assay variability (reproducibility) were set at 15%. While there is no consensus about the validity of these criteria for assay validation in the scientific community, other immunoassays of proven clinical relevance were validated within these target criteria.[8, 9]
Determination of a control range for fTNF-α ELISA
The control range was calculated by evaluation of serum samples from 20 healthy pet cats using the ELISA described here. Each cat was included on the basis of a normal physical examination, and regular vaccination and parasite control. The control range for serum fTNF-α concentration was calculated using the central 95th percentile. This part of the study protocol had been reviewed and approved by the Clinical Research Review Committee at Texas A&M University. All owners whose animals participated in the study signed a consent form.
Thirty-nine left-over serum samples from the Gastrointestinal Laboratory of the Department of Small Animal Clinical Sciences, College of Veterinary Medicine and Biomedical Sciences, Texas A&M University that had been submitted for gastrointestinal function testing were used. Cats considered for inclusion into the study were identified on the basis of a decreased serum concentration of cobalamin (reference interval 290–1499 ng/L), suggesting distal small intestinal disease or exocrine pancreatic insufficiency (EPI). EPI was ruled out by a serum concentration of feline trypsin-like immunoreactivity > 12 μg/L (lower limit of the reference interval). The referring veterinarians who had submitted the samples were contacted and asked to fill out a questionnaire documenting the clinical history for each cat. All 39 cats enrolled in the study showed clinical signs compatible with chronic enteropathy, such as diarrhea, vomiting, and/or weight loss, but some also showed anorexia, abdominal pain, fecal incontinence, and/or depression.
Data were analyzed using a statistical software package (Prism 5, GraphPad Software, Inc., La Jolla, CA, USA). All data sets were tested for normal distribution using the Shapiro–Wilk test. Data sets that were not normally distributed were compared using the Mann–Whitney and the Kruskal–Wallis tests (followed by the Dunn's multiple comparison test). For comparison of proportions, contingency tables were constructed and the Fisher's exact test was used. Statistical significance for all tests was set at P < .05.
Development and validation of the fTNF-α ELISA
Following the purification protocol described above, the final product contained rfTNF-α with a purity of > 95.5% that consisted of 2 proteins with a slightly different molecular mass of 16.9 and 17.6 kDa, respectively (Figure 1). The N-terminal amino acid sequence of 11 amino acids (SSSRTPSDKPV and RTPSDKPVAHV; one-letter code) showed that both purified proteins represented fTNF-α based on a comparison with the published protein sequence available through the NCBI database. The isoelectric point of rfTNF-α was estimated by polyacrylamide gel electrophoresis at 5.3.
Antibody yield in 2 New Zealand White rabbits was 1.63 mg of anti- rfTNF-α IgG/mL of rabbit antiserum. A total of 12.3 mg anti-rfTNF-α IgG was purified, and 2 mg of the pure antibodies were biotinylated for use as a secondary antibody. The biotinylation coefficient was estimated at 6.4 mmol biotin/mmol protein.
The ELISA protocol produced reproducible standard curves. The detection limit of the assay was 19.2 ng/L. As serum samples were diluted 1:2 prior to running the assay, this translated into an effective detection limit in cat serum of 38.4 ng/L.
Observed-to-expected ratios for dilutional parallelism ranged from 75.1% to 111.9% with a mean (± SD) of 98.9% (± 11.4%; Table 1; Figure 2). Observed-to-expected ratios for spiking recovery ranged from 91.3% to 129.7% with a mean (± SD) of 102.3% (± 7.9%; Table 2). Intra-assay variability for 4 different serum samples ranged from 3.9% to 7.6% with a mean (± SD) of 5.0% (± 1.8%; Table 3). Inter-assay variability for 4 different serum samples ranged from 7.8% to 12.5% with a mean (± SD) of 9.5% (± 2.2%; Table 4). Serum fTNF-α was undetectable in 16 of 20 healthy cats, and the control range was determined as < 38.4–223.5 ng/L.
|Sample||Dilution||Observed (ng/L)||Expected (ng/L)||O/E (%)|
|Sample||Spiking conc.||Observed (ng/L)||Expected (ng/L)||O/E (%)|
|Sample||Mean (ng/L)||SD (ng/L)||%CV|
|Sample||Mean (ng/L)||SD (ng/L)||%CV|
Cats with chronic enteropathy
Serum fTNF-α concentrations were measured in 39 cats with chronic enteropathy and the majority of samples from these cats had undetectable serum fTNF-α concentrations with a median of < 38.4 ng/L. Using the Mann–Whitney test, no statistically significant difference between the group of 39 diseased cats and the group of 20 healthy cats could be identified (P = .2181).
Based on their clinical presentation, the 39 diseased cats were further divided into cats with (n = 16) or without a history of diarrhea (n = 23). The median serum fTNF-α concentration in nondiarrheic cats was < 38.4 ng/L (range < 38.4–254.0 ng/L), while it was 134.0 ng/L (range < 38.4–2448.5 ng/L) in diarrheic cats. Using a Kruskal–Wallis test, there was a statistically significant difference between the group of healthy cats and the 2 subgroups of cats with chronic enteropathy (P = .0007). The Dunn's multiple comparison test showed a significant difference between diarrheic and healthy cats (P < .05), and also between diarrheic and nondiarrheic cats (P < .05; Figure 3). However, there was no significant difference in serum fTNF-α concentrations between nondiarrheic and healthy control cats.
To compare the proportion of animals that had a serum fTNF-α concentration above the upper limit of the control range between diarrheic and nondiarrheic cats, a contingency table was constructed. One of 23 (4.3%) nondiarrheic cats and 6 of 16 (37.5%) diarrheic cats had concentrations above the upper limit of the control range. Diarrheic cats were significantly more likely to have increased serum fTNF-α concentrations compared with nondiarrheic ones (odds ratio 13.2; 95% CI 1.4–124.7; P = .0127).
The expression and purification of rfTNF-α from modified E coli have previously been described. However, the production and purification of rfTNF-α could not be reproduced in our laboratory using the published protocol. Using a modified protocol resulted in a high yield of sufficiently pure rfTNF-α (> 95.5%) for antibody production.
The SELDI-TOF-MS method identified 2 distinct protein variants of rfTNF-α with a relative molecular mass of 16.9 kDa and 17.6 kDa, respectively, which is in agreement with the estimated relative molecular mass of rfTNF-α at 17.9 kDa published earlier. N-terminal amino acid sequencing revealed a slight variation of the N-terminal amino acid sequence of the 2 protein variants, consisting of 3 missing serine residues (about 0.32 kDa) explaining in part the size difference of approximately 0.7 kDa. However, we suspect additional differences, such as additional amino acid residues in the unsequenced area of the protein, or differences in glycolysation. Alternatively, these differences could have been due to the thrombin-mediated cleavage of the N-terminus of the protein upstream of the arginine group during recovery of the recombinant protein from the fusion protein. Additional experiments would be needed to confirm or reject this hypothesis. However, the small portion sequenced (11 of 157 amino acids) was 100% homologous with the amino acid sequence of fTNF-α as published in the NCBI database.
A total of 12.3 mg of anti-rfTNF-α antibodies were purified from the serum of an immunized rabbit and an ELISA for the measurement of fTNF-α in serum samples from cats was successfully developed and analytically validated. The O/E ratios between 75.1% and 111.9% ratios in the dilutional parallelism experiment were considered acceptable despite the value of 75.1% being outside the target range, as the low concentration of fTNF-α was close to the detection limit of the assay. The spiking recovery demonstrated that the ELISA has an acceptable accuracy despite the fact that a single value (129.7%) was slightly outside the target range. This value was considered an outlier as all other O/E ratios for this sample ranged between 90% and 110%, well within the acceptable ranges for spiking recovery. The results for both intra- and inter-assay variability were also considered acceptable.
In 20 healthy pet cats, the central 95th percentile of serum fTNF-α concentrations was < 38.4–223.5 ng/L and actually measurable serum fTNF-α was only detectable in 4 cats. Likewise, the range of serum fTNF-α concentrations in a group of 39 cats with clinical evidence of chronic enteropathies was statistically not different from the healthy cats. However, when comparing cats with and without diarrhea, serum TNF-α concentrations were significantly higher in cats with diarrhea than in cats without diarrhea or in healthy cats.
Only 3 of 23 nondiarrheic cats had detectable serum fTNF-α concentrations and only one of these cats had a serum fTNF-α concentration > 223.5 ng/L. In contrast, 10 of the 16 diarrheic cats had detectable serum fTNF-α concentrations, 6 of which were > 223.5 ng/L. The difference in the proportion of cats with concentrations of fTNF-α > 223.5 ng/L was statistically significant.
Because the samples of cats with chronic enteropathy were leftover serum samples, not a lot of clinical data or histopathology data were available. Therefore, due to the retrospective design of this part of the study, specific differences in clinical and/or histopathologic findings between cats that showed increased serum TNF-α concentrations and those that did not could not be further assessed. While it is not clear why cats with diarrhea had increased serum concentrations of fTNF-α as opposed to cats without diarrhea, a plausible hypothesis is that cats with diarrhea suffer a more advanced pathology.
The number of diseases and conditions that affect fTNF-α expression in cats is still poorly documented. Although clinical studies are limited, serum fTNF-α may be increased in cats with a variety of inflammatory conditions, including infectious diseases (eg, feline infectious peritonitis, feline immunodeficiency virus infection), congestive heart failure, and possibly, other conditions.[11-14] Although some of these conditions were not likely in our study population, other disease entities were not conclusively excluded as causes of increased serum fTNF-α concentrations in our study population; therefore, the results of the clinical part of this study should be evaluated with caution. Prospective studies will be required to collect more complete data and better evaluate the usefulness of the measurement of serum TNF-α concentrations for the diagnosis and management of cats with chronic enteropathies.
In this study, an ELISA for the measurement of feline TNF-α in serum of cats was successfully developed and analytically validated, based on the prokaryotic expression of recombinant feline TNF-α. Serum concentrations of feline TNF-α were significantly higher in cats with chronic enteropathies and diarrhea than in cats with chronic enteropathies without diarrhea or in healthy control cats.
The authors thank Dr. Espen Rimstad at the Department of Pharmacology, Microbiology and Food Hygiene of the Norwegian School of Veterinary Science, Oslo, Norway for providing the E coli strain XL1-blue transfected with feline TNF-α, and Dr. Luc Berghman at Texas A&M University for facilitating the transfer of this cell line.
The authors have indicated that they have no affiliations or financial involvement with any organization or entity with a financial interest in, or in financial competition with, the subject matter or materials discussed in this article.