Statistics from Altmetric.com
- ABTS, 2.2`-azino-bis-3-ethylbenzthiazolin-6-sulphonic acid
- AU, arbitrary units
- AUC, area under curve
- BSA, bovine serum albumin
- CD, coeliac disease
- ELISA, enzyme linked immunosorbent assay
- gpl-tTG, guinea pig liver tissue transglutaminase
- HRP, horseradish peroxidase
- h-tTG, human tissue transglutaminase
- IBD, inflammatory bowel disease
- IgA EMA, IgA anti-endomysial antibody
- IgA tTG, IgA anti-tissue transglutaminase antibody
- IIF, indirect immunofluorescence
- PNPP, paranitrophenyl phosphate
- ROC, receiver operating characteristic
- TMB, 3,3`,5,5` tetramethylbenzidine
- tTG, tissue transglutaminase
The identification of autoantibodies strongly associated with coeliac disease (CD; also known as gluten sensitive enteropathy), in particular IgA anti-endomysial antibodies (IgA EMA), has enabled the development of non-invasive serological screening tests for this condition.1–3 The IgA EMA indirect immunofluorescence (IIF) assay has, in subjects with untreated CD, a sensitivity of 84–100% and a specificity of 94–100%, which is superior to the IgA anti-reticulin IIF assay and IgA/IgG antigliadin antibody enzyme linked immunosorbent assays (ELISAs).2
“The use of human tissue transglutaminase has been reported to be associated with fewer false negative and false positive results, and an overall performance closely comparable or equal to the “gold standard” IgA anti-endomysial antibody indirect immunofluorescence assay”
Since Dieterich et al described tissue tranglutaminase (tTG), an 82–85 kDa ubiquitous enzyme, as the major autoantigen target of IgA EMA,4 over 30 publications have appeared using this protein as the basis for an alternative assay to the IgA EMA IIF assay.5–38 Most studies used guinea pig liver tTG (gpl-tTG) in ELISA based assays,5–15,18–38 but purified erythrocyte23 and recombinant human tTG (h-tTG)13–17,24,27,29,35,38 have also been used in ELISA,13,23,24,27,29,38 radioimmunoassay,14–17,35 and dot blot27 assays. Because of its ease of use, potential for automation, objectivity in interpretation, and reduced training requirements, there is growing interest in using an ELISA based IgA anti-tTG antibody (IgA tTG) assay as an alternative to the IgA EMA IIF assay.
Although many studies have concluded that the IgA tTG assay has comparable performance to the IgA EMA IIF assay, several have described false negative IgA tTG results in subjects with IgA EMA positive untreated CD,4,10–16,19–21,25,26,28–30,32,33,34,36,38 and false positive IgA tTG results in the absence of IgA EMA and CD. 5,6,9,10,12–16,18,19,22–29,32,33,36,38 However, most of these studies used gpl-tTG, which has only about 81% homology with h-tTG.39 In contrast, the use of h-tTG has been reported to be associated with fewer false negative and false positive results, and an overall performance closely comparable or equal to the “gold standard” IgA EMA IIF assay.13–15,23,24,29,35,38,40 However, because none of these studies has compared gpl-tTG based ELISAs with two or more h-tTG-based ELISAs, it is unclear whether the use of h-tTG alone results in superior performance to the gpl-tTG-based assays.
We compared 13 commercial IgA tTG ELISA kits, seven gpl-tTG based and six h-tTG based (four recombinant h-tTG), in 49 IgA EMA positive adult patients with CD and 64 adult disease controls to establish the sensitivity and specificity of each kit, and thus determine whether the h-tTG based kits consistently outperformed the gpl-tTG based kits, and produced comparable results to the IgA EMA IIF assay.
One hundred and thirteen sera were selected from samples submitted to: Division of Immunology, Queensland Health Pathology Services, Royal Brisbane and Princess Alexandra Hospitals; Central Sydney Immunology Laboratory; and Department of Immunology, Sullivan Nicolaides Pathology. These comprised sera from the following patients who were aged 21 years or older: (1) 49 patients with typical histological changes of CD on small bowel biopsy,3,41 who had previously been found to have a positive IgA EMA, 38 of whom had never been on a gluten free diet, and 11 of whom were poorly compliant or non-compliant with the diet and had an abnormal small bowel biopsy close to the time of blood sampling; (2) 34 subjects who had been investigated with upper gastrointestinal fibreoptic endoscopy and small bowel biopsy for possible CD and were found not to have histological changes consistent with CD (non-CD controls, with the following results on small bowel biopsy (no evidence of villous atrophy in all cases): normal duodenum (n = 27), duodenal ulcer (n = 3), dilated Brunner's glands (n = 1), non-specific duodenitis (n = 1), fibrotic and thickened small bowel (n = 1), and gastric atrophy (n = 1)); and (3) 30 subjects with biopsy confirmed inflammatory bowel disease (IBD controls).
All sera were retested for IgA EMA at the start of the study to ensure that the sera from patients with CD had not degraded during storage at −70°C. Total serum IgA values were also measured in all 64 non-CD and IBD control sera by nephelometry (Behring Diagnostics, Frankfurt, Germany). All 64 controls had values within the normal range for adults (1.24–4.16 g/litre), thus excluding IgA deficiency as a potential cause for negative results.
IgA EMA IIF assay
The IgA EMA assay was performed by IIF using cryostat sections of monkey oesophagus (The Binding Site, Birmingham, UK), as described previously2 at a screening dilution of 1/4. All slides were viewed by two independent observers and a positive or negative result was determined by consensus.
IgA tTG ELISA
The manufacturer's instructions (table 1) were followed for all 13 IgA tTG ELISA kits. All specimens were tested in duplicate.
Bovine serum albumin and gelatin coated ELISA plates
To investigate the possibility of IgA anti-bovine serum albumin (BSA) antibodies producing false positive IgA tTG results, ELISA plates (Costar, Corning Inc, New York, USA) were coated with 250 μl of 5% BSA (Sigma Chemical Co, St Louis, Missouri, USA) or 1% gelatin (Bio-Rad, Hercules, California, USA). Serum diluted 1/100 in Tween/phosphate buffered saline was incubated for one hour at room temperature. After three washes, horseradish peroxidase (HRP) labelled goat antihuman IgA (Silenus Labs, Melbourne, Australia), at a dilution of 1/500, was added and the plates were incubated for one hour (room temperature). ABTS (2.2`-azino-bis-3-ethylbenzthiazolin-6-sulphonic acid) substrate (Medical Innovations, Sydney, Australia) was added for 15 minutes, and absorbances read at 405 nm.
Cut off values
Both the manufacturers' recommended cut off values and decision thresholds determined by receiver operating characteristic (ROC) plots (see below) were used to calculate the sensitivity and specificity of each assay/kit. The IBD controls were not used in the calculation of specificity because some had not undergone small bowel biopsy to exclude CD.
ROC plot analysis
ROC plot analysis was performed on each kit using the Accuroc software package (Accumetric Corporation, McGill University Health Centre, Montreal, Quebec, Canada) to determine a decision threshold and area under curve (AUC) estimation. The IBD controls were not included in the ROC analysis because some had not undergone small bowel biopsy to exclude CD. The AUC was calculated using the trapezoid rule.42,43 Comparisons between the AUCs of each kit were performed by the non-parametric method for correlated samples, as previously described by DeLong et al.44
The IgA tTG values of the patients with CD and the non-CD and IBD controls measured with the 13 kits are shown in fig 1 (gpl-tTG based kits) and fig 2 (h-tTG based kits) with corresponding ROC curves and AUC estimations. The numbers of sera from patients with CD, and the non-CD and IBD controls that were positive in each assay, using both the manufacturers' and ROC analysis derived decision thresholds, are shown in table 2 (gpl-tTG based kits) and table 3 (h-tTG based kits), with corresponding sensitivities and specificities. Table 4 shows the AUC comparisons between kits, with a significant difference denoted by a p value of < 0.05.
The recombinant h-tTG based Varelisa (Pharmacia & Upjohn Diagnostics, GmbH & Co, Freiburg, Germany) and purified erythrocyte h-tTG based QUANTA Lite (Inova Diagnostic Inc, San Diego, California, USA) kits performed best, with sensitivities of 100% and 98%, specificities of 100% and 100% (using the manufacturers' cut off values), and AUC estimations of 1.000 and 1.000, respectively (fig 2; table 3).
Of the seven gpl-tTG based kits (fig 1; table 2), the QUANTA Lite kit performed best, with 86% sensitivity and 100% specificity using the manufacturer's cut off value of 20 arbitrary units/ml, and an AUC of 0.987. Applying the ROC analysis derived decision threshold of 14.1 arbitrary units/ml improved sensitivity to 92% but reduced specificity to 97%.
To exclude the possibility that some reactions to tTG were really reactions to blocking agents used in the ELISA kits, anti-BSA and antigelatin antibodies were determined (data not shown). Sera from one IBD control and two patients with CD reacted significantly on the BSA coated ELISA plates, suggesting the presence of IgA anti-BSA antibodies. However, none of the non-CD controls reacted significantly on the BSA coated plates and no sera reacted on the gelatin coated plates.
In this comparison of 13 commercial IgA tTG ELISA kits, we found that the human tTG based kits tested generally demonstrated superior performance (especially specificity) to the gpl-tTG based kits (tables 2,3). However, the use of h-tTG alone was insufficient to confer performance equal to the IgA EMA IIF assay, because only two h-tTG based kits (recombinant h-tTG based Varelisa and purified erythrocyte h-tTG based QUANTA Lite) produced closely comparable results to the IgA EMA IIF assay. Furthermore, two of the gpl-tTG based kits (QUANTA Lite and Eurospital (Trieste, Italy)) had AUC estimations that were not significantly different from the h-tTG kits (figs 1, 2; tables 2, 3). This demonstrates that factors other than antigen source are important in determining kit performance.
Using the manufacturers' cut off values, false positive results were found in three of the six h-tTG based and six of the seven gpl-tTG based IgA tTG kits in our study (table 4). This has also been reported in other studies.5,6,9,10,12–16,18,19,22–29,32,33,36,38 However, as in our study, most of these false positive results were detected in gpl-tTG based ELISAs.5,6,9,10,12–15,18,19,22–26,28,29,32–34,36,38 These findings raise the important issue of contaminants in gpl-tTG,24,29,38 which may contain other hepatic proteins.29,38 On sodium dodecyl sulfate-polyacrylamide gel electrophoresis, the gpl-tTG extract (Sigma T5398; Sigma Chemical Co) used in several gpl-tTG based ELISAs5,6,9–11,13–15,18,20–25,29,32,33,36,38 contains multiple bands in addition to the 82–83 kDa tTG band,24,29,38 which only accounted for about 30% of the total protein.29,38 This may be partially overcome by further purification steps, and should be less of an issue with recombinant h-tTG.24,29,38
However, three of the six h-tTG based kits (two recombinant h-tTG based) evaluated also produced false positive results in the absence of IgA EMA and CD, as previously reported.13,14,16,23,24,27 Therefore, other explanations for false positive results are required.
“The methods used to extract and purify tissue derived tissue transglutaminase (tTG), produce and process recombinant tTG, and then coat tTG on to ELISA wells may lead to alterations in the tertiary structure of tTG”
More false positive IgA tTG results were detected in the non-CD controls compared with the IBD controls. A possible explanation may be the presence of IgA anti-BSA antibodies in some of the non-CD control sera, reacting with the BSA used as a blocking agent in some kits.10 However, Lock and colleagues10 did not detect significant IgA anti-BSA antibodies in two disease controls tested, and significant IgA anti-BSA antibodies were not demonstrated in our non-CD controls.
False negative results were found in six of the seven gpl based and five of the six h-tTG based kits in IgA EMA positive patients with CD (figs 1, 2; tables 2, 3), in agreement with previous reports.4,10–16,19–21,25,26,28–30,32–34,36,38 The methods used to extract and purify tissue derived tTG, produce and process recombinant tTG, and then coat tTG on to ELISA wells may lead to alterations in the tertiary structure of tTG. Therefore, conformational epitopes may be lost or formed, with a loss leading to a reduced ability of tTG to bind IgA tTG, thus explaining some of the false negative results.25 Furthermore, the formation of conformational neoepitopes may also result in false positive results (see above).
Take home messages
In general, the human tissue transglutaminase (h-tTG) based kits tested demonstrated superior performance (especially specificity) to the guinea pig liver tTG (gpl-tTG) based kits
Because this was a general and not a universal funding, factors other than antigen source are important in determining kit performance
Most of the kits performed significantly better when the cut off values/decision thresholds were adjusted via receiver operating characteristic plot analysis, which emphasises the importance of cut off point revalidation by laboratories, using appropriate samples from their referral population
The function and tertiary structure of tTG is also altered by the presence of ionised calcium.45 It has been suggested that antibody binding epitopes may be formed or hidden by the presence of ionised calcium in the coating buffer of the IgA tTG ELISA.10,46 Sulkanen and colleagues6 reported that the pretreatment of tTG with calcium (“calcium activation”) dramatically improved the separation between CD and non-CD sera in a gpl-tTG based ELISA, and also increased the binding affinity of tTG to CD sera. However, in our study, the two kits in which the use of “calcium activation” of tTG is recorded (Binding Site gpl-tTG kit and Genesis) did not clearly demonstrate superior performance to the other kits. Furthermore, Lock and colleagues10 found that the addition of calcium to the coating buffer increased both the signal and background values, and therefore produced no overall improvement in the performance of their in house gpl-tTG based IgA tTG ELISA. Nakachi and colleagues46 also reported that the autoantibody binding sites of tTG were formed in a manner that was essentially calcium independent.
Finally, we found that the performances of most of the IgA tTG ELISA kits were significantly improved by adjusting the cut off values/decision thresholds via ROC plot analysis. These discrepancies between the ROC analysis derived decision thresholds and manufacturers' recommended cut off values illustrate the importance of cut off point revalidation by laboratories, using appropriate samples from their referral population. However, the adjustment of the cut off values/decision thresholds via ROC plot analysis would not compensate for less than satisfactory kit performance. Therefore, in selecting an IgA tTG ELISA kit for diagnostic purposes, a laboratory should consider not only the source of tTG antigen, but also the performance of the kit using locally derived cut off values.
We acknowledge K Smithers (Central Sydney Immunology Laboratory, Royal Prince Alfred Hospital) for excellent technical assistance; the Australian distributors of the commercial IgA tTG kits for the generous provision of their kits for assessment in this study; and P Hobson (Department of Immunology, Sullivan Nicolaides Pathology, Brisbane, Australia) for providing some of the sera for the study and performing some of the assays.
A comparison of 13 guinea pig and human anti-tissue transglutaminase antibody ELISA kits
RCW Wong, RJ Wilson, RH Steele, G Radford-Smith, and S Adelstein
Please note that the affiliation of Dr RH Steele should be
South Western Area Pathology Service.
The error is much regretted
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.