In this study, interferon-γ (IFN-γ) responses in whole blood cultures stimulated with tuberculins from different sources were compared with regard to their diagnostic reliability in cattle experimentally and naturally infected with Mycobacterium bovis. The IFN-γ responses to different concentrations of purified protein derivatives (PPDs) from M bovis and Mycobacterium avium were quantified. Significant differences (P<0.05) between sources and concentrations of PPDs used for stimulation were detected, indicating a need for standardisation of PPDs used in the IFN-γ assay. Additionally, a tool named ‘relative potency 30‘ that allows rapid comparison of batches and sources of PPDs was defined.
Statistics from Altmetric.com
BOVINE tuberculosis (TB) caused by Mycobacterium bovis continues to be a problem, even in countries with active control measures. There are many reasons for the failure to eradicate the disease. Limitations in the sensitivity and specificity of diagnostic tests contribute to the persistence of bovine TB. The interpretation and accuracy of test data are confounded by variations in the frequency of tuberculin testing and by the use of different tuberculin preparations (Gormley and others 2006).
The control of bovine TB has been based mainly on a test-and-slaughter approach and/or abattoir surveillance. The BOVIGAM (Prionics) interferon-γ (IFN-γ) assay (Wood and others 1990) is being incorporated into bovine TB eradication programmes in many countries (Vordermeier and others 2006). Compared with the tuberculin skin test, the IFN-γ assay has been shown to offer increased sensitivity, the possibility of more rapid repeat testing, the absence of any requirement for a second visit to the farm, and more objectivity as regards test procedures and interpretation. These advantages have been recognised by regulatory authorities, veterinarians and farmers (De la Rua-Domenech and others 2006, Vordermeier and others 2006). In recent follow-up studies of cattle that tested negative in skin tests for bovine TB, it was shown that animals that tested positive in the IFN-γ assay were more likely to have confirmed bovine TB than those that tested negative in the IFN-γ assay. This finding demonstrates the applicability of the assay as a confirmatory test for bovine TB (Coad and others 2008). In many international studies, the sensitivity of the IFN-γ assay has been shown to range from 80.9 to 100 per cent, and the specificity from 87.7 to 99.2 per cent (Vordermeier and others 2006). Apart from differences in the immune status and potentially confounding mycobacterial background infections or sensitisation of the animals, variations in the accuracy of the test may be partly related to differences in the technical parameters, such as the source and concentration of tuberculin and the interpretation criteria for positive test results. Variations in the performance of tuberculin may be attributable to differences in production methods between sources, or between different batches from a given manufacturer (Bakker and others 2005, Good and others 2008). In repeated skin test assays in guinea pigs and cattle, the potencies of tuberculin combinations including purified protein derivatives (PPDs) from M bovis and Mycobacterium avium, referred to as PPD-B and PPD-A, respectively, varied widely, thereby demonstrating the questionable precision of in vivo tests for determining PPD potency (Good and others 2008).
It was previously assumed that the source of tuberculin used for stimulation in the IFN-γ assay would not affect the diagnostic outcome. The studies of Whipple and others (2001) using cattle sensitised with heat-killed M bovis showed similar performances of PPDs prepared in the USA and in Australia. However, when testing bovine TB-free herds in Italy with the IFN-γ assay, Cagiola and others (2004) reported differences in specificity between various tuberculins, and a lower specificity of the comparative cervical skin test compared with the IFN-γ assay. Considering the lack of global standardisation of PPD sources and concentrations for use in IFN-γ assays, the aims of the present study were, first, to compare the activity levels of PPDs, sourced from different suppliers, in the BOVIGAM assay; secondly, to optimise tuberculins, in terms of the source and concentration, for the in vitro IFN-γ assay; and finally, to establish a reliable tool for determining the in vitro activity levels of different batches of PPD.
Materials and methods
Five male, TB-free, Holstein-Friesian calves were housed in accordance with institutional guidelines at the National Animal Disease Center (NADC), Ames, Iowa, USA, in a biosafety level 3 facility. All animal care and use procedures were reviewed and approved by the NADC Animal Care and Use Committee. All the calves received M bovis strain 95-1315 by aerosol administration at the age of six months, and infection was confirmed at postmortem examination at the age of approximately one year using standard procedures as described previously by Waters and others (2003). Blood samples were taken repeatedly during the first month after inoculation.
Ten cattle that had been naturally infected with M bovis were obtained from herds with a history of bovine TB (as determined by Animal Health in the UK) and positive single intradermal comparative cervical tuberculin test results, and were housed in the animal unit at the Veterinary Laboratories Agency (VLA) – Weybridge, UK. Animal experiments at the VLA were undertaken under a licence granted by the UK Home Office after approval by the local ethical review committee. The animals ranged in age from five months to 2.5 years and were of the Holstein-Friesian breed. TB was confirmed in all 10 animals by postmortem analysis including culture.
Tuberculin preparations from five manufacturers (sources 1 to 5) were used to stimulate in vitro IFN-γ responses in naturally and experimentally infected cattle. Samples from animals naturally infected with M bovis were additionally tested with a second batch of one tuberculin (source 6). All tuberculins were used at eight serial dilutions. Depending on their original concentration, twofold and then (at lower concentrations) fivefold dilutions were prepared (sources 1, 5, 6: 20, 10, 5, 2.5, 1.25, 0.25, 0.05, 0.01 μg/ml; source 2: 15, 7.5, 3.75, 1.875, 0.938, 0.188, 0.038, 0.008 μg/ml; sources 3, 4: 10, 5, 2.5, 1.25, 0.625, 0.125, 0.025, 0.005 μg/ml). The experimentally infected cattle were tested with three serial dilutions (10-fold: 10, 1, 0.1 μg/ml; source 5 was additionally used at 20 μg/ml). The manufacturers' suggested shelf-life for the tuberculins was generally 24 months; one source alone suggested a shelf-life of 36 months for its PPD-A. At the time of the study all the PPDs were within the official shelf-life of 24 months. In a separate study, seven additional batches from source 1 were tested, all of which were beyond their suggested expiry date (these batches were between 28 and 66 months old at the time of this separate study). The suppliers were Prionics, the Tuberculin Production Unit at the VLA – Weybridge, AsureQuality, Istituto Zooprofilatico Perugia, and Lelystad Biologicals.
Pokeweed mitogen (PWM; Sigma) and staphylococcal enterotoxin B from Staphylococcus aureus (SEB; Sigma) were included as positive controls at 5 μg/ml (PWM) and at 1 μg/ml (SEB), to measure the viability of the IFN-γ-producing cells in the blood samples.
Whole blood cultures were performed in 96-well plates by mixing 0.25 ml heparinised blood with 25 μl of antigen-containing solution. The supernatants were harvested after 24 hours of culture at 37°C in air plus 5 per cent carbon dioxide. IFN-γ concentrations were determined using the BOVIGAM ELISA kit. Optical density was determined at 450 nm (OD450). A result was considered positive if the PPD-B OD450 minus PPD-A OD450 was at least 0.1, and the PPD-B OD450 minus the unstimulated OD450 was at least 0.1. Only samples with an unstimulated OD450 of less than 0.2 and PWM/SEB-stimulated OD450 of more than 0.5 were considered valid for analysis. For recombinant antigens, a stimulated OD450 minus unstimulated OD450 of at least 0.1 was considered positive.
Statistical analysis was performed using GraphPad Instat software version 3 (GraphPad). Differences in potency between tuberculins from different suppliers were assessed by parametric analysis of variance and unpaired two-tailed t tests, as the data were normally distributed. Associations between the age (shelf time) and potency of the tuberculins were assessed using non-parametric Spearman correlation. The significance level was set at P<0.05.
Analysis of PPDs in experimentally infected animals
Tuberculins (PPD-B and PPD-A) produced by different manufacturers were used at different concentrations for the stimulation of whole blood taken from five experimentally infected cattle at days 0, 9, 14 and 29 after infection. The interpretation of the IFN-γ assay was determined according to standard criteria as described above. The results are summarised in Table 1. None of the animals tested positive at day 0 (not shown) or at nine days postinoculation (dpi). At 14 dpi, all of the animals showed high production of IFN-γ after PPD stimulation (as shown for source 2 in Fig 1). Despite this substantial IFN-γ production, most of the animals were diagnosed as (false) negative because PPD-A induced higher levels of IFN-γ than PPD-B did, and therefore the standard interpretation was negative. The effect of PPD concentration was most dramatic with the PPD from source 3: at 14 dpi all the animals were false negative with 10 μg/ml while three of five animals were correctly diagnosed at 1 μg/ml (Table 1). The IFN-γ production apparently levelled off when PPD-B was used at 1 and 10 μg/ml but increased when PPD-A was used at the higher concentration (results not shown).
With increasing time after infection, this effect decreased. At 29 dpi, the responses to PPD-B generally exceeded responses to PPD-A; however, exceptions to this were detected that appeared to relate to the source and concentration of the PPD being used for the assay. With PPDs from source 2, all the animals were correctly diagnosed as bovine TB reactors at each concentration evaluated. PPDs from sources 1, 3 and 4 gave false negative diagnoses at 10 μg/ml but the diagnoses were correct when 1 μg/ml was used, suggesting that the strong response to PPD-A is more concentration-dependent than the response to PPD-B. With PPDs from source 5, the test accuracy was dramatically diminished at a concentration of 20 μg/ml, again because the response to PPD-A exceeded that to PPD-B.
Analysis of PPDs in naturally infected animals
In order to extend these findings beyond experimentally infected animals, the various PPDs were assessed in 10 skin-test positive naturally infected cattle. The diagnoses (according to the standard criteria for interpretation of the IFN-γ assay) relative to PPD source and concentration are shown in Table 2. Responses after stimulation with PPDs from sources 1, 2 and 4 were comparable and strongest, and demonstrated a high sensitivity across a range of concentrations. Fewer infected animals were detected with PPDs from sources 5 and 6 as compared with those from sources 1, 2 and 4 (Table 2). Even at a dilution of 10-fold or more, the PPDs from sources 1, 2 and 4 did not lose diagnostic sensitivity, whereas PPDs from the other sources yielded correct diagnoses only at the highest concentration. These differences were evident when comparing the minimum concentrations necessary to detect at least 90 per cent of the animals as being infected, or, in other words, the concentration below which the test sensitivity wanes (Table 2).
Looking at average IFN-γ production at the various PPD-B concentrations (Fig 2), PPDs with a high diagnostic sensitivity were more efficient at inducing IFN-γ production even at low PPD-B concentrations (sources 1, 2 and 4), thereby suggesting that some of the variation in the diagnoses achieved may be related to the potencies of the different PPDs.
Quantitation of PPD activity
A potency index (named relative potency 30 [RP30]) was developed for further quantitative assessment of the observed differences in activity. The RP30 is defined as the protein concentration (μg/ml) or as activity (iu/ml) of a given PPD needed to obtain 30 per cent of the response (RP30) of the peak value of a reference PPD (ODmax) (Fig 3). Data from all 10 naturally infected animals were used to calculate individual RP30 values. The mean RP30 values are shown in Table 3, in terms of iu and also as compared on the basis of protein concentration. On the basis of protein concentration, the RP30 values of the PPDs ranged from 0.12 to 2.02 μg/ml (lower values indicate higher activity). Using the activity levels expressed in iu, as provided by the manufacturers, similar differences between the most active PPDs (sources 1 and 2) and the least active PPD (source 3) were evident. Differences between PPDs with a low RP30 (sources 1 and 2), medium RP30 (source 4) and high RP30 (sources 3, 5 and 6) were are also statistically significant (Fig 3) (t test, P<0.05). PPDs from sources 1, 2 and 4 also gave the best diagnostic results in naturally infected animals (Table 2), suggesting that the RP30 as defined here may be a useful indicator for standardisation and improvement of the IFN-γ assay. RP30 values had a direct relationship with the potency of the respective tuberculin for detecting infected animals. The RP30 values correlated positively with the minimum test concentration (Fig 4> (Spearman r=0.8971, P<0.05) before the test sensitivity waned (see Table 2), a value the authors termed 'minimum diagnostic concentration'.
Shelf-life of PPDs for the IFN-γ assay
Seven different batches of PPDs from source 1 that had been kept on the shelf for various time periods had passed their expiry dates by four to 42 months. The IFN-γ responses elicited by these batches were compared with the responses elicited by a valid batch in whole blood from 10 cattle naturally infected with M bovis. The RP30 values were determined for the different preparations and the values were analysed against the ages of the PPD-B batches (Fig 5a). As expected, there was a significant correlation (Spearman r=0.9341, P<0.005) between loss of activity (as defined by increased RP30 values) and increasing age of the PPD batch. Despite this up to fourfold reduction in RP30 over time, the effect was most evident at low PPD concentrations and could be overcome at higher concentrations (>5 μg/ml; data not shown). Age-related decreases in potency were also evident with PPD-A batches from source 1, especially at <1 μg/ml (data not shown). When PPD-A and PPD-B batches were used at the standard test concentration normally employed in the IFN-γ test (10 μg/ml), all batches were able to correctly classify the reactor animals as test-positive (Fig 5).
Comparison of PPD activity levels using skin test and IFN-γ test
The potency of PPD used in the skin test is estimated by in vivo tests with guinea pigs and/or cattle. To assess whether the activity levels estimated by in vivo models correlate with in vitro activity levels, potency data obtained by both approaches were compared (Table 3). The different PPD-Bs tested were in a range of 25,000 to 50,000 iu/mg (in vivo potency range). PPD-Bs from sources 1 to 3 had similar in vivo potencies (from 25,000 to 35,000 iu/mg), whereas PPD-B from source 4 was significantly more potent (50,000 iu/mg). In vitro, however, sources 1 and 2 were the most active ones (RP30 values of 0.12 and 0.14 μg/ml, respectively), followed by source 4 (RP30 0.26 μg/ml). In contrast, the PPD-B from source 3 was substantially less active in vitro (RP30 2.02 μg/ml).
Diagnostic accuracy is an inherent aspect of test-and-slaughter policies, and diagnostic tests must satisfy national and local requirements for both sensitivity and specificity. These are important factors for disease identification and economic considerations within a bovine TB control programme.
Accurate testing procedures, therefore, may considerably impact the outcome of a TB eradication or control programme. In this study, one of the crucial components was analysed using the IFN-γ assay, that is, PPDs, by introducing a novel tool, RP30, which determines the protein concentration at which a specific PPD preparation has 30 per cent of maximal activity. Analysing the RP30 values of tuberculins from different sources revealed differences of up to 17-fold between the most active and least active preparations. Further, it was possible to detect differences in potency in batches of different ages, even when they were from the same supplier. Most of the differences were more evident at lower protein concentrations, while at high protein concentrations the differences tended to disappear, probably because of saturation of the system. Nevertheless, some differences in sensitivity were also shown in samples from naturally infected animals (Table 2), indicating that the use of suboptimal PPDs may result in a loss of sensitivity, which is the most crucial parameter when testing for the purpose of disease eradication. Similarly, it is important to balance the specificity, because this may determine the economic feasibility of an eradication programme. When the same PPDs were tested in uninfected cattle from Switzerland and the UK, the authors found that the PPDs with the highest potency in TB-infected cattle (indicated by the lowest RP30 values) turned out to be the most specific PPDs in uninfected cattle, due to low cross-reactivity of bovine tuberculin and/or sufficiently high activity of avian tuberculin to detect non-specific IFN-γ production in response to non-tuberculous mycobacteria (data not shown). Careful balancing and standardisation of the assay will help to enhance both its sensitivity and specificity.
The differences that were found between PPDs cannot be explained easily. All PPDs had been produced according to World Organisation for Animal Health (OIE) standards and they had a phenol content of <0.5 per cent (w/v). International guidelines such as those of the OIE (Anon 2008) and European Pharmacopoeia (Anon 2002) regulate key steps of PPD production, giving a basis for standardised tuberculin production. However, some variations in production processes such as culture conditions (media, culture time) and the protein precipitation and filtration steps, may explain the differences in the antigenic profiles and variations in concentration, and may lead to qualitative differences between PPDs (Seibert and DuFour 1940, Landi and McClure 1969, Haagsma and others 1982, Tameni and others 1998). Standardised and reproducible tests are therefore needed in order to monitor the quality of PPD batches reliably. The authors suggest that the RP30 index may be used as a quality control measure for PPDs.
Furthermore, the RP30 offers a technique to optimise and balance the activity of tuberculins for the IFN-γ assay. Currently, the activity of a PPD is measured in iu using an in vivo potency test (which is known be quite variable) in guinea pigs (Good and others 2008). Bovine PPD from source 1 used in the present study failed the guinea pig potency test repeatedly, but was highly active in the cattle potency test (data not shown). In recognition of this discrepancy, a wide potency range is acceptable for skin test tuberculins (Anon 2004). The present study shows that optimising tuberculins and balancing them on RP30 may improve the IFN-γ assay. In this context, the authors emphasise the importance of the method used to determine the protein content during PPD production, given that different assays provide different estimates of protein content. The values estimated by the techniques of Lowry and Kjedahl (the latter technique is the methodology most commonly used by PPD manufacturers and acknowledged to be the most precise) may differ by as much as 25 per cent (Sapan and others 1999; C. Carter, personal observation).
Another variable in accurate diagnosis may be the time point at which the diagnostic test is performed after infection. As demonstrated in the experimentally infected animals in the present study, the responses to avian PPD may exceed those to bovine PPD at an early time point after infection. This effect has also been reported by Rhodes and others (2000). It is not totally clear whether the experimental infection truly mimics natural infection; in particular, the use of high doses of infectious agent in experimental infections might logically be expected have an effect on response times. Nevertheless, the dose-response curve of PPDs was found to be similar in naturally and experimentally infected animals (data not shown), thereby suggesting that the method of infection is not one of the main parameters.
Comparison of the sensitivity of the different PPDs at different concentrations in the 10 naturally infected animals showed that the most active PPDs work over a very broad range in the IFN-γ test. Such PPDs are therefore quite robust against small variations in protein concentration, while PPDs at the limit of their activity spectrum do not have much flexibility. The sensitivity values shown in Table 2 also indicate a tendency for PPDs with good RP30 values to be more sensitive. The authors would therefore predict that the RP30 indicates the performance of a PPD in the field; however, this assumption needs further evaluation.
In conclusion, not all PPDs are equal, but the selection of active PPDs at the appropriate concentration will result in accurate and reliable diagnosis. It will therefore be important to standardise PPDs in order to tailor the diagnostic system to the needs of an eradication or control programme.
The authors thank Jessica Pollock, Rachel Huegel, Bart Olthof and Mike Howard for their excellent technical support. They would also like to express appreciation to the staff of the Animal Service Unit at the VLA, the NADC, and the ETH Forschungsstation Chamau. Part of this study was funded by the UK Department for the Environment, Food and Rural Affairs.
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.