Toxicogenomic markers for corticosteroid treatment in beef cattle: Integrated analysis of transcriptomic data

https://doi.org/10.1016/j.fct.2014.12.001Get rights and content

Highlights

  • This work reports an integrated analysis of DNA-microarray data from 146 bovine muscle samples (novel and previous data).

  • A robust gene expression signature for corticosteroid treatment in beef cattle was defined.

  • A Support Vector Machines classifier built on 73 biomarkers correctly classifies all the samples in the training set.

  • The predicted accuracy on new samples is 0.77 and false positive and false negative percentages are 5% and 6%, respectively.

  • A small set of genes can discriminate between controls and corticosteroid-treated animals, despite biological variation.

Abstract

In the present work, an integrated analysis was performed on DNA-microarray data of bovine muscle samples belonging to controls, animals treated with various growth promoters (GPs) and unknown commercial samples. The aim was identify a robust gene expression signature of corticosteroid treatment for the classification of commercial samples, despite the effects of biological variation and other confounding factors.

DNA-Microarray data from 5 different batches of bovine skeletal muscle samples were analyzed (146 samples). After preprocessing, expression data from animals treated with corticosteroids and controls from the different batches (89 samples) were used to train a Support Vector Machines (SVMs) classifier. The optimal number of gene probes chosen by our classification framework was 73. The SVMs with linear kernel built on these 73 biomarker genes was predicted to perform on novel samples with a high classification accuracy (Matthew's correlation coefficient equal to 0.77) and an average percentage of false positive and false negative equal to 5% and 6%, respectively.

Concluding, a relatively small set of genes was able to discriminate between controls and corticosteroid-treated animals, despite different breeds, animal ages, and combination of GPs. The results are extremely promising, suggesting that integrated analysis provides robust transcriptomic signatures for GP abuse.

Introduction

The European Union has long established a strong policy against the use of growth promoters (GPs) in animal husbandry, to guarantee food safety, animal welfare, and consumer health. The use of hormonal compounds has been banned since 1988 and more recently directive 96/22/EC maintained the prohibition on such substances for growth promoting purposes, extended the ban to beta-agonists; restricted the use of hormones for therapeutic or zoo-technical purposes, and reinforced controls. Directive 96/23/EC also defines standards for official tests. Controls should be performed by routine or field laboratories (RFLs) coordinated and controlled by at least one national reference laboratory (NRL), designated by the national government. Each EU member state is required to adopt a National Monitoring Plan to survey the presence of residues in live animals, their tissues and fluids, and animal products. Finally, based on the Opinion of the Scientific Committee on Veterinary Measures, 1999, Opinion of the Scientific Committee on Veterinary Measures, 2000, Opinion of the Scientific Committee on Veterinary Measures, 2002, the European Commission adopted Directive 2003/74/EC, which amends Directive 96/22/EC in relation to the permanent prohibition for the use of hormones in stock farming.

Despite controls, however, anabolic compounds are still administered to increase growth performances and feed conversion rate. With regard to the illegal use of corticosteroids in beef cattle, results of the Italian Monitoring Plan (PNR) reported 22 non-compliant cases on 3332 (0.7%) analyzed samples for PNR 2011 and 32 on 3694 (0.9%) for PNR 2012. The real impact of illegal use of GPs, however, might be underestimated because various strategies are adopted to elude official controls, such as the administration of different compounds with similar biological action each at very low dose, the use of different classes of compounds with additive or synergic effects at low doses (e.g. corticosteroids and beta-agonists); the development of new chemical species, the use of natural steroids or their precursors, which might be confounded with those of endogenous origin (Courtheyn et al., 2002).

The extremely articulate scenario described above highlights the need to support the existing methods with new analytical approaches to identify putative risks for the consumer's health but also to increase the knowledge of metabolic patterns and associated kinetics of elimination for GPs.

For live animals, animal products and their feed, the Council Directive 96/23/EC still considers only the parental compound or its metabolite as proof for the administration of illegal compounds detected by confirmative methods based on mass spectrometry. Some steps forward have been done for anti-doping enforcement in humans and horses since the presence of markers or scientific indicators is considered a sufficient and officially accepted proof for doping, according to the WADA's Executive Committee, (WADA's Athlete Biological Passport Operating Guidelines, ver- 4.0 2013) to the International Federation of Horseracing Authorities (International Agreement on Breeding, Racing and Wagering, April 2014), respectively.

With respect to indirect biomarkers, histological techniques and analysis of blood chemistry parameters have long been proposed and in some cases officially adopted as alternative screening methods in prevention of GP abuse. The advent of “omics” technologies (genomics, transcriptomics, proteomics, metabolomics) now offers interesting perspectives to discover reliable and quantifiable markers, which however still await for full implementation in routine analysis. Transcriptome analysis using DNA-microarrays has been recently explored as a screening method to reveal the biological effects of different anabolic compounds and thus support the existing tools to combat the use of these substances in beef cattle (Cannizzo et al, 2013, Carraro et al, 2009, De Jager et al, 2011, Pegolo et al, 2012, Rijk et al, 2010).

In the detection of anabolic treatments, some advantages of gene expression profiling with respect to the existing conventional methods have already been discussed (Pegolo et al., 2012). For instance, a trascriptomic approach allowed to identify the effects of GPs administration even when the parental compound or its metabolites were no longer detectable in biological fluids by LC-MS/MS methods or in absence of histological alterations (Cannizzo et al., 2013).

Despite many advantages, gene expression biomarkers proposed to detect animals illegally treated (e.g. Divari et al, 2011, Giantin et al, 2010, Riedmaier et al, 2011) might be influenced by intrinsic or extrinsic factors (e.g. age, breed, diet). Besides, identification of potential biomarkers is carried out in controlled experiments, which often use few animals and consider a small number of factors/variables. As a result, high throughput analyses carried out in different laboratories often give different biomarker lists and different features may be selected under different settings. This is a general problem that has been encountered in the application of gene expression markers in other fields, where possible solutions have been proposed (Abeel et al, 2010, Di Camillo et al, 2012, Sanavia et al, 2012). In particular, it has been suggested to increase the number of cases through meta-analysis or integrated analysis of results from different studies, while at the same time implementing rigorous statistical methods to account for data heterogeneity.

In the present work, an integrated analysis was performed on DNA-microarray data of bovine muscle samples belonging to controls, animals treated with various combination of GPs and also unknown commercial samples; in particular, a Support Vector Machines (SVMs) classifier was trained on data from animals treated with corticosteroids and controls (89 samples). Gene expression profiles for the majority of samples were already presented in previous studies. However, microarray data for two additional groups of samples were produced specifically for the present paper. The first group included younger animals (treated and controls) while the second group consisted of muscle samples preserved under vacuum at 4 °C for 14 days, to mimic meat storage under commercial conditions and to test RNA stability. In fact, the ultimate goal of this study is to explore the feasibility of using gene expression markers as a routine analytical tool on commercial samples, through the identification of a gene expression signature for corticosteroid treatment that is robust despite the effects of biological variation and other confounding factors, but also assessing the possibility to work along the whole supply chain by analyzing samples long after slaughtering.

Section snippets

Sample batches

DNA-Microarray data from 5 different batches of bovine skeletal muscle samples were analyzed for a total of 146 samples. Of these samples, 51 correspond to animals treated with corticosteroids, 32 were treated with different combination of GPs, 38 correspond to control animals, 25 are of unknown class. The 89 samples corresponding to animals treated with corticosteroids (51 samples) and controls (32 samples) were used here to train a Support Vector Machines classifier. The others were

Results

In DNA microarray experiments, data quality is crucial. RNA integrity number (RIN) was used as a measure to standardize the interpretation of RNA quality. In the present study, a conservative threshold was enforced to reduce experimental biases due to poor RNA quality. For newly analyzed batches (batch 4 and 5), only RNA samples with RIN numbers ≥7 were further processed. The same criterion had been used for all other batches as already reported in previous papers. Based on the Agilent quality

Discussion

Disparate gene expression signatures have been proposed to identify animals treated with the same class of anabolic steroids using bovine muscle samples with little agreement in the constituent genes or reduced statistical significance. This can be due to different dosages and/or route of administration, specific compound effect or combination with GPs belonging to different classes (e.g. corticosteroids combined with androgens and/or estrogens), heterogeneity of computational pipelines and

Conclusions

The present work allowed to identify a panel of robust biomarkers for corticosteroid treatment in beef cattle. If gene expression markers are to be used for internal control along the supply chain, efforts should be directed toward the development of a cost- and technically-effective test. For its use as official screening method, however, additional known samples (experimentally treated and especially controls) should be analyzed to set up a more accurate classifier, which could match the

Conflict of interest

The authors declare no competing financial interest.

Transparency document

.

Acknowledgements

This work was financially supported by the Ministero delle Politiche Agricole e Forestali under the project “SAFORISK”, granted to L. Bargelloni (D.M. 2089/09, 29th of January 2009) and by Regione del Veneto under the project “Nuovi approcci genomici e proteomici per lo screening dei trattamenti con promotori di crescita nel bovino da carne”, granted to C. Montesissa (DGR n. 2154, 25th of November 2013).

References (38)

  • CarraroL. et al.

    Expression profiling of skeletal muscle in young bulls treated with steroidal growth promoters

    Physiol. Genom

    (2009)
  • Council Directive 96/22/EC of 29 April 1996 concerning the prohibition on the use in stockfarming of certain substances...
  • Council Directive 96/23/EC of 23 May 1996 on measures to monitor certain substances and residues in live animals and...
  • De JagerN. et al.

    Chronic exposure to anabolic steroids induces the muscle expression of oxytocin and a more than fiftyfold increase in circulating oxytocin in cattle

    Physiol. Genom

    (2011)
  • Di CamilloB. et al.

    Effect of size and heterogeneity of samples on biomarker discovery: synthetic and real data assessment

    PLoS ONE

    (2012)
  • DivariS. et al.

    Corticosteroid hormone receptors and prereceptors as new biomarkers of the illegal use of glucocorticoids in meat production

    J. Agric. Food Chem

    (2011)
  • EdgarR. et al.

    Gene Expression Omnibus: NCBI gene expression and hybridization array data repository

    Nucleic Acids Res

    (2002)
  • FurlanelloC. et al.

    Entropy-based gene ranking without selection bias for the predictive classification of microarray data

    BMC Bioinformatics

    (2003)
  • FurlanelloC. et al.

    Semisupervised learning for molecular profiling

    IEEE/ACM Trans Comput. Biol. Bioinform

    (2005)
  • Cited by (8)

    • Real-time PCR assay for detecting illicit steroid administration in veal calves allows reliable biomarker profiling of formalin-fixed, paraffin-embedded (FFPE) archival tissue samples

      2020, Food Chemistry
      Citation Excerpt :

      Since 2008 a histological method to reveal microscopic lesions related to illicit steroid treatment has been applied in the frame of Italian National Residue Control Plans (NCRPs). Among other biomarker based methods, genome-wide analysis of the transcriptome has helped researchers identify multiple differentially-expressed gene (DEG) patterns (signature patterns) related to growth promoter abuse (Pegolo et al., 2015; Riedmaier et al., 2012). Moreover, the discovery of promising biomarkers has been advanced by proper statistical data treatment to limit confounding variance introduced by species variability, inter-individual variability, experimental design and/or sampling procedures (Riedmaier et al., 2014).

    View all citing articles on Scopus
    View full text