Advertisement

Detecting Staphylococcus aureus in milk from dairy cows using sniffer dogs

Open ArchivePublished:March 01, 2018DOI:https://doi.org/10.3168/jds.2017-14100

      ABSTRACT

      Fast and accurate identification of disease-causing pathogens is essential for specific antimicrobial therapy in human and veterinary medicine. In these experiments, dogs were trained to identify Staphylococcus aureus and differentiate it from other common mastitis-causing pathogens by smell. Headspaces from agar plates, inoculated raw milk samples, or field samples collected from cows with Staphylococcus aureus and other mastitis-causing pathogens were used for training and testing. The ability to learn the specific odor of Staphylococcus aureus in milk depended on the concentration of the pathogens in the training samples. Sensitivity and specificity for identifying Staphylococcus aureus were 91.3 and 97.9%, respectively, for pathogens grown on agar plates; 83.8 and 98.0% for pathogens inoculated in raw milk; and 59.0 and 93.2% for milk samples from mastitic cows. The results of these experiments underline the potential of odor detection as a diagnostic tool for pathogen diagnosis.

      Key words

      INTRODUCTION

      Mastitis is a common and costly disease in dairy industry caused by various infectious pathogens (
      • Harmon R.J.
      Physiology of mastitis and factors affecting somatic cell counts.
      ). Treatment protocols for clinical mastitis include local or systemic use of antimicrobials (
      • Barkema H.W.
      • Schukken Y.H.
      • Zadoks R.N.
      Invited review: The role of cow, pathogen, and treatment regimen in the therapeutic success of bovine Staphylococcus aureus mastitis.
      ). Guidelines for prudent use of antimicrobials advocate an exact, preferably microbiological, diagnosis before treatment (
      • Ungemach F.R.
      • Mueller-Bahrdt D.
      • Abraham G.
      Guidelines for prudent use of antimicrobials and their implications on antimicrobials usage in veterinary medicine.
      ;
      • Teale C.J.
      • Moulin G.
      Prudent use guidelines: A review of existing veterinary guidelines.
      ). One disadvantage of treating cows without knowing the exact pathogen involved is that the treatment chosen may not be specific to the causative pathogen, resulting in discarded milk due to treatment, costs incurred due to incorrect treatment, delayed recovery of the cow, and the possibility of increased antimicrobial resistance due to overuse of an antimicrobials (
      • Cha E.
      • Smith R.L.
      • Kristensen A.R.
      • Hertl J.A.
      • Schukken Y.H.
      • Tauer L.W.
      • Welcome F.L.
      • Gröhn Y.T.
      The value of pathogen information in treating clinical mastitis.
      ).
      The most used test for bacteriological diagnosis of milk samples is culture of the sample on blood agar plates (
      • Oliver S.
      • Gonzalez R.
      • Hogan J.
      • Jayarao B.
      • Owens W.
      Microbiological procedures for the diagnosis of bovine udder infection and determination of milk quality.
      ), which takes usually 24 to 48 h. This presents a dilemma to veterinarians who have to treat clinical mastitis in a timely manner and according to the specific pathogen involved. Also, in 30% of samples from cows with clinical mastitis (
      • Taponen S.
      • Salmikivi L.
      • Simojoki H.
      • Koskinen M.T.
      • Pyorala S.
      Real-time polymerase chain reaction-based identification of bacteria in milk samples from bovine clinical mastitis with no growth in conventional culturing.
      ) and in 49.7% of samples from cows with subclinical mastitis (
      • Makovec J.A.
      • Ruegg P.L.
      Results of milk samples submitted for microbiological examination in Wisconsin from 1994 to 2001.
      ), an infectious agent was not isolated. Furthermore, 40% of positive cultures involve gram-negative species or yeasts (
      • Lago A.
      • Godden S.
      • Bey R.
      • Ruegg P.
      • Leslie K.
      The selective treatment of clinical mastitis based on on-farm culture results: I. Effects on antibiotics use, milk withholding time, and short-term clinical and bacteriological outcomes.
      ). For most coliform mastitis, antimicrobial therapy is either not necessary or not worthwhile because of the high self-cure rate (
      • Roberson J.R.
      • Warnick L.
      • Moore G.
      Mild to moderate clinical mastitis: Efficacy of intramammary amoxicillin, frequent milk-out, a combined intramammary amoxicillin, and frequent milk-out treatment versus no treatment.
      ;
      • Vasquez A.K.
      • Nydam D.V.
      • Capel M.B.
      • Eicker S.
      • Virkler P.D.
      Clinical outcome comparison of immediate blanket treatment versus a delayed pathogen-based treatment protocol for clinical mastitis in a New York dairy herd.
      ).
      Dating back to 400 BC, Hippocrates recognized the diagnostic value of odors, but the industrial and technological revolution of the 18th and 19th centuries changed the history of medical diagnosis, and odors were neglected in this period (
      • Pavlou A.K.
      • Turner A.P.F.
      Sniffing out the truth: Clinical diagnosis using the electronic nose.
      ). More recently, methods such as gas chromatography, mass spectrometry, and electronic nose devices have re-established the value of odors as a potential diagnostic tool. Various infectious agents can be identified by the presence of specific volatile organic compounds (
      • Turner A.P.F.
      • Magan N.
      Electronic noses and disease diagnostics.
      ;
      • Bruins M.
      • Bos A.
      • Petit P.L.
      • Eadie K.
      • Rog A.
      • Bos R.
      • van Ramshorst G.H.
      • van Belkum A.
      Device-independent, real-time identification of bacterial pathogens with a metal oxide-based olfactory sensor.
      ;
      • Weetjens B.J.
      • Mgode G.F.
      • Machang'u R.S.
      • Kazwala R.
      • Mfinanga G.
      • Lwilla F.
      • Cox C.
      • Jubitana M.
      • Kanyagha H.
      • Mtandu R.
      • Kahwa A.
      • Mwessongo J.
      • Makingi G.
      • Mfaume S.
      • Van Steenberge J.
      • Beyene N.W.
      • Billet M.
      • Verhagen R.
      African pouched rats for the detection of pulmonary tuberculosis in sputum samples.
      ;
      • Bomers M.K.
      • van Agtmael M.A.
      • Luik H.
      • van Veen M.C.
      • Vandenbroucke-Grauls C.M.J.E.
      • Smulders Y.M.
      Using a dog's superior olfactory sensitivity to identify Clostridium difficile in stools and patients: Proof of principle study.
      ). These technological approaches, however, are cost intensive and require special equipment and expertise.
      Animals can be trained to identify and discriminate specific odors (
      • Browne C.
      • Stafford K.
      • Fordham R.
      The use of scent-detection dogs.
      ). For rapid diagnosis of tuberculosis, African giant-pouched rats (Crietomys gambianus) were successfully trained to identify Mycobacterium tuberculosis in sputa of patients suspected of having tuberculosis, with a cumulative sensitivity of 86.5% and specificity of 89.1% (
      • Weetjens B.J.
      • Mgode G.F.
      • Machang'u R.S.
      • Kazwala R.
      • Mfinanga G.
      • Lwilla F.
      • Cox C.
      • Jubitana M.
      • Kanyagha H.
      • Mtandu R.
      • Kahwa A.
      • Mwessongo J.
      • Makingi G.
      • Mfaume S.
      • Van Steenberge J.
      • Beyene N.W.
      • Billet M.
      • Verhagen R.
      African pouched rats for the detection of pulmonary tuberculosis in sputum samples.
      ). To identify health-threatening microbial contaminations in hospitals and apartments, dogs have been trained to indicate pathogens such as Streptomyces sp. and Clostridium difficile, respectively (
      • Kauhanen E.
      • Harri M.
      • Nevalainen A.
      • Nevalainen T.
      Validity of detection of microbial growth in buildings by trained dogs.
      ;
      • Bomers M.K.
      • van Agtmael M.A.
      • Luik H.
      • van Veen M.C.
      • Vandenbroucke-Grauls C.M.J.E.
      • Smulders Y.M.
      Using a dog's superior olfactory sensitivity to identify Clostridium difficile in stools and patients: Proof of principle study.
      ).
      The objective of this study was to train dogs to identify Staphylococcus aureus in milk samples from cows with clinical mastitis and to discriminate this pathogen from other relevant mastitis-causing pathogens in dairy cows. Specifically, we set out to demonstrate that (1) dogs can discriminate the odor of Staph. aureus cultivated on blood agar from those of Escherichia coli, Streptococcus uberis, Streptococcus dysgalactiae, Pseudomonas aeruginosa, and Candida albicans; (2) dogs can identify Staph. aureus, Strep. uberis, and Enterococcus inoculated in milk, and (3) dogs can identify Staph. aureus in milk samples from cow with clinical mastitis.

      MATERIALS AND METHODS

      Four experiments were conducted to test the hypotheses. In experiment 1, dogs were trained to identify Staph. aureus on agar plates; in experiments 2 and 3, dogs identified Staph. aureus inoculated in milk; and, in experiment 4, dogs identified Staph. aureus in milk samples from cows with clinical mastitis.

      Dogs and Handlers

      Nine privately owned dogs of various breeds participated in the experiments of this study (Table 1). Dogs and their owners were recruited on a voluntary basis in a dog training center (Tierakademie Scheuerhof, Wittlich, Germany) as a convenience sample. All dogs were clinically healthy. Dogs and trainers participated in advanced training classes, and 5 dogs had former experience in scent work. According to the German Animal Welfare Bill, no part of the study included harm to any animal.
      Table 1Dogs included in this study
      DogBreedSexAge (yr)Experiment
      1234
      1Jack Russell terrierMale, intact4XXXX
      2Border collieMale, intact3XXXX
      3MixMale, neutered10XXXX
      4Poodle mixMale, intact2X
      5Labrador retrieverMale, intact2X
      6Lagotto RomagnoloMale, neutered9XXXX
      7Lagotto RomagnoloMale, intact9XXXX
      8Labrador retrieverMale, neutered5X
      9Lagotto RomagnoloFemale, intact2X

      Bacterial Samples

      Pathogens for training and testing were isolated from milk samples from dairy cows with signs of clinical mastitis; that is, change in milk quality and swelling of the udder. Mastitis samples were sent to a commercial veterinary laboratory (Laboklin GmbH & Co. KG, Bad Kissingen, Germany; accreditation number D-Pl-13186-01-00). Milk samples were cultivated on Columbia Agar with 5% sheep blood; Columbia CNA agar, Endo agar, and Saboraud agar with gentamicin and chloramphenicol (all from Becton Dickinson, Heidelberg, Germany). Columbia 5% blood is not very selective; Pseudomonas spp., Enterobacteriacae, aerobic gram-negative as well as gram-positive pathogens, and yeast can be cultivated; Columbia CNA is selective for gram positive; Endo agar is selective for Enterobacteriaceae and coliforms; and Saboraud agar is selective for fungi (Candida) or dermatophytes. Sampling amount per plate was 10 µL. Additionally, 1 mL of milk samples was placed into thioglycolate broth (in-house production) for enrichment. Plates for bacterial culture were incubated at 36 ± 1°C for 18 to 24 h. Enrichment broth was incubated at 36 ± 1°C for 18 to 24 h and was then plated on Columbia agar with 5% sheep blood and Endo agar for a second incubation at 36 ± 1°C for 18 to 24 h. Agar for yeast isolation was incubated at 36 ± 1°C. A first reading for yeast growth was done after 48 h and a final reading after 1 wk. After bacteriological growth occurred, pathogens were identified first by colony morphology, Gram strain, and biochemical reactions (catalase, hyaluronidase, oxidase, and esculin reaction). Then, identification was confirmed by matrix-assisted laser desorption/ionization time-of-flight (MALDI-ToF; Shimadzu, Duisburg, Germany;
      • Barreiro J.R.
      • Ferreira C.R.
      • Sanvido G.B.
      • Kostrzewa M.
      • Maier T.
      • Wegemann B.
      • Böttcher V.
      • Eberlin M.N.
      • Dos Santos M.V.
      Identification of subclinical cow mastitis pathogens in milk by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry.
      ). Until use, pathogens were stored at −20°C in cryotubes (Pro-Lab Diagnostics, Richmond Hill, ON, Canada). Bacteria were isolated from 85 milk samples, and we received 35 isolates of Staph. aureus and 10 isolates each of E. coli, Strep. uberis, Strep. dysgalactiae, P. aeruginosa, and C. albicans.
      For training and testing, pathogens were cultivated again on Columbia agar with 5% sheep blood for 24 h at 36 ± 1°C. For early trainings steps, some blank agar plates were placed in the same incubator for the same incubation period. Twelve hours before dog training, a sterile cotton swab (5 × 5 cm, Fuhrmann Verbandstoffe GmbH, Much, Germany) was placed in the top of the agar plate to absorb the bacterial odor or, for the negative control, the odor of the blank agar plate at 36 ± 1°C. The swab had no direct contact with the blood agar. The swabs were then placed in a plastic container (1 L, Jockey Plastik, Wippeführt, Germany) with a perforated lid. To avoid cross contamination, a container was used for one type of pathogen only. For training, the containers including the swab were used for a maximum of 2 wk. Between training sessions, they were kept refrigerated at 6°C. For dog testing, new samples were produced by culturing pathogens on blood agar and exposing cotton swabs to the culture odor for 12 h. The swabs were placed into new containers.
      For experiment 2, bacterial colonies from same species as in experiment 1 were captured with an inoculation loop from the plate (1,000 cfu/mL) and dissolved in 2 mL of fresh bulk tank milk with SCC <120,000/mL. Of this milk, 1 mL was pipetted onto a cotton swab and placed in the bottom of the plastic container. In experiments 2 and 3, bacterial concentrations were 102 and 108 cfu/mL, respectively.
      Bacterial cell pellets were produced for experiment 3 by incubating 20 strains of Staph. aureus, Strep. uberis, and Enterococcus spp. each (all isolated from milk samples of cows with clinical mastitis) for 24 h at 36 ± 1°C in 10 mL of brain-heart broth. From this broth, 100 µL was transferred into 10 mL of sterile brain-heart broth and again incubated at 24 h at 36 ± 1°C under aerobic conditions. After incubation, 2 mL of broth of each strain (approximately 2 × 108 cfu/mL) was transferred into a reaction tube (Eppendorf, Wezzlingen-Berzdorf, Germany) and centrifuged (600 × g) for 15 min. The cell pellet was solved in 100 µL of H2O-free glycerin and the cell glycerin suspension stored in cryotubes at −20°C. For dog training and testing, the cell glycerin suspension was solved in 2 mL of fresh bulk tank milk (SCC <120,000/mL; total plate count <10,000 cfu) and 1 mL was pipetted on a cotton swab. We used bulk milk instead of sterile milk in order to train the dogs to ignore low concentrations of ubiquitous bacteria. The cotton swabs were placed into plastic containers as described above.
      For experiment 4, milk samples from cows with signs of clinical mastitis (see above) were used. Milk sample were sent to the Freie Universität Berlin from various dairy farms in Brandenburg, Germany. Aliquots of 10 mL were shipped to 3 mastitis diagnostic laboratories and 1 aliquot was refrigerated at −20°C. Samples were only included in the experiment if at least 2 laboratories made the same bacteriological diagnosis. Samples with Staph. aureus (n = 11), E. coli (n = 5), Strep. uberis (n = 8), and Trueperella pyogenes (n = 1) were included. For dog testing, 1 mL of the milk sample was pipetted on a cotton swab and placed into a new plastic container. Up to 10 test samples were generated from 1 aliquot.

      Dog Training

      Training was conducted according to a planned training protocol comprising 10 steps (Table 2) under the supervision of a dog trainer, a veterinarian with a postgraduate degree in animal behavior science (V. Theby; www.tierakademie.de). Dogs were trained from May 2012 to February 2013 for experiment 1, from May to July 2014 for experiment 2, and from June to December 2015 for experiment 3. There was no extra training for experiment 4. The duration of each training step and training progress was documented. Training was based on positive reinforcement, with a clicker as secondary reinforcer and food rewards as primary reinforcer according to
      • Pryor K.
      Reaching the Animal Mind: Clicker Training and What It Teaches Us About All Animals.
      .
      Table 2Training protocol in 10 steps to train dogs to discriminate Staphylococcus aureus from other common infectious agents
      StepDescription
      1Present a swab with Staph. aureus odor to the dog. Reward the dog when it sniffs at the swab. Repeat 5× before proceeding to next step.
      2Present container holding the swab, and train dog to touch it with the nose. Present container to dog in hand. Have dog touch container 5×.
      3Slowly move container toward floor in 10-cm steps and place at different locations on the ground for the dog to touch. Continue with next step when dog has indicated container on ground with nose with 1-s latency 10 times out of 10.
      4Add second container without a swab at distance of 50 cm. Ensure that the dog is more likely to reach positive sample first. Change randomly from left to right. Reward as soon as dog touches positive sample. Repeat 10× before proceeding to next step.
      5Present both containers next to each other in random order. Reward as soon as dog reaches correct container. Repeat 10× before proceeding to next step.
      6Add a third container without a swab in random order. Reward dog when it indicates correct sample for 3 s. Repeat 10× before proceeding to next step.
      7Add swabs with odor of other pathogens into the empty container (sample to discriminate). Repeat 10× before proceeding to next step.
      8Present 3 containers to discriminate and 3 positive containers in random order. Dog should indicate positive sample for 3 s. Repeat 10× before proceeding to next step.
      9Same as step 10, without handler knowing the location of the positive sample. Dog must indicate positive container for 3 s.
      10Increase number of containers to 10 (1 positive and 9 for discrimination) with handler blinded.

      Dog Testing

      In experiments 1 and 2, dogs had to indicate 1 container holding a swab with Staph. aureus odor out of 10 containers holding swabs with odor of 1 of 4 other bacterial species and 1 yeast. The composition of the 9 containers was random. Randomization both for samples within a trial and the order of trials were generated with the random number function of Excel (Microsoft Corp., Redmond, WA).
      For experiments 1 and 2, the prevalences of different pathogens within the 10 containers were Staph. aureus (10%); E. coli (5%), Strep. uberis (27%), Strep. dysgalactiae (28%), P. aeruginosa (20%), and C. albicans (10%). Therefore, the chance of identifying the correct container by chance was 1:10. Each dog performed 10 trials with 10 containers each, resulting in 100 samples searched per dog. The protocol was modified for experiment 3. The number of containers was reduced to 7 per trial. Dogs performed 15 trials for a total of 105 samples searched. In experiment 3, the containers to discriminate were holding swabs with odor of Strep. uberis (23%), Enterococcus (24%), or bulk milk without bacteria (39%). Two trials served as negative controls without a container with Staph. aureus, and 2 trials were done with 2 containers with Staph. aureus. The order of trials was randomized, and the prevalence of Staph. aureus was 15%.
      In the fourth experiment, 10 trials were performed with 7 containers each (i.e., total of 70 samples). The container to differentiate included Strep. uberis (31%), Strep. dysgalactiae (3%), E. coli (10%), T. pyogenes (2%), Enterococcus (16%), and bulk milk (25%). The prevalence of Staph. aureus in this test was 13%.
      Containers were positioned in a circle with a radius of approximately 80 cm, with each container approximately 70 cm apart. For every trial, unused containers were included. Dogs, handlers, and any other persons in the experimental room were blinded to the position of the sample to avoid hidden clues. The test was videotaped and transferred to another room where the experimenter was seated. Dog handlers were instructed to make sure that the dogs examined every container, even if the dog had already indicated one as being positive. After the dog had searched each container, the dog handler announced the number of the positive container to the experimenter via video transmission. The experimenter gave feedback to the dog handler so that the dog was eventually rewarded. Indications were documented as correct positive, correct negative, false positive, or false negative.
      Test characteristics including number of dogs, samples, trials, and odorant used in each experiment are summarized in Table 3.
      Table 3Test characteristics of dogs for detecting Staphylococcus aureus in experiment 1 (blood agar), experiment 2 (inoculated milk, 103 cfu/mL), experiment 3 (inoculated milk, 108 cfu/mL), and experiment 4 (mastitis milk)
      CharacteristicExperiment
      1234
      Dogs (no.)8265
      Positive:negative containers1:101:101:71:7
      Samples/dog (no.)10050–10010570
      Trials (no.)105–101510
      Correct indications (%)97.286.695.989.4
      Sensitivity (%)91.35583.359
       95% CI83–9634.2–74.174–8943.4–73.9
      Specificity (%)97.99598.093.2
       95% CI96.5–98.790.7–97.396.1–98.289.9–95.5

      Statistical Analyses

      The analyses were performed with Excel (Office 2010, Microsoft Deutschland GmbH, Munich, Germany) and PASW Statistics for Windows (version 19.0, SPSS Inc., Munich, Germany). Sensitivity and specificity were calculated by using cross tables. Confidence intervals for sensitivity and specificity were calculated with the Wilson score method without continuity correction (
      • Newcombe R.G.
      Two-sided confidence intervals for the single proportion: Comparison of seven methods.
      ). Influences of content of container to discriminate and the individual dog, handler, or trial were calculated with Fisher exact test.

      RESULTS

      Required Training Time

      Eight dogs finished the training procedure for experiment 1. Dogs were trained from 27 to 38 training days over 9 mo. The overall training duration (time that dogs worked in training sessions) ranged from 111 to 290 min (163.4 ± 70.2 min). Two dogs were not able to complete the training protocol because they could not progress beyond step 5 in the training protocol. In experiment 2, 5 dogs were trained. Dogs were trained on 17 to 22 d over 3 mo. Total training time varied from 97 to 191 min (154 ± 35.7 min). Six dogs finished training for experiment 3. They were trained on 19 to 22 d over a period of 6 mo, with an average training time of 98 to 177 min (141.6 ± 27.5 min). Training time did not differ between the experiments. No additional training was performed for experiment 4.

      Dog Testing

      Sensitivity, specificity, and percentage of correct indications of all dogs for all experiments are summarized in Table 3. Sensitivity and specificity in experiment 1 were 91.3% (95% CI: 83–96%) and 97.9% (95% CI: 96.5–98.7%), respectively. Four dogs did not make any mistakes (i.e., sensitivity and specificity = 100%; Table 4). In experiment 2, 4 dogs showed behavioral signs of stress, such as panting or barking, and were excluded from the test. The 2 remaining dogs completed all 10 trials of experiment 2 (Table 5). Sensitivity and specificity in experiment 2 were 55% (95% CI: 34.2–74.1%) and 95% (95% CI 90.7–97.3%), respectively. In experiment 3, testing was completed by all 6 dogs trained, and they identified Staph. aureus with a sensitivity and specificity of 83.3% (95% CI: 74.0–89.0%) and 98% (95% CI: 96.1–98.2%), respectively (Table 6). Five dogs completed experiment 4; test sensitivity was 59% (95% CI: 43.4–72.9%) and specificity was 93.2% (95% CI: 89.9–95.5%) (Table 7). Type of container content to discriminate, or individual dog, handler, or trial had no influence on a false-positive indication.
      Table 4Test characteristics of sniffer dogs discriminating Staphylococcus aureus from Escherichia coli, Streptococcus uberis, Streptococcus dysgalactiae, Pseudomonas aeruginosa, and Candida albicans (experiment 1)
      CharacteristicDog
      12345678Total
      Correct positive (no.)10971010981073
      Correct negative (no.)9088809090898890705
      False positive (no.)02100012015
      False negative (no.)013001207
      Sensitivity (%)1009087.5100100908010091.3
      Specificity (%)10097.797.610010098.897.710097.9
      Correct indication (%)1009787100100989610097.2
      Table 5Test characteristics of dogs in experiment 2, identifying Staphylococcus aureus (103 cfu/mL inoculated in 2 mL of bulk milk) against Streptococcus uberis and Enterococcus spp. (both 103 cfu/mL)
      CharacteristicDog
      12457Total
      Correct positive (no.)6
      Test was not completed.
      511
      Correct negative (no.)8685171
      False positive (no.)459
      False negative (no.)459
      Sensitivity (%)605055
      Specificity (%)9594.495
      Correct indication (%)929091
      1 Test was not completed.
      Table 6Test characteristics of dogs in experiment 3, identifying Staphylococcus aureus (108 cfu/mL inoculated in 2 mL of bulk milk) against Streptococcus uberis andEnterococcus spp. (both 108 cfu/mL)
      CharacteristicDog
      124689Total
      Correct positive (no.)11131313141175
      Correct negative (no.)889089868888529
      False positive (no.)20142211
      False negative (no.)42221415
      Sensitivity (%)73.386.686.686.693.37383.3
      Specificity (%)9710098.99597.797.798
      Correct indication (%)94.398.197.194.397.194.395.9
      Table 7Test characteristics of sniffer dogs discriminating Staphylococcus aureus in mastitis milk samples against Streptococcus uberis, Streptococcus dysgalactiae, Trueperella pyogenes, Enterococcus, and Escherichia coli (experiment 4)
      CharacteristicDog
      12468Total
      Correct positive (no.)4463623
      Correct negative (no.)5558585861290
      False positive (no.)6535221
      False negative (no.)5334116
      Sensitivity (%)44.457.166.742.985.759.0
      Specificity (%)90.287.195.192.196.893.2
      Correct indication (%)84.388.691.487.195.789.4

      DISCUSSION

      Four experiments were conducted to study the ability of dogs to identify odors related to Staph. aureus, a relevant mastitis pathogen. In experiment 1, we demonstrated that dogs can discriminate the odor of Staph. aureus cultivated on blood agar plates from 5 other bacteria and 1 yeast with a sensitivity of 91.3% and a specificity of 97.9%. Four dogs identified all samples correctly.
      • Lim S.H.
      • Mix S.
      • Anikst V.
      • Budvytiene I.
      • Eiden M.
      • Churi Y.
      • Queralto N.
      • Berliner A.
      • Martino R.A.
      • Rhodes P.A.
      Bacterial culture detection and identification in blood agar plates with an optoelectronic nose.
      used an electronic nose device to identify 15 pathogenic bacterial species by their volatile organic compounds; the sensitivity and specificity of the electronic nose was 91 and 99.4%, respectively. Pathogens produce a specific odor as a result of metabolism of substrates available in growth media (
      • Lough F.
      • Perry J.D.
      • Stanforth S.P.
      • Dean J.R.
      Detection of exogenous VOCs as a novel in vitro diagnostic technique for the detection of pathogenic bacteria.
      ). Our data indicate that dogs could not only identify this odor but also differentiate it from that of other pathogens. Furthermore, this is the first report of dogs being trained to identify the odor of Staph. aureus isolated from mastitis milk samples.
      The sensitivity and specificity in experiment 1 were similar to results of a previous study (
      • Johnen D.
      • Heuwieser W.
      • Fischer-Tenhagen C.
      Canine scent detection—Fact or fiction?.
      ), in which dogs differentiated tea against water, which we assume is a relatively easy scent task. Thus, we speculate that Staph. aureus has a distinctive smell that makes it easy for the dog to differentiate it from other pathogens. This is in agreement with the results of
      • Bruins M.
      • Bos A.
      • Petit P.L.
      • Eadie K.
      • Rog A.
      • Bos R.
      • van Ramshorst G.H.
      • van Belkum A.
      Device-independent, real-time identification of bacterial pathogens with a metal oxide-based olfactory sensor.
      , who identified Staph. aureus strains with a specificity of 88% using an electronic nose device, and with those of
      • Bomers M.K.
      • van Agtmael M.A.
      • Luik H.
      • van Veen M.C.
      • Vandenbroucke-Grauls C.M.J.E.
      • Smulders Y.M.
      Using a dog's superior olfactory sensitivity to identify Clostridium difficile in stools and patients: Proof of principle study.
      , who described similar results for detecting Clostridium difficile with a dog.
      Milk is a complex matrix comprising a colloid of butterfat, globules, and water with dissolved carbohydrates and protein complexes. The flavoring components vary according to the sample's state such that raw milk elicits a distinct odor (
      • d'Acampora Zellner B.
      • Dugo P.
      • Dugo G.
      • Mondello L.
      Gas chromatography–olfactometry in food flavour analysis.
      ). It is well known that bacteria grow in and spoil dairy products, which manifests as changes in taste and odor (
      • Bekker A.
      • Jooste P.
      • Steyn L.
      • Bothma C.
      • Hugo A.
      • Hugo C.
      Lipid breakdown and sensory analysis of milk inoculated with Chryseobacterium joostei or Pseudomonas fluorescens.
      ). The bacteria are suspended in solution as well as entrapped and absorbed on proteins, micelles, and fat globules (
      • Mortari A.
      • Lorenzelli L.
      Recent sensing technologies for pathogen detection in milk: A review.
      ). In experiments 2 and 3, we tested whether the odor of Staph. aureus changes when inoculated in milk.
      In experiment 2, the dogs' detection performance deteriorated. Only 2 of 5 dogs were able to complete all test trials, and the other dogs showed behavioral signs of stress. Scent detection performance of dogs depends on the concentration of the target odor in the sample (
      • Walker D.B.
      • Walker J.C.
      • Cavnar P.J.
      • Taylor J.L.
      • Pickel D.H.
      • Hall S.B.
      • Suarez J.C.
      Naturalistic quantification of canine olfactory sensitivity.
      ).
      • Barkema H.W.
      • Schukken Y.
      • Lam T.
      • Beiboer M.
      • Wilmink H.
      • Benedictus G.
      • Brand A.
      Incidence of clinical mastitis in dairy herds grouped in three categories by bulk milk somatic cell counts.
      considered 100 cfu/mL of Staph. aureus sufficient to cause an IMI in their study on the incidence of clinical mastitis in dairy herds.
      • Grönlund U.
      • Hultén C.
      • Eckersall P.D.
      • Hogarth C.
      • Waller K.P.
      Haptoglobin and serum amyloid A in milk and serum during acute and chronic experimentally induced Staphylococcus aureus mastitis.
      found more than 105 cfu/mL of Staph. aureus in milk samples 24 h after experimental infection of dairy cows. In experiment 3, the concentration of Staph. aureus was 103 cfu/mL in each milk sample, indicating that this bacterial concentration or the volume of 2 mL was below the detection threshold for the dogs. Therefore, we repeated the experiment with 108 cfu/mL inoculated in 2 mL of bulk milk, which allowed all dogs to complete the test trials without signs of stress.
      The dogs discriminated Staph. aureus with sensitivity and specificity of 83.3 and 98%, respectively, in experiment 3. To our knowledge, this is the first report to indicate that dogs can be trained to identify Staph. aureus in milk samples. For raw milk samples, only a biosensor for odor detection based on an anti-protein antibody test has been shown to be effective for the detection of Staph. aureus (
      • Esteban-Fernández de Ávila B.
      • Pedrero M.
      • Campuzano S.
      • Escamilla-Gómez V.
      • Pingarrón J.M.
      Sensitive and rapid amperometric magnetoimmunosensor for the determination of Staphylococcus aureus.
      ).
      In experiment 4, we tested whether dogs could identify Staph. aureus in milk samples collected from cows with signs of clinical mastitis. Milk samples from mastitic cows and milk samples inoculated with mastitis pathogens differ in composition of volatile organic compounds (i.e., methyl esters of free fatty acids) in the headspace as measured by solid-phase microextraction and GC-MS (
      • Hettinga K.A.
      • Van Valenberg H.
      • Lam T.
      • Van Hooijdonk A.
      The origin of the volatile metabolites found in mastitis milk.
      ). In milk samples collected from cows with clinical mastitis caused by Staph. aureus, the sensitivity and specificity of detecting this pathogen were 59 and 93.2%, respectively, which was lower than in experiment 3. One dog, however, identified Staph. aureus with a sensitivity of 85.7% and a specificity of 96.8%. Inflammation of the udder can lead to considerable changes in milk appearance with flakes, blood, or pus. These components may affect the ability of the dogs to smell the odor of Staph. aureus. We did not conduct specific training for experiment 4 because the number of mastitis samples with matching results from at least 2 laboratories was limited (n = 25). Another limitation was that only samples that had not been used in previous training could be used for testing to avoid the possibility that dogs would remember the odor of individual samples instead of indicating the odor of Staph. aureus (
      • Johnen D.
      • Heuwieser W.
      • Fischer-Tenhagen C.
      An approach to identify bias in scent detection dog testing.
      ). It is unclear whether additional training for the detection of Staph. aureus in field cases of mastitis would have improved test characteristics. The good performance of a single dog warrants a larger proof-of-concept study to test this hypothesis.

      CONCLUSIONS

      We demonstrated that Staph. aureus emits a specific odor in the headspace of blood agar plates, inoculated milk samples, and milk samples of cows with clinical mastitis. The training and use of sniffer dogs for scent detection in a laboratory is, however, critical for hygiene or as an operating diagnostic tool. Therefore, developing technical solutions such as electronic nose devices as real-time diagnostic methods for mastitis pathogens is considered worthwhile.

      ACKNOWLEDGMENTS

      This work was supported by Laboklin GmbH & Co. KG (Bad Kissingen, Germany). In particular, we want to thank Julia Elze and Babette Klein for a great collaboration and providing us with pathogens. We also thank the dog owners (and their dogs), who did a great job and spent a considerable amount of time on training and participating in this study.

      REFERENCES

        • Barkema H.W.
        • Schukken Y.
        • Lam T.
        • Beiboer M.
        • Wilmink H.
        • Benedictus G.
        • Brand A.
        Incidence of clinical mastitis in dairy herds grouped in three categories by bulk milk somatic cell counts.
        J. Dairy Sci. 1998; 81 (9532494): 411-419
        • Barkema H.W.
        • Schukken Y.H.
        • Zadoks R.N.
        Invited review: The role of cow, pathogen, and treatment regimen in the therapeutic success of bovine Staphylococcus aureus mastitis.
        J. Dairy Sci. 2006; 89 (16702252): 1877-1895
        • Barreiro J.R.
        • Ferreira C.R.
        • Sanvido G.B.
        • Kostrzewa M.
        • Maier T.
        • Wegemann B.
        • Böttcher V.
        • Eberlin M.N.
        • Dos Santos M.V.
        Identification of subclinical cow mastitis pathogens in milk by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry.
        J. Dairy Sci. 2010; 93: 5661-5667
        • Bekker A.
        • Jooste P.
        • Steyn L.
        • Bothma C.
        • Hugo A.
        • Hugo C.
        Lipid breakdown and sensory analysis of milk inoculated with Chryseobacterium joostei or Pseudomonas fluorescens.
        Int. Dairy J. 2016; 52: 101-106
        • Bomers M.K.
        • van Agtmael M.A.
        • Luik H.
        • van Veen M.C.
        • Vandenbroucke-Grauls C.M.J.E.
        • Smulders Y.M.
        Using a dog's superior olfactory sensitivity to identify Clostridium difficile in stools and patients: Proof of principle study.
        BMJ. 2012; 345 (23241268): e7396
        • Browne C.
        • Stafford K.
        • Fordham R.
        The use of scent-detection dogs.
        Ir. Vet. J. 2006; 59: 97-103
        • Bruins M.
        • Bos A.
        • Petit P.L.
        • Eadie K.
        • Rog A.
        • Bos R.
        • van Ramshorst G.H.
        • van Belkum A.
        Device-independent, real-time identification of bacterial pathogens with a metal oxide-based olfactory sensor.
        Eur. J. Clin. Microbiol. Infect. Dis. 2009; 28 (19190942): 775-780
        • Cha E.
        • Smith R.L.
        • Kristensen A.R.
        • Hertl J.A.
        • Schukken Y.H.
        • Tauer L.W.
        • Welcome F.L.
        • Gröhn Y.T.
        The value of pathogen information in treating clinical mastitis.
        J. Dairy Res. 2016; 83 (27845019): 456-463
        • d'Acampora Zellner B.
        • Dugo P.
        • Dugo G.
        • Mondello L.
        Gas chromatography–olfactometry in food flavour analysis.
        J. Chromatogr. A. 2008; 1186 (17915233): 123-143
        • Esteban-Fernández de Ávila B.
        • Pedrero M.
        • Campuzano S.
        • Escamilla-Gómez V.
        • Pingarrón J.M.
        Sensitive and rapid amperometric magnetoimmunosensor for the determination of Staphylococcus aureus.
        Anal. Bioanal. Chem. 2012; 403 (22290389): 917-925
        • Grönlund U.
        • Hultén C.
        • Eckersall P.D.
        • Hogarth C.
        • Waller K.P.
        Haptoglobin and serum amyloid A in milk and serum during acute and chronic experimentally induced Staphylococcus aureus mastitis.
        J. Dairy Res. 2003; 70 (14649408): 379-386
        • Harmon R.J.
        Physiology of mastitis and factors affecting somatic cell counts.
        J. Dairy Sci. 1994; 77 (7929968): 2103-2112
        • Hettinga K.A.
        • Van Valenberg H.
        • Lam T.
        • Van Hooijdonk A.
        The origin of the volatile metabolites found in mastitis milk.
        Vet. Microbiol. 2009; 137 (19200667): 384-387
        • Johnen D.
        • Heuwieser W.
        • Fischer-Tenhagen C.
        Canine scent detection—Fact or fiction?.
        Appl. Anim. Behav. Sci. 2013; 148: 201-208
        • Johnen D.
        • Heuwieser W.
        • Fischer-Tenhagen C.
        An approach to identify bias in scent detection dog testing.
        Appl. Anim. Behav. Sci. 2017; 98: 1-12
        • Kauhanen E.
        • Harri M.
        • Nevalainen A.
        • Nevalainen T.
        Validity of detection of microbial growth in buildings by trained dogs.
        Environ. Int. 2002; 28: 153-157
        • Lago A.
        • Godden S.
        • Bey R.
        • Ruegg P.
        • Leslie K.
        The selective treatment of clinical mastitis based on on-farm culture results: I. Effects on antibiotics use, milk withholding time, and short-term clinical and bacteriological outcomes.
        J. Dairy Sci. 2011; 94 (21854917): 4441-4456
        • Lim S.H.
        • Mix S.
        • Anikst V.
        • Budvytiene I.
        • Eiden M.
        • Churi Y.
        • Queralto N.
        • Berliner A.
        • Martino R.A.
        • Rhodes P.A.
        Bacterial culture detection and identification in blood agar plates with an optoelectronic nose.
        Analyst. 2016; 141 (26753182): 918-925
        • Lough F.
        • Perry J.D.
        • Stanforth S.P.
        • Dean J.R.
        Detection of exogenous VOCs as a novel in vitro diagnostic technique for the detection of pathogenic bacteria.
        Trends Anal. Chem. 2017; 87: 71-81
        • Makovec J.A.
        • Ruegg P.L.
        Results of milk samples submitted for microbiological examination in Wisconsin from 1994 to 2001.
        J. Dairy Sci. 2003; 86 (14672176): 3466-3472
        • Mortari A.
        • Lorenzelli L.
        Recent sensing technologies for pathogen detection in milk: A review.
        Biosens. Bioelectron. 2014; 60 (24768759): 8-21
        • Newcombe R.G.
        Two-sided confidence intervals for the single proportion: Comparison of seven methods.
        Stat. Med. 1998; 17 (9595616): 857-872
        • Oliver S.
        • Gonzalez R.
        • Hogan J.
        • Jayarao B.
        • Owens W.
        Microbiological procedures for the diagnosis of bovine udder infection and determination of milk quality.
        National Mastitis Council, Verona, WI2004
        • Pavlou A.K.
        • Turner A.P.F.
        Sniffing out the truth: Clinical diagnosis using the electronic nose.
        Clin. Chem. Lab. Med. 2000; 38 (10834396): 99-112
        • Pryor K.
        Reaching the Animal Mind: Clicker Training and What It Teaches Us About All Animals.
        Scribner, New York, NY2009
        • Roberson J.R.
        • Warnick L.
        • Moore G.
        Mild to moderate clinical mastitis: Efficacy of intramammary amoxicillin, frequent milk-out, a combined intramammary amoxicillin, and frequent milk-out treatment versus no treatment.
        J. Dairy Sci. 2004; 87 (15202642): 583-592
        • Taponen S.
        • Salmikivi L.
        • Simojoki H.
        • Koskinen M.T.
        • Pyorala S.
        Real-time polymerase chain reaction-based identification of bacteria in milk samples from bovine clinical mastitis with no growth in conventional culturing.
        J. Dairy Sci. 2009; 92 (19447993): 2610-2617
        • Teale C.J.
        • Moulin G.
        Prudent use guidelines: A review of existing veterinary guidelines.
        Rev. Sci. Tech. 2012; 31 (22849288): 343-354
        • Turner A.P.F.
        • Magan N.
        Electronic noses and disease diagnostics.
        Nat. Rev. Microbiol. 2004; 2 (15040263): 161-166
        • Ungemach F.R.
        • Mueller-Bahrdt D.
        • Abraham G.
        Guidelines for prudent use of antimicrobials and their implications on antimicrobials usage in veterinary medicine.
        Int. J. Med. Microbiol. 2006; 296 (16520092): 33-38
        • Vasquez A.K.
        • Nydam D.V.
        • Capel M.B.
        • Eicker S.
        • Virkler P.D.
        Clinical outcome comparison of immediate blanket treatment versus a delayed pathogen-based treatment protocol for clinical mastitis in a New York dairy herd.
        J. Dairy Sci. 2017; 100 (28161180): 2992-3003
        • Walker D.B.
        • Walker J.C.
        • Cavnar P.J.
        • Taylor J.L.
        • Pickel D.H.
        • Hall S.B.
        • Suarez J.C.
        Naturalistic quantification of canine olfactory sensitivity.
        Appl. Anim. Behav. Sci. 2006; 97: 241-254
        • Weetjens B.J.
        • Mgode G.F.
        • Machang'u R.S.
        • Kazwala R.
        • Mfinanga G.
        • Lwilla F.
        • Cox C.
        • Jubitana M.
        • Kanyagha H.
        • Mtandu R.
        • Kahwa A.
        • Mwessongo J.
        • Makingi G.
        • Mfaume S.
        • Van Steenberge J.
        • Beyene N.W.
        • Billet M.
        • Verhagen R.
        African pouched rats for the detection of pulmonary tuberculosis in sputum samples.
        Int. J. Tuberc. Lung Dis. 2009; 13 (19460250): 737-743