Subjects Nine (5 male) African Giant Pouched rats (C. ansorgei) averaging 3.7 years at the start of the experiment were sourced from APOPO’s rat colony in Morogoro, Tanzania. Sample size was determined based on prior work.23 One (female) rat failed to complete the test at the conclusion of Experiment 1 and was therefore excluded from Experiment 2. All rats had similar experiences serving as subjects in prior experiments with unrelated odors. Rats were single- or pair-housed in cages that contained clay pots for sleeping, wood shaving substrate, and a chewing/climbing stick made from untreated wood. Home cages were located within a vivarium maintained between 21 and 34° C with a natural light:dark cycle of approximately 12:12 hours. All sessions were conducted during the light period. Rats were maintained at their ad libitum weight by providing opportunity to earn up to 36 banana-flavored pellets (5 g, 5TCY OmniTreatTM) during experimental sessions, which was supplemented at the end of the working day with 20 g of commercially available rodent chow (Specialty Feeds, Glen Forrest, Western Australia). Prior to weekends, rats were fed 65 g of chow along with locally sourced fresh foods including 10 g of sundried sardines and one produce item, (e.g., mango, avocado, banana, etc.). Rats always had access to clean drinking water while in their home and transport cages.
Prior to the start of the experiment, all rats were pre-screened for Brucella infection. Blood samples were collected from the tail vein of each rat using plain vacutainer tubes, and then centrifuged to obtain a serum. The resulting serum samples were subjected to a Rose Bengal Test (RBT) for Brucella antibodies according to the procedure recommended by the OIE.4 This test was repeated after the project was completed. Both pre- and post-screening tests generated negative results.
Apparatus A custom engineered line cage measuring 247 x 41 x 34 cm mounted on four 96 cm high legs, with hinged top panels and glass walls was used for all experiments. Ten circular holes, measuring 30-mm in diameter, were evenly spaced on the floor of the cage. Each hole was fitted with a through-beam (infrared) photoelectric sensor (ENV-254, MedAssociates) and covered by an aluminum plate that slid open via underlying solenoids. Breaks in the infrared beam (when the rat inserted its nose into the hole) were accompanied by a continuous beep as auditory feedback to the rat. The first hole opened at the start of the session and each subsequent hole opened after the infrared beam was broken in the hole immediately preceding it. Light emitting diodes (LEDs) located inside the Perspex wall of the cage alongside each hole turned on and off in correspondence with the opening and closing of the holes. The aluminum plate covering each hole closed 500 milliseconds (ms) after the rat removed its nose from the hole. On the left side of the cage, a pellet dispenser (ENV-203-94, MedAssociates, Georgia, VT) delivered pellets via a 20 cm long plastic tube attached to a 6 x 6 cm square receptacle located flush with the chamber floor. Timing and duration of beam breaks, and delivery of all auditory feedback, and food pellet reinforcement was controlled using custom designed software (MS Visual Basic).
Aluminum cassettes measuring 192 x 8 x 45 cm, with 10, 40-mm diameter holes positioned in correspondence with the 10 holes in the floor of the cage, were loaded with sample containers before the first daily session. Sample containers were 40 ml plastic pots measuring 5.6 cm tall by 3.5 cm wide. The pre-loaded cassettes fit into a hinged bracket that swung up and locked into position underneath the cage.
Samples Cultured Brucella were prepared in a laboratory located on the campus of SUA and were brought to the training facility daily. The Brucella used in the present study were obtained from previous project that collected samples from Kitengule Ranch, located in Kagera region, Tanzania. A total of 11 sample types were used during this project; eight were used during initial training while three were reserved for testing experiments that followed (Table 1). Individual samples were prepared the morning of their use. After preparation, the containers were labeled (ID and type, e.g., 10+ or 10-, see Table 1), placed in a plastic bag, and transported to APOPO training facility via a cool box. After each training session, all samples were collected in autoclavable bags and returned to the facility where they were autoclaved for 45 minutes before disposal.
Table 1. Samples used during rat training and tests. All samples were heat-inactivated prior to behavioral sessions. Brucella-positive samples served as scent targets for rats, while all Brucella-negative samples served as non-reinforced controls.
|
Sample
|
Brucella
|
Experiment
|
Description
|
|
5+
|
Positive
|
Exp 2: Novel trials only
|
5-day old Brucella culture isolates with nutrient agar
|
|
7+
|
Positive
|
Exp 1, 2, and 3
|
7-day old Brucella culture isolates with nutrient agar
|
|
10+
|
Positive
|
Exp 1, 2, and 3
|
10-day old Brucella culture isolates with nutrient agar
|
|
10BA+
|
Positive
|
Exp 1: Phase I only
|
10-day old Brucella culture isolates with blood agar media
|
|
12+
|
Positive
|
Exp 1, 2, and 3
|
12-day old Brucella culture isolates with nutrient agar
|
|
7-
|
Negative
|
Exp 1, 2, and 3
|
Blank nutrient agar incubated for 7 days
|
|
10-
|
Negative
|
Exp 1, 2, and 3
|
Blank nutrient agar incubated for 10 days
|
|
12-
|
Negative
|
Exp 1, 2, and 3
|
Blank nutrient agar heat inactivated after 12-day period
|
|
E. Coli
|
Negative
|
Exp 1, 2, and 3
|
E. coli culture in nutrient agar
|
|
5BA-
|
Negative
|
Exp 2: Novel trials only
|
5-day old blood agar media
|
|
Feces+
|
Positive
|
Exp 3
|
fecal sample spiked with 10-day old Brucella -positive solution
|
|
Feces-
|
Negative
|
Exp 3
|
fecal sample spiked with 10-day old Brucella -negative solution
|
For the preparation of Brucella positive samples, isolated Brucella bacteria from placental and milk samples was cultured in Farrell's medium, as described by Mathew et al.30 The plates were observed for any growth before and after harvesting for training samples beginning on Day 3 through Day 14; those that did not show any growth after 12 days were discarded. After the pre-determined incubation period, isolates from the growth were picked using sterile wire loop. Ten colonies were mixed with 10 mls of distilled water and the resulting solution was heat inactivated at 56°C for 3 hours in a water bath and then allowed to cool. On the day of their use, 500 μl of the solution was pipetted into a sample container filled with 5 mls of either nutrient or blood agar (see procedures below).
Brucella negative samples were handled identically to positive samples, including culture conditions, heat inactivation, and storage, but contained no Brucella bacteria. For their preparation, 10 mls of distilled water was mixed with 10 loopfuls of Farrell’s medium, incubated both aerobically and anaerobically and 500 μl of the solution was pipetted into a sample container filled with 5 mls of nutrient or blood agar on the day of their use.
Specimens from ten pre-grown colonies of Escherichia coli (E. coli) were mixed with 10 mls of distilled water to form a solution. The resulting solution was heat inactivated at 56°C for 3 hours in a water bath and then allowed to cool. On the day of their use, 500 μl of the E. coli solution was pipetted into sample containers filled with 5 mls of nutrient agar.
Spiked fecal samples were introduced in Experiment 3. Fecal samples were collected in stool containers from animals that screened Brucella-negative using RBT. Samples were collected from both Kitengule Ranch and from Magadu Farm at SUA and were kept refrigerated at 4°C. A spatula was used to pick feces which was mixed with 1 ml of distilled water and the resulting solution was placed into a sample container. These samples were then spiked with 500 μl of either Brucella-positive or Brucella-negative solution (incubated for 10 days, as described above) and pipetted into each container before they were transported for rat evaluation sessions later the same day.
Data Analysis All data were analyzed using SPSS 20 (IBM). Individual rats occasionally failed to evaluate all samples assigned to a daily training or test session. Results from these rats were excluded from analysis, as noted below. Baseline detection accuracy for evaluating test performance was calculated using the last five training sessions in which the specified criterion was met.
Behavioral Procedures All training and test sessions (collectively referred to as evaluation sessions) across experiments were conducted five days per week, excluding weekends and public holidays. We adopted a within-subject design in which each rat underwent identical training (as described below) to determine if rats can be trained to reliably detect the scent of cultured Brucella by comparing individual rat behavior in response to target (Brucella-positive) and non-target (Brucella-negative) control samples. All rat evaluation sessions lasted a maximum of 30 minutes. If a rat took more than 30 minutes to complete a session, it was removed from the cage, the trial at which the rat stopped was noted, and the session was not repeated. If at any time during an evaluation session, three minutes elapsed without a trial being completed due to inactivity or hyperactivity, the rat was removed from the apparatus, returned to its transport cage, and given another opportunity to resume training after the last rat had finished. Repeated sessions started with the trial the rat had previously failed to complete.
Trials within a session were initiated and terminated by the rat’s behavior. When the rat approached the first hole, the trainer used the software to start the session, thereby triggering the first hole to open. When the rat inserted its nose into the hole (termed nose poke), the duration of the beam break (nose poke) was recorded, and the next hole opened. If the duration of the nose poke met or exceeded a pre-determined indication threshold on a hole containing a Brucella-positive sample, the response was classified as a “hit” and resulted in reinforcement (delivery of flavored food pellets). There were no programmed consequences if the indication threshold was exceeded over a Brucella-negative sample, but the nose poke indication was categorized as a “false positive.” If the indication threshold was not met, the response was categorized as a “miss” when it occurred in a hole containing a Brucella-positive samples, or a “correct reject” if the hole contained any Brucella-negative sample, including E. coli controls. The indication threshold for all rats was initially set at 2000 ms but was adjusted for each rat depending on individual performance according to the guidelines described below. When all samples in the cassette had been evaluated, trainers removed the cassette and inserted the next one, while the rat remained in the cage.
Exp 1: Does Brucella emit a detectable odor signature?
Training occurred in five phases with progression through the phases determined by the performance of the entire team of rats. Criteria for phase advancement was pre-determined using the published sensitivity and specificity of current diagnostic methods as a guide.31 Phase advancement never occurred after a weekend or holiday to ensure performance was not impacted by the break in training.
Phase I: Indication 12 samples (8 sessions). A clear Perspex glass divider was positioned between holes three and four, thus restricting access to only the first three holes. Rats were presented with 12 Brucella-positive samples (6 each of 10+ and 10BA+) per session and were required to meet the indication threshold to receive reinforcement. To advance to Discrimination Training, 7 of the 9 rats were required to correctly indicate (exceed threshold) on at least 10 samples within a single session.
Phase II: 3-Hole Discrimination 30 samples (12 sessions). The divider remained in the same position, however non-target samples were introduced. Rats were presented with 30 samples per session, including 10 samples of 10+ and 20 Brucella-negative samples (10-). At the start of this phase, the indication threshold was set to 1500 ms for all rats, which served as the minimum threshold for the rest of training. If, after four days of training, a rat committed more than 10 false positives for two consecutive sessions, the indication threshold was increased by 500 ms. Indication thresholds continued to be adjusted following these procedures for all subsequent phases. At least 7 rats were required to hit at least 9 of the 10+ target samples (≥90% sensitivity) and commit no more than 5 false positives (≥75% specificity) for two consecutive training sessions to advance to the next phase.
Phase IIIa: Discrimination 30 samples (6 sessions). The Perspex divider was removed, allowing access to all 10 sample holes within the cage. As with Phase II, rats were presented with 30 samples per session, including 10 samples of 10+ and 20 Brucella-negative samples (10-). After two consecutive sessions in which at least 7 rats hit at least 9 targets while committing no more than 5 false positives, target and non-target samples of the 7- and 12-day samples were introduced.
Rats continued to be presented with 30 samples per session; however, the 10 positive samples included four samples of 10+ plus 3 samples each of the new 7+ and 12+ targets. Likewise, the 20 negative samples included 8 samples of 10- and 6 samples each of the new 7- and 12- non-targets. At least 7 rats were required to hit 9 positives and commit no more than 5 false positives for two consecutive sessions before advancing to Phase IIIb.
Phase IIIb: Discrimination 50 samples (9 sessions). Rats were presented with 50 samples per session, including 10 positive samples (4 samples of 10+ and 3 samples each of 7+ and 12+). Two blind trials were included per session. These samples were coded as negative so even the trainers conducting the sessions were unaware that these samples contained Brucella. Blind trials served two purposes: 1) they ensured that the rats’ detection behavior was driven by the odors emitted by the samples themselves, rather than any extraneous cues, and 2) they prepared the rat for potential future screening work in which the true status of any given sample is unknown until after the rat has performed the evaluation. In this scenario, the rat cannot be rewarded for indicating the sample. Thus, blind trials accustom the rat to working under conditions of partial reinforcement in which some correct indications are not reinforced. Two different blind samples were chosen at random each day. Additionally, 40 negative samples were included in each session, including 14 samples of 10- and 13 samples each of 7- and 12-.
When at least 7 rats hit at least 9 targets (≥90% sensitivity) and had no more than 10 false positives (≥75% specificity) for two consecutive days, a novel negative sample (E. coli) was introduced.
Phase IV: E. coli Controls (5 sessions). The inclusion of cultured E. coli samples ensured that the rats learned to specifically identify Brucella bacteria rather than any bacteria in comparison to bacteria-free culture medium. E. coli controls therefore served a crucial function in determining to what extent Brucella abortus emits a unique odor signature. Rats continued training with 50 samples per session, including 10 Brucella-positive targets (4 samples of 10+ and 3 samples each 7+ and 12+) from which two different blind samples were randomly chosen daily. The 40 negative samples included 9 samples of 10-, 8 samples each of 7- and 12-, and 15 E. coli samples. To advance to the next phase, at least 7 rats were required to hit at least 9 positives while committing no more than 10 false positives for two consecutive days.
Phase V: Discrimination 100 samples (19 sessions). Rats were presented with 100 total samples per session: 10 positive (4 samples of 10+ and 3 samples each of 7+ and 12+; two different blinds chosen randomly each day) and 90 negative samples (two types from among 10-, 7-, 12-, and E. coli randomly selected each day to have 22 samples while the two remaining types each had 23 samples). After at least 7 rats correctly indicated at least 9 positive samples and committed no more than 12 false positives for at least 4 out of 6 consecutive sessions, the rats advanced to test.
Test (1 session). This test ensured extraneous cues inadvertently introduced during sample preparation were not guiding rat detection by enlisting a naïve researcher to prepare all samples. The new sample preparer was only instructed how to prepare the samples but not in which order. This effectively randomized potential cues that might have resulted from the training sample preparer or method of preparation during training.
Exp 2: Can rats generalize to younger cultures?
Experiment 2 began the day immediately following Experiment 1 Test.
Baseline (2 sessions). To ensure the Experiment 1 Test did not disrupt baseline scent detection performance, two sessions identical to Experiment 1, Phase V Discrimination Training were conducted. At least 7 rats correctly indicated at least 9 positive samples and committed no more than 12 false positives during these two sessions.
Novel samples (2 sessions). A novel Brucella-positive sample was introduced to the rats while blood agar media (heat inactivated) was also re-introduced, this time as an un-spiked Brucella-negative sample. The novel Brucella-positive samples were prepared as before, except they contained Brucella culture harvested after just 5 days of incubation (5+). As with Experiment 1, a total of 10 targets were presented, including 3 novel target samples (5+, with one blind sample), 3 samples of the previously trained 10+ (with one blind sample) and 2 samples of each remaining familiar target (7+ and 12+). Among the 90 non-target samples, rats encountered 18 samples of each familiar type (7-, 10-, 12-, and E. coli) plus 18 samples of un-spiked blood agar (5BA-). Although rats had previously encountered Brucella-positive samples with blood agar media during Phase I of Experiment 1, Brucella-negative blood agar was novel to the rats and served as control for potential novelty bias that might otherwise confound interpretation of 5+ detection accuracy.
Exp 3: Can rats detect cultured Brucella within field-relevant media?
Each evaluation session included 100 total samples comprised of 2 samples each of familiar Brucella-positive targets (7+, 10+, 12+), plus one blind sample randomly selected from among these types. Additionally, 3 fecal samples (including 1 blind sample) were spiked with Brucella-positive culture solution (Feces+) to serve as novel targets. Among the 90 non-target samples, rats encountered 18 samples of each familiar control type (7-, 10-, 12-, and E. coli) plus 18 novel fecal samples spiked with the Brucella-negative control solution (Feces-). Feces- samples not only controlled for any inherent bias for or against novelty but also any potential bias for the fecal samples themselves. Sessions were conducted for five consecutive days.