*Article* **Canine Olfactory Detection of a Non-Systemic Phytobacterial Citrus Pathogen of International Quarantine Significance**

**Timothy Gottwald 1,\*, Gavin Poole 1,2, Earl Taylor 1, Weiqi Luo 1,2, Drew Posny 1,2, Scott Adkins 1, William Schneider <sup>3</sup> and Neil McRoberts <sup>4</sup>**


Received: 19 August 2020; Accepted: 27 October 2020; Published: 9 November 2020

**Abstract:** For millennia humans have benefitted from application of the acute canine sense of smell to hunt, track and find targets of importance. In this report, canines were evaluated for their ability to detect the severe exotic phytobacterial arboreal pathogen *Xanthomonas citri* pv. *citri* (Xcc), which is the causal agent of Asiatic citrus canker (Acc). Since Xcc causes only local lesions, infections are non-systemic, limiting the use of serological and molecular diagnostic tools for field-level detection. This necessitates reliance on human visual surveys for Acc symptoms, which is highly inefficient at low disease incidence, and thus for early detection. In simulated orchards the overall combined performance metrics for a pair of canines were 0.9856, 0.9974, 0.9257 and 0.9970, for sensitivity, specificity, precision, and accuracy, respectively, with 1–2 s/tree detection time. Detection of trace Xcc infections on commercial packinghouse fruit resulted in 0.7313, 0.9947, 0.8750, and 0.9821 for the same performance metrics across a range of cartons with 0–10% Xcc-infected fruit despite the noisy, hot and potentially distracting environment. In orchards, the sensitivity of canines increased with lesion incidence, whereas the specificity and overall accuracy was >0.99 across all incidence levels; i.e., false positive rates were uniformly low. Canines also alerted to a range of 1–12-week-old infections with equal accuracy. When trained to either Xcc-infected trees or Xcc axenic cultures, canines inherently detected the homologous and heterologous targets, suggesting they can detect Xcc directly rather than only volatiles produced by the host following infection. Canines were able to detect the Xcc scent signature at very low concentrations (10,000× less than 1 bacterial cell per sample), which implies that the scent signature is composed of bacterial cell volatile organic compound constituents or exudates that occur at concentrations many fold that of the bacterial cells. The results imply that canines can be trained as viable early detectors of Xcc and deployed across citrus orchards, packinghouses, and nurseries.

**Keywords:** early detection; Asiatic citrus canker; latent class; information theory; field diagnostic; scent signature; direct assay; deployment

#### **1. Introduction**

Asiatic citrus canker (ACC) is a fruit, foliar, and twig lesion disease that has significant international, national, and local quarantine implications and has been the focus of multiple extensive eradication

programs. The disease is caused by the bacterium *Xanthomonas citri* pv. *citri* (Xcc). When pathogen infection is severe, leaf drop and up to 69% crop loss can occur due to fruit drop [1]. The bacteria disperse via meteorological events ranging from gentle rain to tropical storms and hurricanes. The more severe the meteorological event, the more effective the inoculum dispersal [2,3]. Hurricanes and tropical storms have been associated with long-range dissemination as well as local increase of the bacterium [2,4–6]. Xcc inoculum can also be transmitted mechanically by machinery and humans if the foliage is wet. Inoculum begins to exude from Xcc lesions within 1–5 min of becoming wet [4,6–8]. The maximum concentration of bacteria is exuded within the first 1–2 h period following the beginning of the rainfall event or wetting of the lesion, although inoculum is produced continuously at a lower concentration for the duration of the wetting period or storm [2–4]. Infection takes place when inoculum-laden water passes through stomata of foliage, fruit, or green twigs. Infection can occur through wounds as well, and it is highly exacerbated by the Asian citrus leafminer (*Phyllocnistis citrella* Stainton), whose feeding galleries create a labyrinth of wounds that expose susceptible leaf mesophyll tissues to splashed inoculum, greatly increasing the probability of infection by Xcc [9,10].

After infection occurs, bacteria propagate within the plant tissues, eventually forming small blister-like protrusions that become visible 5–7 days post-infection with close examination augmented by 10× magnification. Within 12–14 days, these protrusions erupt through the epidermis, forming 1–2 mm light brown erumpent lesions. As the lesions age, they darken to brown and develop water-soaked margins with a surrounding chlorotic halo [7]. The lesion center develops a raised spongy and corky appearance on adaxial and abaxial foliar surfaces. Older lesions can reach 1 cm in diameter and can coalesce to form mass infections.

Pathogen detection in the field is almost exclusively by human visual inspection. Confirmation of Xcc infection can be accomplished by serological or Polymerase Chain Reaction (PCR) assays, both of which require infected tissue. Therefore, such assays are of minimal use for a field survey of a pathogen that causes only local lesions, i.e., non-systemic infections. Visual inspection is tedious, labor intensive, and highly variable due to the elusiveness of Xcc lesions, especially when high in the canopy. Since mature citrus trees can have in excess of 100,000 leaves, finding initial infections visually in such a large canopy is challenging and uncertain. As a consequence, multiple infection cycles must occur before symptoms are of sufficient prevalence to permit visual detection. During a major epidemic in Florida (1996–2006), the detection time for Xcc by trained regulatory inspectors averaged 106 days post-infection [11]. In one regulatory exercise, 14 teams of two inspectors per team examined the same infected orchard in succession. No two teams found the same infected trees, each successive team found new infected trees previously undetected by prior teams, and no team found all known infections. [Riley, unpublished results]. Since it is highly improbable that all infected trees in an orchard will be found by visual inspection, the true incidence of infection cannot be determined, and thus, calculating the detection accuracy of visual inspection is not possible.

The unpromising results for human visual detection suggest the need for the deployment of a detection technology with much better performance. Genetic analyses indicate that modern-day canines (*Canis lupus familiaris*) were the first domesticated animal, arising from two separate wolf (*Canis lupus*) populations ≈ 15,000 years ago [12–17]. The mammalian (including canids) olfactory system is antediluvian, having evolved from early chemotactic receptors during the Precambrian over 600 million years ago. Acute olfaction evolved to enhance finding food, mates, detecting danger, avoiding predators, etc. [18]. An overarching advantage to the human use of canine olfaction is that all canids non-destructively interrogate their environment for a scent signature of interest. This scent signature is composed of a specific volatile organic compound (VOC) or complex of VOCs. In contrast, commonly used molecular or biochemical assays often require destructively sampling a small proportion of the host or environment and are specific and do not detect complex VOC composites.

Wolves employ acute vision and hearing when prey is in close proximity. However, when tracking prey, often over great distances, wolves resort to olfactory cues to locate widely dispersed and often low-density prey [19], as is true for domestic canines [20]. To find rare targets of human interest, we exploit the tracking expertise and acute sense of smell handed down to modern canines from wolves and their evolutionary predecessors. Nonetheless, scent detection is pervasive through the animal kingdom. Other vertebrates and invertebrate animals have been explored recently in the burgeoning science of scent detection research with diverse practical applications [21]. However, canines dominate as olfactory detector tools because of their unsurpassed domestication and easily deployed tracking skills.

Canines are used broadly to detect and locate a wide array of organic and inorganic odors, (e.g., explosives, drugs, accelerants, pollutants, toxins, pesticides), tracking humans and game animals, finding cadavers, and matching the scent of criminal perpetrators to crime scenes [22]. Canines are also proficient at medical human disease detection, especially cancer (malignant melanoma, non small-cell, small-cell lung, breast, prostrate, bladder, ovarian, and colorectal cancers) often with equivalent or superior sensitivity to current medical assays and can detect epileptic seizures prior to onset [22–25].

Relative to agricultural diseases, canines were recently trained extensively to detect *Candidatus* Liberibacter asiaticus (CLas), which is the causal pathogen of citrus Huanglongbing with greater than 99%, 96%, and 92% accuracy in field trials, commercial citrus orchards and citrus in residential properties, respectively [26]. In the same study, canines detected CLas in the Asian citrus psyllid, *Diaphorina citri,* that vectors the bacteria and in bacterial co-cultures devoid of plant or animal host cells. Additionally, canines have been trained to detect plum pox virus in commercial peach (*Prunus persica*) orchards [Gottwald unpublished] and the fungal pathogen *Ra*ff*aelea lauricola*, which is the cause of Laurel wilt disease of avocado (*Persea americana*) [27]. Thus, canines have been demonstrated to detect bacterial, viral, and fungal pathogens in plant hosts.

Canine detection of the plant pathogens as indicated above represents a novel extension of previous detection targets. For drugs, explosives, pollutants, etc., canines are trained to detect specific VOC scent signatures of the target compounds. For biologicals such as plants, plant parts, humans, and other animals, the canines are trained to recognize the volatilome, e.g., unique VOC complex, emitted by the target. Conversely, when trained to detect pathogenic organisms, the canines must detect one organism (bacterium, virus, or fungus) within another organism, i.e., the host plant or animal. The three plant pathogens discussed above, CLas, plum pox virus and *R. lauricola,* are all systemic within the vascular system of the host, and therefore, they can be completely or incompletely distributed within the plant as the infection progresses. In contrast, Xcc is non-systemic and causes only local lesions, which can range in incidence from a single lesion to thousands per tree and can be rare, highly aggregated, or diffused within the host.

In this study, we demonstrate proof of concept that canines can discriminate and detect Xcc and can be trained as viable detectors of the pathogen in agricultural environments. Here we document the detection of Xcc, an exotic bacterial pathogen by use of canine olfactory surveillance in orchards and packinghouses. Finally, we determine that canines can directly detect the target bacterium *in planta* and *in vitro*.

#### **2. Materials and Methods**

#### *2.1. Initial Sensitization Training of Canines for Xcc Scent Signature Recognition*

Eleven canines of various breeds were utilized for the various trials throughout the course of the study (Table 1).


**Table 1.** Canines trained to detect *Xanthomonas citri* pv.*citri* (Xcc) scent signature and utilized in this study.

At the onset of this study, Xcc was considered a quarantine pathogen in Florida [3,6,7,11], and neither the bacteria nor infected plant material could be transported to new locations for experimental purposes. Therefore, we used a Scent Transfer Unit (STU) (Model STU-100, Tolhurst Big "T" Enterprises, Lockport, New York, NY, USA, 14094) vacuum device commonly used in canine detection work to draw air at a constant rate (≈300 L/min) to collect volatile samples containing the "scent signature" of the target, in this case Xcc-infected plants, and deposit the scent onto a cotton "scent pad" [28]. Since the scent pad does not contain the pathogen (Xcc is only splash dispersed, generally by rain and not dispersed in dry air) nor any potentially infected plant material, it can be safely transported to non-endemic areas for testing. Scent pad samples were collected from non-infected and Xcc-infected Ruby Red grapefruit (*Citrus paradisii*) trees using the STU. A sterile 12.7 × 22.9 cm (5 × 9 inch) cotton pad was placed into the STU for 10 min by setting the STU within the canopy of a non-infected or Xcc-infected tree 5–10 cm from leaves, fruit and branches while air was pulled through the pad. Following each 10 min sampling period, scent pads were aseptically removed, placed in a volatile-proof plastic bag (K-pak), and heat sealed. The STU was disinfected with ethyl alcohol between each use. Multiple heat-sealed samples from infected and non-infected trees were segregated, placed in zip lock plastic bags, and stored at −20 ◦C until transported (Figure 1A–C).

Canine detection companies are for-profit ventures, and therefore, they are often reticent to undertake non-profitable and time-consuming basic research. Fortuitously, a canine detection company in California was willing to dedicate two canines to do an initial testing of canine detection of Xcc. In preliminary studies, it was not feasible to bring canines to the Xcc-infected sites due to quarantine constraints. Therefore, scent pad samples were collected in South Florida, sealed in a styrofoam shipping box with ice packs, and transported overnight to the California training facility.

One Belgian Malinois (Kimba) and one German Shepard (Tank), previously trained for criminology scent detection, were trained to the Xcc scent signature by imprinting, during which the canine was introduced to the target and other neutral scents. Relative to scent signature training, when the canine becomes interested in/reacts to the correct target, the experience is immediately encouraged with verbal and play rewards [18]. In our case, canines were imprinted to the scent signature presented on Xcc-infected scent pads to discriminate the scent from scent pads without the scent signature. To take advantage of the candid passion to search and track, a mixed population of mostly non-infected with a few Xcc-infected scent pad samples were arrayed in a series of rows of metal cans placed outdoors on metal stands (Figure 1D). Correctly alerting on the Xcc-infected pad placed in the bottom of a can resulted in the canine receiving verbal praise and a few moments of play with the handler and a ball

or "kong" (hard rubber ball on a short rope). Initial training used non-frozen scent pads. Canines subsequently detected both non-frozen and frozen scent pads from Xcc-infected trees. Thereafter, scent pads were stored frozen until use in training and assessment.

**Figure 1.** Initial training of canines for detection of the phytobacterial arboreal pathogen *Xanthomonas citri* pv. *citri* (Xcc), the causal agent of Asiatic citrus canker (Acc). (**A**) Xcc-infected red grapefruit fruit and (**B**) foliage. (**C**) Scent Transfer Unit (STU) used to draw in Xcc volatiles and deposit on cotton scent collection pad. (**D**) Canine "Kimba" training by interrogating a row of metal cans containing Xcc-positive and negative scent pads. (**E**,**F**) Detector canine NDD-1 and NDD-3 alerting on boxes containing Xcc-infected foliage at the USDA, APHIS, National Detector Dog Training Center.

Following initial scent training, a series of preliminary studies were conducted in which canines interrogated scent pads exposed to Xcc-infected red grapefruit for periods of 1, 5, 10, or 30 min versus pads exposed to non-infected trees for similar times, to determine if there was a lower threshold of exposure necessary to create scent pad training materials. Additionally, once the scent pads were removed from the heat-sealed plastic bags, the canines interrogated them every few days to determine the temporal viability of the scent pads as a training tool.

## *2.2. Training Canines for Detection of Xcc-Infected Plant Material*

During 2004–5, in collaboration with the U.S. Department of Agriculture, Animal and Plant Health Inspection Service (USDA, APHIS), National Detector Dog Training Center (NDDTC) in Orlando, Florida, we trained canines to detect Xcc-infected Duncan grapefruit seedlings. At the USDA, ARS laboratory in Fort Pierce, Florida, quarantine greenhouse, seedlings were infected by placing drops of Xcc inoculum, <sup>≈</sup>10<sup>6</sup> to 10<sup>8</sup> cfu/mL prepared from pure cultures on young leaves and using a sterile needle to wound the leaf lamina, causing 5–10 wounds/leaf of five leaves/seedling. Seedlings were incubated in the greenhouse for ≈2–3 weeks until symptomatic and then transported under regulatory permit in sealed containers to the NDDTC in Orlando along with non-infected seedlings as training

materials where they were maintained until use. At the NDDTC, two beagles (NDD-1 and NDD-2), one German pointer (NDD-3), and one Labrador retriever (NDD-4) were trained to recognize the Xcc scent signature by a similar imprinting/reward method as described above. Scent detection proficiency was tested by placing excised branches or whole potted Xcc-infected or non-infected trees in closed cardboard boxes and arraying the boxes on the floor of the training facility (Figure 1E,F). In a second NDDTC study, the trained Labrador was transported to an Xcc-infected orchard in Indian River County, Florida. To avoid disease quarantine issues, leaves were collected from Xcc-infected and non-infected trees and placed into separate sealed plastic pipe containers with holes drilled in the sides to allow the volatiles but not infected plant material to escape for canine interrogation. The canine interrogated the containers, which were placed on the ground at the edge of the orchard. When the trial ended, the infected and non-infected plant materials were removed from the containers, discarded in the infected field, and the containers were disinfected. Due to quarantine issues, intermittent and uncoordinated testing by NDDTC personnel in absence of the authors, and uncoordinated replacement of canines assigned to the study, data were not consistently taken with the exception of an overall assessment of performance by NDDTC personnel.

In all subsequent studies, we collaborated with three commercially certified canine detector companies who specialize in trained detector canines for military, police and domestic clientele, for the detection of an array of targets including explosives, drugs, bed bugs, etc. This was important because professional canine handlers are themselves trained not to give any voluntary or involuntary cues to the canines while testing their performance. A single handler was used for all replications of most studies to avoid bias or variation due to handler. These companies trained nine canines (7 were used in this study) for the detection of Xcc via imprinting on scent pads from Xcc-infected trees or Xcc-infected grapefruit seedlings until canines gained proficiency at differentiating Xcc-infected from non-infected samples (Table 1).
