Biological organisms as volatile compound detectors: A review

Forensic Science International 232 (2013) 92–103

Contents lists available at SciVerse ScienceDirect

Forensic Science International journal homepage: www.elsevier.com/locate/forsciint

Review article

Biological organisms as volatile compound detectors: A review Olivia Leitch a,b,*, Alisha Anderson b, K. Paul Kirkbride c, Chris Lennard a a

National Centre for Forensic Studies, University of Canberra, Canberra, ACT 2617, Australia CSIRO Division of Ecosystem Sciences and Food Futures Flagship, Canberra, ACT 2601, Australia c School of Chemical and Physical Sciences, Flinders University, Bedford Park, SA, 5042, Australia b

A R T I C L E I N F O

A B S T R A C T

Article history: Received 10 October 2012 Received in revised form 30 June 2013 Accepted 5 July 2013 Available online

The detection and identification of volatile compounds is essential to the successful undertaking of numerous forensic analyses. Biological olfactory systems possess the extraordinary ability to not only detect many thousands of distinct volatile compounds (odors) but also to discriminate between them. Whole-organism biological sensors, such as detection canines, have been employed in forensic science as volatile compound detectors for many years. A variety of insects including bees, wasps, and moths, which have also been shown to detect volatile compounds of forensic significance, have been investigated for their potential application in field-based detection systems. While the fundamental aim for many developers of portable instruments is to replicate the remarkable ability of biological olfactory systems, such analytical equipment is yet to possess the detection and discriminatory powers achieved by biological sensors. Recent literature reveals an increasing interest in olfactory receptors – the biological components that impart olfactory ability – for detecting volatile compounds associated with forensically significant substances such as explosives and illicit drugs. This paper reviews the literature regarding the current, and potential future, use of biological organisms as sensors for forensic science applications. ß 2013 Elsevier Ireland Ltd. All rights reserved.

Keywords: Volatile compound detection Biological sensors Detection canines

Contents 1. 2.

3.

4. 5. 6.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . Vertebrates. . . . . . . . . . . . . . . . . . . . . . . . . Canines . . . . . . . . . . . . . . . . . . . . . . 2.1. Rats . . . . . . . . . . . . . . . . . . . . . . . . . 2.2. Invertebrates . . . . . . . . . . . . . . . . . . . . . . . Wasps . . . . . . . . . . . . . . . . . . . . . . . 3.1. Honeybees. . . . . . . . . . . . . . . . . . . . 3.2. Moths . . . . . . . . . . . . . . . . . . . . . . . 3.3. Detection devices employing live insects . Detection using isolated olfactory organs. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

1. Introduction Volatile compounds (VCs) are significant constituents within a range of substances commonly encountered in forensic science

* Corresponding author at: P.O. Box 1700, Canberra, ACT 2601, Australia. Tel.: +61 400 183 618. E-mail address: [email protected] (O. Leitch). 0379-0738/$ – see front matter ß 2013 Elsevier Ireland Ltd. All rights reserved. http://dx.doi.org/10.1016/j.forsciint.2013.07.004

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . .

92 93 93 95 96 96 98 98 99 100 101 101

(e.g. explosives, illicit and abused drugs, ignitable liquids, etc.). The detection of VCs may be exploited for a variety of purposes, such as: security screening (e.g. detection of illicit drugs, explosives and/ or other prohibited items in baggage), assisting in investigator efforts (e.g. locating clandestine graves, ignitable liquid residues at fire scenes, etc.), or chemical analyses (e.g. analysing the headspace of fire scene debris for the presence of ignitable liquid residues). In many instances, the primary objective is to detect the active or most abundant VC components (e.g. the active drug in a drug

O. Leitch et al. / Forensic Science International 232 (2013) 92–103

sample); however, it is often the detection of by-products or other manufacturing artifacts on which the success of particular techniques rely. The extraordinary ability of biological organisms (i.e. animals) to detect, recognize, locate, and discriminate between target materials via VCs is well documented, having been researched thoroughly in various species [1–5]. Whole-organism biological sensors have been employed as VC detectors in various ways for numerous years [2,4,6–9]. For example, canaries were the original chemical sensors utilized for detecting carbon dioxide in mines [10]. Additionally, the detection canine (the most well-known and widely used biological sensor) is successfully employed for rapidly searching large scenes for the presence of target VCs in a variety of forensic areas [7,9,11–14]. There is also significant interest in the potential employment of other biological organisms, such as rats [15,16], wasps [17,18], and bees [6,19]. The advantages of biological organisms as VC detectors have been reported by many authors [11,20–22]; however, their limitations are also well documented [6,7,18,23,24]. In comparison to more conventional physico-chemical detectors (e.g. flame ionization or surface acoustic wave detectors) the actual operational employment of whole organisms as VC sensors, with the exception of canines, is relatively low. However, because of the significant advantages of biological organisms, such as their sensitivity and selectivity, there has been continued interest in their wider application for forensic purposes. This paper reviews the published literature regarding the wider applications of whole organisms and biologically derived VC sensors and highlights their current or potential future use in forensic fields. 2. Vertebrates Chemical cues (odors) provide information about food, mates, offspring, predators, prey, and pathogens; they are also employed for communication purposes [21,22]. Therefore, well developed olfactory abilities are essential for survival for the majority of animals [20]. The olfactory systems of vertebrates are highly sophisticated, imparting such discriminatory capabilities that thousands of VCs are perceived as being distinct odors [21]. Highly sensitive and selective olfactory abilities have been demonstrated in a variety of vertebrates [25–27]; however, their use as chemical sensors has been limited primarily to canines, potentially due to a lack of knowledge regarding their trainability as well as limitations with their physical employment. Only recently has there been interest in other vertebrates such as rats [15,16,28]. 2.1. Canines The most well-known and widely employed biological sensor is the common canine, Canis lupus var. familiaris. Reported as being significantly more sensitive and selective than human olfaction,

93

the canine’s extraordinary sense of smell has been utilized in numerous areas of law enforcement for the detection of VCs of forensic interest. For example, detection canines have been used for detecting illicit drugs, land mines (Fig. 1a) [14], guns, ignitable liquid residues [12,24], explosives [11,23], clandestine burials [13,29], and controlled goods such as illegally imported food or ivory [13,30]. Canines have also been used for tracking purposes, such as for locating criminals, missing persons and disaster victims, as well as for providing a visual deterrent for potential illegal activities [7,9,14]. As a result, the detection canine and its handler offer significant contributions to the real-time detection of VCs of interest. Training detection canines typically involves: imprinting the odors the canine is supposed to detect; using representative odors that the canine should not alert to but which are likely to be encountered in their working environment; and performing regular ‘‘refresher’’ training to ensure on-going reliability of results [31]. Indications as to the presence of the target substance include either passive or active behavioral alerts. Passive alerts involve the canine sitting near the odor source, whereas active alerts (also referred to as aggressive alerts) involve the canine scratching at the odor source. Active alerts are considered by some handlers to provide greater ‘‘pin-point’’ accuracy and are used typically for the detection of illicit drugs. However, such alerts are highly dangerous if explosives and/or landmines are being targeted; therefore passive alerts are more appropriate for such circumstances. As reported in the literature, the primary advantages associated with the use of detection canines include their agility [30], their ability to rapidly and thoroughly search large areas [32,33], their olfactory sensitivity that allow them to detect and discriminate between target and non-target substances even at low concentrations [24,31], and their scent-to-source capabilities that allow them to pinpoint areas of highest concentration [9,11,31]. Due to these advantages, the deployment of detection canines is unlikely to decrease in favor of field portable instrumentation in the near future. However, a study of the literature also reveals several limitations. For example, successful detections are highly variable from canine to canine [7,29], and depend significantly on their training [11,30], hormonal changes, possible infections, illness [9,11], and environmental conditions [32,34]. Additionally, detection canines are expensive to train [9,33,35] and require ongoing maintenance and refresher training [14]. They can also suffer from olfactory fatigue, which is the loss of sensitivity and selectivity due to repeated and prolonged exposure to the same odor [29,30,34]. While the deployment of detection canines is widely accepted in the forensic and law enforcement communities, questions regarding their accuracy, reliability, and validity have been raised and are regularly debated [9,36,37]. Research regarding the accuracy of individual detection canines in a range of detection scenarios can illustrate great variability, with some authors reporting successful detections ranging between 40% and 100%

Fig. 1. Vertebrate volatile compound detectors. (a) Detection canine in training. (b) African pouch rat trained for landmine detection. Images reproduced with permission from APOPO [57].

94

O. Leitch et al. / Forensic Science International 232 (2013) 92–103

[7,24,29,32]. However, on average, accuracy rates of studied canines are well over 80%. For example, of the 39 accelerant detection canines investigated by Tindall and Lothridge [24], 34 exhibited accuracy rates of 80%, with 21 canines achieving 100% correct detections with mixed scent matrices. The results of Osterhelweg et al. [13] showed that tested cadaver dogs positively alerted to decomposition 94–98% of the time, while Harvey and Harvey [7] reported an overall success rate of 96% for mature, welltrained canines for human scent trailing in simulated operational scenarios. For explosives detection canines, Gazit and Terkel [38] found detection rates of 87.8% and 93.8% for uncontrolled outdoor searches conducted in low and high lighting conditions respectively. The variability observed across some of the individual canines illustrates the differences that can exist from canine to canine and may be the result of any of the previously mentioned factors on which successful detections depend (i.e. training, environmental conditions, age, etc.). Nevertheless, as outlined by Furton and Myers [11], the application of proficiency tests and certification standards [39] arguably maintain proficiency standards equal, if not superior, to those for instrumental methods. Ultimately, while there is limited scientific peer-reviewed data available across the range of applications for which canines are employed, the research completed thus far highlights the extraordinary abilities of well-trained canine/handler teams. Significant enhancement of a detection canine’s performance was found to occur when a ‘‘performance check’’ was completed prior to the canine working [9,35,40]. As described by Schoon [35], a ‘‘performance check’’ involves assessing the canine’s ability and motivation to perform their detection duties. It was found that the level of performance directly influenced the reliability of the results; there were more correct alerts and less false positive alerts when the canines were motivated. Therefore, such an assessment allows those canines not sufficiently able or willing to work to be identified and removed from that particular task. Essentially, it is a method for eliminating unnecessary mistakes by unmotivated canines. However, while such ‘‘performance checks’’ may be highly valuable prior to canine deployment, the time required for implementation and the consequences of reducing the number of operational canines means that checks will not be practical for all forensic and law enforcement situations. Laboratory studies determining the sensitivity of canine olfaction have focused on both individual compounds [41–47] and whole samples [24,48–50]. For example, Ashton et al. [41] and Moulton et al. [42] reported thresholds for eight separate fatty acids (not specifically from a forensic context), while Becker et al. [43] examined canine olfactory sensitivity to individual explosive compounds. Regarding whole samples, Kurz et al. [24,48,49] and Tindall and Lothridge [24] examined the ability of accelerant detection canines to detect small volumes of various ignitable liquids. Studies demonstrating the limits of detection of canine olfaction (i.e. where approximately 50% of tested canines alert) have indicated thresholds ranging from parts-per-million (ppm) to part-per-trillion (ppt) [45,46,51,52]. For example, Johnston [45] reports canine sensitivity to the compounds methyl benzoate, cyclohexanone and nitroglycerine to be in the 10 ppb range, with much greater sensitivity to 2,3-dimethyldinitrobutane (DMNB; a marker chemical added to plastic and sheet explosives) at 500 ppt. Results obtained by Lorenzo et al. [46] showed that 50% of canines alerted to 2,4-dinitrotoluene (2,4DNT) at 100 ppm. One particular study by Macais et al. [44] directly compared the sensitivity of detection canines and solid phase microextraction-ion mobility spectrometry to piperonal (a predominant odor of MDMA) and found that the limits of detection were 1 ng and 2 ng, respectively. This research reveals that canine sensitivity is within the same order of magnitude of

analytical instrumentation, thus demonstrating the advantage of using detection canines in field based examinations (e.g. comparable detection thresholds with increased portability). However, odors of forensically significant substances are not typically single chemicals but consist of various chemical constituents that make up an odor signature. It is this odor signature that is thought to be utilized by detection canines to alert to target substances. Consequently, studies of sensitivity must take into account the odor signature. From an evaluation of canine sensitivity for differing quantities of several ignitable liquids, Kurz et al. [48] reported successful detections for as little as 0.01 mL. This finding is supported by Tindall and Lothridge [24], who in a study of five detection canines observed that all had accurate detection abilities for 0.01 mL, while four of the five could detect gasoline at a quantity of 0.005 mL. The ability to detect the presence of 0.01 mL of an ignitable liquid was also reported as being equal, or superior, to the traditional dynamic headspace sampling with subsequent gas chromatography-flame ionization detector (GC-FID) analysis. A later study conducted by Kurz et al. [49] placed more emphasis on the ability of detection canines to reliably alert to differing quantities of a commonly encountered ignitable liquid (50% evaporated gasoline) in the presence of distracting pyrolysis products and/or background hydrocarbons. Accurate detections were reported for nearly all of the canines at quantities of 5 mL. However the success rate reduced significantly when amounts of 0.2–0.05 mL were employed; less than half of the detection canines studied alerted to samples doped with these quantities. A significant number of alerts to samples containing only background compounds were also observed. The background found to cause the greatest amount of interference with respect to successful canine detections was burnt carpet with foam backing. The ability to differentiate between target and non-target substances is very important for any detection method. Studies by Kurz et al. [48], Kurz et al. [49], and Lasseter et al. [32] indicated that detection canines have the ability to differentiate between target and non-target substances. However significant numbers of false positives have also been observed [32,49]. Very little is known or understood regarding the specific combination or class of VCs responsible for triggering an alert and the effects of closely related substances on detection [11,14,36,46,51,52]. Studies examining the particular VCs emitted from various forensic evidence types (and thus are available for detection) such as explosives [46,51,52], human decomposition [46,53,54], and illicit drugs [46,55], have been completed. While the research is limited, several studies have begun the process of identifying key VCs to which detection canines respond. For example, Williams et al. [52], Harper et al. [51] and Lorenzo et al. [46] combined analytical techniques with olfactory testing of canines to determine the dominant and active (i.e. elicit canine alerts) headspace chemicals of various explosives, illicit drugs and human decomposition tissues. Interestingly, while the substances studied emit complex mixtures of VCs, only a few of the individual compounds within that mixture have been shown to elicit canine responses comparable to those obtained for the whole target substance. While all three studies identified 2-ethyl-1hexanol and cyclohexanone as the dominant headspace chemicals of C-4, field trials with certified explosive detection canines revealed 2-ethyl-1-hexanol as the key compound to elicit positive responses, with up to 70% of tested canines alerting [46]. Limited responses were obtained for cyclohexanone [46,51]. 2,4-DNT was also identified as the active odor for TNT preparations [46,51,52]. Additionally, when examining various concentrations and ratios of multiple compounds from MDMA and methamphetamine, Lorenzo et al. [46] found that none of the tested canines altered to the mixtures and ratios investigated; the majority of canines (83%) alerted to the individual compound piperonal. These studies begin to explain how canines recognize the various substances they are

O. Leitch et al. / Forensic Science International 232 (2013) 92–103

trained to detect. What these studies also illustrate, however, is that the signature chemicals used for detection are not the same for all canines involved in the studies. Further research in this particular area will be essential for understanding canine olfaction, interpreting results and, importantly, could help optimize canine performance. It is acknowledged by various authors that the effectiveness of a detection canine is dependent not only on the level of training, environmental conditions, maintenance, and temperament, but also on the ability of the handler [24,29,32,35,49,56]. Field studies conducted by Lasseter et al. [32] found that 10% of canine alerts went unrecognized by their handler. Similar results were also reported by Komar [29]. However, because the researchers were aware of the locations of the target substances, they could feasibly have interpreted the canine’s behavior as a positive alert when there actually was not one. Lasseter et al. [32] reported handlermissed positive alerts that, despite the canine exhibiting behavior not typical of trained positive alerts, were nevertheless considered by the authors’ as positive signals. It should be expected that handlers possess greater awareness of their particular canine’s behavioral responses due to the extensive training completed together; therefore, it is possible that reported detection errors attributed to the handler are the result of erroneous interpretations by the authors, not the handlers. In addition to unrecognized alerts, a number of positive detections and false alerts have also been attributed to the handler. Lasseter et al. [32] found that 5% of positive detections were actually the result of the handler themselves locating the target substances. Kurz et al. [49] found that, because the handler knew the target substances were present, they would consciously or unconsciously encourage the canine to alert despite them not having detected anything. Similar reports supporting these occurrences were also made by Lasseter et al. [32]. Ultimately, it has been speculated that errors attributed to the handler are a result of a lack of training, confidence, and/or practical experience. 2.2. Rats Rats are model organisms for the study of mammalian olfaction with numerous studies having been completed [22,58]. As a result, knowledge of their olfactory abilities and underlying olfactory mechanisms is quite extensive. However, research on their application as biological sensors is limited. Nolan et al. [59] first researched the potential use of rats as biological sensors for the detection of explosives in the late 70s. Initial training methods involved electric brain stimulation and conditioning chambers [59,60]. Researchers claimed they could obtain up to eight hours of continuous attention (i.e. they could work without breaks) from the rats with this type of training [60]. Modern training methods are both laboratory- and simulated field-based and use a combination of clicker signals and food rewards as motivation. A detailed description of current training methods is provided by Poling et al. [16]. Alerts given when rats detect a target substance involve either natural scratching and biting behaviors or a trained rearing response (i.e. the rat raises onto its back legs) [15,16,31]. While active scratching alerts are generally not employed for mine detection canines due to associated dangers, the small size and weight of detection rats means they are unlikely to detonate the mines [61]. Literature regarding the use of detection rats indicates that they are effective animals for VC detection purposes. Reported advantages include their size, weight, rapid training times, relatively inexpensive maintenance, high breeding rates, potential for semi-automated training, ability to maintain sizable colonies on-site, and they are not dependent on single hander-biological sensor combinations [15,16,62,63]. Additionally, Verhagen et al.

95

[28] report that novice trainers can easily pick up the requirements for the training and handling of detection rats. Due to these advantages, detection rats appear to have great potential for deployment in a wide range of detection areas. However, they also have several limitations including short working times (20– 40 min, which differs from [60] because of modern training and employment methods), and that, similar to the detection canine, successful detections are variable between individual rats, depend heavily on environmental conditions, and their behavior appears to be context specific [15,16,28]. With exception to their use by the Anti-Persoonsmijen Ontmijnene Product Ontwikkeling (APOPO), rats are not widely used operationally as biological sensors (they are a novel unvalidated detection technique that lack social acceptance). The APOPO is a non-profit organization dedicated to the detection and safe removal of landmines that has been deploying African giant pouch rats (Cricetomys gambianus) (Fig. 1b) for landmine detection since 2003. Despite this, literature regarding their accuracy, reliability, sensitivity and selectivity is severely lacking. Otto et al. [15] reported consistent success rates of over 90% for Sprague-Dawley rats; however, these results were obtained after a series of progressive training stages through which the rats only advanced if 100% accuracy was obtained. Verhagen et al. [28] also reported high accuracy rates of 80% in dry conditions, although this dropped to 80% in wet conditions. This study also revealed significant variability between individual rats, ranging from 67% to 100%, and that success depended heavily on the evaluation area (i.e. the area around the landmine in which a rat must alert for it to be a positive or false-positive alert), time of day and ambient temperature [28]. Additionally, Poling et al. [16] reported that the high accuracy of rats in simulated field-based training is not always replicated in operational searches. While reports regarding accuracy of detection show less overall variability than those for detection canines, it should be noted that the number of studies completed, and the number of rats examined, is significantly lower than those of detection canines. As a result, conclusions regarding the true accuracy of rats and comparisons with those of detection canines are difficult to make. While there is little scientific data regarding the performance of detection rats, reliability is arguably ensured with the compulsory accreditation testing and ongoing training of operational detection rats deployed by APOPO [16]. Poling et al. [16] reported that APOPO detection rats must achieve a standard of 100% correct alters to training aids and less than 5% false positives in training before they are advanced to operational use. However, information regarding the number of rats that successfully completed training from those that initially commenced the program was not provided nor was the range of success rates observed for rats that did not advance. While research has examined the ability of rats to detect whole substances, very little information regarding their ability to detect target odors at varying concentrations has been published. Marshall et al. [64] examined the detection threshold of rats (Sprague-Dawley albinos) to ethylene glycol dinitrate (EGDN; a major constituent of dynamite) and three homologous perfluorocarbons (that at the time were being evaluated as potential additives to explosives). Trials with descending odor concentrations indicated an average threshold (the concentration at which 75% of the rats correct alerted) to be in the ppm range. Performance was found to be variable in response to EGDN, however, distinct dose-response curves (i.e. as odor concentration was lowered, the performance of the rats declined) were apparent for the three perfluorocarbons compounds. Weinstein et al. [65] investigated the ability of rats to detect different concentrations of pure TNT vapour following training with saturated levels of the compound (rather than specifically to low concentrations). Results indicated that correct detection rates of 60% were maintained at 1.07 ng l 1.

96

O. Leitch et al. / Forensic Science International 232 (2013) 92–103

While these studies are encouraging, it is clear that further research is required in order to accurately assess rat sensitivity. Another particular challenge with assessing the ability of rats as operational biological sensors is the lack of data regarding their specificity. The majority of rats tested were trained to detect a single compound, being 2,4,6-trinitrotoluene (TNT) [16,28]. Only one reference was found regarding research completed with additional target and non-target odors [15]. While APOPO detection rats are exposed to multiple environmental odors during initial ‘‘socialisation’’ [16,28], they are not trained in the presence of potentially confounding odors or with odor mixtures (although later training stages do involve whole mines). The rationale behind this is that TNT is the major explosive component of landmines therefore this is the compound that rats should be trained to detect. However, TNT has a very low vapour pressure; therefore, it is not always the predominant chemical within the headspace [62,63] and is not always readily detectable [61]. This was illustrated by Verhagen et al. [28] who found that successful detections decreased significantly in damp conditions. Similar to detection canines, if rats are trained only on a single compound or substance, their success and utility for operational purposes will be limited. A major advantage of rats as biological sensors is their size, which provides an ability to search very small and tight spaces that would be impenetrable with detection canines and humans with conventional instrumentation. However, as observable behavioral responses are employed as indications for the presence of target substances, biological sensors must remain in sight. Otto et al. [15] tested a sensor device (MiniBird system, Ascension Technology Corporation) that monitors the location and alerting behavior from a distance. Such a device would allow for continual remote monitoring even when a detection rat is out of sight, which means that the biological sensor can be unaccompanied and the risk of harm to handlers reduced. However, the particular monitoring device used by Otto et al. [15] requires an attachment cable that limits the working distance and, ultimately, the range of searches that could be conducted (e.g. inside fallen buildings). Additionally, the limited experimental conditions to which this device has been exposed means it is not yet known if this device would be suitable for searches under various environmental and ecological settings. Further refinement and research with modern information and communication technology would likely provide more compact and effective monitoring methods. 3. Invertebrates The ability to learn to detect volatile odors is not limited to vertebrates. Invertebrates have been shown to be highly sensitive and selective to a range of compounds both native and foreign to

their natural environments [2,3,18,66,67]. Additionally, the lack of social bonds with humans means invertebrates would be relatively unaffected by the presence of handlers and thus unlikely to sense cues as reportedly occurs with canines. Such information presents the possibility of exploiting insects as a novel, potentially more ‘‘objective’’, approach to the detection of volatile compounds of forensic significance. Some of the insects investigated for use as biological chemical sensors include the parasitic wasp Microplitis croceipes C., the honeybee Apis mellifera L., and the hawkmoth Manduca sexta L. (Fig. 2a–c). 3.1. Wasps M. croceipes, which are larval parasitoids (they lay their eggs inside the living bodies of other insect larvae) [67], learn to recognize new chemical signals indicating the best host habitat and food as environmental and food conditions change [70–72]. It has been found that this learning ability can be trained within a laboratory environment in association with either food or hosts [8,18,73,74]. Training methods employed are relatively simple and easy to implement, and generally involve exposing wasps to the odor of interest simultaneously with, or immediately followed by, natural food or hosts. Typically, wasps are starved for approximately two days prior to food-association training. Olson et al. [8] and Rains et al. [67] provide detailed descriptions of training methods. The advantage of M. croceipes is that rearing of the insects is easy and inexpensive, hundreds can be reared at one time, training can be completed in a variety of ways (by hand with individual insects, automated or in a mass training system) and generally within 5 min [8], they have the ability to discriminate a target odor from mixtures [18], can be trained with a diverse array of compounds [8,75], can be used to assess the relative proximity of the target odor (e.g. exhibited behaviors intensify progressively) [76], and give resource specific behavioral responses [8,66]. However, the primary limitations are reported to include maintenance requirements greater than traditional physico-chemical detection techniques (e.g. their physiological state needs to be maintained within set requirements, thus requiring a high level of monitoring) [67], and a short adult lifespan of only approximately two to three weeks [4]. Thus, wasps would have to be continually reared and trained. Additionally, like almost all whole-organism biological chemical sensors, indications that a specific volatile chemical is present involve behavioral responses. Therefore, constant observation is required. The reliance on human observation means that some changes in behavior may be missed, particularly at lower concentrations when the behavioral responses may be more subtle. Behavioral responses indicating the presence of a target odor can include coiling, antennating (rubbing antennae on the surface),

Fig. 2. Invertebrates investigated for use as biological chemical sensors. (a) Microplitis croceipes, a parasitic wasp, can be trained to associate target compounds with food and/ or hosts. Image reproduced with permission from Olson et al. [8]. (b) Apis mellifera (honeybee) – natural foraging behavior can be exploited for the detection of target odors. Image reproduced with permission from Greb [68] (c) Manduca sexta – hawkmoth – can be trained to associate target odors with food. Image reproduced with permission from Cohen [69].

O. Leitch et al. / Forensic Science International 232 (2013) 92–103

Fig. 3. Conditioned responses of M. croceipes. (a) Food associated ‘‘seeking’’ behavior whereby the conditioned wasp searches the area where the odor is emanated. (b) Host associated coiling behavior where female wasps rise on their hind legs and characteristically bend their antennae. Images reproduced with permission from Olson et al. [8].

head sticking (when the wasp sticks its head through a hole emitting the target odor), directed flight, area-restricted searching, and crowding (where several wasps exhibit area-restricted searching in the same area) [8,17,73,77]. These behaviors are not specifically trainable but are natural responses to the organism’s expectation of food or a host. A major advantage of female M. croceipes is that the behaviors exhibited in anticipation of food or hosts are specific to the respective resource [8,74]. As a result, it is possible to distinguish between two different target odors based on the type of behavior exhibited. For example, Olson et al. [8] used seeking flight behavior (exhibited in the presence of food) and reflexive coiling behavior (exhibited when attacking and laying eggs in a host) to demonstrate that conditioned female wasps exhibit specific behavior, seeking or coiling, exclusively in accordance with the conditioned target odor (Fig. 3). Training M. croceipes to associate two different odors with the same food or host can be completed simultaneously [66] or successively [78]. This indicates that they are able to ‘‘store’’ information for future use. However, Takasu and Lewis [78], who examined the optimal training age and protocol for female M. croceipes, found that, while conditioned odors were remembered for at least two days after training, repeated exposure to the target odor in the absence of the food or host ultimately led to the cessation of behavioral responses when presented with the target odor. This means that repeated conditioning is required. Nevertheless, Tertuliano et al. [73], who also investigated various training protocols, found that the level of behavioral response would increase significantly after only a single reinforcement of the target odor in association with food. A particular advantage of this ability to re-associate target odors to the food or host reward is that the target odor could be repeatedly changed. In conjunction with the rapid training and ability to learn a variety of odors, this would allow for the same wasp(s) to be used for a range of different applications as the need arises. Recent research has shown that M. croceipes learn and respond behaviourally to a range of ‘‘non-native’’ volatile compounds significant to forensic science, such as 2,4-DNT [8,79], 3,4dinitrotoluene (3,4-DNT) [8], cyclohexanone [8,80], methyl benzoate [80], cadaverine, and putriscene [67,75]. While it has been shown that M. croceipes can learn a wide and varied range of chemicals, including ones foreign to their natural environment, it cannot be assumed that they can be successfully trained to any compound of interest. For example, Meiners et al. [75] were unable to condition M. croceipes to decanol or nonanol, despite the fact that these wasps respond physiologically to these particular odors [81]. Additionally, differences in response affinities have also been observed. Meiners et al. [75] found that behavioral responses were

97

stronger for some compounds than for others, and the ability to distinguish between similar compounds depended on chain length and the type and position of the functional groups. There is limited information regarding the range of concentrations over which M. croceipes can operate. Studies conducted by Rains et al. [67] to determine the level of sensitivity of M. croceipes to four separate odor compounds (3-octanone, mycrene, cadaverine and putrescine) indicated minimum thresholds of 10 7 mol L 1. For the specific compounds tested, this sensitivity is reportedly approximately over 10-fold better than the commercially available electronic nose to which it was directly compared [67]. However, the response time of the wasps (i.e. the length of time the wasps exhibited the behavioral response indicating the presence of the target compound) could not be used as the basis for estimating a compound’s concentration, only that the compound was present. This is because the average amount of time that the wasps exhibited positive responses to the different amounts of substance used could not be statistically differentiated (i.e. the duration of the response was statistically similar over a wide concentration range) [67]. It has also been found, however, that the positive responses exhibited by conditioned wasps decreased at quantities other than that used for training. For example, Tomberlin et al. [79] trained wasps to detect 2,4-DNT using 0.5 mg of the substance and found that, while 80% of the conditioned wasps responded to the target odor at that particular quantity, the response decreased to 70% when exposed to greater amounts and to 30% and 10% when exposed to lesser amounts (0.25 mg and 0.05 mg, respectively). This suggests that, if possible, wasps should be trained using the target odor at various quantities thus exposing them to a range of concentrations. While it is possible that the wasps cannot detect the target odor at the lower concentrations, the decreased responses could also be due to the type of behavior being used to indicate positive responses and/or the observer. Tomberlin et al. [79] reported that the particular behavior(s) used to indicate the presence of target odors influenced the ability of the observer to distinguish between the target odors and their concentrations. For example, it was reportedly found that associating target odors with hosts, rather than with food, resulted in more dependable behavioral responses, particularly when distinguishing between target and non-target odors of similar molecular structures [79]. Additionally, it was also reported that, using host associated behaviors, the wasps proved to be extremely sensitive. The potential for behavioral responses to go unobserved is supported by the examinations of Olsen et al. [8] and Lewis et al. [66,76] who found that exhibited behaviors intensified when the wasps where close to the odor source and became more subtle and harder to recognize when they were further away. Consequently, to obtain more accurate and precise indications of the sensitivity and specificity of M. croceipes, observer bias needs to be eliminated and more objective and sensitive methods employed for detecting behavioral responses. Several attempts have been made to objectively quantify the behavioral responses exhibited by wasps in the presence of target compounds [73,77,82]. Most recently for use with wasps, Utley et al. [77] assembled a computer vision system using the software package Visual Cortex (developed from LabView). Video footage of conditioned wasps is recorded remotely and subsequently analyzed with the software program to determine if the target odor is present. This system proved to be rapid – results were determined within 20–30 s after sample exposure – and capable of detecting and distinguishing crowding responses to 3-octanone at 5.5 and 1.1 mg L 1. However, while the system can detect positive indications as to the presence of the target odor, the concentration of that odor was unable to be inferred from the magnitude or slope of the response curve [77]. Additionally, such a system does not allow for the wasps to be free-moving but requires them to be

98

O. Leitch et al. / Forensic Science International 232 (2013) 92–103

contained in order to capture the necessary footage. The use of restrained biological organisms in detection instruments is discussed below. 3.2. Honeybees The learning behavior of A. mellifera – honeybees – (Fig. 2b) has received significant attention over the years (e.g. [83–87]) and, similarly to wasps, bees have been found to be capable of olfactory memory. Bromenshenk et al. [6] developed a technique for training honeybees that involves injecting trace amounts of a target compound into the honeybees’ feeder. The honeybees would then seek out sources of food with the same odor when next foraging [6,88]. The use of honeybees as a detection technique for VCs has particular advantages in that, unlike detection canines, the honeybees do not require a handler, they will not cause the detonation of a mine or an improvised explosive device, they can be trained in approximately one to two days [88], they have the ability to search large areas relatively quickly [19,62], they have low maintenance costs [6], and their trainers do not require specialized skills [4]. However, a significant limitation to the employment of honeybees is that their flying behavior is greatly affected by environmental conditions [62]. For example, they do not fly at night, in heavy rain or wind, or in temperatures below approximately 4.5 8C [6]. Additionally, while attempts have been made to detect and track the movement of honeybees away from the hive [2,88,89], an appropriate method for use under all search conditions is yet to be developed. There is also limited knowledge regarding their utility in areas other than clear, open fields [62]. Studies completed by Bromenshenk et al. [2,6] focused exclusively on the ability of honeybees to detect and locate landmines. The technique exploits the natural foraging behavior of the honeybees, which can frequently cover areas up to several kilometers around the hive in search of nectar and pollen [6,88]. Indications as to the presence of the target odor are made by the number of bees following the vapour plumes toward and over the source [2]. The number of bees over the source is integrated over time and compared to the number of bees over the rest of the area. The density of the bees in a particular area is mapped either visually, via a camera, or with laser-assisted counts. Field trials conducted by Bromenshenk et al. [6] indicated that honeybees are capable of detecting vapour compounds at concentration levels of ppb to ppt. For example, the bees were able to positively indicate the presence of 2,4-DNT generated in an estimated vapour concentration of 50–80 ppt. However, this detection threshold dropped significantly under moist conditions [6]. An additional finding by Bromenshenk et al. [6] is that the number of bees hovering over the target source correlates with the concentration of the odor being detected. The primary strategy for using bees as biological sensors involves monitoring their movement and recording locations in which they swarm. One particular method for tracking the movement of free flying honeybees is by employing ‘‘light detection and ranging’’ (LIDAR) measurements. LIDAR is based on the properties of scattered light and involves sweeping laser light over the target foraging area. Light that strikes an object, in this case bees, is scattered back and the object’s distance from the laser is determined [4,88]. Shaw et al. [88] examined the use of scanning polarized LIDAR to monitor the movement of honeybees trained to detect landmines. During blind field trials, LIDAR bee detection proved highly successful in locating explosive residues in an area cleared of mines [88]. Positive detections within the cleared control areas, which were later confirmed by chemical analysis, provide encouraging evidence that explosive residue detection by bees is highly effective. However, the primary

limitations of polarized LIDAR are that it is not ‘‘bee-specific’’ and it has difficulty distinguishing between the bees and interfering objects such as vegetation [61,88,89]. The requirement for a clear line of sight and the need to sweep the laser at heights that avoid vegetation (but which bees frequently fly under) [88] means that the operational deployment of this technique is significantly impaired. Another tracking method is harmonic radar. Harmonic radar involves placing a transponder on the object to be tracked and capturing the radar signal emitted. Riley and Smith [90] developed a miniaturized harmonic radar system capable of recording the flight trajectories of low-flying insects, such as honeybees, over hundreds of meters. Practical and monetary considerations were taken into account during the development and configuration process, thus the system was based on existing, and sometimes mass-produced, radar technologies that have previously proven to be rugged and relatively inexpensive [90,91]. Building such tracking systems from scratch or modifying radars from specialist applications would be highly expensive and may limit their deployability. The advantage of this technique is that the issue of ‘‘clutter’’ produced by interfering objects is overcome. However, while the transponder has been significantly miniaturized and is suitable for use with honeybees, it has limited application for other insect species because it requires insects to be of a certain size (i.e. insects over approximately 50 mg in weight) [90]. Additionally, this technique also requires a clear line of sight between the radar and the transponder and individual transponders cannot be distinguished; therefore, flight trajectories can become confused if more than one tagged honeybee is in flight. 3.3. Moths As with the other insects discussed above, M. sexta (noctuid hawkmoths) (Fig. 2c) has also proved to be trainable through association of a target odor with food [3,92,93]. Determining positive responses of M. sexta when exposed to the target odor involves monitoring the activity of the feeding muscle (the cibarial pump muscle) via electromyography (EMG) before, during and after training [92]. Briefly, conditioning involves restraining the moth and threading the extended proboscis through a length of tube. An electrode is then brought into contact with the cibarial pump muscle, located under the head capsule between the compound eye and sagittal mid-line [3,92]. Interpretations of responses are based on the measured electrical activity. A significant advantage of M. sexta is that conditioning to a target odor appears to last for the duration of the moths’ adult life – approximately one to two weeks [93]. However, due to the current method required for response detection, moths are unable to be used in free flying mode at this stage. Daly et al. [3] examined the ability of M. sexta to discriminate odors and found that their success depended on reinforcement training, not on the particular target odor. For example, it was found that systematically presenting cyclohexanone (the target odor) with food reinforcement followed by a structurally similar compound (non-target odor) without reinforcement resulted in M. sexta exhibiting excitatory responses to cyclohexanone when later presented with this odor alone. When the target and non-target odors were reversed (e.g., cyclohexanone was not reinforced), no excitatory responses to cyclohexanone were observed. Therefore, food reinforced odors produce greater responses than nonreinforced odors. Such differential conditioning potentially presents the opportunity to ‘‘tune’’ the specificity of the moth’s response [3,93,94]. The number of false positives could also potentially be reduced by employing multiple moths conditioned to different odors [3,93,94].

O. Leitch et al. / Forensic Science International 232 (2013) 92–103

While M. sexta has been shown to be capable of detecting explosive compounds [3,93,94], their sensitivity to such odors has not been reported. However, studies investigating conditioned behavioral and physiological responses of this moth to individual compounds, such as 2-hexanone (a compound detected in the headspace of human decomposition), suggests detection thresholds of 0.01 mg [95]. Such sensitivity was found to be independent of the initial conditioning concentration, however, it was dependent on the specific odor; therefore, such detection thresholds cannot be expected for all odors. 4. Detection devices employing live insects Due to the advantages of insects as VC sensors but the difficulties in monitoring them while free-flying, the development and refinement of portable detection units employing confined whole-organisms has been attempted [67,93,96]. Initial attempts by Raman and Gerhardt [97] utilized the innate behavioral responses of six confined gravid face flies. Positive responses exhibited by the flies to certain VCs were based on sound, which was recorded with a built-in microphone. Since this initial effort, researchers have focused on exploiting the remarkable learning abilities insects possess. The Wasp Hound1, developed by Rains et al. [17,67], is a handheld device that employs five female parasitic M. croceipes wasps contained within a ventilated, LED lit, transparent chamber (Fig. 4a). The transparent chamber is of a sufficient size that the wasps can move freely but small enough to keep them in close vicinity of the air sample inlet to enable timely responses. Operation of the Wasp Hound1 involves slowly drawing air samples into the chamber and monitoring the wasps behavioral responses with a modified version of the Visual Cortex developed by Utley et al. [77] (discussed previously). Video image capture (at a rate of four frames per second) is used to monitor the arearestricted food-searching behavior of the wasps. The Visual Cortex software subsequently transduces these images into numerical data that results in a response curve. When the slope of this response exceeds a threshold (which at this stage is determined by the user through calibration experiments with trained and untrained wasps) the operator interprets the response as positive with regard to the presence of the target compound. Experiments completed by Rains et al. [17] found that the Wasp Hound1 could numerically describe the searching behavior of trained wasps and successfully detected the positive behavioral responses to the target odor even when it was in a background of potentially confounding substances. While assessments can be made relatively quickly, for the Wasp Hound1 to work effectively headspace samples need to provide a continuous flow of the target

99

odor for approximately 20–30 s [17,18]. If the target odor is intermittent then there will be plateaus in the response slope making interpretation of results more complicated. At this stage, the Wasp Hound1 can only provide a positive or negative indication as to the presence of the target odor – no indication of concentration is possible. Additionally, there is currently no information regarding the rate of false-positives given by the detection system. As the Wasp Hound1 is not yet commercially available, it is anticipated that these limitations, as well as the question of obtaining trained wasps (e.g. if they are supplied with the device or require training after purchase, and/or if a continual supply can be provided or if the user must maintain their own colony), will be addressed as development and refinement continues. Other types of portable detection devices that employ restrained whole-organisms have also been developed [93,96,98]. For example, King et al. [93] developed a sensor that monitors the electromyography signals (feeding response) of ten M. sexta moths (Fig. 4b). Subsequent to training, the moths are confined within a small wind tunnel with a fan to draw air samples over the moths’ antennae. A connected detector monitors the moths’ feeding response behavior. Of the ten moths used in this device, only five are trained to detect the target odor. The remaining five are un-trained. The basis for this is to allow for the determination of any false positives exhibited by the trained moths. That is, theoretically, if both groups of moths respond then it is interpreted as a false positive; but if only the target-trained moths respond, then a true positive result has occurred. Preliminary results from this detection system reveal that it has the ability to positively detect the presence of cyclohexanone (a compound used in the purification of RDX and one of the two most predominant volatiles within the headspace of the explosive C4) [46] when the compound is released in front of the prototype system. However, while the device shows promise for the detection of volatile compounds associated with explosives, its size and weight (0.25 m3 and 17 kg) preclude it from being a portable device. The combined wind tunnel and moth chamber also appear to be in a fixed position somewhat elevated above the ground. This makes the device less flexible with respect to sampling location and could prove difficult for direct headspace sampling. Additionally, King et al. [93] found that bumping the device and sudden changes in light intensity affected both groups of moths, thereby making the sensor less rugged. Inscentinel Ltd. (Hertforshire, UK), a commercial enterprise, developed a trace vapour detection system using 36 conditioned honeybees housed within the company’s patented cassettes (Fig. 4c) [96]. The head of each bee protrudes from the cassette, air samples are drawn over them, and measurements of the

Fig. 4. Wasp, moth and bee volatile compound detection devices. (a) The Wasp Hound1 containing an air intake fan, monitoring camera, and holding cartridge for five conditioned wasps. Wasp Hound1 image reproduced with permission from Rains et al. [17]. (b) Volatile compound detector that monitors the feeding response of ten Hawkmoths (copyrightß IEEE. All rights reserved. Reprinted, with permission, from King et al. [93]). (c) Inscentinel’s trace vapour detector employing trained honeybees. Images reproduced with permission from Inscentinel Ltd. [99].

100

O. Leitch et al. / Forensic Science International 232 (2013) 92–103

learned behavior (proboscis extension), which indicate the presence of the target odor, are acquired electronically and displayed graphically [99]. This device is reportedly successful at detecting traces of explosives, such as Semtex and TNT, at concentrations as low as parts per trillion in the laboratory [17,99]. During prototype testing alongside an ion-mobility spectrometer (IMS) detector, the bee sensor detected explosive traces (20– 50 ppb) even when the IMS detector could not [99]. At this stage, the aforementioned detection systems are only capable of detecting one, or at most two, target compounds. Although this makes these detection systems highly specific, it does not allow for the detection of complex odors made up of a complex mixture of compounds. Often a substance is not detected simply by identifying the chemical compounds in its vapour and their abundance, but is determined by the complex interaction between compound abundance, physiological sensitivity of the animal for particular compounds, and the specific training received. For example, detection canines are trained on the relevant actual substances, not one or two compounds present in their vapour, and therefore are exposed to a substance’s ‘‘odor signature’’. Furton and Myers [11] reported that odor signatures change over time and that, as mentioned previously, individual canines employ different odor signatures for detecting target substances, despite being nominally trained using the same methodology. While this makes determining the specific compounds utilized for substance detection (and the canines’ sensitivities to those compounds) difficult, it does allow for the odor space associated with the target substance (and its potential changes due to low concentrations and interfering backgrounds) to be adequately covered. 5. Detection using isolated olfactory organs All of the VC detection systems discussed above involve behavioral responses, which in any animal is variable and highly dependent on various internal and external conditions. Recent literature reveals an increasing interest in olfactory receptors – the biological components that impart olfactory ability – for detecting VCs of forensic significance. This leads to the idea that biological sensors for VC detection could potentially be based on measurement of cellular responses in olfactory organs [98,100–102]. Electroantennograms (EAGs) are bioassays used widely to detect the volatile compounds perceived by the olfactory organs of insects [103]. Recognition of a volatile compound is represented by electrical voltage fluctuations of olfactory receptor neurons (ORNs). Housed within ORNs are olfactory receptors (ORs), and it is the interaction between these receptors and VCs (odors) that form the basic mechanism of olfactory systems. Park et al. [98] utilized EAGs of five different insect species that possess different EAG response profiles – Drosophila melanogaster (vinegar fly), Heliothis virescens (moth), Helicoverpa zea (moth), Ostrinia nubilalis (moth), and M. croceipes (wasp) – to a wide variety of VCs. Within the detection device, rather than using the whole organism, just the excised antennae of the five species were employed. Experiments completed with the antennal array found that this device was capable of detecting and discriminating 20 different VCs (some of which are associated with explosives) under simulated field conditions [98]. There have also been several attempts to make single antenna biological sensors [104–106]; however, while they show reasonable sensitivity to target volatiles, they lack the discriminatory ability provided using additional antennae – that is, they are unable to distinguish the target compound from unknowns within the background. By employing more antennae, Park et al. [98] showed that selectivity and sensitivity could be obtained. Nevertheless, Park et al. [98] also showed that compounds of

similar structure (e.g. 1-heptanol and 1-octanol) were not easily distinguished even with a greater number of antennae as they evoked almost identical EAG response profiles. Employing a different combination of antennae could alleviate this limitation. The use of excised antennae reduces the lifespan of the sensor significantly. While adult moths live approximately one to two weeks, excised antennae will typically only last up to two hours (or less than 30 min depending on the EAG protocols) [98]. While Park et al. [98] focused on the summed response of all ORs within the antenna, efforts have also been made to identify the individual ORs responsible for responding to compounds associated with substances of forensic significance. For example, Radhika et al. [107] identified an OR in the rat that responds to 2,4-DNT. Additionally, a number of studies [101,102,108,109] have identified individual ORs from invertebrates such as Caenorhabditis elegans (nematode) and D. melanogaster that are capable of detecting VCs associated with explosives, illicit drugs and decomposition. Liao et al. [101] showed that C. elegans ORs are highly sensitive and selective. For example, behavioral assays revealed that these worms have a limit of detection in the ppm range for cyclohexanone. However, other experimental results and knowledge of other C. elegans ORs suggests that the limit of detection could be much lower – potentially in the ppt range [101]. Additionally, there are strong indications that C. elegans ORs are highly selective whereby only one or two receptor types respond to an individual compound [101,110]. Conversely, D. melanogaster has been shown to possess both broadly tuned (respond to odorants from multiple chemical classes) and narrowly tuned (respond to odorants within a single chemical class) ORs [102,108,109]. For example, research completed by Hallem and Carlson [108] showed that, of the 110 odorants tested, some ORs responded to approximately 30% of the odors (i.e. broadly tuned) while other ORs responded strongly to only one (i.e. narrowly tuned). This reveals that odor detection is accomplished with a number of ORs in a combinatorial fashion. Thus, in order to achieve the sensitivity and selectivity of biological olfaction in a detection instrument, the sensors selected would need to be used in a manner that mimics the in vivo system; that is, an array of differently tuned ORs would need to be employed. While the aforementioned studies indicate that C. elegans and D. melanogaster ORs would be extremely useful within a biologically derived sensor, it cannot be assumed that these organisms have ORs capable of detecting all substances of interest. For example, of the 35 compounds tested by Marshall et al. [102] with D. melanogaster (which included toxic gases and precursors, explosives and illicit drugs as well as their precursors and contaminants) only 13 elicited positive responses. Similarly, not all of the compounds associated with explosives that were tested by Liao et al. [101] were detected by C. elegans. These results are not unexpected because many of the compounds are not derived from sources that these organisms have evolved to detect. That is, many of the VCs of forensic significance are unlikely to be biologically relevant; therefore, the positive responses obtained surpass what would be expected for normal ecological purposes. Such responses may therefore be representative of broader abilities to sense their surrounding environments (e.g. locating specific predators, hosts, food, etc. within highly variable and dynamic foraging surroundings). The ability to detect such a broad array of ‘‘non-native’’ chemicals is unlikely to be restricted to the aforementioned organisms; therefore, numerous opportunities for finding ORs suitable for use in a sensor array are available. For example, a vast number of putative ORs have been identified in other organisms, including 79 in Anopheles gambiae (malaria mosquito) [111], 170 in A. mellifera (honeybee) [112], 66 in Bombyx mori (silkworm moth) [58], and 131 in Aedes aegypti (dengue mosquito) [113]. Some of

O. Leitch et al. / Forensic Science International 232 (2013) 92–103

these invertebrates have already been shown to be capable of detecting VCs of interest (e.g. A. mellifera can detect explosive residues) whereas others have evolved specifically to detect others (e.g. A. gambiae detects living humans). Along with the varying number of ORs, each of these organisms will also possess its own unique olfactory response repertoire. As a result, the potential to build OR-based sensor arrays tailored to specific detection areas is immense. In order to use ORs in a detection device, it must be possible to recognize and interpret the response obtained in the presence of sampled substances. Several proof-of-concept studies investigating methods for transducing the signal from OR–ligand binding have been completed. For example, Radhika et al. [107] constructed a system utilizing yeast cells that encoded and expressed an isolated rat OR that responds to 2,4-DNT. The response of the OR to the odor was observed via fluorescence. However, due to the fragility of cell-based systems, the development of a stable detector would be difficult with this technology. Thus, a cell-free assay, such as the one developed by Dacres et al. [114], would be highly beneficial. The assay is based on bioluminescence resonance energy transfer and detects conformational changes in C. elegans ORs that occur subsequent to ligand binding. This assay proved to be rapid, extremely sensitive (parts per quadrillion), and was specific to the OR–ligand combination examined [114]. At this stage, no detection systems employing ORs as VC sensors are known to be in the laboratory prototype stage and are certainly not commercially available. However, with the increasing speed with which ORs from a variety of biological organisms are being isolated and de-orphaned (i.e. the ligands to which they respond are identified), the advancing knowledge of potential transduction systems, and the technological proof-ofconcept studies, it is likely that an OR-based detection instrument will emerge in due course. 6. Conclusions The successful employment of biological organisms – in the case of detection canines – and the feasibility of their use for VC detection – in the case of invertebrates – have been demonstrated. However, there are a significant number of factors that can impact on the ability of biological organisms to detect VCs of forensic interest. The benefit of free-moving biological sensors (i.e. whole organisms allowed to move without constraint) is that they are able to move toward the source of the odor – i.e. they can follow the scent from least to highest concentration. Many of the limitations of whole organisms are innate; therefore, they are unlikely to ever be overcome, although there are a number of options to alleviate such limitations. For canines, the significant dichotomy of exploiting their abilities for detection purposes and their historical association with humans as friendly, reliable, family companions makes the objective assessment of their utility difficult. In contrast, rats and invertebrates do not form social relationships with humans, therefore their responses are theoretically easier to assess. Additionally, their potential loss would be less poignant and, in practical terms, they are easier and less expensive to replace. It is unlikely that the other whole-organism biological sensors discussed in this paper will ever replace the detection canine; however, they do present a viable, potentially cost-effective and complementary option. They would be highly beneficial in situations deemed too dangerous and/or unethical for the detection canine. Volatile compound detectors employing biological sensors possess some major advantages (e.g. sensitivity, selectivity, rapid training/preparation, etc.); however, all devices developed thus far are still in the laboratory, or very early prototype, phase. Further

101

research investigating the true capability of such detectors (e.g. detection thresholds, rate of false-positives and false-negatives, etc.) in ‘‘real-world’’ field scenarios is required. Additionally, some of the unique challenges of employing such devices, such as desensitization to target odors without reinforcement, continuous sampling requirements that mean holding chambers and insects need to be replaced regularly, and the organism’s susceptibility to changing environmental conditions that could affect the insects performance and/or longevity, need to be fully studied and potential limitations alleviated in order for such devices to be adopted operationally. Due to the behavioral variations and fatigue associated with whole-organism biological sensors, an artificial system capable of providing the same performance would be highly desirable. Ultimately, further research is required before the performance, effectiveness and potential broader employment of biological and biologically derived sensors is fully understood and accepted for mainstream applications. Acknowledgements Olivia Leitch’s PhD candidature is financially supported by the Australian Federal Police, Forensic and Data Centres, Canberra, ACT, Australia. References [1] R. Benton, On the ORigin of smell: odorant receptors in insects, Cell. Mol. Life Sci. 63 (14) (2006) 1579–1585. [2] J.J. Bromenshenk, et al., Can honey bees assist in area reduction and landmine detection? J. Mine Action (2003) 273–283. [3] K.C. Daly, M.L. Durtschi, B.H. Smith, Olfactory-based discrimination learning in the moth, Manduca sexta, J. Insect Physiol. 47 (4–5) (2001) 375–384. [4] G.C. Rains, J.K. Tomberlin, D. Kulasiri, Using insect sniffing devices for detection, Trends Biotechnol. 26 (6) (2008) 288–294. [5] R. Smart, et al., Drosophila odorant receptors are novel seven transmembrane domain proteins that can signal independently of heterotrimeris G proteins, Insect Biochem. Mol. Biol. 38 (8) (2008) 770–780. [6] J.J. Bromenshenk, C.B. Henderson, G.C. Smith, Appendix S: biological systems, Alternatives for Landmine Detection, Montana, 2003, pp. 273–283 (paper II). [7] L.M. Harvey, J.W. Harvey, Reliability of blodhounds in criminal investigations, J. Forensic Sci. 48 (4) (2003) 1–6. [8] D.M. Olson, et al., Parasitic wasps learn and report diverse chemicals with unique conditionable behaviors, Chem. Senses 28 (6) (2003) 545–549. [9] G.A.A. Schoon, The effect of the ageing of crime scene objects on the results of scent identification line-ups using trained dogs, Forensic Sci. Int. 147 (1) (2005) 43–47. [10] P.L. Schmidt, Companion animals as sentinels for public health, Vet. Clin. North Am.: Small Anim. Pract. 39 (2) (2009) 241–250. [11] K.G. Furton, G.S. Myers, The scientific foundation and efficacy of the use of canines as chemical detectors for explosives, Talanta 54 (3) (2001) 487–500. [12] M. Nowlan, et al., Use of solid adsorbent and an accelerant detection canine for the detection of ignitable liquids burned in a structure fire, J. Forensic Sci. 52 (3) (2007) 643–648. [13] L. Oesterhelweg, et al., Cadaver dogs – a study on detection of contaminated carpet squares, Forensic Sci. Int. 174 (1) (2008) 35–39. [14] M. Williams, J. Johnston, Training and maintaining the performance of dogs (Canis familiaris) on an increasing number of odor discriminations in a controlled setting, Appl. Anim. Behav. Sci. 78 (1) (2002) 55–65. [15] J. Otto, M.F. Brown, W.I. Long, Training rats to search and alert on contraband odors, Appl. Anim. Behav. Sci. 77 (3) (2002) 217–232. [16] A. Poling, et al., Using giant African pouched rats (Cricetomys gambianus) to detect landmines, Psych. Rec. 60 (4) (2010) 715–728. [17] G.C. Rains, S.L. Utley, W.J. Lewis, Behavioral monitoring of trained insects for chemical detection, Biotechnol. Prog. 22 (1) (2006) 2–8. [18] J.K. Tomberlin, G.C. Rains, M.R. Sanford, Development of Microplitis croceipes as a biological sensor, Entomol. Exp. Appl. 128 (2) (2008) 249–257. [19] P.J. Rodacy, et al., Training and deployment of honeybees to detect explosives and other agents of harm, Proc. SPIE 474 (2002) 4742. [20] U. Kaupp, Olfactory signalling in vertebrates and insects: differences and commonalities, Nat. Rev.: Neurosci. 11 (2010) 188–200. [21] B. Malnic, et al., Combinatorial receptor codes for odors, Cell 96 (5) (1999) 713– 723. [22] P. Mombaerts, et al., Seven-transmembrane proteins as odorant and chamosensory receptors, Science 286 (1999) 707–711. [23] I. Gazit, et al., A simple system for the remote detection and analysis of sniffing in explosive detection dogs, Behav. Res. Methods Instrum. Comput. 35 (1) (2003) 82–89.

102

O. Leitch et al. / Forensic Science International 232 (2013) 92–103

[24] R. Tindall, K. Lothridge, An evaluation of 42 accelerant detection canine teams, J. Forensic Sci. 40 (4) (1995) 561–564. [25] K.M. Dorries, E. Adkins-regan, B.P. Halpern, Olfactory sensitivity to the pheromone: androstenone, is sexually dimorphic in the pig, Physiol. Behav. 57 (2) (1995) 255–259. [26] C. Wellerdieck, et al., Functional expression of odorant receptors of the zebrafish Danio rerio and of the nematode C. elegans in HEK293 cells, Chem. Senses 22 (4) (1997) 467–476. [27] M. Laska, A. Seibt, A. Weber, ‘Microsmatic’ primates revisited: olfactory sensitivity in the squirrel monkey, Chem. Senses 25 (1) (2000) 47–53. [28] R. Verhagen, et al., Chapter 4: Programme reviews. Part 2: Preliminary results on the use of Cricetomys rats as indicators of buried explosives in field conditions, Mine Detection Dogs: Training, Operations and Odour Detection, Geneva International Centre for Humanitarian Demining, Switzerland, 2003, pp. 175– 193. [29] D. Komar, The use of cadaver dogs in locating scattered, scavenged human remains: preliminary field test results, J. Forensic Sci. 44 (2) (1999) 405–408. [30] T. Cafe, Aids used for detecting accelerants at fire scenes, 2007 Available from: http://www.tcforensic.com.au/docs/article4.html (12.04.08). [31] K.G. Furton, R. Harper, Detection of ignitable liquid residues in fire scenes: canine (ADC) teams and other field tests, in: J. Almirall, K.G. Furton (Eds.), Analysis and Interpretation of Fire Scene Evidence, CRC Press, Boca Raton, FL, 2004, pp. 75–96. [32] A. Lasseter, et al., Cadaver dog and handler team capabilities in the recovery of buried human remains in the Southeastern United States, J. Forensic Sci. 48 (3) (2003) 1–5. [33] A. Pert, M. Baron, J. Birkett, Review of analytical techniques for arson residues, J. Forensic Sci. 51 (5) (2006) 1033–1049. [34] D. France, et al., A multidisciplinary approach to the detection of clandestine graves, J. Forensic Sci. 37 (6) (1992) 1445–1458. [35] G.A.A. Schoon, A first assessment of the reliabilty of an improved scent identification line-up, J. Forensic Sci. 43 (1) (1998) 70–75. [36] S. Katz, C. Midkiff, Unconfirmed canine accelerant detection: a reliability issue in court, J. Forensic Sci. 43 (2) (1998) 329–333. [37] T. Tomaszewski, P. Girdwoyn, Scent identification evidence in jurisdiction (drawing on the example of judicial practice in Poland), Forensic Sci. Int. 162 (1–3) (2006) 191–195. [38] I. Gazit, J. Terkel, Domination of olfaction over vision in explosives detection by dogs, Appl. Anim. Behav. Sci. 82 (1) (2003) 65–73. [39] Scientific Working Group on Dog and Orthogonal detector Guidelines, 2013 Available from: http://swgdog.fiu.edu/ (cited 20.05.13). [40] G.A.A. Schoon, Scent identification lineups by dogs (Canis familiaris): experimental design and forensic applications, Appl. Anim. Behav. Sci. 49 (3) (1996) 257–267. [41] E. Ashton, J. Eayrs, D. Moulton, Olfactory acuity in the dog, Nature 179 (1957) 1069–1070. [42] D. Moulton, E. Ashton, J. Eyrs, Studies in olfactory acuity. Part 4. Relative detectability of n-aliphatic acids by the dog, Anim. Behav. 8 (3–4) (1960) 117–128. [43] R. Becker, J. King, J. Markee, Studies on olfactory discrimination in dogs. II. Discriminatory behaviour in a free environment, J. Comp. Physiol. Psychol. 55 (5) (1962) 773–780. [44] M.S. Macais, et al., Detection of piperonal emitted from polymer controlled odor mimic permeation systems utilizing Canis familiaris and solid phase microextraction-ion mobility spectrometry, Forensic Sci. Int. 195 (1) (2010) 132–138. [45] J. Johnston, Canine Detection Capabilities: Operational Implications of Recent R&D Findings, Institute for Biological Detection Systems, Auburn University, Alabama, 1999. [46] N. Lorenzo, et al., Laboratory and field experiments used to identify Canis lupus var. familiaris active odor signature chemicals from drugs: explosives, and humans, Anal. Bioanal. Chem. 376 (8) (2003) 1212–1224. [47] D. Moulton, D. Marshell, The performance of dogs in detecting a-ionone in the vapor phase, J. Comp. Physiol. 110 (3) (1976) 287–306. [48] M. Kurz, et al., Evaluation of canines for accelerant detection at fire scenes, J. Forensic Sci. 39 (6) (1994) 868–873. [49] M. Kurz, et al., Effect of background interference on accelerant detection by canines, J. Forensic Sci. 41 (5) (1996) 868–873. [50] J. Johnston, et al., Canine olfactory capabilities for detecting NG smokeless powder, in: C. Midkif (Ed.), Proceedings of the 5th International Symposium on Analysis and Detection of Explosives, Bureau of Alsohol Tobacco and Firearms, Washington, DC, 1995. [51] R.J. Harper, J.R. Almirall, K.G. Furton, Identification of dominant odor chemicals emanating from explosives for use in developing optimal training aid combinations and mimics for canine detection, Talanta 67 (2) (2005) 313–327. [52] M. Williams, et al., Canine detection odor signatures for explosives, in: SPIE Conference on Enforcement and Security Technologies, 1998. [53] A.A. Vass, et al., Odor analysis of decomposing buried human remains, J. Forensic Sci. 53 (2) (2008) 384–391. [54] E.M. Hoffman, et al., Characterisation of the volatile organic compounds present in the headspace of decomposing human remains, Forensic Sci. Int. 186 (1) (2009) 6–13. [55] D.-T. Vu, SPME/GC–MS characterization of volatiles associated with methamphetamine: toward the development of a pseudomethamphetamine training material, J. Forensic Sci. 46 (5) (2001) 1014–1024. [56] L. Lit, J.B. Schweitzer, A.M. Oberbauer, Handler beliefs affect scent detection outcomes, Anim. Cogn. 14 (3) (2011) 387–394.

[57] APOPO Detection Rats Technology, 2012 Available from: www.apopo.org/ (cited 03.08.12). [58] K. Touhara, L. Vosshall, Sensing odorants and pheromones with chemosensory receptors, Annu. Rev. Physiol. 71 (2009) 307–332. [59] T. Nolan, S. Weinstein, C. Weinstein, Electroencephalographic studies of specifically-conditioned explosives detecting rats, in: Proceedings of the New Concepts Symposium and Workshop on Detection and Identification of Explosives, Reston, VA, USA, 1978. [60] S. Weinstein, C. Weinstein, R. Drozdenko, The challenge of biodetection for screening persons carrying explosives, in: Proceedings of the 1st International Symposium on Explosive Detection Technology, FAA Technical Center: Atlantic City International Airport, NJ, USA, 1992. [61] M.K. Habib, Controlled biological and biomimetic systems for landmine detection, Biosens. Bioelectron. 23 (1) (2007) 1–18. [62] J. MacDonald, J.R. Lockwood, Chapter 2: innovatice mine detection systems, Alternatives for Landmine Detection, 2003, pp. 15–47. [63] J. Yinon, S. Zitrin, Modern Methods and Applications in Analysis of Explosives, Wiley, Chichester, UK/NY, USA, 1996. [64] D.A. Marshall, et al., Odor detection thresholds in the rat for the vapors of three related perfluorocarbons and ethylene glycol initrate, Chem. Senses 6 (4) (1981) 421–433. [65] S. Weinstein, R. Drozdenko, C. Weinstein, Detection of extremely low concentrations of ultra pure TNT by rat, in: Proceedings of the Third Symposium on Analysis and Detection of Explosives, 1989. [66] W.J. Lewis, K. Takasu, Use of learned odors by a parasitic wasp in accordance with host and food needs, Nature 348 (257–259) (1990) 635–636. [67] G.C. Rains, et al., Limits of volatile chemical detection of a parasitoid wasp, Microplitis croceipes, and an electronic nose: a comparative study, Trans. Am. Soc. Agric. Eng. 47 (6) (2004) 2145–2152. [68] P. Greb, Honey Bee (Apis mellifera), Bugwood.org, USDA Agricultural Research Service, 2007. [69] E.A. Cohen, Manduca sexta, 2012. [70] C.M. de Moraes, et al., Herbavore-infested plants selectively attract parasitoids, Nature 393 (1998) 570–573. [71] W. Lewis, J.H. Tumlinson, Host detection by chemically mediated associative learning in a parasitic wasp, Nature 331 (1988) 257–259. [72] T.C.J. Turlings, J.H. Tumlinson, W.J. Lewis, Exploitation of herbivore-induced plant odors by host-seeking parasitic wasps, Science 250 (1990) 1251–1253. [73] M. Tertuliano, et al., Influence of handling and conditioning protocol on learning and memory of Microplitis croceipes, Entomol. Exp. Appl. 110 (2004) 165–172. [74] F. Wa¨ckers, C. Bonifay, W.J. Lewis, Conditioning of appetitive behavior in the Hymenopteran parasitoid Microplitis croceipes, Entomol. Exp. Appl. 103 (2) (2002) 135–138. [75] T. Meiners, F. Wa¨ckers, W.J. Lewis, The effect of molecular structure on olfactory discrimination by the parasitoid Microplitis croceipes, Chem. Senses 27 (9) (2002) 811–816. [76] W. Lewis, et al., The role of kairomones and other behavioral chemicals in host finding by parasitic insects, Behav. Biol. 16 (3) (1976) 267–289. [77] S.L. Utley, G.C. Rains, W.J. Lewis, Behavioral monitoring of Microplitis croceipes: a parasitoid wasp, for detecting target odorants using a computer vision system, Trans. Am. Soc. Agric. Biol. Eng.: ASABE 50 (5) (2007) 1843–1849. [78] K. Takasu, W. Lewis, The role of learning in adult food location by the larval parasitoid Microplitis croceipes (Hymenoptera: Braconidae), J. Insect Behav. 9 (2) (1996) 265–281. [79] J.K. Tomberlin, et al., Conditioned Microplitis croceipes Cresson (Hymenoptera: Braconidae) detect and respond to 2,4-DNT: development of a biological sensor, J. Forensic Sci. 50 (5) (2005) 1187–1190. [80] K. Takasu, G.C. Rains, W.J. Lewis, Comparison of detection ability of learned odors between males and females in the larval parasitoid Microplitis croceipes, Entomol. Exp. Appl. 122 (3) (2007) 247–251. [81] Y. Li, J. Dickens, W. Steiner, Antennal olfactory responsiveness of Microplitis croceipes (Hymenoptera: Broconidae) to cotton plant volatiles, J. Chem. Ecol. 18 (10) (1992) 1761–1771. [82] T. Balch, Z. Khan, Veloso, Automatically tracking and analyzing the behavior of live insect colonies, in: Proceedings of the 5th International Conference on Autonomous Agents, ACM Press, New York, NY, 2001, pp. 521–528. [83] M. Bitterman, et al., Classical conditioning of proboscis extension in honeybees (Apis mellifera), J. Comp. Psychol. 97 (2) (1983) 107–119. [84] J. Carcaud, et al., Odour aversion after olfactory conditioning of the sting extension reflex in honeybees, J. Exp. Biol. 212 (Pt 5) (2009) 620–626. [85] M. Hammer, R. Menzel, Learning and memory in the honeybee, J. Neurosci. 15 (3) (1995) 1617–1630. [86] V.H. Meller, R.L. Davis, Biochemistry of insect learning: lessons from bees and flies, Insect Biochem. Mol. Biol. 26 (4) (1996) 327–335. [87] K. Takeda, Classical conditioned response in the honey bee, J. Insect Physiol. 6 (3) (1961) 168–179. [88] J.A. Shaw, et al., Polarization lidar measurements of honey bees in flight for locating land mines, Opt. Express 13 (15) (2005) 5853–5863. [89] S. Bender, et al., Tracking honey bees using lidar (light detection and ranging) technology, in: Sandia Report SAND2003-0184, Sandia National Laboratory, Albuquerque, NM, 2003p. 87185. [90] J.R. Riley, A.D. Smith, Design considerations for an harmonic radar to investigate the flight of insects at low altitude, Comput. Electron. Agric. 35 (2–3) (2002) 151– 169. [91] J. Riley, et al., Tracking bees with harmonic radar, Nature 379 (1996) 29–30.

O. Leitch et al. / Forensic Science International 232 (2013) 92–103 [92] K.C. Daly, B.H. Smith, Associative olfactory learning in the moth Manduca sexta, J. Exp. Biol. 203 (Pt 13) (2000) 2025–2038. [93] T.L. King, et al., Explosives detection with hard-wired moths, IEEE Trans. Instrum. Measure. 53 (4) (2004) 1113–1118. [94] K.C. Daly, et al., The generalization of olfactory-based conditioned response reveals unique but overlapping odour representations in the moth, Manduca sexta, J. Exp. Biol. 204 (Pt 17) (2001) 3085–3095. [95] K.C. Daly, L.A. Carrell, E. Mwilaria, Detection versus perception: physiological and behavioural analysis of olfactory sensitivity in the Moth (Manduca sexta), Behav. Neurosci. 121 (4) (2007) 794–807. [96] P. Davis, L. Wadhams, J. Baylis, Detection of odors using insects, UK, 2005. [97] D.R. Raman, R.R. Gerhardt, An insect-based detector of bovine manure odors, Trans. Am. Soc. Agric. Eng.: ASAE 40 (6) (1997) 1699–1707. [98] K.C. Park, et al., Odor discrimination using electroantennagram responses from an insect antennal array, Chem. Senses 27 (4) (2002) 343–352. [99] Incentinel Ltd., 2012 Available from: http://www.inscentinel.com/index.html (cited 24.03.12). [100] R. Glatz, K. Bailey-Hill, Mimicking nature’s noses: from receptor deorphaning to olfactory biosensing, Prog. Neurobiol. 93 (2) (2011) 270–296. [101] C. Liao, et al., Behavioural and genetic evidence for C. elegans’ ability to detect volatile chemicals associated with explosives, PLoS ONE 5 (9) (2010) e12615. [102] B. Marshall, C.G. Warr, M. de Bruyne, Detection of volatile indicators of illicit substances by the olfactory receptors of Drosophila melanogaster, Chem. Senses 35 (7) (2010) 613–625. [103] Syntech1, Electroantennagraphy: A Practical Introduction, Hilversum, The Netherlands, 2004.

103

[104] Y. Kuwana, I. Shimoyama, A pheromone-guided mobile robot that behaves like a silkworm moth with living antennae as pheromone sensors, Int. J. Robot. Res. 17 (9) (1998) 924–933. [105] M. Scho¨ning, et al., A BioFET on the basis of intact insect antennae, Sens. Actuators B 47 (1–3) (1998) 234–237. [106] S. Schu¨tz, et al., Detection of volatiles released by diseased potato tubers using a bio-sensor on the basis of intact insect antennae, Biosens. Bioelectron. 14 (2) (1999) 221–228. [107] V. Radhika, et al., Chemical sensing of DNT by engineered olfactory yeast strain, Nat. Chem. Biol. 3 (6) (2007) 325–330. [108] E. Hallem, J. Carlson, Coding of odors by a receptor repertoire, Cell 125 (1) (2006) 143–160. [109] S. Kreher, J. Kwon, J.R. Carlson, The molecular basis of odor coding in the Drosophila larva, Neuron 46 (3) (2005) 445–456. [110] J. Chou, et al., Olfactory recognition and discrimination in Caenorhabditis elegans, in: Symposia on Quantitative Biology, Cold Springs Habor, 1996, pp. 157–164. [111] C. Hill, et al., G protein-coupled receptors in Anopheles gambiae, Science 298 (5591) (2002) 176–178. [112] H. Robertson, K. Wanner, The chemoreceptor superfamily in the honey bee Apis mellifera: expansion of the odorant, but not gustatory, receptor family, Genome Res. 16 (11) (2006) 1395–1403. [113] L. Kent, K.K. Walden, H.M. Robertson, The Gr family or candidate gustotory and olfactory receptors in the yellow-fever mosquito Aedes aegypti, Chem. Senses 33 (1) (2008) 79–93. [114] H. Dacres, et al., Greatly enhanced detection of a volatile ligand at femtomolar levels using bioluminescence resonance energy transfer (BRET), Biosens. Bioelectron. 29 (1) (2011) 119–124.