The “olfactory fingerprint”: can diagnostics be improved by combining canine and digital noses?

Introduction

The domestic dog (also known as Canis lupus familiaris) is a member of the genus Canis (canines), belonging to the family of wolf-like canids [1]. According to the most recent theory, dogs very likely descended from some extinct late Pleistocene grey wolf (Canis lupus) exemplars, which tended to live in East Asia approximately 100,000 years ago [2]. A remarkable post-divergence gene flow has since occurred between wolves and dogs, accompanied by expansion, contraction and replacement of the two populations around the world, so that modern wolves shall no longer be considered direct ancestors of domestic dogs which, in turn, have rapidly evolved taking profit from a mutually beneficial relationship with humans [3].

Although a definitive date for the first dog domestication cannot be accurately set, it is commonly thought that the human-dog relationships commenced some 35,000–37,000 years ago, much earlier than any other animal or plant domestication [4]. The most accredited scenario is that of grey wolf ancestors being attracted by human meal leftovers, which led to a gradual engagement in following human hunters for purposes of easier pickings. The increasing presence of these animals near human settlements became a deterrent for other dangerous carnivores from straying close to humans, thus generating a kind of mutually beneficial partnership for wolves and for their new (human) companions. Such an unconscious process of “proto”-domestication has then been followed by a more strict and real domestication, which probably commenced some 10,000–14,000 years ago [4]. During this second domestic process, an intense selective pressure was started, which led to the generation of smaller sized animals, with partially mutated phenotypes and expressing more friendly and docile behaviors [2]. A third domestication phase then initiated a few hundred years ago (nearly 200–300), characterized by systematic breeding practices which led to the appearance of hundreds of different breeds. Notably, genome-wide single nucleotide polymorphism (SNP) array studies now show that the modern dog breeds carry a kaleidoscope of differing genomic signatures compared to their ancestral progenitors, as well as an enhanced haplotype homozygosity and linkage disequilibrium, whose characteristics have enriched the Dog 10K Genomes Project and Dog Genome SNP Database (DoGSD). Overall, the broad degree of genetic diversity that has finally emerged across different dog populations drives the conclusion that dogs shall no longer be regarded as a single group when performing demographic inferences, whereby they can vary for sheer size, body, limb and skull proportions [5]. It is hence not surprising that dogs now display a much larger phenotypic heterogeneity than that observed throughout the whole order of carnivores, whereby their breeds span from the dolichocephalic, short-limbed dachshund to the wolf-like Alaskan malamute, from the 0.5 kg poodle to the over 90 kg mastiff, and even from the 20-cm-high Chihuahua to the over 80-cm-high Great Dane.

Regardless of their structural (and functional) heterogeneity, there is now little doubt that dogs have become “man’s best friend” and have been – and are still involved – in many different activities alongside their human companions. These basically include being “simple pets”, or else working as guard dogs, hunting dogs, herding dogs, guide dogs (e.g. for blind people), police dogs and – last but not least – sniffer (detection) dogs [6].

The mammal olfactory system

One of the most widely accepted definitions of “olfaction” is that of a sense mediating the perception of volatile chemicals, whose combination can also be referred to as “volatilome”. Irrespective of structural heterogeneities, a chemosensory system can be found in almost every animal species and has the main function of detecting changes in the biochemical structure of odorant molecules, in their concentration, as well as in their specific combination, thus enabling to garner vital information on the surrounding environment [7]. Briefly, the olfactory system in mammals is conventionally organized in two distinct organs within the nasal cavity, i.e. the main olfactory epithelium (MOE), containing olfactory receptors of class I and II (ORs class I and ORs class II), which mostly reacts to volatile odorants and is hence responsible for the conscious odor perception, along with the vomeronasal organ (VNO), containing vomeronasal receptors type 1 and 2 (V1R and V2R) and formyl peptide receptor-like proteins (FPRs), which preferentially reacts to pheromones and mediates a large number of neuroendocrine and behavioral responses [8]. Despite a basically similar structure, the olfactory system has evolved independently and rather differently in the various animal species, so that the functional odor perception now varies rather widely. A comparison of the main characteristics of olfactory system in humans and dogs is summarized in Table 1 [9], [10]. Compared to humans, dogs possess ~3-fold more genes encoding for olfactory receptors, an up to 30-fold larger olfactory epithelium extension, a nearly 50-fold higher number of olfactory receptors and a 3-fold bigger olfactory bulb size. Yet, dogs are not the best-smelling species in the animal kingdom, whereby the ratio between olfactory bulb to total encephalic volume is ~2% in mice, then followed by dogs (~0.3%) and goats (~0.2%). Notably, that same olfactory bulb to encephalic volume ratio is only ~0.1% in macaques and ~0.01% in humans, respectively [11].

The potential of olfactory receptors in recognizing a kaleidoscope of different odorant molecules is mostly attributable to the presence of hyper-variable domains located within some of their transmembrane domains [12]. Another interesting aspect is that the different dog breeds share some notable difference in the structure of their olfactory system. For example, the number of olfactory receptors approximates 150 million in the Dachshund, around 225 million in Beagles and German Shepherds, but is as high as 300 million in Bloodhounds. Not only the number but also the structure of olfactory receptors varies across dog breeds. Robin et al. sequenced nearly 100 olfactory receptor genes, which were representative enough of the entire canine repertoire [13]. Overall, a consistent number of SNPs was found in over 50% olfactory receptor genes in six different dog breeds, whereas the remaining genes were free of SNPs or were only modestly polymorphic. Interestingly, up to 25% of the observed SNPs were classified as being breed-specific, and half of these encoded for receptors with a modified ligand-binding capacity. Further studies confirmed that the genetic heterogeneity is associated with different smelling capabilities not only among the different dog breeds, but also in dogs belonging to the same breed [14], [15], [16].

Comparing the analytical sensitivity and instrumental approach of “electronic” and dog noses

The electronic nose, or e-Nose, was first formally defined in 1994 by Gardner and Bartlett, as “an instrument which comprises an array of electronic chemical sensors with partial specificity and appropriate pattern recognition system, capable of recognizing simple or complex odors” [17]. In defining this term, the authors discussed assay intricacy, limits of detection (LoD), and time-constraints on using conventional measurement techniques (e.g. gas chromatography-mass spectrometry, GC-MS) for the analysis of complex odor profiles. This defined a need to mimic the mammalian olfactory system in a manner which provided a rapid and responsive answer to the defined analytical question. To date, an array of technologies has been developed for mimicking the olfactory system to detect volatile chemicals including but not limited to acoustic, electrochemical, infrared and optical sensors [18]. Many of these developments have been applied within settings such as environmental and food science (e.g. [19], [20]); however, there has been a perpetuating interest over recent years in the human volatilome, with contributions in its make-up generated from exhaled breath, sweat, saliva, urine and feces [21]. To date, the study of this human “smell print” has been widely applied in applications as diverse as clinical investigations [22], psychological stress [23], sport and exercise science [24], and even for the potential purpose of locating trapped survivors of natural disasters [25]. These interests have led to a rapid rise in the application of technical “smelling” devices (i.e. e-Noses), with a 5-fold increase in annual publications on this topic in the previous decade (source: PubMed, December 2019).

Much of the research into human volatilome has been performed using offline MS-based assays (e.g. GC-MS) or have employed online/real-time measurements using large footprint and expensive MS instruments such as proton transfer reaction (PTR)-MS and selected ion flow tube (SIFT)-MS [26]. However, a recent push toward transportable MS systems for online measurements has provided excellent specificity and sensitivity without time-consuming sample preparation, chromatographic separation and high cost instrumentation, thus providing a sample delivery more akin to that entering a biological nose and thus more translatable to that which is inhaled by the “sniffer dog”. Examples of this include compact quadrupole mass spectrometers for direct analysis of exhaled breath [27], headspace analysis (i.e. volatile ‘smells’ above a sample) [28] and a direct-off-surface sampling approach for volatile compounds present in dried blood spot samples [29]. These compact MS instruments (e.g. Advion Expression CMS) can be set up on mobile trolley systems with the necessary infrastructure (e.g. N₂ generator and rotary pumps) to be transported to the clinical bedside or patient residences and are currently being assessed for clinical suitability of in-situ sampling and analysis [30]. In addition to this, developments of innovative MS ionization techniques have provided novel avenues for the direct measurement of non-volatile and semi-volatile molecules in biofluid samples, with semi-volatile molecules shown to be highly detectable by the canine nose and therefore relevant to the responses observed from bio-detection dogs [31]. A prominent example of this is paper spray ionization-MS where a biofluid can be spotted onto a filter paper “triangle”, and a high voltage is then applied which desorbs the molecules off the paper and into an adjacent MS instrument [32]. This technique is currently being applied to clinical cohorts for potential translation to everyday workflows [33], and can be combined to miniature MS instruments that can be carried around by a handle [34]. Furthermore, probe electrospray ionization-MS has shown excellent capability of instantaneous and direct measurement of biofluids for clinical purposes [35]. This technique uses an electrospray probe as a “dipstick”, causing molecules from the biosamples to adhere to the probes’ surface which are subsequently desorbed rapidly into the MS system following the application of a high voltage [36]. This approach has the potential for rapid diagnostic information on easily obtainable biofluids such as urine, saliva and sweat, thus allowing the expansion to investigate beyond the analysis of only volatile chemicals [37]. For these reasons, the application of a full spectrum analysis of chemical compounds, i.e. volatile, semi-volatile and non-volatile, using adapted e-Nose-style MS systems provides a fantastic analytical opportunity to both complement and add to the exciting prospect of resident “medical-detection” dog companions.

One of the most essential aspects that needs to be clarified before defining the analytical sensitivity of the canine olfactory apparatus is that the olfactory detection thresholds for all the different odorant molecules is strongly dependent on a genetic basis, as previously discussed [12]. The sensibility of different animal species to different chemical substances is hence extremely heterogeneous. For example, it has been now clearly established that dogs are extremely sensitive to certain amino acids (e.g. cysteine and methionine), aliphatic aldehydes (e.g. bourgeonal), acetate esters (e.g. amyl acetate) and sulfur-containing volatiles (e.g. butanethiol and phenylethyl sulfide), whilst they are virtually insensitive to other compounds that instead are strongly perceived by humans (e.g. aliphatic alcohols and benzene derivatives) [7]. Therefore, although it is common perception that dogs may be 10,000–100,000 more efficient than humans in their general sniffing capacity [38], it is not essentially true that dogs are always more sensitive than humans in smelling substances. It may also be hypothetically misleading to set a universal sensitivity threshold for the canine olfactory apparatus, whilst it seems more appropriate to use some practical examples for establishing “how low dogs could go” in detecting certain compounds. Walker et al. published the results of an interesting study, where the olfactory sensitivity of two different dogs (i.e. a Standard Schnauzer and a Rottweiler) for n-amyl acetate was comprehensively assessed [39]. After systematically lowering the n-amyl acetate concentration within Teflon boxes during several weeks, the lowest sniffing sensitivity (i.e. equaling the LoD of conventional laboratory methods) of the two dogs for the chemical compound was found to be as low as 1.14 and 1.90 parts per trillion (ppt), respectively. Considering that the molecular weight of n-amyl acetate is ~130 Da, these values will roughly correspond to a LoD comprised between 8.8 and 14.6 fmol/L. A virtually identical lower limit of sensitivity (i.e. 1.5 ppt, corresponding to ~11.5 fmol/L) has also been more recently identified by Concha et al. using 10 different dogs (3 Labrador Retrievers, 3 Working Cocker Spaniels, 2 English Springer Spaniels and 2 Border Collies) [40], though the animals displayed a high inter-individual response. Such an extraordinary analytical sensitivity almost translates into the capability of identifying one drop of a given liquid within nearly 20 Olympic-size (i.e. 760 m²) swimming pools [41]. Recent technical and analytical developments in MS-based applications have now made it possible to detect (and measure) metabolites at concentrations lower than pmol/L, approximating the femtomolar level (Figure 1) [42].

Figure 1:

Lower limit of detection for certain compounds of some laboratory techniques compared with dogs.

ELISA, enzyme-linked immunosorbent assays; NMR, nuclear magnetic resonance; GC-MS, gas chromatography coupled with mass spectrometry; LC-MS, liquid chromatography coupled with mass spectrometry.

Taken together, this evidence would lead us to conclude that the “analytical sensitivity” of canine olfactory apparatus for certain compounds is perhaps as low as that of the most sensitive laboratory techniques, thus opening intriguing scenarios for many civilian, military and forensic applications of canine olfactory detection, whereby the so-called “sniffer” (detection) dogs can be trained for detecting substances such as foods, humans, explosives, illegal drugs, wildlife scat, currency, contraband electronics (e.g. illicit mobile phones) and – last but not the least – human diseases.

Clinical trials with sniffer dogs

A sniffer dog is an animal trained to use its olfactory sense for detecting a vast array of substances, mostly volatile organic compounds (VOCs; mostly represented by low molecular weight chemicals, which evaporate easily under normal temperature and pressure conditions), also including those exceptionally or exclusively generated in patients with specific human diseases, whose variable association contribute to generate multiple specific volatilomes [41], [43]. Although the use of the so-called “bio-detection” or “medical-detection” dogs is just in embryo, being still far from regarded as routine replacement for other and more validated diagnostic investigations, the results of some encouraging studies will be reviewed and summarized in the following part of this article. Notably, some consistent evidence has already been garnered concerning some specific human diseases such as cancer, diabetes and seizures, promising findings are emerging for others such as narcolepsy and migraine, whilst preliminary evidence has also been recently published that sniffer dogs may be used for screening infectious diseases (Table 2).

Conclusions

The many important findings described in the previous sections of this article would hence cumulatively contribute to depict the dog smelling capacity applied to diagnostic medicine as a kind of “olfactory fingerprint” loop, which can be initiated at virtually any stage, as shown in Figure 2. It may start with the dog, as the evidence that the animal can sniff (and thereby anticipate or detect) a certain human pathology provides implicit evidence that the disease is associated with a specific volatilome, which can in turn be identified and dissected, paving the way to develop laboratory techniques for its routine assessment. On the other hand, the evidence that specific pathologies are associated with enhanced generation of certain VOCs may serve as a reliable basis for specifically training dogs to detect these compounds, even (or especially) in patients at an asymptomatic phase of disease. Due to their extremely fine analytical sensitivity, dogs may hypothetically be able to identify human diseases even earlier than present routine diagnostic investigations could do, using a substantially non-invasive approach, overcoming the many challenges associated with pre-analytical biospecimen management, while contextually providing companionship, loyalty and full devotion to their owners [67], [68]. Dogs have also very limited needs in terms of environment adaptation and are “fully portable”, even more easily than some point-of-care instruments. Unlike current laboratory techniques, their diagnostic efficiency is perhaps magnified by the fact that they would be capable of recognizing an entire olfactory fingerprint rather than a single compound. Dogs can then use a rich variety of signals to figure out what is probably happening to their owners. Not only can they accurately detect an “abnormal volatilome” as an index of something wrong occurring to their human companions, but they can also perceive some other visual, vocal and behavioral signals, which altogether would contribute to raise their alertness. This seems especially frequent in certain pathologies such as diabetes and seizures, whereby changes of posture or expression of owners are clear signs that the dog perceives and, perhaps unconsciously, translates into the conclusion that something bad or harmful will be happening soon [52]. A deep investigation of this innate dog’s ability can hence substantially expedite the development of new laboratory technology and instrumentation (i.e. biosensors and online MS-based e-Noses) aimed at reproducing the animal detection performance (Figure 2, Table 3). Although many future studies would be needed for more accurately identifying what trained and non-trained dogs are actually smelling (or – more generally – perceiving) on their owners, some automatic devices have been recently constructed for facilitating the training of dog sniffing capacity [69], and which may be ultimately effective in accelerating and improving the accuracy of detection.

Figure 2:

Synergy between dogs and laboratory testing for improving medical diagnostics.

That said, the most logical question here is whether we can envision a future with “bio-detection” or “medical-detection” (sniffer) dogs populating our hospitals, wearing lab equipment and performing complex diagnostics tasks (Figure 3). Neither would a crystal ball help predicting the possible scenarios, but some additional reflections can be made. First, the training of sniffer dogs, especially of “bio-detection” animals, is still a challenging and expensive activity. There are no official courses that have been established, nor have official guidelines been published on how these pets shall be educated. Then, despite some breeds being unquestionably better than others in smelling capability and the consequent disease detection (e.g. breeds selected for scent works outperform non-scent and short-nosed) [70], the inter-individual performance is also extremely variable, with some dogs capable of identifying cancerous specimens with a 1.00 sensitivity, whilst others performing only modestly better than chance.

Figure 3:

Exemplar of “bio-detection”, “medical-detection” (sniffer) dog.

It seems hence reasonable to conclude that sniffer dogs will never be completely “diagnostic” for a given human disease, similar to the fact that when police dogs point a specific baggage this does not automatically translate into a final sentence of guilt, at least until illicit or unfair goods (drugs, tobacco, food, money, etc.) are objectively detected by policemen. Similarly, dogs could only serve for screening diseases, as additional diagnostic investigations would then be needed for confirming the final diagnosis and for staging the pathology. Nevertheless, there is now virtually incontrovertible evidence that man’s best friends can really improve the quality of life of many diseased people, whereby accurately trained dogs would be extremely efficient in alerting their owners on some imminent life-threatening conditions, such as hypo- or hyperglycemia crises, seizures, and even on migraine episodes. Although no clinical trials have been published so far, to the best of our knowledge, on dogs’ capability to anticipate an acute ischemic event, the results of a recent survey on as many as 181,696 patients with a previous episode of acute myocardial infarction (AMI) and 154,617 patients with a previous episode of ischemic stroke [71], revealed that patients owning dogs had a ~30% lower risk of death after hospitalization for AMI (hazard ratio, 0.67; 95% confidence interval, 0.61–0.75) or ischemic stroke (hazard ratio, 0.73; 95% confidence interval, 0.66–0.80). This would finally reinforce the concept that dogs’ friendship, loyalty and devotion, coupled with their extraordinary sensory capability, shall be seen as an exceptional boon for human health.