Immunosensor for Assessing the Welfare of Trainee Guide Dogs

Cortisol is a well established biomarker hormone that regulates many processes in the body and is widely referred to as the stress hormone. Cortisol can be used as a stress marker to allow for detection of stress levels in dogs during the training process. This test will indicate if they will handle the stress under the training or if they might be more suitable as an assistant or companion dog. An immunosensor for detection of cortisol was developed using electrochemical impedance spectroscopy (EIS). The sensor was characterized using chemical and topographical techniques. The sensor was calibrated and its sensitivity determined using a cortisol concentration range of 0.0005 to 50 μg/mL. The theoretical limit of detection was found to be 3.57 fg/mL. When the immunosensor was tested on canine saliva samples, cortisol was detected and measured within the relevant physiological ranges in dogs.


Introduction
Guide dogs have an important role as they are the eyes of their visually impaired companion. However, not all dogs are capable or suitable to function as a guide dog. A guide dog's ability to guide an individual can be affected by external factors in their surroundings, such as loud noises, social interactions and environmental factors [1]. Certain dogs can deal with stressful situations better than others, and thus it is important for both organizations and trainers to better understand the level of stress that a guide dog is experiencing. Stress results in a higher level of cortisol present in their saliva [2]. Several factors can affect canine cortisol levels including both physical and cognitive activity. Cortisol levels in dogs may also indicate if a dog is suffering from an illness. Hypercortisolism and hypocortisolism are medical conditions resulting from excess and deficient cortisol levels in dogs, respectively [3]. Cortisol levels in working dogs have been found to increase as a result of their working activities [1], in addition to health conditions such as those noted above. Hence, this parameter could be useful towards informing the selection of dogs which undergo the expensive and resource consuming guide-dog training process, in order to optimize the success rate for selecting dogs appropriately. Biosensors have been tested with saliva and other body fluids such as blood, interstitial fluid and urine, all of which will give the amount of cortisol present [4]. The stress response of a number of animals has been assessed through saliva analysis, owing to the link between saliva and plasma cortisol concentrations. Moreover, saliva sampling offers a non-invasive means of assessing animal behavior [5][6][7]. There are a number of methods and indications that can be used to assess a canine's stress levels [8]; however, salivary cortisol is one of the more commonly used techniques to measure the stress levels in animals.
Several studies using human salivary have been developed. Kämäräinen et al. [9] developed an SPE coated with a cortisol-alkaline phosphatase conjugate and using square wave voltammetry; the sensor showed a working concentration range of 0.2 to 44.6 ng/mL. The results were highly reproducible within the desired physiological range for human salivary cortisol and showed a 0.90 correlation in results when compared with ultra-high pressure liquid chromatography tandem mass spectrometry [9]. Recent research states the potential use of dual electrochemical sensor systems (for cortisol and lactate) using an electro-reduced graphene oxide screen-printed electrode, also showing high sensitivity and specificity [10].
Enzyme-linked immunosorbent assay (ELISA) is the most used technique for detection of cortisol levels because of the sensitivity and versatility to evaluate protein concentrations [11]. The development of a reliable, non-invasive point of care is needed to help animals with stress-related problems. These types of point of care can deliver quick real-time cortisol values [12].
Saliva has been used in a number of studies to evaluate dog and animal welfare [5,[13][14][15][16]. Cobb et al., documented a cortisol concentration range of 0 to 337.9 ng/mL, with a mean of 4.5 ng/mL and a median of 1.5 ng/mL on studying meta-analysis of salivary cortisol levels [1]. Plasma cortisol and salivary cortisol levels in dogs are very closely related and it has been demonstrated that if the collection time takes less than 4 min, there will be no handling effect on the cortisol concentration [17]. Cobb et al. stated that no significant variations were evident in salivary cortisol concentrations of different dogs based on body weight or coat color or type (i.e., pet, assistance or therapy dog, military etc.), but found that unneutered females generally have higher cortisol concentrations than neutered females, males and unneutered males [1]. In an investigation by Batt et al. [18] cortisol concentrations were measured using an enzyme immunoassay from the saliva samples of potential guide dogs at 6 months of age and 14 months of age (before training commenced). Once a dog was determined to be suitable to work as a guide dog, cortisol concentrations were measured again using the same procedure. The study noted that at 6 months of age, dogs with higher cortisol levels were found to be stressed when in an unfamiliar environment. Furthermore, once training was complete the cortisol levels were found to be significantly higher in comparison to measurements obtained prior to training. However, it was unclear what exactly this elevation reflected. It may have been a result of maturation or the effect of prolonged kenneling during the training period.
Another study examined cortisol levels in animal-assisted activity dogs to monitor the welfare of these dogs [19]. Cortisol levels were found to be significantly raised when the dogs were exposed to a novel setting (3.97 ng/mL) compared to a home setting (2.13 ng/mL) or after activity without interaction with a stranger (2.57 ng/mL). The methodology applied in that research is not dissimilar to this study-cortisol samples were collected before, during and after 60-min training sessions (across three different settings). For this study, the samples were collected from the trainee guide dogs also before and after a behaviour assessment session. Immunosensors have a number of advantages over conventional analytical methods, including increased detection ranges, higher sensitivities in addition to rapid selective detection. Hence, they are an attractive and appropriate technique for evaluating stress levels [20,21].
Electrochemical sensors have been reported in the literature for the detection of cortisol, but only few of these are related to dogs [14]. Both antibodies and aptamers can be used to assemble cortisol sensors as can be seen in Table 1 [9,.
In this work, we describe a successful label-free method for ultrasensitive cortisol detection using an electrochemical impedance immunosensor in canine saliva obtained from Irish Guide Dogs for the Blind (IGDB). Here, we show that polyaniline (PANI) electrodeposited on graphene can be used for EIS cortisol analysis in both buffer and dog saliva samples. A linear response to cortisol concentrations between 0.0005 and 50 µg/mL, and an LOD of 3.57 fg/mL was achieved.  Electrochemical impedance spectroscopy (EIS) and Cyclic voltammetry (CV) were done with a potentiostat, Metrohm Autolab (Utrecht, The Netherlands), MAC90389, in connection to a laptop and measurements were carried out using Nova software. Graphene-SPEs were obtained from DropSens (DRP-110GPH). Raman measurements were done with a Renishaw InVia system with a wavelength of 514 nm and power of 7 mW at 100%, using an objective magnification of 50. SEM were completed with an FEI Quanta 650 SEM (ThermoFischer Scientific, Waltham, MA, USA).

Fabrication of the Immunosensor
PANI deposition was performed as described in our previous work [45]. Briefly, PANI was obtained by electropolymerization of 0.1 M aniline, with a potential sweep between −0.1 and +1.2 V, 50 mV/s for 20 cycles. The working electrode (WE) was rinsed with PBS prior to adding 8 µL of anti-cortisol on the surface (incubating a 100 µg/mL antibody solution in 25 mmol/L EDAC and 50 mmol/L of NHS). The immobilization of antibodies using the NHS/EDAC method does not ensure that the antibody is in the correct orientation after immobilisation-some of binding sites may not be available for cortisol-but helps to improve the binding. In addition, the cortisol should bind to the binding sites of antibodies at their Fab parts, not between them. The sensor was then left at room temperature for 2 h. BSA (0.5 mg/mL) was also added to the modified WE for 30 min at room temperature to prevent unspecific surface binding (Figure 1). The modified surface was stored at 4 • C until testing commenced. Electrochemical impedance spectroscopy (EIS) and Cyclic voltammetry (CV) were done with a potentiostat, Metrohm Autolab (Utrecht, The Netherlands), MAC90389, in connection to a laptop and measurements were carried out using Nova software. Graphene-SPEs were obtained from DropSens (DRP-110GPH). Raman measurements were done with a Renishaw InVia system with a wavelength of 514 nm and power of 7 mW at 100%, using an objective magnification of 50. SEM were completed with an FEI Quanta 650 SEM (ThermoFischer Scientific, Waltham, MA, USA).

Fabrication of the Immunosensor
PANI deposition was performed as described in our previous work [45]. Briefly, PANI was obtained by electropolymerization of 0.1 M aniline, with a potential sweep between −0.1 and +1.2 V, 50 mV/s for 20 cycles. The working electrode (WE) was rinsed with PBS prior to adding 8 μL of anti-cortisol on the surface (incubating a 100 μg/mL antibody solution in 25 mmol/L EDAC and 50 mmol/L of NHS). The immobilization of antibodies using the NHS/EDAC method does not ensure that the antibody is in the correct orientation after immobilisation-some of binding sites may not be available for cortisol-but helps to improve the binding. In addition, the cortisol should bind to the binding sites of antibodies at their Fab parts, not between them. The sensor was then left at room temperature for 2 h. BSA (0.5 mg/mL) was also added to the modified WE for 30 min at room temperature to prevent unspecific surface binding (Figure 1). The modified surface was stored at 4 °C until testing commenced.

Characterization and Measurements
The cortisol levels in the dogs' saliva samples and the surface functionalization process were evaluated and performed by EIS and CV. CV measurements were performed in 5.0 mmol/L of [Fe(CN) 6 ] 3− /[Fe(CN) 6 ] 4− at 50 mV/s from −0.7 V to +0.7 V. EIS were tested using an amplitude of 100 mV at a potential of +0.10 V, and a frequency equivalent to 50 Hz, with a frequency range of 1000 Hz-0.05 Hz. The impedance data were fitted to a R(C[R(W)]) Randles equivalent circuit using the Nova Software.
The WE was covered with 8 µL of sample solution and incubated at room temperature for 15 min. Cortisol was also tested in canine saliva samples.

Canine Saliva Handling Process
Sample were collected from the dogs at Irish Guide Dogs for the Blind, then centrifuged at 2000 rpm for 10 min. From this, the supernatant solution was collected and 8µL was placed onto the WE sensor. Samples were kept at −20 • C until use.

Characterization of the Immunosensor Surface Chemistry
The morphology of the sensor was assessed during the fabrication process. Raman measurements were taken before and after PANI modification to better understand the structure and defects of the graphene layer ( Figure 2) [46]. The control graphene displays three peaks at 1300, 1600 and 2700 cm −1 , which correspond to the D, G and 2D peaks, respectively [47]. Graphene modified with PANI ( Figure 2B) and graphene modified with PANI and Ab ( Figure 2C) show a similar spectra compared to the control graphene, with an increased intensity on the D peak. The introduction of a PANI layer on the surface of graphene produces this type of increase on the D peak [48], which is a known Raman spectrum attributed to PANI, indicating that PANI has successfully deposited onto the graphene layer. Other distinct bands attributed to PANI were also visible [47] and overlapping with the standard graphene spectra, but the sp 3 graphene peak becomes more intense Graphene modified with PANI ( Figure 2B) and graphene modified with PANI and Ab ( Figure 2C) show a similar spectra compared to the control graphene, with an increased intensity on the D peak. The introduction of a PANI layer on the surface of graphene produces this type of increase on the D peak [48], which is a known Raman spectrum attributed to PANI, indicating that PANI has successfully deposited onto the graphene layer. Other distinct bands attributed to PANI were also visible [47] and overlapping with the standard graphene spectra, but the sp 3 graphene peak becomes more intense after PANI electropolymerization.
When the anti-cortisol binds, an additional intensification in the peak intensity of the original D and G peaks ( Figure 2C) can be observed, showing that anti-cortisol has bound to the PANI graphene layer.

Scanning Electron Microscopy (SEM)
SEM images of the control graphene ( Figure 3A) and PANI modified graphene ( Figure 3B) surfaces. This imaging allows for visualisation of the surface morphology and further proves that a uniform PANI deposition on the conductive graphene layer was successful. The control graphene shows a smooth surface with some significant topographical features. When PANI is electropolymerized, a rougher surface is introduced, which shows the crystalline form of the polymer [49].  Figure 4A,B shows the modification layers at the sensor surface. Electrochemical data showed an Rct increased ( Figure 4B) compared to control sensor which is observed on the CV data obtained ( Figure 4A). The immunosensor shows an increase in the current after the PANI deposition, as seen by [50], and a decrease when Ab and BSA are added onto the surface, because of the conductivity of the same decreases compared to the PANI modified surface, showing a decrease in the potential peak separation ( Figure 4A), which is related with the decreased conductivity of the Ab and BSA compared to PANI graphene, resulting in a lower current flow [51].  Figure 4A,B shows the modification layers at the sensor surface. Electrochemical data showed an R ct increased ( Figure 4B) compared to control sensor which is observed on the CV data obtained ( Figure 4A). The immunosensor shows an increase in the current after the PANI deposition, as seen by [50], and a decrease when Ab and BSA are added onto the surface, because of the conductivity of the same decreases compared to the PANI modified surface, showing a decrease in the potential peak separation ( Figure 4A), which is related with the decreased conductivity of the Ab and BSA compared to PANI graphene, resulting in a lower current flow [51]. Figure 4C shows the impedance plots of the fabricated immunosensor with the different cortisol concentrations, and Figure 4D shows the matching calibration curve. The concentration range of cortisol used for the calibration curve was between 0.0005 and 50 µg/mL. On the EIS spectrum, no diffusion-controlled effect was observed [52]. The Rct increased with the increase of cortisol concentration. One of the justifications for this behaviour is that the proteins' structures when attach to the surface will act as a barrier to the electrical transfer. The immunosensor presents a sensitivity of 0.52 KΩ and a R 2 of 0.97. The theoretical limit of detection was 3.57 fg/mL. showed an Rct increased ( Figure 4B) compared to control sensor which is observed on the CV data obtained ( Figure 4A). The immunosensor shows an increase in the current after the PANI deposition, as seen by [50], and a decrease when Ab and BSA are added onto the surface, because of the conductivity of the same decreases compared to the PANI modified surface, showing a decrease in the potential peak separation ( Figure 4A), which is related with the decreased conductivity of the Ab and BSA compared to PANI graphene, resulting in a lower current flow [51].

Controls
Controls were performed to understand the importance of specific binding for the analyte in question. For that, three different processes were conducted, a base graphene, a PANI graphene and an anti-IgG antibody. Figure 5 shows that there is a significant difference (p < 0.0001) between the anticortisol antibody sensor, the base graphene SPE and the PANI graphene SPE detection. The results confirm the affinity of the analyte towards the antibody, and its capability for signal amplification and improved sensitivity [53]. The use of bare graphene-SPE and PANI/graphene-SPEs is not adequate for the detection and quantification of cortisol protein. This is due to the inability of the data to provide an appropriate calibration curve. Between an anti-cortisol antibody sensor and an anti-IgG antibody sensor, a significant difference (p < 0.0001) can be seen, confirming the specificity of the cortisol antibody and the capacity to distinguish between different analytes in biological samples. If we assume that the response of each sensor was irreversible because of the bond between antibody antigen interaction, we can access the precision of the sensor comparing values from different sensors. For this, relative impedance was considered and the mean (±standard deviation) of the slope and intercept were 1.6985 (±3.04 × 10 2 ) and 0.4379 (±1.34 × 10 2 ) for Ab and PANI, respectively. Each sensor is single use, so after calibration, it is not activated anymore, so a direct comparison between them can be done only in relation with the mean relative Rct values (R', relative to the blank signal of each device). The mean calibration data in graphene and anti-IgG Ab were R' = 0.1036 × log(cortisol, ng/mL) + 0.9559 (R2 = 0.5238) and R' = 0.1079 × log(cortisol, ng/mL) + 0.4936 (R2 = 0.9083), respectively. The calibration performed with anti-cortisol Ab is the greatest in terms of analytical performance.
Ab and PANI, respectively. Each sensor is single use, so after calibration, it is not activated anymore, so a direct comparison between them can be done only in relation with the mean relative Rct values (R', relative to the blank signal of each device). The mean calibration data in graphene and anti-IgG Ab were R' = 0.1036 × log(cortisol, ng/mL) + 0.9559 (R2 = 0.5238) and R' = 0.1079 × log(cortisol, ng/mL) + 0.4936 (R2 = 0.9083), respectively. The calibration performed with anti-cortisol Ab is the greatest in terms of analytical performance.

Interference of the Immunosensor
Selectivity was performed to evaluate the immunosensor to cortisol and understand non-specific binding; a different hormone, progesterone was used. The calibration curve results are shown in Figure 6. A linear reply from 0.0005 to 50 µg/mL, with a slope of 1.1 kΩ/[cortisol, µg/mL] were detected. Comparing both the cortisol 1.1 kΩ/[cortisol, µg/mL], and progesterone 0.64 kΩ/[progesterone, µg/mL], we can observe a decrease in sensitivity when using the interfering, suggesting that the immunosensor has good selectivity.

Interference of the Immunosensor
Selectivity was performed to evaluate the immunosensor to cortisol and understand non-specific binding; a different hormone, progesterone was used. The calibration curve results are shown in Figure 6. A linear reply from 0.0005 to 50 μg/mL, with a slope of 1.1 kΩ/[cortisol, μg/mL] were detected. Comparing both the cortisol 1.1 kΩ/[cortisol, μg/mL], and progesterone 0.64 kΩ/[progesterone, μg/mL], we can observe a decrease in sensitivity when using the interfering, suggesting that the immunosensor has good selectivity.

Evaluation of Cortisol Levels in Canine Saliva Sample
Values of cortisol in saliva samples from dogs were determined using impedance values on calibration of BSA/anti-cortisol/PANI/graphene-SPE with saliva samples collected from guide dogs. The samples were collected on two different days, before and after training, as shown in Table 2. The samples tested show an increase of approximately two-fold in salivary cortisol concentrations after the training session. However, one dog (Flora) was the only one that appears to have had considerably higher cortisol levels, and indeed this dog was a very stressed pup, according to the trainer, which makes this result accurate with the condition of the dog. All other samples were not meaningfully above or below the average cortisol level, suggesting little to no stress in the dogs. Cobb et al. [3], documented a cortisol concentration range of 0 to 337.9 ng/mL, with a mean of 4.5 ng/mL and a median of 1.5 ng/mL on studying meta-analysis of salivary cortisol levels.

Evaluation of Cortisol Levels in Canine Saliva Sample
Values of cortisol in saliva samples from dogs were determined using impedance values on calibration of BSA/anti-cortisol/PANI/graphene-SPE with saliva samples collected from guide dogs. The samples were collected on two different days, before and after training, as shown in Table 2. The samples tested show an increase of approximately two-fold in salivary cortisol concentrations after the training session. However, one dog (Flora) was the only one that appears to have had considerably higher cortisol levels, and indeed this dog was a very stressed pup, according to the trainer, which makes this result accurate with the condition of the dog. All other samples were not meaningfully above or below the average cortisol level, suggesting little to no stress in the dogs. Cobb et al. [3], documented a cortisol concentration range of 0 to 337.9 ng/mL, with a mean of 4.5 ng/mL and a median of 1.5 ng/mL on studying meta-analysis of salivary cortisol levels.

Conclusions
In this study, we present a simple modification approach to a graphene immunosensor for antibody binding with a suitable orientation for antigen binding. The resistance of the immunosensor increases with the increase of log(cortisol) concentration. The sensors demonstrated a good analytical performance in addition to measuring a cortisol concentrations applicable to the physiological levels present in dog saliva with a linear detection range from 0.0005 to 50 µg/mL and a theoretical limit of detection of 3.57 fg/mL. This immunosensor can be used as a POC device to assess welfare in dogs.