VSDI data were acquired from 32 adult male Wistar rats (Charles River Laboratories, Germany). Water and food were offered ad libidum. In vivo experiments were approved by the German Animal Care and Use Committee (reference number: 9.93.2.10.32.07.022) and were performed in accordance with the Deutsches Tierschutzgesetz (§ 8 Abs. 1 Tierschutzgesetz) and the NIH guidelines. All experiments involving animals were carried out in accordance with the European Union Community Council guidelines.

The surgery was performed as previously described (Rothermel et al., 2011). In brief, animals were initially anesthetized by an injection with chloral hydrate (intra peritoneal, 25% solution in saline, 400 mg kg⁻¹). Lidocaine (10 mg/ml) was applied to all incisions and pressure points. During experiments, animals were fixed in a stereotactic apparatus. They were tracheotomized and artificially ventilated (50–75 cycles/min, 5–6.5 ml tidal volume; UGO BASILE, Italy). Anesthesia was maintained using isoflurane (1%). An electrocardiogram, the exhaled CO₂ level, and the rectal temperature were continuously monitored (core temperature was held at 37.5°C). A craniotomy was performed and the hemispheres were gently removed to access the TG at the base of the skull (Stüttgen et al., 2006). Since the decerebration causes unconsciousness, the concentration of isofluran was decreased to less than 1% prior to starting the measurements in order to exclude interference with nociceptive channels (Matta et al., 2008). Measurements were performed for up to 8 h after decerebration. Animals were euthanized with an overdose of anesthetic after the experiments.

Recordings were performed using the voltage-sensitive dye RH-1838 (Optical Imaging Inc., USA) that binds to the external surface of cellular membranes without interfering with the cells physiological functions. It has an absorption maximum within the range of 550 and 570 nm and alters the intensity of emitted light (680–690 nm) as a function of changes in membrane potential. For loading of the TG, a micropipette, pulled from borosilicate glass capillaries (1.17 × 1.50 × 100 mm, Science Products, Germany) using a horizontal puller (Zeitz Instruments, Germany), was filled with saline solution containing the voltage-sensitive dye RH-1838. Afterwards, small amounts of dye were injected into the ganglion at 15–20 locations (total injected volume ~500 and 1000 nl) evenly distributed across the ganglion by manually applying slight pressure via a syringe directly connected to the micropipette.

In order to monitor optical signals arising from the TG, we used an Imager 3001 (Optical Imaging Inc., USA). All signals were collected at 100 Hz by a Dalstar 1M60 CCD camera (Dalsa, USA), coupled with a tandem lens macroscope (50 mm/1.2 toward camera and 50 mm/1.2 toward subject). The TG was illuminated with 630 ± 10 nm light and emitted light was high-pass filtered with a cutoff at 665 nm using a dichroid filter system. Data acquisition was triggered by coincident heart beat and respirator phase to minimize artifacts occurring from small movements induced by these events.

For nasal administration of volatile stimuli, we used an OL022 olfactometer (Burghart Medizintechnik, Germany). This olfactometer containing four vapor lines (flow rate 0.4 l/min). One of them was used as a control line (saturator tube contained aqua dest.), one line contained pure CO₂ (33% final concentration), which was used as a positive stimulus. Saturator tubes of the remaining channels were filled with varying substances. Vapor lines merge with a moistened background channel (flow rate 1 l/min) terminating in a single tube. This was gently adjusted to the nostril ipsilaterally to the imaged ganglion, in order to ensure that stimuli were exclusively administered to the animal's nose. A reverse tracheotomy was adapted as previously described for mice (Silver and Moulton, 1982; Wachowiak and Cohen, 2001), in order to ensure a constant nasal passage of the stimulating air stream thus minimizing mechanically and thermally induced trigeminal activations by alterations in the air stream. Animals were ventilated via the lower tracheotomy tube. Stimuli were applied in a randomized order for 4.8 s with inter-stimulus intervals of 20 s. For mechanical stimulation, we used forceps to slightly pinch the animal's nose (~0.5–1 s) once during recording from the TG.

First, each pixel was divisively normalized to its mean value during the pre-stimulus period (i.e., F/F₀). Afterwards, activation of each condition was compared to the mean of the two blank conditions and was expressed as the relative change in fluorescence (ΔF/F = F/F₀ − F_B/F_0B). Significance of signals across trials was expressed as z-score, where z = ΔF/F/√(SEM(F/F₀)² + SEM(F_B/F_0B)²) (SEM = standard error of the mean). Activity patterns of individual subjects were obtained by averaging the main response interval (2–4.8 s after stimulus onset). The stability of activity maps across all 29 animals which were stimulated with CO₂ (the remaining three were mechanically stimulated) was verified by statistical comparison of CO₂ evoked activity monitored from the animals number 1–15 with activity monitored from the animals number 16–29 by a Mann–Whitney-U-test. Onset latencies were defined as the time points at which signals exceeded a z-score ≥2. Unless otherwise stated all VSDI-data are presented as the median +75th percentile/−25th percentile. Normal-distribution was tested by using the Kolmogorov-Smirnov test (KS-test). Since not all data sets were normally distributed, Mann-Whitney-U-test was used for statistical comparison of averaged signal-intensities and onset-latencies. Statistical testing and the PCA were performed by using standard algorithms in Matlab (The MathWorks, USA). Countour plots were created with SigmaPlot (Systat software, USA), box plot diagrams with Origin (OriginLab Corporation, USA). Numbers of repeated experiments are indicated as n = X animals.

In this study, we used VSDI to monitor bioelectrical activity at the level of the TG of adult male wistar rats upon nasal stimulation with the volatile chemicals ethanol (EtOH), menthol, CO₂, eugenol, allylisothiocyanate (AITC), geraniol, vanillin, helional, sandalore, and 2-phenylethanol (PEE). Compounds were prepared as 1 mM in aqueous solution, except of AITC (150 μM), EtOH (99.99%), and CO₂ (33%).

The substances EtOH (n = 5), menthol (n = 6), AITC (n = 5), geraniol (n = 7), and eugenol (n = 6) were chosen for this experiment since they induce trigeminal sensations by activation or modulation of ligand- or voltage-gated channels (Trevisani et al., 2002; Clapham, 2003; Behrendt et al., 2004). Sandalore (n = 4) and helional (n = 5) are frequently used in fragrances, although until today no study described their effects on the TS. The substances vanillin (n = 4) and PEE (n = 5) were previously classified as pure olfactory stimuli (pure odors) reported to lack any trigeminal component (Doty et al., 1978; Radli and Wysocki, 1998; Cometto-Muñiz et al., 2005; Daiber et al., 2012). Interestingly, nasal administration of vanillin was shown to induce bioelectrical activity at the level of the TG in vivo (Rothermel et al., 2011) and was tested in order to further substantiate this finding. PEE was therefore used as a negative control. CO₂ (n = 29) was shown to activate exclusively the TS and was therefore administered in each experiment as a positive control (Doty et al., 1978).

In general, a high signal-to-noise ratio was attained in each experiment and signals were judged as statistically significant once exceeding a z-score ≥2. Signal onset latencies were defined as the interval between stimulus onset and detection of significant activity within a given range of interest (ROI). Interestingly, nasal application of all tested substances induced significant bioelectrical activity within the TG at the given concentrations. Exemplary z-score maps of bioelectrical activity including local-time courses of the monitored activity and z-score values in selected regions of interest are depicted in Figures 1A–J (II–III). In rare cases, we observed deflections in the local-time courses obtained from individual animals (Figures 1C,F) which seemed to be driven by the breathing or the respiration rhythm. These animals retained autonomous or spontaneous breathing, even after removal of the hemispheres. Breathing frequency in animals that revealed autonomous/spontaneous breathing occurred with an irregular frequency. The small movement correlating with single gasps may lead to these deflections that, in contrast to regular movements resulting from artificial respiration, could not be successfully filtered.

In order to generalize data from multiple animals, we measured maximum signal intensities (ΔF/F) and the respective onset latencies in 70 ROIs within activity maps obtained from single animals. The uniform distribution of these ROIs is depicted in Figure 2A (red marks). Each ROI covered an area of 300 μm². To exclude staining-specific intensity differences across animals, activity within individual ROIs was normalized to the ROI value representing the strongest response to CO₂, in the same animal (max-normalized). Finally, the datasets from multiple animals challenged with the same substance (containing signal intensities or onset latencies within these 70 ROIs), were averaged and median values of signal intensities and onset latencies were depicted as contour plots in Figures 3A–J (I and II).

Generally, strongest activity was detected in postero-medial regions (Figures 1A–J, 3A–J; I). The exact localizations of these spots were stimulus dependent, resulting in different stimulus-specific activity patterns, which are further described in the following section. Concurrently, strongest signals (40.78–109.57%) monitored within these areas displayed the shortest onset latencies (90–380 ms), which highly varied between the activity evoked by individual substances (Figures 3A–J; II). These regions were recently reported to house mainly nasal TG neurons (Lazarov, 2002; Rothermel et al., 2011). Originating from these regions, activity seemed to expand to more anterior and lateral parts of the ganglion, still reaching significant values although with significantly lesser intensities (34.89–76.9%; p = 5.74 ^* 10⁻⁵ to 2.25 ^* 10⁻³) and longer latencies (500–3200 ms; p = 3.32 ^* 10⁻⁶ to 6.56 ^* 10⁻³). This activity spread was not restricted to nasal stimulation with volatile substances but also occurred upon mechanical stimulation of the animal's nose (n = 3) thereby excluding that it is based on retronasal stimulation (Figures 4A–C). Alike chemical stimulation, mechanically-evoked activity occurred within postero-medial regions of the TG, thereby further emphasizing that chemical stimulation activates sensory neurons innervating nasal regions (Figures 4A–C).

Together, we can show that all tested substances induced significant bioelectrical activity in the TG in vivo and could thereby confirm that VSDI can be used to monitor bioelectrical activity arising from the rat TG in vivo, evoked by volatile chemicals including vanillin (Rothermel et al., 2011). Differences between monitored activity patterns, signal intensities, and onset latencies of the detected bioelectrical activity will be analyzed separately in the following.

Activity patterns elicited by nasal administration of the tested substances were monitored across a total of 29 animals. In order to ensure the stability of stimulus-specific patterns recorded across different animals, we used the CO₂-evoked patterns from all animals and divided them into two groups. Then, all values measured within each single ROI of the animals from group 1 were compared to the corresponding values measured within animals of group 2 by performing Wilcoxon-ranksum-tests. At a level of p < 0.05, we observed only in 1/70 areas (at the antero-lateral edge of the TG) a significant discrepancy between the two groups, indicating the stability of the monitored signals across different animals in all other areas.

In order to characterize spatio/temporal dynamics of the response patterns, we aimed to analyze the spatial distribution of stimulus-evoked activity independent from absolute signal intensities. For this, all ROI values were normalized relative to the mean of the ten highest ROI values within the same condition (within-normalized). Afterwards, within-normalized datasets elicited by the same stimulus in different animals could be averaged and depicted as contour plots (Figure 5A).

We then tested, to which extend these within-normalized activity patterns correlate to each other. Therefore, we used two mathematical approaches, namely calculation of the Spearman's rank correlation coefficient [Spearman's rho (ρ)] and a PCA. For both calculations, we used the within-normalized 70 point datasets, respectively. These approaches allowed sorting the monitored activity patters into three different groups (Figure 5D). Group 1 (highlighted in green) contained menthol and PEE, group 2 (highlighted in yellow) was the biggest group and contained the substances EtOH, eugenol, CO₂, geraniol, AITC, and vanillin, while group 3 (highlighted in blue) contained the substances sandalore and helional. Significant correlations (p < 0.05) between substances are represented in Figure 5D, which are indicated by lines between substances (gray: positive; red dotted: negative). The dimension of each correlation is given by the respective rank correlation coefficient (ρ). Correlations between stimulus-specific activity patterns were further substantiated by the performed PCA (Figures 5B,C). Again, the within-normalized and averaged stimulus-specific activity patterns were used for the calculation.

Second, intensities of stimulus-specific signals were compared among each other by calculating significance levels between all datasets (Wilcoxon-ranksum-test; p < 0.05), containing all averaged max-normalized median values measured within the 70 ROIs, respectively. In this way, the tested substances could be sorted in the following order: menthol > EtOH > eugenol > CO₂ > geraniol > AITC > vanillin > helional > sandalore > PEE (values can be derived from Figure 6A). Signals evoked by geraniol, AITC, vanillin, helional, sandalore, and PEE were significantly weaker than signals evoked by CO₂, eugenol, EtOH, and menthol. No significant differences were found between the intensities of signals evoked by menthol and EtOH and between signals evoked by EtOH and eugenol although those evoked by menthol were significantly stronger than those evoked by eugenol. Furthermore, signals induced by menthol, EtOH, and eugenol were stronger than those evoked by CO₂ (see Figure 6A). Next, we wanted to investigate how signal intensities correlate with onset latencies. Sorting the stimuli according to the means of the median onset-latencies revealed the following order: EtOH < CO₂ < eugenol < AITC < helional < sandalore < vanillin < menthol < geraniol < PEE (values can be derived from Figure 6B). Significances were compared as previously described for the comparison of signal intensities. Onset latencies of signals evoked by EtOH were significantly shorter than those by the other stimuli. No significant differences were observed among latency times of signals evoked by CO₂, eugenol, AITC, helional, and sandalore as well as between those of signals evoked by sandalore and vanillin. Furthermore, onset latencies of signals evoked by vanillin were not significantly different from the ones evoked by menthol and geraniol. Longest onset latencies were monitored upon application of PEE, which were significantly longer than those of signals evoked by all other substances tested (Figure 6B).

These results confirm the appearance of stimulus-specific patterns of bioelectrical activity at the level of the TG. Activity-patterns and signal-intensities seem to occur independent from each other [e.g., eugenol and geraniol elicit significantly correlating activity patterns (p = 1.51 ^* 10⁻¹²; ρ = 0.7234 (Figure 4B)] although with different intensities (p = 1.2 ^* 10⁻⁴), possibly indicating that the same group of sensory afferents is activated by the two substances but with different efficacies. Interestingly, we found no direct correlation between the intensity and the onset latency of stimulus-induced signals. Assuming that the onset latency of bioelectrical activity in the TG correlates with the perception of a sensation induced by a given substance, stimulus-specific onset latencies carry further information which is independent from stimulus-evoked activity patterns and the intensity of neuronal activity.

The previous findings revealed activity patterns characteristic for the monitored signals. Next, we wanted to investigate how the concentration of an administered stimulus correlates with the intensity of signals arising from the TG. Therefore, the olfactometer was equipped with aqueous solutions with various concentrations (0.01, 0.1, 1 mM) of menthol (n = 3–6), geraniol (n = 4–7), and sandalore (n = 4–5), respectively (Figure 7).

As described above, signal intensities (ΔF/F) were measured within 15 out of 70 ROIs uniformly distributed across the TG. These ROIs are indicated by numbers from 1 to 15 in Figure 2B. Data were normalized relative to the strongest ROI-value evoked by CO₂ in the same animal. Again, datasets were averaged across animals and the respective median values were depicted as bar diagrams (Figure 8), thereby allowing the direct comparison of signal amplitude and onset latency within the same ROIs.

During these recordings, all substances evoked significant signals in a concentration dependent manner (Figure 7). Again, strongest activity was observed in postero-medial parts of the TG (black arrows) while signals monitored in more antero-medial (red arrows) and lateral regions (blue arrows) displayed smaller amplitudes (Figures 7, 8).

On average, signals monitored upon application of 0.01 mM menthol significantly increased upon elevating the concentration to 0.1 mM menthol (p = 1.2 ^* 10⁻⁶) and to 1 mM (p = 3.39 ^* 10⁻⁶). The same applied to signals monitored upon application of geraniol (p_0.01/0.1= 2.6 ^* 10⁻⁴; p_0.1/1 = 0.014) and sandalore (p_0.01/0.1 = 0.0011; p_0.1/1 = 4.8 ^* 10⁻³) (values can be derived from Figure 9A). The increase of the amplitudes was accompanied by a decrease of the signal's onset latencies. Those of menthol-induced signals decreased significantly upon increasing the concentration from 0.01 to 1 mM (p_0.01/1 = 0.0462). A similar decrease of onset latencies was observed upon increasing the concentration of geraniol from 0.01 to 0.1 mM (p_0.01/0.1 = 1.2 ^* 10⁻⁴), although no further increase was evoked upon administration of 1 mM geraniol (p_0.1/1= 0.5896). Onset latencies of signals evoked by 0.01 mM signals were not significantly different from those evoked by 0.1 mM sandalore (0.1; p_0.01/0.1= 0.3757), although they significantly decreased upon administration of 1 mM sandalore (1 mM; p_0.1/1 =0.0028) (values can be derived from Figure 9B).

These findings demonstrate that signal amplitude and onset latency highly depend on the concentration of nasally administered stimuli. Together with the previous findings, our results indicate that different characteristics of a given substance, namely kind and concentration, are already represented at the level of the TG. We show for the first time that this becomes apparent by stimulus-specific spatial activity patterns, differences in signal intensities and onset latencies, together forming a complex representation of volatile chemicals at the level of the TG.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.