Wine experts occasionally identify a broad variety of aromas such as »bacon fat, pain grillé, black raspberries, cassis, white flowers, and Provençal olives« when they taste, for example, a vintage of Côte Rôtie 1. Even an expert's nose, however, can get surprisingly confused in the presence of misleading visual cues. In a noteworthy study performed at the University of Bordeaux, Gil Morrot and collaborators asked 54 undergraduate students from the Faculty of Oenology to compare a glass of white Bordeaux wine (containing Sémillon and Sauvignon grapes) to a glass of exactly the same wine, which had been coloured by a mixture of red anthocyanins. Although the added anthocyanins had no perceptible taste or odour, the subjects identified remarkably different sets of aromas in both glasses 2. Despite this surprising lack of discriminatory skills, it seems our sense of smell has other proficiencies, such as a remarkable power of attributing emotional qualities to objects that we see. The attributive nature of the sense of smell is illustrated by the fact that we need so-called descriptors to communicate odours 3. What alternatives do we have but to portray an olfactory perception as »black raspberries-like«, or »Provençal olives-like«? None of these scents can be addressed directly, without reference to its respective source 4. Other perceptual qualities can often be named independently. We can speak, for example, of »red« and »sweet« without employing metaphors and similes such as »anthocyanin-like« or »sucrose-like«. Odours, on the other hand, can affect our emotional state profoundly 567. A gustatory stimulus, such as »too salty«, effectively keeps us from drinking sea water, but compared to bad odours, the aversion evoked by the taste of salt is less affective. This is vividly illustrated by an authentic report by the early Gabriel Garcia Marquez, who describes the fate of a shipwrecked sailor who drifts on his raft through the Caribbean sea. After four days of suffering the agonies of thirst, he decides to drink salt water. "This water does not satisfy your thirst", the salvaged sailor later testifies, "but it refreshes" 8. Apparently, the repulsion evoked by the salty taste is based on reflex and rational thinking rather than on emotional disgust. By the same token, sweetness alone will rarely make us long for artificial sweeteners and it is unlikely to evoke a similar savouriness as the scent of freshly baked bread. In contrast, the repulsiveness of the nauseating stench of rotting meat is so effective that the use of this smell was suggested as a powerful stimulus for therapeutical conditioning of sexual offenders 910.

It seems that an important role of the human sense of smell is adding emotional qualities to situations and objects that we see, and not so much to making an essential contribution to finding or identifying things. Of course we can recognize typical odours even with our eyes shut, but the scent of ripe bananas, for example, rather influences our appetite than being necessary to distinguish bananas from pineapples. This observation raises questions about the biological purpose of the olfactory sense in a visually dominated species, such as man. What is so important about watering our mouth that it exerts sufficient selection pressure to sustain several hundreds of functional olfactory receptor genes through evolution? Why do we even maintain the ability to regenerate olfactory neurons from adult stem cells, just to make sure that our sense of smell stays alive as long as we do?

The functional anatomy of the mammalian nose and its close contact to the oral cavity point to possible answers to these questions. Odour molecules approach the olfactory epithelium along two principle routes: When we sniff or breathe with our mouth closed, the inhaled air enters the lumen of the nasal cavity through external nostrils. Food almost touches our nares before it enters the oral cavity, and it is almost impossible not to stick our nose into a glass of wine that we taste. When we chew, we break down the food into a mush from which favours exhale. These volatile odour molecules then travel up the retro-nasal passage and reach the olfactory epithelium through internal nostrils, which connect the pharynx to the nasal cavity 11. The phylogenetic emergence of these internal nostrils (or choanae) was an important step in the adaptation of the olfactory system to air breathing 12. It now enables us to employ olfactory perception in the assessment of nutrients at every level of uptake. We can now even evaluate compounds in our food that are made accessible only by mechanical exposure. This close involvement of the olfactory system in the analysis of food along with the attributive nature of olfactory perception allow us to employ the sense of smell in aquiring refined food preferences. In this respect, humans are uniquely flexible. There might be some inherently aversive odours (such as the smell of rotting meat), but depending on trends and cultural background, we can enjoy almost everything, from fried locusts spiced with salt and lime (as served in the Mexican region of Oaxaca) to live oysters (which are perceived as a delicacy in a lot of wealthy Western societies). We can even "learn to enjoy things that we should not enjoy" 13. Used as flavoring agents, bitter compounds, including plant-derived phenols and polyphenols, flavonoids, catechins, and caffeine, enhance the sensory appeal of beverages and food, such as beer, coffee, chocolate, tea, and tonic water 14. This flexibility effectively supports opening up new food resources. In this context, olfactory conditioning helps us to pass on approved food preferences to our infant offspring, while at the same time protecting them from exploring potentially noxious material while unattended (long before they understand the advise that we give). Cooking-ambitious parents often experience how cautious and how annoyingly conservative their children can be, when it comes to trying out unknown delicacies. The juveniles may well obstinately insist on fish sticks with mashed potatoes when served cod steaks cooked in an »interesting spice mixture of ginger, cloves, chilies, cilantro leaves, and lemon juice« 15. Training pays out, however, and we can learn to expand our pool of nutrients to include surprisingly variable ingredients.

For free living rats, the importance of social learning of food preferences has also been demonstrated 16, and olfaction certainly plays an important role in this process 17. Mice prefer sweet over bitter, even if the sweet perception is triggered erroneously by a bitter compound 18. Humans, in contrast, are capable of expanding their menu well beyond sweet and non-bitter aliments, and they seem to do so by prioritizing hedonic odour qualities over gustatory perception, and by dynamically adapting their food preferences to changing age, resources, and trends.

The production of aromas and scents has long been an important economical factor 19 and an interesting perspective for biotechnological fabrication 20. Olfactory signals can act as mediator of social interactions 21 and the potential of olfactory cues as marketing tools in social and economical context has long been recognized 222324. Nevertheless, empirical research in the field of how we perceive odours is still fairly new, although we know more today about our brain's strategy in computing odour perception 25262728 than we did only a few years ago. Innovative techniques such as genetic tracing of neural circuits 293031 and functional magnetic resonance imaging (fMRI) 3233343536 produced exciting new results. It turns out, that, even in insects, coding and representation of odours is highly complex and dynamic 2728373839, but significant progress is made, and, for example, we now begin to understand why a typical white wine tastes like a typical red wine, just because we added a tiny pinch of neutral anthocyanins. In a striking fMRI study, Jay Gottfried and Raymond Dolan recently found evidence of why "the nose smells what the eye sees". Their data indicate that the "human hippocampus mediates a reactivation of crossmodal semantic associations, even in the absence of explicit memory processing" 3233. Apparently, »black raspberries and Provençal olives« are imprinted in our brain, while the nose just pulls the trigger to unfold them in our mind.

The initial event in all these fascinating aspects of odour sensation is the binding of odorant molecules 40 to olfactory receptor proteins 4142. The released binding energy is transduced by a complex chain of molecular interactions into electrochemical membrane potentials, which dynamically modulate neuronal odour representation 194344454647484950515253545556. How do these primary olfactory processes work and where do they occur?

A crux of this matter are the bipolar chemosensory neurons in the olfactory epithelium of the nose. Olfactory neurons are based at the front line between the brain and the odour-containing air that we breathe. Their surface membrane can be divided into two major spatial compartments, which are separated by a belt of tight junctional proteins 57. The basolateral surface membrane lies well protected inside of the olfactory epithelial tissue. It covers large part of the apical dendrite, the cell body, and an unbranched axon that projects to the brain. On the other hand, there is an apical surface membrane compartment, the size of which is comparable to that of the basal compartment. It is located externally, directly exposed to the air, and attached to the rest of the cell only through the nexus of a thin apical dendrite, which reaches into the lumen of the nasal cavity with its terminal knob. The apical surface membrane is embedded into mucus and airway surface fluid. These represent ionic and biochemical compartments, which are important for olfactory signal transduction 5859, as well as for innate defense of all airway epithelial surfaces 60. Disturbances of the composition of these fluids, e.g. by mutation of ion transport systems, can cause severe pathologies, including cystic fibrosis 61.

Through their apical membrane compartment, olfactory sensory neurons contact the ambient air. At this surface, they collect information about volatile odour molecules, and here odours are transduced into neuronal signals, which the brain can read and analyse 626364. The molecular machinery that mediates olfactory signal transduction has been thoroughly characterized. Briefly, volatile odour molecules bind to odorant receptors 65 and trigger the activation of G-proteins (Gα_olf) 47. This in turn stimulates adenylyl cyclase (type III) activity 66, and the resulting rise of intracellular adenosine 3',5'-cyclic monophosphate (cAMP) opens cyclic nucleotide-gated (CNG) cation channels 67. The elicited influx of Ca²⁺ from the extracellular space depolarizes the ciliary membrane and prompts a secondary opening of Ca²⁺-gated chloride channels. These are responsible for large part of the depolarizing current across the olfactory ciliary membrane 6869. The depolarizing nature of the induced chloride current is based on an unusually high intracellular chloride concentration in the cytoplasm and cilia of olfactory sensory neurons, which is replenished by sodium-potassium-chloride co-transporters (NKCC-1) in the basolateral surface membrane compartment 6869. Odour-induced signals are thought to be terminated as a result of multiple factors and mechanisms 70, although recent evidence indicates that the dwell-time of the receptor-odorant complex might be too small for some inactivation processes to occur under physiological conditions 71.

Beside this main stream model of olfactory signal transduction survived a scuff resistant hypothesis, proposing InsP₃ as an alternative second messenger in olfaction 507273. The possibility of a supplementary signalling pathway in olfactory neurons was initially raised by a study, in which the authors found that in contrast to most odours, some potent odorants failed to induce cAMP when applied to isolated olfactory cilia in biochemical assays 74. It was later found that these "non cAMP odours" induced InsP₃ instead 7576, and that odorant receptors coupled to either cAMP or InsP₃ when they were functionally expressed in insect cells, depending on the receptor's specificity for either "cAMP-" or "IP3-odours" 77. Despite these findings, the resulting speculation around a possible duality of signalling pathways in olfactory neurons remained controversial 4978. The goose of InsP₃ as an alternative second messenger in olfactory neurons got cooked, when mutant mice that lacked functional expression of Gα_olf 79, adenylyl cyclase (type III) 80, and cyclic nucleotide-gated ion channels 81 came out anosmic for both cAMP- and InsP₃-odours. Consequently, many researchers began to accept that cAMP might be the sole excitatory second messenger in olfaction 828384. The evidence linking phosphoinositide-related signalling to the mammalian olfactory epithelium, however, remained orphaned.

The apical surface membrane of olfactory sensory neurons is a platform that links the physical world of odour molecules to the mental world of odour perception. At this interface, the physiochemical properties of volatile compounds are transduced into the electrochemical membrane potentials, which modulate the information from which the brain constructs, for example, the aromas of »black raspberries« and »pain grillé«. The apical surface membrane compartment therefore represents a most important cell organelle. It is significantly enlarged by cilia or microvilli, which emerge from the terminal knob of the apical dendrite to protrude into the mucus 85. What is the biological significance of this enlargement? Why do some sensory neurons sustain this extension with cilia, while others possess microvilli or both?

A requirement for high sensitivity is often regarded as a sufficiently meaningful biological reason for the enlargement of the surface area of sensory membranes. In photoreceptor cells, this argument intuitively makes sense. A photon on its path through a photoreceptor cell is either captured by a photosensitive rhodopsin molecule or the information it carries will be lost by absorption in non sensory material. Therefore, photoreceptor cells increase their quantum efficiency by forming multilayered stacks of membranes that are tightly packed with photosensitive rhodopsin molecules. This functional cytoarchitecture can likewise be observed in vertebrate 86 and invertebrate eyes 87, where stacks of membrane discs or rhabdomeric microvilli ensure high quantum efficiency of each photosensitive cell. In the case of olfactory sensory neurons, however, the situation is slightly different. Compared to photons, odour molecules are stable and could theoretically be trapped and retained in a lipophilic environment until receptor activation has occurred. Furthermore, the physiological purpose of increasing sensitivity is also not as obvious as a first glance might suggest. It has been reported that olfactory neurons are sufficiently sensitive to detect even single odour molecules 88, but the biological significance of this finding was challenged 498990. Extraordinary sensitivity has undoubtedly been observed in the olfactory system of moths 91, but since only a few mammals fly around in the dark with a need to smell their food or mating partner from miles away, it remains unclear how solitary odour molecules could carry meaningful information and why they should be detected at all. Quantal sensitivity of olfactory neurons, similar to that observed in photoreceptor cells 9293, could also provide confusing sensory input to the brain. Olfactory neurons express only one or a few olfactory receptors 9495 and many olfactory receptors are broadly tuned to overlapping sets of qualitatively distinct odour molecules 626364. Thus, two identical solitary odour molecules could trigger different sensory input into the olfactory system, depending on which receptor they hit by chance. Such an arrangement does not appear like a plausible approach to ensure reliable and reproducible sensory input, particularly at extremely low levels of odorant concentration. Accordingly, in contrast to phototransduction, olfactory signal transduction lacks amplification at one of its very basic levels, namely that of receptor-G-protein-activation. The life-time of the receptor-odorant complex was found to be so short, that the complex might often dissociate even before a single G-protein was activated. Consequently, signal amplification by one active receptor triggering multiple downstream G-protein/effector enzyme molecules seems unplausible 71.

Alternative explanations for the observed enlargement of the apical membrane compartment of olfactory neurons include the outstanding ability of these cells to provide information that allows the detection of even minute differences between odour concentrations, or, equivalently, differential affinities of odour molecules to distinct receptor populations.

Even humans are capable of extracting spatial information from smell 969798. Dogs just need five footsteps to determine the direction of an odour trail 99 and a rat can determine the location of an odour source by stereo-localisation in a single sniff 100. To enable such impressing performances, it will most likely be necessary to maximize the dynamic sensitivity of olfactory neurons. The dynamic sensitivity represents the ability of a signalling system to respond to a small change of the input signal by modulating the output signal in a statistically significant manner. The dynamic sensitivity of an olfactory neuron is directly influenced by the size of its apical membrane surface compartment, because the latter is correlated to the number of signalling proteins (i.e. receptor molecules, G-proteins, and ion channels) and thus to the total number of elementary responses that can simultaneously be generated in each cell.

The requirements to a high performance odorant detector can be illustrated by comparing it to a light detecting device, such as a charge-coupled device (CCD) camera. Here, each photosensitive element (analogous to an olfactory neuron in the nose) is characterized by its background noise, its quantum efficiency, and its full well capacity 101. In a high performance CCD camera, noise reduction is achieved by cooling the sensor to low temperatures. Quantum efficiency of a photon detector describes the ratio between the number of photons registered by the detector to the number of photons entering the device. Lastly, the full well capacity of the photon detector describes the total number of photons that can be registered within a single readout cycle. To improve the dynamic sensitivity of any charge-coupled device, it is necessary to combine high quantum efficiency with a large full well capacity. If the ratio between full well capacity and quantum efficiency of the detector is low, the slope of its response curve will be shallow. This, in turn, makes it difficult to reliably detect minute differences in signal amplitude, particularly when background noise is high. Therefore, high performance light detectors require the optimization of all three parameters combined: low background noise, high quantum sensitivity, and large full well capacity.

For olfactory sensory neurons, similar requirements might apply. In the olfactory sytem, noise reduction is a multi level process 102. It begins with a nonlinear signal amplification by Ca²⁺-activated chloride currents 103104, continues by the convergence of dozens of cilia onto the terminal knob of the sensory neuron's apical dendrite, and is further sophisticated by the convergence of hundreds of olfactory neurons on a single glomerulus in the olfactory bulb 105106107. This multilevel integration with the possibility of adjusting thresholds at each level enables effective filtering of background noise, which has been observed as an intrinsic feature of olfactory signal transduction 102108109110111. To represent high performance detectors for chemical constituents of the air, olfactory neurons might also have to improve quantum efficiency and full well capacity by enlarging their apical membrane compartment. At a given odour concentration and affinity of the odorant to its receptor, the quantum efficiency of the olfactory neuron as a whole will directly depend on the total number of receptors present. The same holds true for the neuron's "full well capacity", which represents its capacity to register odorant-to-receptor-docking-events within a single readout cycle (which in this case might be equivalent to the duration of a sniff). Since there is a limit to packaging density of receptors and other signaling molecules within the surface membrane, these parameters will also depend on the total apical surface membrane area. Olfactory neurons therefore enlarge this area by the folding the apical membrane into cilia and/or microvilli 85. Interestingly, it seems that olfactory neurons further improve their physiological performance by additional measures. As mentioned above, the receptor-odorant dwell-time is very short, enabling repeated binding of odorant molecules to the same receptor. Furthermore, the quantal responses of olfactory neurons are very small. At the same time, the density of receptors in the ciliary membrane is so high that their response domains overlap, which causes non linear summation of the odour-induced unitary responses 71. Since linearity is neither necessary for detecting differential activation patterns (i.e. specific odours), nor for sensing relative differences in odour concentration, this strategy seems to improve the dynamic resolution of each olfactory neuron over a broad range of odour concentration levels and thus perfectly serves to the needs of a biological high performance chemodetector.

The olfactory epithelium of mammals consists of a relatively small number of cell types 132. A monolayer of supporting or sustentacular cells separates the nasal cavity from the sensory neuroepithelium. This palisade-like monolayer covers a pseudostratified layer of bipolar sensory neurons, which basally project an axon to the olfactory bulb (Figure 1). Apically, the olfactory neurons contact the lumen of the nasal cavity with the terminal knob of their apical dendrite (Figure 2). Between the cell bodies of the chemosensory neurons and the basal lamina is a thin layer of so-called globose and horizontal basal cells, which contains a population of adult neuronal stem cells 133134135136. These stem cells are important to maintain function and histology of the olfactory epithelium, because based on their exposed situation in contact to ambient air, olfactory neurons are prone to damage and infection. Consequently, olfactory neurons have a limited lifetime before they are periodically replaced 137138139140141.

In addition to these three well established major olfactory epithelial cell types, i.e. ciliated olfactory sensory neurons, microvillous supporting or sustentacular cells, and basal cells, at least five classes of so-called olfactory microvillar cells exist 132. To prevent confusion, Bert Menco and Edward Morrison suggested to systematically re-classify olfactory microvillar cells, and to combine all olfactory epithelial cells that have microvilli under the term "microvillous cells" 132. The function of this peculiar type of cells is totally unknown.

Some microvillous cells have been proposed to represent a second class of bipolar sensory neurons. Based on an electron microscopic study of the olfactory epithelium of rat, François Jourdan first made this suggestion in 1975 130. In his thesis, he described a microvillar type of bipolar olfactory cells (type B cells), the morphology of which, in his view, was that of a typical bipolar receptor neuron (Figure 3). Similar cells were later reexamined by David Moran, Carter Rowley, and Bruce Jafek 131, who also found evidence that supported the existence of a second morphologically distinct class of bipolar sensory neurons in the mammalian olfactory epithelium 142143. The authors injected horseradish peroxidase into the olfactory bulb of rat and detected retrograde labelling of olfactory epithelial microvillar cell bodies. They concluded that microvillar cells are connected by an axon to the olfactory bulb, and that the rat olfactory system will thus "need to be expanded to include two morphologically distinct classes of sensory receptors, ciliated olfactory receptors and microvillar cells" 143. This hypothesis seemed plausible on the basis of the observation that ciliated and microvillar olfactory neurons occur in many other species throughout the clade of vertebrates 85144. However, the evidence that microvillar cells are neuronal remained controversial. When Virginia Carr and collaborators stained rat olfactory epithelia with monoclonal antibody 1A-6, they labelled microvillar cells that were not immunoreactive with SUS-1 antibodies (which is a marker for supporting cells). The 1A-6 immunoreactive cells, however, neither expressed olfactory marker protein (OMP), nor did they possess an identifiable axonal process. Since the cells also survived bulbectomy-induced degeneration of olfactory sensory neurons, the authors concluded that olfactory microvillar cells are of non-neuronal nature 145. The absence of OMP in microvillar cells was confirmed by Edward Johnson and collaborators 146. These authors also acknowledged some arguments that speak for a non-neuronal function of microvillar cells, but pointed out that the peroxidase backfill experiments 143 and the expression of Spot-35 protein 147 are in favour of positioning microvillar cells "in with known microvillar olfactory receptor cells of other vertebrates". They carefully concluded, that the function of the olfactory microvillar cell remains enigmatic 146.

Recently, we have examined the distribution of phosphoinositide-related signalling proteins in the mammalian olfactory epithelium and found that some elements of InsP₃-mediated signal transduction were found exclusively in olfactory microvillar cells 19148. These cells that we call "Jourdan cells", were labelled with antibodies against phospholipase C beta-2 (PLC β2), type 3 InsP₃-receptors (InsP₃R-III), type 6 transient receptor potential channels (TRPC6), and other proteins, including neuronal marker protein MAP2b 148. In contrast, no co-expression with OMP, cyclic nucleotide-gated channels, or adenylyl cyclase could be detected. When dissociated and exposed in vitro to depolarizing concentrations of potassium chloride, Jourdan cells reacted with a transient increase of their intracellular Ca²⁺ concentration. A similar reaction was observed, when the cells were exposed to odours. They also possessed a basal axon-like protrusion, which raised the possibility that they in fact represent a second class of olfactory sensory neurons. In contrast to ciliated olfactory sensory neurons 149150151, however, microvillar cells did not degenerated upon bulbectomy. This has been interpreted as one argument for a non neuronal nature of olfactory microvillar cells 145152.

Is there a duality of signalling pathways and sensory neurons in the mammalian olfactory epithelium, similar to the existence of ciliary and rhabdomeric photoreceptor cells? Unfortunately, it is too early to submit such an admittedly attractive proposal. At present, only a few arguments that can be brought in position to support any concept on what olfactory microvillar cells might do. First, these cells look like sensory neurons 130. They show a strong bipolar asymmetry, with a distinctive array of microvilli protruding towards the nasal cavity at their apical pole, and an axon-like process, the distal extension of which often crosses the basal lamina and seems to join the submucosal tracts of the olfactory nerve 148. This bipolar architecture of microvillar cells would at least be tailor made to enable the cells to collect signals at the olfactory mucosal surface and transmit this information across the epithelium. This view is supported by the localization of PLC β2 in the apical microvilli. PLC β2 is a G-protein-activated signal transduction enzyme and thus points towards the presence of yet unidentified upstream G-protein-coupled receptors. Moreover, the cells express neuronal marker protein MAP2b 148, Ca²⁺-binding protein SPOT-35 153, and they are depolarized by high potassium chloride 148. All of these findings are compatible with a role of olfactory sensory neurons, but none of them is a persuasive argument. To determine, whether or not olfactory microvillar cells in fact represent sensory neurons, it will be necessary to characterize them by electrophysiological recordings and to analyze their cell morphology and gene expression profile in greater detail.

Albeit their (un)neuronal nature, we have now found evidence that points towards a possible function of olfactory microvillar cells. As pointed out above, olfactory sensory neurons periodically die by apoptosis, being constantly regenerated from a distinct population of adult neuronal stem cells. To avoid unproportional growth or degeneration of the olfactory epithelial tissue throughout life, both processes – cell death and regeneration – must be precisely coordinated over many decades. It has been demonstrated by Donna Hansel, Betty Eipper, and Gabriele Ronnett, that among the factors, which contribute to this control, are amidated neuropeptides 154. In the adult olfactory neuroepithelium, neuropeptide Y is released from an uncharacterized set of cells 155. We were now able to show that the NPY-expressing cells are identical to the olfactory microvillar cells that we had previously characterized 156. This finding raises the possibility, that olfactory microvillar cells link signals on the surface of the olfactory mucosa to proliferation and differentiation of olfactory stem cells at the basis of the epithelium. Whether or not olfactory microvillar cells indeed represent a second class of olfactory sensory neurons remains elusive. In analogy to melanopsin-expressing retinal ganglion cells, they could be derived from an ancient type of microvillar olfactory neurons, which in the course of evolution have aquired a new physiological function. These thoughts, however, are totally speculative. Our efforts to understand their biological significance have just begun. A first step must be to unravel their input specificity by identifying the G-protein-coupled receptors, which act upstream of PLC β2.

JP outlined the manuscript and wrote the initial draft. RE significantly contributed to its contents, figures, and to finalizing the manuscript. Both authors read and approved the final manuscript.

The authors have no competing interests.