CD36 is involved in oleic acid detection by the murine olfactory system

Olfactory signals influence food intake in a variety of species. To maximize the chances of finding a source of calories, an animal’s preference for fatty foods and triglycerides already becomes apparent during olfactory food search behavior. However, the molecular identity of both receptors and ligands mediating olfactory-dependent fatty acid recognition are, so far, undescribed. We here describe that a subset of olfactory sensory neurons expresses the fatty acid receptor CD36 and demonstrate a receptor-like localization of CD36 in olfactory cilia by STED microscopy. CD36-positive olfactory neurons share olfaction-specific transduction elements and project to numerous glomeruli in the ventral olfactory bulb. In accordance with the described roles of CD36 as fatty acid receptor or co-receptor in other sensory systems, the number of olfactory neurons responding to oleic acid, a major milk component, in Ca2+ imaging experiments is drastically reduced in young CD36 knock-out mice. Strikingly, we also observe marked age-dependent changes in CD36 localization, which is prominently present in the ciliary compartment only during the suckling period. Our results support the involvement of CD36 in fatty acid detection by the mammalian olfactory system.


Introduction
Mammals have a demand to consume food with high fat content, possibly due to an evolutionary pressure-derived propensity to hoard energy. Behavioral studies in rodents have provided evidence for a role of olfaction in preference for fatty foods and triglycerides (Ramirez, 1993;Kinney and Antill, 1996;Takeda et al., 2001). Humans are also capable of detecting fatty acids by their odor (Bolton and Halpern, 2010;Wajid and Halpern, 2012;Chukir et al., 2013). Even in combination with other odorants present in food, humans can detect fat content by the sense of smell alone (Boesveldt and Lundstrom, 2014). The molecular basis of fat detection by olfactory sensory neurons is unknown so far.
Canonical sensory neurons in the main olfactory epithelium express one allele out of ∼1200 olfactory receptor genes (Buck and Axel, 1991). Upon stimulation, olfactory receptors activate a specific Gαs isoform (Gα olf ; Belluscio et al., 1998), leading to cAMP increase by activation of type III adenylyl cyclase (ACIII; Wong et al., 2000). The membrane depolarizes through Na + and Ca 2+ influx due to opening of cyclic nucleotidegated (CNG) channels (Brunet et al., 1996) and subsequent activation of calcium-activated chloride channels (Reisert et al., 2005). Additional subpopulations of olfactory neurons are defined by specific expression of non-canonical receptors and/or distinct signaling mechanisms. Neurons expressing trace amineassociated receptors (TAARs) lack olfactory receptors, but share a Gα olf -dependent signaling cascade. The axons of TAAR neurons project to a specific set of glomeruli in the dorsal region of the olfactory bulb (Johnson et al., 2012;Pacifico et al., 2012). Another neuronal subpopulation expresses the receptor guanylyl cyclase GC-D and recognizes natriuretic peptides and small gaseous molecules (CO 2 and CS 2 ). These neurons project their axons to a specific region of the caudal olfactory bulb, and form approximately 12 "necklace glomeruli" (Hu et al., 2007;Leinders-Zufall et al., 2007;Munger et al., 2010).
In our study, we found that up to ∼8% of mature olfactory neurons express the transmembrane glycoprotein cluster of differentiation 36 (CD36). CD36 is a well described receptor for fatty acids (Ibrahimi et al., 1996) that was so far unidentified in the peripheral mammalian olfactory system, although it has already been detected in central parts (e.g., piriform cortex, nucleus of the lateral olfactory tract; Glezer et al., 2009). A CD36 homolog is expressed in a subpopulation of insect olfactory neurons and plays an essential role in the detection of the fatty acid-derived pheromone cis-vaccenyl acetate (cVA; Benton et al., 2007). We here describe that CD36 is localized in cilia of olfactory neurons during the first weeks of life. Mice lacking CD36 display reduced response rates when challenged with oleic acid, a well described CD36 ligand.

Animals
All animal procedures were in compliance with the European Union legislation (Directive 86/609/EEC) and FELASA recommendations. The study was approved by the Lageso (Landesamtes für Gesundheit und Soziales Berlin, T0264/11). C57BL/6 (Charles River Laboratories, Sulzfeld, Germany) and CD36 knockout mice (Febbraio et al., 1999) were housed at RT in a 12:12 h light dark cycle with food and water ad libitum. Mice were used for analysis at ages between P6 and P14, unless indicated otherwise.

Standard Immunostaining Protocol
Tissue was rinsed two times in PBS +/+ followed by incubation in fresh blocking solution [1x PBS +/+ , 0.1% Triton X-100, 1% gelatin (Sigma, G7765)] for at least 1 h. Primary antibodies were diluted in blocking solution and put on the tissue over night at 4 • C. After five times rinsing in PBS +/+ for 5 min each, the tissue was incubated in secondary antibody (1:500) solution for 1 h. The tissue was rinsed three times for 5 min each and mounted in ProLong R Gold antifade reagent (Life Technologies).

OE Membrane Preparation and Western Blotting
Membrane preparation was performed on ice to avoid protein degradation. Olfactory epithelium was collected, homogenized, and centrifuged for 30 min at 20,000 g at 4 • C. After solubilization, protein samples were mixed with Laemmli buffer and separated in a 10-15% SDS gel. Proteins were transferred to nitrocellulose membrane. Membranes were blocked in 5% dry milk (Bio-Rad) in TBST and incubated with the antibodies that were detected using ECL Select ™ and ECL Hyperfilm (Amersham).

Olfactory Neuron Dissociation
Dissociation of olfactory sensory neurons was performed using the Miltenyi neural tissue dissociation kit for postnatal neurons. Main olfactory epithelium was collected in pre-heated enzyme mix 1 (10 µl solution 1 + 1950 µl solution 2) and cut into smaller pieces with spring scissors. After 15 min incubation at 37 • C on the MACSmix ™ Tube Rotator (Miltenyi), enzyme mix 2 (30 µl solution 3 + 15µl solution 4) was added and incubated for 10 min at 37 • C on the Tube Rotator.
For immunohistochemistry, samples were centrifuged for 2 min at 800 rpm and resuspended in 400 µl fixative solution. The cells were dispersed on adhesion slides (Polysine R slides, Thermo Scientific) on which an area was encircled with a hydrophobic marker pen (PAP pen, Sigma-Aldrich). The cells were incubated in fixative solution for 10 min at room temperature without moving the slides to let the OSNs settle down on the glass. All washing and staining steps were performed very carefully, as OSNs and their connection to the glass slide were very fragile. Solutions were sucked off by pipetting and new solutions were applied by slowly pipetting them in one corner of the encircled area. Primary antibody solution was applied for 1-3 h, secondary antibody solution for 45 min.

Transcriptome Sequencing
For transcriptome sequencing, olfactory epithelium from wild type P8 pups was collected and dissociated as described above. CD36 expressing cells were separated using the magnetic cell separation system from Miltenyi Biotec (MiniMACS ™ Separator with Anti-Rat IgG MicroBeads). RNA from the CD36-enriched fraction and from the cells passing the column was isolated using the RNeasy Micro Kit (Qiagen). Sequencing libraries were generated by Eurofins MWG Operon (Ebersberg, Germany). Briefly, a 3 ′ -fragment cDNA library with fragment sizes of 200-450 bp was created for every sample. Libraries were sequenced on Illumina HiSeq 2000 with chemistry v3.0 and with 1 × 100 bp single read module. Reads were mapped on a "proteincoding unspliced-genes" reference from Ensemble (GRCm38.p1, created by Eurofins MWG Operon) using the software BWAbacktrack. 55.06 and 53.97% reads were mapped of 25607716 and 28205993 total reads for the CD36-enriched fraction and the CD36-depleted fraction, respectively. The mean read coverage (covered reference bases/sum of all reference bases) accounted for 148.5 for the CD36-enriched and 185.7 for the CD36depleted cell fraction. Samples were normalized using the Trimmed Mean of M-values (TMM) method (Robinson and Oshlack, 2010), and analyzed using the Integrative Genomics Viewer (Robinson et al., 2011). Mapped reads were termed by gene accession numbers. To assign gene names to provided accession numbers, we used a self-made "WebRequest" tool, designed and kindly provided by Sven Fabricius (http://www. designerco.de/). Gene names were automatically requested from the NCBI database (http://www.ncbi.nlm.nih.gov/), exported to an overview table and manually imported in Excel. Gene function was evaluated manually by PubMed searches. Olfactory receptor transcripts were counted when detected by at least 10 raw counts.

Cryosections
The head was fixed for 2 h at 4 • C in fixative solution, cryoprotected in 30% sucrose, and embedded in Leica tissue freezing medium (Leica Microsystems). Sections of 12 µm thickness were cut on a Leica CM 1900 UV cryostat (Leica Microsystems), adhered to Super Frost R Plus cryoslides (Thermo Scientific) and stored at −25 • C.

Vibratome Sections
Pups were decapitated and the skin, lower jaw, palate, and front teeth were removed from the head. The back part of the head was removed and the front part was glued to the specimen plate of a VT1200S microtome (Leica Microsystems) and embedded in 2% low-gelling temperature agarose. Sections of 100-200 µm thickness were cut and collected in ice-cold ACSF in a 24 well plate. Sections were fixed for 20 min at room temperature and blocked for 1 h in blocking solution. Antibody staining was either performed for 3 h with a CD36 antibody conjugated to Alexa Fluor 488 or following the standard immunostaining protocol.

Whole-mount and En Face Preparation
Mice were decapitated and skin and bones removed. The head was split with a razor blade 1 mm beside the sagittal midline to expose one nasal cavity. Turbinates, nasal septum or bulbs were carefully removed, and fixed for 10-30 min at room temperature. The tissue was collected in Ringer's solution without touching the epithelial surface. Epithelium sheets were carefully transferred to an adhesion slide (Polysine R slides, Thermo Scientific) by moving it with fine forceps and fixed for 10 min at room temperature. Top view preparations were handled very cautiously to prevent damage of fine cilia structures. All solutions were applied by pipetting next to the tissue (Oberland and Neuhaus, 2014). Antibody staining was performed for 3 h or o.n.

Cilia Characteristics
En face preparations were used to characterize cilium characteristics of CD36+ and mOR EG+ OSNs. Analyzed regions were randomly chosen with bright field microscopy, confocal images of both fluorescence stainings were taken. ImageJ (Schneider et al., 2012) was used to mark all cilia of one OSN. A freehand line was drawn for every cilium from its base at the knob to its end. Selections were collected in the ROI manager (regions of interest) and the length was measured by ImageJ. Data was transferred to Excel and further processed.

Microscopy
Widefield fluorescence microscopy of cryosections and for olfactory neuron counting was documented using bright field microscopy on a Leica DMI6000 B system (Leica microsystems). Whole-mount stainings of turbinates and olfactory bulbs and vibratome sections were analyzed with the fluorescence stereo microscope M205FA (Leica Microsystems). Documentation of stainings in cryosections, en face preparations, and vibratome sections was predominantly performed using confocal microscopy with TCS SPE and TCS SP5 systems (Leica microsystems). For Ca 2+ imaging in main olfactory epithelium slices, we used an upright confocal microscope DM6000CFS (Leica Microsystems) equipped with a 20x 1.0 NA objective. Images were further processed using LAS AF, ImageJ, and Photoshop. To analyze protein distribution in more detail we used stimulated emission depletion (STED) microscopy (Dyba et al., 2003). We used the TCS STED system on a confocal TCS SP5 II platform with STED CW laser (Leica microsystems), as described before (Henkel et al., 2015).

RNA In situ Hybridization
RNA in situ hybridization was performed using non-fluorescent digoxigenin labeled probes. CD36 riboprobes cover around 1400 bp and were prepared as described (Glezer et al., 2009). RNA in situ primers (sequence 5 ′ to 3 ′ ): CD36_IS_Rive_fw GTAAAGCTTATGGGCTGTGATCGGAAC TGTGGGCG CD36_IS_Rive_rv GTAGAATTCTTATTTTCCATTCTTGGA TTTGCAAGCAC PCR products were cloned into the pcDNA3 vector. The vector was linearized according to the manufacturer instructions (Roche) at the 3 ′ or the 5 ′ end of the insert, to start in vitro transcription with T7 or SP6 RNA Polymerase to get the negative sense probe or the positive antisense probe, respectively (DIG RNA Labeling Mix (11277073910), transcription buffer (11465384001), Protector RNase Inhibitor (03335399001), T7 or SP6 RNA polymerase (10881767001, 10810274001). After 2 h incubation at 37 • C, the reaction was inactivated by adding 0.2 M EDTA (pH 8.0). RNA was cleaned up using the RNeasy Mini Kit and stored at −80 • C.
Sections were rinsed with PBST three times for 20 min each and equilibrated in buffer B3 (0.1 M Tris-Cl, 0.1 M NaCl, 50 mM MgCl2, 0.1% Tween-20) for 10 min. The final developing step was performed using freshly prepared buffer B4 (NBT/BCIP readyto-use tablets, Roche 11697471001) for at least 30 min up to 3 days, depending on the staining intensity. Reaction was stopped by washing the sections with PBST for 5 min. After rinsing the slices in H 2 O, they were mounted in ProLong R Gold antifade reagent (Life Technologies, P36930).

Electro-olfactogram Recordings
Mice were decapitated and skin and lower jaw removed. The head was split with a razor blade 1 mm beside the sagittal midline to expose one nasal cavity. Turbinates were removed to uncover the olfactory epithelium of the nasal septum. The head was fixed in a recording chamber half-filled with agarose and connected to an indifferent electrode. The electro-olfactogram (EOG) recording setup consists of a glass pipette electrode holder fixed to a probe that is connected to an IDAC-4 data-acquisition amplifier (Ockenfels Syntech GmbH). The glass pipette is filled with Ringer's solution and gets in contact to the probe by an inserted tungsten wire. The stimulus controller (CS-55, Ockenfels Syntech GmbH) generates a continuous humidified air flow onto the epithelium with triggered stimulus pulses, and was connected to the amplifier. Odors were loaded on a filter paper, inserted in a modified syringe and introduced in the stimulus air stream. We used a mixture of 100 odorants (Henkel 100,1:200 in H 2 O, Henkel KGaA) for stimulation (Wetzel et al., 1999). Stimulus air flow was applied for 0.1 s. Signals were recorded and analyzed with the software AutoSpike (Ockenfels Syntech GmbH).
Acute main olfactory epithelium slices were prepared from young mice of either sex between postnatal day 6 and 9. Mice were sacrificed by decapitation. Acute coronal main olfactory epithelium tissue slices were prepared as described (Fluegge et al., 2012). Briefly, the skin, lower jaw, teeth and palate were removed. The anterior aspect of the head was dissected, embedded in 4% low-gelling temperature agarose, placed in ice-cold oxygenated S2, and coronal slices (250 µm) were cut on a VT1000S vibratome (Leica Microsystems, Nussloch, Germany). Slices were transferred to a submerged, chilled, and oxygenated (S2) storage chamber until use. Slices were incubated in fluo4/AMcontaining S1 (10 µM; 30 min; RT; Molecular Probes), washed, transferred to a Slice Mini Chamber (Luigs and Neumann, Ratingen, Germany), anchored (0.1 mm thick lycra threads), and superfused with oxygenated S2 (∼3 ml/min; gravity flow). Fluo-4 was excited at 488 nm and images were acquired at 1.0 Hz (∼10 µm optical thickness).
Time-lapse experiments were analyzed using LAS-AF and MM-AF software (Leica Microsystems). All cells in a field of view were marked as ROIs and relative fluorescence intensities were calculated as arbitrary units. Data are reported as means from at least three independent experiments and statistical analyses were performed using a chi-squared (Fisher's exact) as dictated by data distribution and experimental design. p-values that report statistical significance (≤0.05) are individually specified in figure legends. In fluorescence imaging experiments, ROIs were defined to encompass single cells based on visually identified morphological features of dye-loaded cells at rest. An increase in fluorescence intensity was judged as a stimulus-dependent response if the following two criteria were both fulfilled: (a) the peak intensity value exceeded a given threshold that was calculated as the average baseline intensity before stimulation plus an intensity value corresponding to twice the baseline intensity standard deviation (I resp > I baseline + 2 × SD(I baseline ) (Riviére et al., 2009;Ferrero et al., 2011;Cichy et al., 2015); (b) the increase in light intensity was observed within 10 s after stimulus application. Frontiers in Cellular Neuroscience | www.frontiersin.org

CD36 Is Expressed in a Subset of Neurons in the Olfactory Epithelium
We detected CD36 in olfactory epithelia of mice by previous proteomic analysis (Rasche et al., 2010), CD36 protein has also been found to be highly abundant in the proteome of rat cilia (Mayer et al., 2009). We therefore analyzed the expression pattern of CD36 in the olfactory system. Quantitative real-time PCR revealed that the expression level of CD36 was significantly higher compared to olfactory receptors Olfr73 (mOR-EG) and Olfr124 (CD36: 0.161, mOR-EG: 0.009, Olfr124: 0.003; Figure 1A), similar as the protein amounts detected in previous proteomic approaches. We also detected the described 88 kD glycosylated form of CD36 protein (Talle et al., 1983;Abumrad et al., 1993;Han et al., 1997) by western blot analysis of whole olfactory epithelium and of olfactory epithelium membrane preparations, but not in the supernatant fraction containing soluble cytosolic proteins ( Figure 1B).
Immunostaining of coronal olfactory epithelia cryosections revealed CD36 localization in a subset of mature sensory neurons, identified by expression of the olfactory marker protein (OMP; Figure 1C). By contrast, we did not find CD36 expression in the accessory olfactory system ( Figure 1D). We then used a previously generated CD36 knockout mouse strain (CD36 −/− ; Febbraio et al., 1999) to confirm antibody specificity. Using OMP for counterstainings to visualize mature olfactory sensory neurons we found CD36 staining to be completely absent from the olfactory epithelium of CD36 −/− mice (Figures 1E-G).
Using RNA in situ hybridization we moreover detected CD36 mRNA in sensory neurons of P8 wild-type mice (Figure 1H), but not in olfactory epithelium from CD36 −/− animals ( Figure 1I).

CD36 does not Show Zonal Expression Pattern
Most olfactory sensory neurons that express a specific odorant receptor are restricted to one of four zones of the olfactory epithelium in the nasal cavity in the mouse (Ressler et al., 1993;Vassar et al., 1993). Also TAAR genes are expressed in unique populations of sensory neurons spread in single zones of the olfactory epithelium (Liberles and Buck, 2006). By contrast, CD36 expressing cells were not confined to a specific epithelial region (Figure 2). Analysis of cryosections from both rostral and caudal areas of the olfactory epithelium (Figure 2A) and wholemount turbinate preparations ( Figure 2C) revealed non-zonal CD36 expression along the septum and on all turbinates.

CD36-positive Glomeruli in the Olfactory Bulb
Next, we immunostained whole-mount olfactory bulb preparations, and screened the surface for CD36-positive glomeruli. A large subset of glomeruli was labeled at different staining intensities (Figure 3A), clearly outnumbering two glomeruli per bulb typically observed for canonical olfactory neurons with defined receptor identity (Vassar et al., 1994). CD36-positive glomeruli were analyzed in more detail using vibratome slices (Figures 3B-F). CD36 was detected in nerve fibers and axon termini ( Figure 3B). We analyzed the distribution of CD36-positive glomeruli from rostral to more caudal parts of the olfactory bulb and found CD36-positive glomeruli predominantly in the ventral and ventromedial region. CD36-positive glomeruli were essentially absent in the most caudal sections of the olfactory bulb (Figures 3C-F).

CD36 Neurons Express Olfactory Signaling Proteins
We tested whether and, if so, which members of the olfactory transduction machinery are expressed in CD36-positive sensory neurons. Immunostaining of four canonical olfactory transduction proteins, Gα olf , ACIII, CNGA2, and ANO2, in dissociated olfactory neurons demonstrated co-expression in CD36-positive cells (Figures 4A-D). To further uncover the expression of signaling proteins in CD36-positive neurons, we performed quantitative real-time PCR and transcriptome sequencing. Dissociated neurons were sorted via a magnetic cell separation system, yielding in a fraction enriched in CD36 expressing neurons ( Figure 4E). Quantitative real-time PCR of the sorted neurons confirmed the expression of Gα olf , ACIII, CNGA2, and ANO2 ( Figure 4F). Also transcriptome sequencing of the CD36-enriched fraction showed transcripts for Gα olf , ACIII, CNGA2, and ANO2 together with mRNAs from other elements of the canonical olfactory signaling cascade (Gβ1, Gγ13, CNGA4, CNGB1, PDE1C, PDE4A, NKCC1, NCKX4, RTP1, RTP2, REEP1; Supplementary Table 1). To exclude proteins with low abundance, analysis was restricted to transcripts that showed both an expression level of at least 0.1% of ACIII. The transcriptome dataset from CD36 expressing neurons further showed expression of ORs, but not of TAARs or GUCY2D, the particulate guanylate cyclase expressed in CO 2 responsive olfactory neurons.

CD36 Neurons Express ORs
To further analyze potential co-expression of ORs and CD36, we raised an antibody (panOR) against an amino acid sequence motif that is shared by most ORs. Similar to previous results, the non-transmembrane motif with the highest degree of conservation is MAYDRYVAIC at the transition between transmembrane domain 3 and intracellular loop 2. Although many class A GPCRs share similarity in the DRY motif, the MAY and VAIC sequences were specific for ORs and are not found in other GPCRs. Western blot analysis revealed a strong band in olfactory epithelium membrane preparations, but not in liver (Figure 5A). Immunostaining in cryosections showed labeling of the ciliary layer and many mature olfactory neurons ( Figure 5B). The immunostaining partially overlapped with CD36 immunostaining: some neurons revealed strong OR and CD36 staining (Figures 5C-E). We then investigated expression of mOR-EG, Olfr71, and Olfr726 with specific antibodies, but never observed co-localization with CD36 (Figures 5F-H), indicating that individual ORs may be specifically co-expressed in CD36-positive neurons.

CD36 is Localized in Olfactory Cilia
Since CD36 was co-expressed with canonical signal transduction proteins, we investigated whether CD36 was also localized in olfactory cilia. We used a preparation that allows an en face view onto the epithelial surface (Oberland and Neuhaus, 2014). CD36 was localized in a subset of olfactory knobs and cilia spreading from these ( Figures 6A,B), but did not co-express with the olfactory receptor mOR-EG ( Figure 6C). Moreover, labeling of CD36 and mOR-EG on the same preparation revealed morphological differences between both stained cell populations. To examine these differences, we analyzed 55 CD36positive and 55 mOR-EG-positive olfactory neurons in costainings (from 3 animals) by counting cilia and measuring cilia length (Figures 6D,E). We counted 14 cilia per mOR-EG expressing neuron, which is comparable to earlier scanning electron microscopy studies showing an average of 17 cilia, each up to 60 µm long (Menco et al., 1997). CD36 expressing neurons extend only half as many cilia (CD36: 7.6 ± 0.4 m, Figure 6F). Moreover, mOR-EG-positive cilia were significantly longer ( Figure 6G). CD36 expressing neurons therefore seem to differ from other olfactory neurons by morphological characteristics.

CD36 Showed a Receptor-like Expression Pattern in Cilia from Olfactory Neurons
Little is known about the spatial distribution of receptor proteins within olfactory cilia. Using stimulated emission depletion (STED) super-resolution microscopy, we recently showed that CNG channels and calcium-activated chloride channels of the Anoctamin family are localized to discrete microdomains in the ciliary membrane (Henkel et al., 2015). We here used STED microscopy to analyze mOR-EG and CD36 expression patterns in top view en face preparations (Figure 7). Similar to the ion channels, mOR-EG dissolved into distinct foci along the olfactory cilia under STED conditions (Figures 7A,C,E), while acetylated tubulin showed homogenous protein distribution (Figures 7G,H). This localization pattern provided first optical evidence for distinct spatial organization of olfactory receptors in cilia of sensory neurons. When analyzing CD36 staining pattern with STED microscopy, we found it also to be localized to discrete foci in the ciliary membrane (Figures 7B,D,F). Together, STED microscopy clearly provided morphological evidence for the organization of olfactory receptors in distinct domains along olfactory cilia. Since also the ion channels involved in odorant detection showed clustered localization pattern (Henkel et al., 2015), the presence in such microdomains seems to be a common feature of signaling proteins in olfactory cilia. The similarity in localization of CD36 is an indicator for a potential role of this fatty acid receptor in olfactory signal transduction.  (1), membrane preparation (2), and supernatant (3) of three P60 mice showing panOR specificity to olfactory epithelium. Adult liver (4) was used as negative control; absence of GAPDH shows successful membrane preparation; mOR-EG demonstrates an olfactory receptor-specific band of same size like the panOR band. (B) Confocal image (maximum projection) of a cryosection immunostained with the panOR antibody (P8). Scale bar: 20 µm. (C-E) Confocal images (maximum projections) of a P8 cryosection immunostained for CD36 (green) and panOR (red). (F-H) Confocal images of P8 cryosections immunostained for CD36 (green) and specific olfactory receptors Olfr726 (F), Olfr71 (G), and mOR-EG (H) (red). Scale bars: 20 µm (B), 10 µm (C-H).

Absence of CD36 Has No Effect on General Morphology of the Olfactory System
We further analyzed the potential role of CD36 in olfactory signaling in CD36 knockout (CD36 −/− ) mice (Febbraio et al., 1999). The gross anatomy of the nasal cavity and the olfactory bulb was unaltered (Figures 8A-D). Moreover, the sensory neuron morphology showed no differences compared to wild type mice (Figure 8E). We counted all OMP-positive sensory neurons in a 600 µm region of septal epithelium and detected no differences between wild type (224 neurons; n = 10 septum regions from 5 animals) and CD36 −/− mice (222 neurons; n = 8 septum regions from 4 animals). Ciliary morphology of mOR-EG expressing neurons was not altered in knockout animals ( Figure 8F). To determine whether the amount or the localization of signaling proteins is changed in the CD36 −/− mice, we stained the olfactory epithelium for the olfactory signal transduction proteins G olf , ACIII, CNGA2, and ANO2 (Figures 8G-N) and performed quantitative PCR (data not shown). Both experiments revealed no changes in CD36 −/− mice compared to wild type mice.
Responses to Oleic Acid are Reduced in CD36 −/− Animals The olfactory function in wild type and CD36 −/− mice was first analyzed by recording epithelial field potentials (electroolfactogram, EOG), elicited by a complex odorant mixture. The signals were indistinguishable between wild type and CD36 −/− mice (Figures 9A-C), indicating that CD36 −/− mice are not generally anosmic.
By using confocal imaging in acute slices (Fluegge et al., 2012), we then recorded Ca 2+ signals in response to odorant mixtures ( Figure 9D). Brief odor stimulation triggered Ca 2+ transients in a subpopulation of olfactory neurons in both, wild type and CD36 −/− mice, whereas K + -mediated depolarization activated most neurons. No difference in the number of responding cells was observed between wild type and CD36 −/− mice ( Figure 9K). This observation confirms field potential recordings, which did not reveal a general olfactory deficit of CD36 −/− mice.
Together, our results show that loss of CD36 does not impair the development or the overall function of the olfactory system, but instead indicates a specific role for the CD36 expressing population of olfactory neurons. Given that CD36 has been described as a receptor for fatty acids (Baillie et al., 1996;Ibrahimi et al., 1996) we investigated responses to the known CD36 ligand oleic acid. Since oleic acid and other free fatty acids are rapidly oxidized, we kept the substances under inert gas to prevent formation of oxidation products. First, we analyzed whether olfactory neurons respond to fatty acids, which has not been shown before. We stimulated olfactory epithelium slices with oleic acid (2 nM and 20 µM) and recorded Ca 2+ signals by confocal imaging (Figure 9E). A subset of ∼10-15% of neurons from wild type mice responded to the ligand (Figure 9K), showing that oleic acid can indeed cause Ca 2+ increases in olfactory neurons of young mice. These responses appear to be dose-dependent (Figures 9F,G). To analyze the role of CD36 in oleic acid-induced Ca2+ elevations we also recorded signals in neurons from CD36 −/− animals ( Figure 9E) and compared various signal characteristics as well as the proportion of oleic acid-sensitive neurons in CD36 −/− to data from wildtype mice (Figures 9H-K). Oleic acid-induced Ca 2+ responses not only displayed several differences in signal shape (amplitude, rise, half-width) as compared with odor-dependent signals (Figures 9H-J). Some oleic acid-specific signal parameters also changed with genotype ( Figure 9J). Moreover, we found that the number of responding cells was significantly reduced in CD36 −/− mice (Figure 9K), indicating that CD36 plays a role for the detection of oleic acid by the main olfactory epithelium.

Plasticity of CD36 Expression during Postnatal Development
As oleic acid is a major milk component, we next analyzed ciliary localization of CD36 from prenatal to adult age (Figures 10A-F). Preparations were co-immunostained for mOR-EG to ensure integrity of the tissue. CD36 was localized to a subset of olfactory knobs and cilia spreading from these. While mOR-EG was localized to cilia at all ages, clear ciliary localization of CD36 was only observed from P8-P21 (Figures 10C-E). At P21, punctate CD36 staining appeared around several olfactory knobs, indicating that CD36 labeled cilia were disintegrated. The cilia of co-stained mOR-EG expressing neurons appeared unchanged, ruling out a general disassembly or reorganization of cilia at this age (higher magnification in Supplementary Figure 1). Some regions in the epithelium even completely lacked CD36 ciliary staining at P21. Around P28, punctate labeling declined strongly as well (data not shown), and completely disappeared in older animals. Ciliary CD36 staining was lacking in P90 animals ( Figure 10F). A strong, relatively uniform staining with CD36 antibodies in adult animals, however, was derived from labeled microvilli of sustentacular cells (Figure 10G and RNA in situ hybridization, data not shown). Although ciliary CD36 labeling was progressively lost between P21 and P28, CD36 was still present within somata and dendrites of single neurons until adulthood ( Figure 10G). We counted the relative amount of CD36 expressing neurons among mature OMP-positive neurons from E18 to P180. Analyzing whole cryosections, we found on average ∼1.6% CD36 expressing neurons (11 ± 1/677 ± 44 per section) in E18 animals, ∼8.0% at P8-P14 (328 ± 29/4098 ± 436), and gradually decreasing numbers during and post-puberty ( Table 1). Although we counted 161,353 sensory neurons in total (11,304 of them CD36-positive), it was not feasible to perform statistical analyses with whole slice countings, as the total number of analyzed cryosections was too low with 1-3 sections per age. For statistical analysis, we therefore defined a 600 µm epithelial stretch along the septum and quantified CD36-positive relative to OMP-positive neurons ( Figure 10H). Together, the percentage of CD36 expressing neurons significantly increases after birth until P8, and decreases moderately during further aging. Nevertheless, CD36 completely disappeared from the cilia when mice were weaned. We therefore hypothesize that the presence of CD36 in cilia of olfactory neurons during the suckling period plays an important role for adaptation to the particular needs during this phase of life.

Discussion
Olfactory cues from foods are important for survival of mammals. We here describe a population of neurons in the olfactory epithelium of mice, which is characterized by the expression of the fatty acid receptor CD36 and has distinct morphological characteristics. Knockout of CD36 does not cause gross morphological or functional alterations in the olfactory system, but reduces the amount of oleic acid-sensitive olfactory neurons.
CD36 is expressed in the cell soma, dendrite, axon termini of olfactory neurons. Also the olfactory receptor proteins are localized not only in cilia, but also in axon termini in the olfactory bulb (Strotmann et al., 2004). Most notably, its expression in olfactory cilia guarantees direct contact to the inhaled air. High resolution STED microscopy showed a scattered distribution of CD36, similar to the localization of mOR-EG. Tubulin, on the other hand, was homogeneously distributed along the cilia. Previous Ca 2+ measurements in frog cilia revealed discrete signaling events, providing evidence for transduction domains along the ciliary membrane (Castillo et al., 2007). Using high resolution STED microscopy, our observation of scattered mOR-EG expression supports the notion of signaling microdomains in olfactory cilia. An anti-TRPM5 antibody also labeled discrete spots in olfactory cilia (Lin et al., 2007), and CNGA2 had been shown to have a punctate distribution by electron microscopy (Matsuzaki et al., 1999) and by STED microscopy (Henkel et al., 2015). The fact that CD36 was localized in discrete spots in the cilia supports our hypothesis of it having a role in olfactory signal transduction.
CD36 is a membrane glycoprotein, which belongs to the scavenger receptor family. CD36 is not specific for the olfactory system but has pleiotropic physiological functions in immune defense, metabolism, and angiogenesis (Silverstein and Febbraio, 2009). CD36 is described as a fatty acid receptor and/or transporter of long-chain fatty acids. Oleic acid is one of the best described CD36 ligands exhibiting nanomolar affinity (Baillie et al., 1996). We show here that oleic acid can induce Ca 2+ signals in olfactory neurons, which, in part, depend on CD36 expression. The fact that all key elements of canonical signal transduction, including olfactory receptors, were identified in CD36 neurons suggests a co-receptor function. We did not find evidence for TAAR expression in CD36 neurons (Supplementary Table 1), which is in accordance with the observation that TAARexpressing sensory neurons converge on glomeruli within a dorsomedial domain of the olfactory bulb (Johnson et al., 2012;Pacifico et al., 2012), distinct from the domain where CD36 expressing neurons project to. CD36 has already been implicated in fatty acid signaling in other sensory cells. Exposure to long-chain fatty acids causes a CD36-dependent Ca 2+ rise in taste cells (El-Yassimi et al., 2008), resulting in attraction to lipid-rich food (Laugerette et al., 2005). In addition, the two G-protein coupled receptors GPR40 and GPR120, which are activated by medium and long chain fatty acids, are expressed in the taste buds, and knock-out mice showed a diminished preference for linoleic acid and oleic acid (Cartoni et al., 2010). Recent work on gustatory perception of dietary lipids suggests that CD36 could act as co-factor for these G protein-coupled receptors that mediate fatty acid preference (Gilbertson and Khan, 2014). Both receptors that have been described in the taste buds have not been detected in CD36 expressing neurons (Supplementary Table 1), or in whole olfactory epithelium (Ibarra-Soria et al., 2014;Kanageswaran et al., 2015). These receptors therefore likely do not contribute to fatty acid responses of olfactory neurons.
In addition, CD36 is involved in insect pheromone signaling. Many insect species express SNMP, a CD36 ortholog, as part of a multi-protein signaling complex on dendrites of those antennal neurons that recognize the fatty acid-derived pheromone cisvaccenyl acetate (Rogers et al., 1997;Benton et al., 2007;Jin et al., 2008). SNMP seems to couple lipid-based extracellular ligands to signaling receptors in chemosensory communication (Benton et al., 2007), since SNMP is required by the receptor dimer Or67d/Orco to detect cis-vaccenyl acetate (Benton et al., 2007;Jin et al., 2008), and by the receptor dimer Or83c/Orco for normal responses to the fatty alcohol farnesol (Ronderos et al., 2014). It has been suggested that SNMP/CD36 may couple lipid-based extracellular ligands to signaling receptors in chemosensory communication by capturing fatty acids on the surface of OSN cilia and facilitating their transfer to the odorantreceptor (Benton et al., 2007). CD36 was also previously shown to act as a co-receptor for Toll-like receptors (Hoebe et al., 2005;Jimenez-Dalmaroni et al., 2009), suggesting that CD36-related proteins could have transmembrane partners in all their cellular roles. Alternatively, CD36 may function itself as a receptor by interacting with non-receptor tyrosine kinases (Moore et al., 2002). So far, the precise molecular function remains unclear in all sensory systems studied so far. The mechanistic basis of CD36 ligand interactions and signaling is still poorly understood in any biological system. It is tempting to speculate that the co-expressed olfactory receptors may be involved in fatty acid discrimination since humans, and possibly other mammals, can discriminate vapor-phase long-chain fatty acids (Bolton and Halpern, 2010). Whether CD36 also facilitates the recognition of other fatty acids remains to be elucidated.
Remarkably, CD36 is markedly localized to olfactory cilia during the first days of life, and largely disappears from cilia after weaning. This temporally restricted localization of CD36 in the signaling compartment coincides with the presence of high ligand concentrations, since oleic acid is an abundant fatty acid in breast milk. The exact concentration of oleic acid in vapor phase entering the nasal cavity during suckling is unknown. The concentration of oleic acid in milk (∼200 mM, roughly inferred from approximately 6-8% oleic acid in milk, depending on the maternal diet, Baillie et al., 1996) is far above the concentrations that elicited oleic acid responses in our study (2 and 20 nM). During the application of oleic acid to the main olfactory epithelium (Ca 2+ imaging experiments) we cannot prevent access of the ligand to ambient oxygen, which leaves the possibility that small amounts of the applied oleic acid could be oxidized. Since the substance was always kept under inert gas until the final use in the experiments, we are convinced that oxidation products of oleic acid are only FIGURE 10 | CD36 localization is regulated. (A-F) Confocal images of en face olfactory epithelium preparations for age stages E18-P90 immunostained for CD36 (green) and mOR-EG (red) showing localization changes of CD36 in olfactory cilia. mOR-EG immunostainings validate successful sample preparation and show clear ciliary localization at all age stages. CD36 ciliary localization is restricted to P8 (C) and P14 (D). Before, it is found in single olfactory knobs, but barely in cilia (E18: A; P1: B) and later, ciliary localization decreases (P21: E) and is absent in adult animals (P90: F). (G) Confocal image (maximum projection) of a P90 cryosection immunostained for CD36 (green) and OMP (red). CD36 is still localized in OSN soma and denrite in adult animals. Additionally, CD36 is expressed in the microvilli layer of sustentacular cells. (H) Cell counting of CD36 and OMP expressing neurons in a stretch of 600 µm epithelium lining the septum. Cell numbers for CD36 were normalized to OMP expressing cells. If possible, both sides of every septum were counted (n = 2) to get a total number of 7-10. Significance was calculated in comparison to P8 using two sample t-test. Error bars represent SEM (*p ≤ 0.05). Scale bars, 10 µm (A-F). present in trace amounts. Interestingly, oleic acid triggers Ca 2+ signals that appear somewhat different in shape that responses to conventional odors. It is thus possible that CD36-dependent transduction varies from canonical odor signaling. Future studies will have to address such potential mechanistic differences.
The temporally restricted localization in the signaling compartment is reminiscent to recently published data showing Data from single cryosections are presented. Cell counting in whole cryosections was performed using a cell counting plugin for ImageJ. OMP-and CD36-expressing neurons were manually marked and counted by the software. We normalized the amount of CD36 neurons to mature OMP expressing cells.
that daily postnatal exposure to lyral induces plasticity in the population of OSNs expressing MOR23 (Cadiou et al., 2014). Although intracellular protein distribution was not analyzed, MOR23 neurons have higher levels of olfactory receptor transcripts density after odorant exposure (Cadiou et al., 2014). It is possible to investigate in future studies whether developmentally regulated splicing accounts for the variable ciliary localization, since several splice variants of CD36 affecting the coding region of the protein have been described (Tang et al., 1994;Andersen et al., 2006). Olfactory neurons expressing the same olfactory receptor, although randomly dispersed within a broad zone of the epithelium, project axons to two spatially invariant glomeruli in each olfactory bulb (Mori and Sakano, 2011). A marked difference was observed with regard to the number of labeled glomeruli since we found approximately 60-80 CD36 glomeruli clustered on the ventromedial part of the olfactory bulb. Similarly, GC-D expressing neurons terminate in more than two (about 10-20) "necklace" glomeruli encircling the caudal main olfactory bulb (Juilfs et al., 1997;Hu et al., 2007). Previous analysis of the rat pup olfactory bulb during suckling behavior revealed an increase in 2-deoxy-glucose incorporation in a subset of unusually shaped glomeruli, suggesting that receptors projecting to these specialized glomeruli are responsive to a specialized suckling odor cue (Teicher et al., 1980). Although, the position of these glomeruli is not identical to those positive for CD36, there could be substantial overlap. In both cases, significant labeling is observed on the ventromedial bulb. Interestingly, glomeruli innervated by different mOR37positive neurons are also clustered in the ventral domain of the olfactory bulb (Strotmann et al., 2000). However, using transcriptome profiling, we did not find Olfr37 family members. Also TRPM5-expressing OSNs project axons to glomeruli in the ventral olfactory bulb (Lin et al., 2007), and apparently respond to social or sexual chemosignals, but not to regular odors (Lopez et al., 2014). It has therefore been suggested that the ventromedial/ventrolateral area of the olfactory bulb, in part, constitutes an "olfactory fovea" critical in processing information on semiochemicals (Xu et al., 2005;Lin et al., 2007).
High percentages of completely anosmic mice die shortly after birth, but the survival rate of these mice can be enhanced by reducing the litter sizes (Brunet et al., 1996;Belluscio et al., 1998;Wong et al., 2000). CD36 −/− pups do not die due to malnutrition or dehydration since initiation of suckling is dependent on variable blends of maternal "signature odors" that are learned and recognized prior to first suckling (Logan et al., 2012), but not single odor cues such as oleic acid.

Conclusions
We here identified a novel subset of sensory neurons in the main olfactory epithelium of mice, characterized by expression of CD36. CD36 shows a receptor-like spatial arrangement in olfactory cilia of nursed mice in high resolution STED microscopy, but is redistributed to intracellular sites after weaning. In accordance with the described roles of CD36 as fatty acid receptor or co-receptor in other sensory systems, absence of CD36 causes defects in the recognition of oleic acid by olfactory neurons in young mice. The described dramatic intracellular re-localization of CD36 after weaning constitutes a remarkable example of the main olfactory system plasticity in the first days of life.

Author Note
Recently, another study also reported expression of CD36 mRNA and protein in olfactory sensory neurons and sustentacular cells of 8-12 week old wild-type mice (Lee et al., PLoS ONE 10:e0133412).