Lipidomic profile of human nasal mucosa and associations with circulating fatty acids and olfactory deficiency

The nasal mucosa (NM) contains olfactory mucosa which contributes to the detection of odorant molecules and the transmission of olfactory information to the brain. To date, the lipid composition of the human NM has not been adequately characterized. Using gas chromatography, liquid chromatography coupled to mass spectrometry and thin layer chromatography, we analyzed the fatty acids and the phospholipid and ceramide molecular species in adult human nasal and blood biopsies. Saturated and polyunsaturated fatty acids (PUFAs) accounted for 45% and 29% of the nasal total fatty acids, respectively. Fatty acids of the n-6 family were predominant in the PUFA subgroup. Linoleic acid and arachidonic acid (AA) were incorporated in the main nasal phospholipid classes. Correlation analysis revealed that the nasal AA level might be positively associated with olfactory deficiency. In addition, a strong positive association between the AA levels in the NM and in plasma cholesteryl esters suggested that this blood fraction might be used as an indicator of the nasal AA level. The most abundant species of ceramides and their glycosylated derivatives detected in NM contained palmitic acid and long-chain fatty acids. Overall, this study provides new insight into lipid species that potentially contribute to the maintenance of NM homeostasis and demonstrates that circulating biomarkers might be used to predict nasal fatty acid content.


Results
Clinical characteristics of the patients. Clinical characteristics of human subjects are presented in Table 1. The median age of the donors was 46 years (interquartile range = 34-52.5). Seven patients were smokers, and 4 patients had a BMI > 30. Assessment of olfactory performance by the ''Sniffin' Sticks'' test revealed that 7 patients were hyposmic. The total cholesterol concentration in the plasma ranged from 0.72 to 2.22 mg/mL (median concentration = 1.51 mg/mL).  Table 2. Fatty acid composition of total lipids from human nasal mucosa (% of total fatty acids). IQR interquartile range (Q1-Q3), AA arachidonic acid, ALA α-linolenic acid, DHA docosahexaenoic acid, dma dimethyl acetal, DPA docosapentaenoic acid, LA linoleic acid, MUFAs monounsaturated fatty acids, PUFAs polyunsaturated fatty acids, SFAs saturated fatty acids. a n = 20 subjects. www.nature.com/scientificreports/ phatidylinositol (PI) and lower quantities of lysophosphatidylcholine (LPC) and phosphatidylglycerol (PG) were also detected. The phosphatidylserine (PS) class could not be detected in human NM by HPLC coupled to Corona-CAD. However, PS molecular species were identified and quantified by HPLC coupled to mass spectrometry (MS), as shown below.   Table S1).
Ceramide and glycosphingolipid species. HPLC under the HILIC conditions was used to separate the Cer and glycosphingolipid (hexosyl-ceramides (HexCer), and dihexosyl-ceramides (Hex2Cer)) classes from the human NM lipid extract based on an increase in polarity. Coupling of this method to a high-resolution mass spectrometer (Orbitrap Fusion) enabled identification of various molecular species of these lipid classes. The relative distribution of these molecular species is presented in Fig. 4. The Cer class exhibited the highest diversity of molecular species and included 22 identified species. The glycosphingolipid profile was characterized by a lower diversity than that of the Cer class and included 12 and 11 molecular species of HexCer and Hex2Cer classes, respectively. Species composed of sphingosine (d18:1) and palmitic acid (16:0) were the most abundant in the Cer class and in both glycosphingolipid classes. The composition of the most abundant species of these molecules, concerning the sphingosine and the fatty acid, was similar to that of the most abundant species of SM class (d34:1) as shown in Fig. 3. Substantial levels of d18:1/24:1 and d18:1/24:0 (or d41:1) were also detected in the Cer class and in glycosphingolipids.
Fatty acid composition of the blood samples. The complete FA compositions of plasma, plasma-PL, plasma-cholesteryl esters (plasma-CE), and erythrocytes obtained by standard GC are presented in Table 3. The plasma samples were characterized by high levels of palmitic acid (16:0), oleic acid (18:1n-9), and LA (18:2n-6) and by moderate levels of stearic acid (18:0) and AA (20:4n-6). PUFAs of the n-3 family were present at lower levels, and DHA (22:6n-3) was the most abundant. A very high level of 16:0 and lower levels of 18:0, 18:1n-9, LA, and AA were detected in PL extracted from the plasma (plasma-PL). n-3 PUFAs accounted for approximately 5% of total FAs. The CE samples extracted from the plasma (plasma-CE) were characterized by a very high level of LA and by lower levels of 18:1n-9, 16:0, and AA. Finally, erythrocytes were found to contain high levels of 16:0, 18:0, and 18:1n-9 and lower levels of LA and AA. Notably, variability of the FA profiles in the blood compartments was lower than that in the nasal tissue.
Associations between nasal and blood fatty acids. All Spearman correlation coefficients between nasal and blood FAs are shown in Supplementary Table S2. The main statistically significant associations are presented in Table 4 Additionally, we evaluated correlations between nasal AA-containing PL species and blood n-6 PUFAs and several significant moderate associations were found. Analysis showed that four PL species, including PE      www.nature.com/scientificreports/

Associations between nasal FAs and individual characteristics. Several significant associations
between nasal FAs and individual characteristics were observed ( Supplementary Fig. S1). After adjustment for covariates (age, sex, smoking status, BMI, and olfactory diagnosis), nasal AA, dma 16:0 and dma 18:0 levels were significantly greater in hyposmic subjects than those in normosmic subjects (Fig. 5). Conversely, the nasal 20:1n-9 level was lower in hyposmic subjects. On the other hand, smoking subjects displayed significantly reduced levels of nasal 16:1n-9 and 16:1n-7 compared to those in nonsmokers.

Discussion
The present study aimed to describe the lipid composition of human NM. We analyzed the complete FA profile of this tissue and the composition of three major lipid groups (PL, Cer, and glycosphingolipids). In addition, we identified significant associations between nasal and blood FAs and between nasal FAs and individual characteristics of the donors. Our data showed the presence of substantial amounts of n-6 PUFAs (mainly LA and AA) in nasal biopsies and indicated that a high level of nasal AA might be associated with hyposmia (p = 0.024). LA, the precursor of the n-6   www.nature.com/scientificreports/ family, and AA, a long-chain member of the n-6 family, were detected in equivalent amounts. In contrast to LA, AA is relatively rare in the diet. AA is generated from LA by consecutive elongation and desaturation catalyzed by delta-5-and delta-6-desaturases (FADS1 and FADS2) and an elongase (ELOVL5) 15 . The liver is the primary site for PUFA metabolism; however, it can also take place in other tissues 16,17 . Several desaturases and elongases have been detected in human NM 18 , suggesting that at least partial AA synthesis may occur in this tissue. AA can then be stored in PLs and neutral glycerides or undergo further elongation/desaturation steps to generate adrenic acid (22:4n-6) and n-6 docosapentaenoic acid (22:5n-6). Upon release from PLs by phospholipases A2 (PLA2s), AA can be converted to various bioactive eicosanoids, depending on the activities of cyclooxygenases, lipoxygenases, or cytochrome P450 monoxygenases 19,20 . There is strong evidence that AA-derived eicosanoids are the central components of the proinflammatory processes. n-3 PUFAs EPA and DHA can also be metabolized by the same enzymes, resulting in compounds that have lower proinflammatory properties 21,22 . Hence, the local concentrations of these PUFAs ultimately govern the types and concentrations of eicosanoids that can be formed during inflammation processes. Higher concentrations of AA than those of EPA and DHA in human NM might contribute to perpetual inflammation in peripheral olfactory tissues of individuals exposed to environmental agents and chemicals. Positive association between nasal AA and hyposmia supports this hypothesis. However, the small number of hyposmic subjects included in the present study is a weakness and do not allow to draw firm conclusions. Similar analyses should be done in a larger number of patients with olfactory deficiency to confirm that a high level of nasal AA can be related to hyposmia. LA and AA were found to be incorporated in the main PL classes, PC and PE, and in PI. In agreement with the PL structures observed in most tissues, these PUFAs were esterified at the sn-2 position of the glycerol moiety, while the sn-1 position was mainly occupied by SFAs (16:0 and 18:0). LA and AA can be released from PLs by a number of PLA2s, including secreted PLA2s, cytosolic Ca 2+ -dependent PLA2s, and Ca 2+ -independent PLA2s 23,24 . Several of these PLA2s have been detected in human NM; PLA2s expressed at the highest levels include PLA2G6 (a cytosolic Ca 2+ -independent PLA2), PLA2G12A (a secreted PLA2), and PLA2G16 (a lysosomal PLA2) 18 . These findings indicate that human NM has enzymatic machinery to liberate PUFAs from PLs. In the present study, we also demonstrated that PlsEs are an important subgroup of the PE class (28% of total PE species) in human NM and that the major PlsE species contained AA. AA-containing PlsCs were also detected. Plasmalogens constitute a special group of PLs characterized by the presence of fatty alcohols (mainly 16 and 18 carbon chains) with a vinyl-ether bond at the sn-1 position of the glycerol backbone 25 . Due to the vinyl ether bond, these compounds play a critical role in the cell membranes, providing unique structural attributes and protecting membrane lipids from oxidation 26 . These compounds were also suggested to function as reservoirs for lipid mediators involved in inflammatory responses 19,20 . The high AA content in the PlsE species and notable expression of PLA2G6, a PLA2 with preferential activity toward plasmalogens, suggest that this PL subgroup may play a crucial role in the control of AA availability in human NM.
In addition, we observed that NM from smokers had significant lower levels of 16:1n-7 and 16:1n-9 than those from nonsmokers (p = 0.030 and p = 0.022, respectively). The biological properties of these MUFAs are incompletely understood. In vivo and in vitro studies have shown that these compounds possess anti-inflammatory activity 27,28 . 16:1n-7 is also involved in other cellular functions, such as growth, proliferation and endoplasmic reticulum stress 29 . It is difficult to explain these deficiencies based on currently available information. Substances in cigarette smoke may impair the activity of stearoyl-CoA desaturase, the enzyme that catalyzes the introduction of a double bond in several saturated fatty acyl-CoAs 30 . Reduced stearoyl-CoA desaturase activity may result in lower levels of 16:1n-7 and 16:1n-9.
We did not detect an association between smoking and olfactory alterations in our study. This result may be related to small number of subjects, which could have impacted the statistical power of correlation analysis and/or to the fact that relatively high doses of cigarette smoke over a prolonged period of time are needed to www.nature.com/scientificreports/ www.nature.com/scientificreports/ produce olfactory dysfunction 31 . A number of studies have assessed whether smoking is a risk factor for poor olfaction and demonstrated a significant association with smoking; however, other studies reported no associations. Nevertheless, recent investigations (a meta-analysis and large population-based studies) have provided strong evidence for an association between chronic smoking and olfactory dysfunction [32][33][34] .
Notably, the amount of n-3 PUFAs (3-4% of total FAs) detected in human nasal samples was considerably lower than that detected in rodent olfactory mucosa 8 . This result suggests that nasal n-3 PUFA content might be species-specific. Nevertheless, this discrepancy can also be due to the nature of collected samples. In rodents, olfactory mucosa is clearly defined and can be harvested without contamination by respiratory tissue. In contrast, in humans, the boundaries between olfactory and respiratory mucosa are irregular, and patches of respiratory tissue can be interspersed in olfactory mucosa [11][12][13] . Although the biopsies were taken from an area rich in olfactory tissue, we cannot be certain that the samples contained only olfactory tissue. The characterization of the tissues in each biopsy could not be carried out because most of the samples were too small to perform both a comprehensive lipid analysis and a molecular RNA analysis. Nevertheless, RNA analyses performed on some biopsies indicated that respiratory genes were expressed in these samples. We therefore assume that the nasal samples collected in the present study may contain both olfactory and respiratory tissues. If this is the case, the lipid profiles were representative of both tissues, which makes it difficult to compare the lipid composition of human nasal samples with that of rodent olfactory mucosa.
Additionally, the present study analyzed the FA composition of nasal ceramides. Ceramides are a family of lipid molecules known to be involved in major cellular events, including growth, differentiation, and cell death. Ceramides are also precursors of complex lipids that play important structural and regulatory roles in the cells 35 . Ceramides consist of a sphingoid long-chain base linked to a FA via an amide bond. Glucose or galactose can be added to ceramide molecules, resulting in glucosylceramides or galactosylceramides (abbreviated as Hex-Cer in the present study). Subsequent attachment of a galactose molecule to glucosylceramides results in lactosylceramides (abbreviated as Hex2-Cer in the present study), which are precursors of complex glycosphingolipids, such as gangliosides. Ceramides can contain a wide spectrum of saturated or monounsaturated FAs, ranging from 14 to 26 carbon atoms in length. During de novo synthesis, acylation of the sphingoid long-chain base is carried out by a family of ceramide synthase enzymes (CerS). Each CerS has a high degree of FA specificity 36 . Some CerS isoforms are ubiquitously expressed, and other isoforms have a very specific distribution in tissues. A transcriptomic analysis performed on human NM indicated the presence of 6 CerS isoforms in this tissue, with CerS2, CerS4, and CerS5 being expressed at the highest levels 18 . These isoforms are responsible for the generation of Cer d18:1/16:0 and very long-chain ceramides Cer d18:1/24:1 and Cer d18:1/24:0 36 , the main species identified in human NM. In addition to de novo synthesis, ceramides can be generated by degradation of SM or by recycling glycosphingolipids (salvage pathway). Two sphingomyelinase isoforms, SMPD1 and SMPD4, and a ceramidase isoform (ASAH1) involved in the salvage pathway are expressed at a substantial level in human NM 18 , suggesting that these recycling pathways can contribute to the formation of nasal ceramides. The sphingomyelinase pathway was reported to enable rapid and transient production of ceramides, while the de novo pathway, which requires multiple enzymatic steps, is responsible for slow and robust accumulation of ceramides over a period of several hours 37 .
Ceramides and glycosphingolipids contribute to the structural organization and dynamic properties of the membranes 38 . Together with cholesterol and SM, glycosphingolipids form highly organized microdomains or "lipid rafts", which are involved in various cellular functions, including cell-cell interactions and cell signaling 39 . Activation of sphingomyelinases in response to diverse stimuli induces the generation of ceramides within the rafts 40 . Then, ceramide molecules spontaneously associate to form ceramide-enriched membrane platforms that are involved in reorganization and clustering of proteins involved in intracellular signalosomes 40 . An increase in the ceramide levels can mediate many cell stress responses, including the regulation of cell growth, differentiation, senescence, proliferation, and cell death 37,41 . Ceramides and their phosphorylated derivatives are also implicated in inflammation 42 . These compounds can induce the expression or activation of the proinflammatory transcription factor NF-kappaB, which regulates the expression of many genes involved in inflammatory responses, such as cytokines and enzymes catalyzing the formation of prostaglandins 43 . The 16:0-ceramide species were shown to be mostly proapoptotic, whereas very long-chain 24-and 24:1-ceramides promote cellular proliferation [44][45][46] . Opposite effects of these ceramides contribute to the maintenance of cellular homeostasis. The role of ceramides detected in human NM remains to be explored.
Cholesterol would be also an important component of human sinus mucosa 10 . Several CEs, including cholesteryl linoleate and cholesteryl arachidonate, have been detected in this tissue 10 , indicating that LA and AA can be incorporated in this lipid group in addition to PLs. We did not examine CEs in the present study because our HPLC conditions (HILIC) were not adapted for analyzing these non-polar molecules. Likewise, our analytical conditions did not allow to analyze cardiolipins because these molecules have rather an apolar character. Moreover, gangliosides could not be characterized because their analysis by HPLC-HILIC/MS is influenced by ion suppression effects due to co-elution with more abundant PLs. Purification of gangliosides prior to HPLC-HILIC/MS analysis would enable to circumvent this problem 47 but unfortunately, nasal samples did not contain enough biological material to carry out this purification step.
Like many tissues, NM is not easily accessible. To circumvent this issue, a number of studies have investigated relationships between circulating FAs and the FA composition of tissues, such as the brain and retina [48][49][50] . Blood fractions such as erythrocytes, plasma and PL or CE extracted from these components might be reliable biomarkers of tissue FA status but there is currently no scientific consensus regarding what fractions could be the most relevant biomarkers. The FA composition of plasma likely reflects the lipid status of other tissues and also represents the endogenous processing of the lipids and their recent dietary intake 51 . FA compositions in plasma CEs and PL are considered fairly independent of fasting status mainly because of their relatively slow turnover 52 . Similarly, the FA composition of erythrocytes is representative of long-term dietary intake because  53 . Therefore, in the present study, we assessed the correlations between various circulating FA fractions and the FA nasal content. Moderate positive association between the AA content in the nasal tissue and that in the plasma is one of the major results of our study. Interestingly, significant correlations were observed in the case of total plasma and the plasma CE and PL fractions. Our results are in agreement with the data of a previous study that demonstrated a strong correlation between FA proportions in CEs and PLs in various Swedish populations, suggesting that these two plasma fractions can be used as interchangeable biomarkers 54 . In the present study, similarities between total plasma and plasma fractions (CE and PL) may be explained by the fact that FA analyses were performed in the fasting plasma samples. Notably, a quite strong association was observed in the case of AA esterified in plasma-CE (r = 0.68), suggesting that this marker can be a good indicator of nasal AA levels. This observation is consistent with a recent study showing that plasma CEs might be relevant biomarkers of retina PUFAs 55 . Taken together, these results indicate that the PUFA content of plasma-CE might reflect that of neuronal tissues such as the NM and retina. Our finding also suggests that plasma lipoproteins (such as LDL and HDL) that contain AA esterified within CE or PL may supply AA to NM. Correlations between the plasma AA level and nasal AA level suggest an efficient uptake of AA from the circulating lipid pools by NM cells.
In conclusion, the present study showed that human NM is characterized by a high content of n-6 PUFA species, which might have repercussions on olfactory function, as suggested by significant associations between nasal AA (20:4n-6) levels and olfactory deficiency. Multiple ceramide species involved in cell death and proliferation were also detected. Our findings suggest that these lipid compounds may be involved in the control of NM homeostasis. Identification of plasma CE as a circulating biomarker might be useful for prediction of nasal AA levels.

Material and methods
Patients. The study was performed according to the guidelines of the Declaration of Helsinki and has been formally approved by the Dresden Hospital Ethics Committee (ethics protocol number EK118032015). Twentythree patients (9 females and 14 males) who underwent septoplasty or other nasal operations under general anesthesia were recruited. All subjects gave written informed consent to participate in the study. Various clinical characteristics, such as sex, age, body mass index and smoking status, were recorded (Table 1). Before surgery, patients were tested for odor identification by a validated psychophysical olfactory test using 12 different odors, which is known as the ''Sniffin' Sticks'' test 56 . The cutoff score between normosmia and hyposmia is the 10th percentile of the normal distribution of results obtained in the odor identification test 57 . Biopsies. Nasal tissue samples were acquired from the upper portions of the middle turbinate and the uncinate process because olfactory mucosa was shown to be the predominant tissue in these nasal regions 58,59 . Biopsies were collected from only one side of the nasal cavity. Blood was sampled by venipuncture after fasting and collected in heparinized tubes; then, erythrocytes were separated from the plasma by centrifugation at 3,000 rpm for 10 min at + 4 °C. Erythrocytes were washed three times with an isotonic saline solution. All samples were immediately stored at − 80 °C until further analyses. Lipid extraction from the plasma, erythrocytes, and nasal mucosa. Total lipids were extracted from the plasma and erythrocytes according to the method of Moilanen et al. 60 . Briefly, 10 mL of CHCl 3 /CH 3 OH (1/1, v/v) was added to 1 mL of sample. After shaking and centrifugation (1500g, 3 min), 3 mL of CHCl 3 and 1.8 mL of a solution containing 17 mM NaCl and 1 mM H 2 SO 4 were added to the liquid phase. After vigorous shaking, the mixture was centrifuged (1500g, 3 min), and the lower phase was collected and evaporated under a stream of nitrogen. Total lipids were redissolved in CHCl 3 /CH 3 OH (1:1, v/v) and stored under nitrogen at − 30 °C until further analyses.
Total lipids from NM were extracted according to the Folch method 61 . Briefly, each sample was homogenized in 2 mL of 0.73% NaCl using a tissue lyser (Qiagen, Venlo, Netherlands, 2 cycles of 7 min at 30 Hz). Total lipids were then extracted with 10 mL of CHCl 3 /CH 3 OH (2:1, v/v). The mixture was centrifuged (1500g, 3 min), and the lower organic phase was collected and evaporated under a stream of nitrogen. Total lipids were redissolved in CHCl 3 /CH 3 OH 1:1 (v/v) and stored under nitrogen at − 30 °C until further analyses.
Preparation of plasma phospholipids and cholesteryl esters. PLs and CEs were separated from the lipid extracts of total plasma by thin-layer chromatography (TLC) as described previously 49  www.nature.com/scientificreports/ fluorescein, the fluorescence signal of the silica gel plates was visualized under UV light (360 nm), and the spots corresponding to PLs and CEs were scraped off the plate. The silica scrapped off the TLC plate was then submitted to the transesterification procedure and analyzed by gas chromatography coupled to flame ionization detection (GC-FID) following the FA analysis procedure.
Fatty acid composition. Lipid extracts were evaporated to dryness under nitrogen. Then 0.7 ml of boron trifluoride-methanol (7%) and 0.3 ml of toluene were added to 1-20 mg of lipid. The tube was closed with the screw cap, and then heated at 95 °C using an oven for two hours, cooled and opened. After that, 5 ml of NaHCO 3 solution and 5 ml of hexane were added. The mixture was shaken, centrifuged (1500g, 3 min) and the FA methyl esters were extracted from the upper phase (hexane). After transmethylation, FA methyl esters were analyzed by GC-FID as described previously 8 . Briefly, GC was performed on a Hewlett Packard model 5890 gas chromatograph (Palo Alto, CA) equipped with a CP-Sil 88 column (Varian, Les Ulis, France) and a flame ionization detector. Hydrogen was used as the carrier gas. FA methyl esters were identified by comparison with commercial and synthetic standards. The data were processed using EZChrom Elite software (Agilent Technologies, Massy, France) and reported as the percentage of total FAs. FA analyses were performed on 23 blood samples but only on 20 nasal samples because some nasal biopsies did not contain enough biological material to carry out this analysis.
Plasma cholesterol measurement. Total cholesterol was extracted from the plasma samples and quantified using the method described by Bretillon et al. 62 . Briefly, total lipids extracted from 50 µL of the plasma were subjected to alkaline hydrolysis in 1 M KOH (5 mL) for 2 h. After the addition of phosphoric acid (65 µL), the sterols were extracted with CHCl 3 (9 mL) in the presence of 0.9% sodium chloride (3 mL). The organic phase was evaporated, and 80 µg of 5α-cholestane was added as an internal standard. Sterols were derivatized to trimethylsilyl ethers by heating at 60 °C in the presence of pyridine (200 µL) and N,O-bis(trimethylsilyl)trifluoroacetamide (200 µL), and then analyzed by GC-FID (Hewlett-Packard HP4890A) using conditions described previously in detail 62 . The data were processed using EZChrom Elite software (Agilent Technologies) and reported as mg cholesterol/ mL plasma.

Separation and quantification of nasal phospholipid classes by liquid chromatography coupled to charged aerosol detector (Corona-CAD).
The phosphorus content of the total lipid extracts from nasal samples was determined according to the method developed by Bartlett and Lewis 63 as described previously 64 . After dilution to a concentration of 500 µg/mL in CHCl 3 /CH 3 OH 1:1 (v/v), PL classes were separated by LC under the HILIC conditions and detected using a Corona™ ultra RS™ charged aerosol detector (CAD, Thermo Scientific, USA. Separation of PL classes was based on their increasing polarities (supplementary Fig. S3). PL classes were determined relative to total PLs in nasal sample. Analysis methods, including detector parameters, instruments, elution gradient, and other chromatographic separation conditions, were described in detail by Le Bon et al. 8 . PL classes were analyzed in only 19 nasal biopsies because some samples did not contain enough biological material to perform this analysis.
Structural analysis of nasal phospholipids by liquid chromatography coupled to mass spectrometry. PL classes were separated from lipid extracts by HPLC, under HILIC conditions, according to their increasing polarities. PL molecular species were characterized in the elution zone corresponding to each PL class, using a high-resolution MS. Then, the characterized species were quantified by targeted analyses using a triple quadrupole MS (see the workflow in supplementary Fig. S2). The analytical conditions of these methods were as follows.
Chromatographic separation of phospholipids. HPLC separation was performed using an UltiMate™ 3000 LC pump equipped with a dual-gradient pump and an UltiMate™ 3000 autosampler from Thermo Scientific (USA). The injection volume was 10 µL. PL classes were separated by LC under the HILIC conditions as described previously 8 (supplementary Fig. S3). The HILIC conditions were similar to those used for the separation of PL classes when coupling the chromatographic system to Corona. The flow from LC was split using an analytical fixed flow splitter (split ratio = 1:1, postcolumn) from Analytical Scientific Instruments (El Sobrante, CA, USA).
The LC system was controlled by Standard Instrument Integration software based on Dionex Chromeleon TN 7. Lipid extracts were diluted in CHCl3/CH3OH (1:1, v/v) and the concentration of the injected lipids for characterization purposes was 0.5 mg mL -1 .
Characterization of phospholipid species. An Orbitrap Fusion™ Tribrid mass spectrometer (Thermo Scientific, USA) was used for high-resolution analyses of PL species in nasal tissues. The instrument was equipped with an EASY-Max NG™ ion source (heated electrospray ionization (H-ESI)). H-ESI source parameters were optimized and set as follows: ion transfer tube temperature of 285 °C, sheath gas flow rate of 35 a.u., auxiliary gas flow rate of 25 a.u., sweep gas flow rate of 1 a.u., and vaporizer temperature of 370 °C. Positive and negative ions were monitored alternatively by switching the polarity with a spray voltage set to 3500 V in the positive and negative ion modes. An Orbitrap mass analyzer was used to obtain all mass spectra in the full scan mode within a normal mass range and a target resolution of 120,000 (full-width at half maximum (FWHM) at m/z 200). All MS data were recorded using a maximum injection time of 50 ms, automatic gain control at 4.10 5 , and one microscan. An intensity threshold filter of 1.10 3 counts was applied. For tandem mass spectrometry (MS/MS) analyses, datadependent mode was used to characterize the PL species. Precursor isolation was performed in a quadrupole analyzer with an isolation width of m/z 1.6. Higher-energy collisional dissociation was used for fragmentation www.nature.com/scientificreports/ of PL species with an optimized stepped collision energy of 30% (± 5%). Linear ion trap was used to acquire spectra of the fragment ions in the data-dependent mode. The automatic gain control target was set to 2.10 4 with a maximum injection time of 50 ms. All MS and MS/MS data were acquired in the profile mode. Orbitrap mass spectrometer was controlled by Xcalibur™ 4.1 software. All PL species were identified using high-accuracy data based on the information collected from the fragmentation spectra by LipidSearch™ software and the LIPID MAPS™ database (supplementary Fig. S3).
Quantification of phospholipid species. Samples were diluted to a concentration of 25 µg mL −1 PLs in CHCl 3 / CH 3 OH (1:1, v/v) for analysis. The LC system under the previously described HILIC conditions was coupled to a triple quadrupole mass spectrometer (Thermo Finnigan TSQ Quantum) equipped with a standard electrospray ionization source to quantify PL molecular species. Specific acquisition methods in the positive and negative ion modes were optimized and used according to the studied compounds (supplementary Fig. S3). PC, PlsC and SM species were quantified in the positive ion mode by precursor ion scanning of m/z 184 amu, which corresponds to the choline head group. PE species lose their ethanolaminephosphate head group as a neutral fragment of 141 Da. Therefore, neutral loss scanning of 141 Da in the positive ion mode was used for specific detection and quantification of PE. Similarly, PS species lose their serine-phosphate head group as a neutral fragment of 185 Da. Therefore, neutral loss scanning of 185 Da in the positive ion mode was used to quantify PS species.
In the negative ion mode, PI species show a fragment at m/z 241 amu identified as inositolphosphate minus one molecule of H 2 O. This fragment was used for precursor ion scanning to quantify these compounds. The quantification of PlsE was performed in the negative ionization by multiple reaction monitoring mode (MRM) of the parent/fragment transition for each selected plasmalogen. The amount of PL species in each PL class was expressed relative to the total lipids in this class. The data were processed using Xcalibur software. In addition, isotopic overlap corrections were applied to the data 65 . Additional details about the electrospray source parameters and methods of mass spectrometry are available in Le Bon et al. 8 .
Analysis of nasal ceramides, hexosyl-ceramides, and dihexosyl-ceramides by liquid chromatography coupled to mass spectrometry. Cer, HexCer, and Hex2Cer classes were separated from lipid extracts by HPLC, under modified HILIC conditions, according to their increasing polarities. The amounts of Cer, HexCer, and Hex2Cer molecular species was determined using a high-resolution MS (see the workflow in supplementary Fig. S2). The analytical conditions of these methods were as follows.
Chromatographic separation. HPLC separation of the Cer, HexCer, and Hex2Cer classes was performed using a Dionex UltiMate™ 3000 LC pump from Thermo Scientific equipped with an UltiMate™ 3000 autosampler.
Knowing that Cer, HexCer, and Hex2Cer are less polar than PL, the separation of these sphingolipid classes was achieved under adapted parameters of the HILIC method already used to analyze PL classes using an Accucore™ HILIC LC column (150 × 2.1 mm, 2.6 µm). The injection volume was 5 µL, and the column was maintained at Characterization and quantification of Cer, HexCer, and Hex2Cer species. An Orbitrap Fusion™ Tribrid mass spectrometer equipped with an H-ESI source was used for the analysis of Cer, HexCer, and Hex2Cer molecular species from nasal samples. Identification of these compounds is based on their retention time using lipid standards, on the data acquired at a high resolution, and on tandem MS experiments. The retention time of lipid standards on the total ion current chromatogram enabled identification of elution zones for each sphingolipid class (supplementary Fig. S3). High-resolution full scan experiments were used to acquire exact masses of analyzed species, and information on FAs and long-chain bases (sphingosines) of Cer, HexCer, and Hex2Cer was obtained from the MS/MS spectra. Data processing of high-resolution MS and MS/MS experiments was performed by Xcalibur™ software with the help of LipidSearch™ software and the LIPID MAPS™ database. Additional details of the source parameters and the MS and MS/MS methods used for identification are described in the section "Characterization of phospholipid species". The quantitative analysis of Cer, HexCer, and Hex2Cer molecular species within a specific sphingolipid class was performed using the high-resolution full scan data acquired by the Orbitrap Fusion mass spectrometer in the positive ion mode. Sphingolipids were detected as protonated ions [M + H] + by positive ionization. The proportion of individual species of a specific (Cer, HexCer, or Hex2Cer) class was calculated as the ratio of its MS signal to the sum of all detected peak signals in this class.
Statistical analyses. The Spearman correlation coefficients were calculated to determine associations between nasal and blood FA profiles using XLSTAT v2019.1.2 (build 57072) software (Addinsoft). The relationships between nasal FA levels and individual characteristics were analyzed using R software (version 3.5.2) 66 . The Spearman correlation test was used for continue variables (stat_cor, ggpubr package), and the Wilcoxon test was used to compare the mean values of the discrete variables (stat_compare_means, ggpubr package). Type II ANCOVA (glm, stats package) was used to test the influence of various factors (sex, smoking status, and olfactory diagnosis) and covariates (age and BMI) on FA levels and to calculate the p-values for each main effect corrected for all other effects. For all data analyses, the correlations were considered significant at p < 0.05. Schober www.nature.com/scientificreports/ classification thresholds 67 were followed to interpret the correlation coefficients: 0.10 to 0.39, weak; 0.4 to 0.69, moderate; 0.70 to 0.89, strong; and ≥ 0.9, very strong.