Characterisation of the volatile profile of microalgae and cyanobacteria using solid-phase microextraction followed by gas chromatography coupled to mass spectrometry

Microalgae and microalgae-derived ingredients are one of the top trends in the food industry. However, consumers’ acceptance and purchase intention of a product will be largely affected by odour and flavour. Surprisingly, the scientific literature present a very limited number of studies on the volatile composition of microalgae and cyanobacteria. In order to fill the gap, the main objective of the present study was to elucidate the volatile composition of seven microalgal and cyanobacterial strains from marine and freshwaters, with interest for the food industry while establishing its potential impact in odour. Among the seven selected strains, Arthrospira platensis showed the highest abundance and chemical diversity of volatile organic compounds (VOCs). Aldehydes, ketones, and alcohols were the families with the highest diversity of individual compounds, except in Arthrospira platensis and Scenedesmus almeriensis that showed a profile dominated by branched hydrocarbons. Marine strains presented a higher abundance of sulfur compounds than freshwater strains, while the ketones individual profile seemed to be more related to the taxonomical domain. The results of this study indicate that the VOCs composition is mainly driven by the individual strain although some volatile profile characteristics could be influenced by both environmental and taxonomical factors.


Hydrocarbons.
A total of 12 linear, 4 aromatic, 40 branched, and 5 alicyclic hydrocarbons were identified in the studied microalgal and cyanobacterial strains. These hydrocarbons showed a moderated variation in both the number of individual compounds and percentage of relative abundance among strains. The most remarkable variability was in branched hydrocarbons which only occurred in two freshwater species (SA and AP). These species presented 40 and 39 branched compounds, respectively, whereas the maximum number detected in the other species ranged between 2 and 7 individual compounds. Similarly, branched hydrocarbons represented 2-9% of total VOCs relative abundance in most strains, although the abundance of these compounds in SA and AP was much higher (P ≤ 0.05) (46.2 and 37.3% of total VOCs, respectively; Table 2). It is important to highlight that, in the present work, some branched hydrocarbons were only identified to chemical family level due to the similarity among mass spectra and linear retention index (LRI) between branched compounds, the limited bibliographic information on LRI values, and the scarce availability of commercial high purity compounds. This noteworthy VOCs present in AP was overlooked in the scarce bibliography available, despite the presence of  IG  NG  TS  SA  CV  SY  AP   1175  57,43,85,71  Unknown  ND  ND  ND  8.82 b  ND  ND  24.6 a   1195  17312-80-0  43,85,57,41  Undecane, 2,4-dimethyl-ND  ND  ND  3.32 b  ND  ND  9.61 a   1198  17301-23-4  57,71,43,41  Undecane, 2,6-dimethyl-ND  ND  ND  19.1 b  ND  ND  65.8 a   1209  17301-33-6  43,71,57, Means with different superscripts indicate statistically significant (P ≤ 0.05) differences among strains. 1 1-cyclohexene-1-carboxaldehyde, 2,6,6-trimethyl-; 2 1,3-cyclohexadiene-1-carboxaldehyde, 2,6,6-trimethyl-; 3 2-cyclohexen-1-one, 3,5,5-trimethyl-; 4 2-cyclohexene-1,4-dione, 2,6,6-trimethyl-; 5 3-buten-2-one, 4-(2,6,6-trimethyl-2-cyclohexen-1-yl)-; 6 2-cyclopenten-1-one, 2-(2-butenyl)-4-hydroxy-3-methyl-, (Z)-; 7 3-buten-2-one, 4-(2,6,6-trimethyl-1-cyclohexen-1-yl)-; 8 cyclohexanol, 5-methyl-2-(1-methylethyl)-, acetate; 9 1,2-benzenedicarboxylic acid, diethyl ester; 10   www.nature.com/scientificreports/ branched hydrocarbons have been highlighted in superior algae 17 such as Capsosiphon fluvescens 18 and Undaria pinnatifid 19 or even in other microalgae such as Nostoc sp. 20 . On the other hand, it is also remarkable the similarity of the VOCs profile of AP and SA, even though SA is a eukaryotic microalga and AP a prokaryotic cyanobacterium (Fig. 1). The individual chromatographic VOCs profile of each strain is provided as Supplementary  Table S1. Among individual compounds, a larger proportion of large linear hydrocarbons (more than 13 carbon atoms) were found in prokaryotic cyanobacteria (Table 1) as reported previously 16 . In this respect, some of these linear hydrocarbons could have contaminated the samples during the production and freeze-drying process of strains in the laboratory, or during storage of the samples in plastic containers. The origin of branched hydrocarbons has been previously related with the oxidation of branched-chain fatty acids in other food matrices 21 . However, it should be noted that in some microalgae species such as AP, very low branched-chain fatty acid contents have been reported 22 . Therefore, it is likely that these compounds are coming from a secondary route, but further research is needed to understand the origin of these compounds. In general, acyclic hydrocarbons have been described to have no significant flavour contribution in food matrices 13 and it was confirmed by the odour impact ratio (OIR) values estimated in the selected strains (Table 3). Moreover, the contribution of branched hydrocarbons to flavour is still quite uncertain as there are no OT values available for these compounds in specialised databases or the scientific literature.
Aromatic hydrocarbons can be assumed to be generated mainly from the degradation of aromatic amino acids 23 . They are usually known as important aromatic compounds. However, in the present work, the only one that seemed to impact in a moderate way (OIR values from 168 to 655) was benzene, ethenyl- (Table 3).   www.nature.com/scientificreports/ Ketones. Acyclic (18) and cyclic (8) ketones were deemed a very representative chemical class in some of the studied strains. In this regard, IG and CV were the microalgae with the highest number of individual acyclic ketones (15 each) and also the highest percentage of RA (P ≤ 0.05) (15.9 and 13.9%, respectively). In turn, cyclic ketones were more representative of the VOCs profile in TS, SA, and AP in both number of compounds and percentage of RA (Table 2). In macroalgae or seaweed, acyclic ketones are usually related with less evolved brown seaweeds, while cyclic ketones are related with more evolved species 19 . In the current study, cyanobacterial strains showed a lower number of cyclic ketones than most microalgae ( Table 2). The relative abundance of acyclic ketones was significantly higher in IG when compared to the other strains (Table 1), mainly due to the extremely high abundance of 3,5-octadien-2-one (E,E)-and (Z,Z)-in the former. Both configuration isomers were also present as major ketones in SA and CV and they were detected exclusively in eukaryote microalgae whereas among prokaryote strains, the major ketone was 2-propanone (Table 1). Van Durme et al. 9 reported 1-penten-3-one, 3-pentanone and 2-butanone, 3-methyl-as major ketones in microalgae species. In addition, acyclic ketones were very similar between marine and freshwater strains in both, relative abundance and number of individual compounds.
Branched cyclohexanones and cyclopentanones were major compounds among cyclic ketones in all the studied strains. Cyclohexanone, 2,2,6-trimethyl-has been previously reported as a characteristic ketone in cyanobacteria 23 . Isophorone was present in both cyanobacteria and 4-oxoisophorone was a major ketone in CV and SA. Furthermore, 1,3-cyclopentanedione, 2-methyl-showed significantly (P ≤ 0.05) higher RA in IG, TS, and SA than in the other strains. In line with the high relative abundance (P ≤ 0.05) of β-ionone found in AP, cyanobacteria were previously featured as a rich source of carotenoids 23 . However, the results of the present work indicated that the presence of nor-carotenoids was strain-dependent, since β-ionone was not detected in the SY.
The origin of acyclic ketones is variable, while linear ketones are derived from lipid oxidation, some methyl ketones may result from β-oxidation of the fatty acids and subsequent decarboxylation and others are mainly products of oxidative cleavage of carotenoids as above-mentioned for β-cyclocitral 25 . In general, acyclic ketones are related with desirable odours in food. Saturated ketones are related with sweet, floral, and fruity odour notes, while unsaturated ketones are responsible for green odour notes 10 . Moreover, Van Durme et al. 9 related the seafood-like odour in microalgae to their high content in diketones such as 2,3-pentanedione and 3,5-octadien-2-one. In the present work, these compounds were found in considerable amounts in microalgae, while the only diketone found in cyanobacteria was 3,5-heptadien-2-ona, 6-methyl-(exclusively in AP). Prokaryotic chlorophytes such as AP have been previously related with high amounts of methyl-ketones such as 2-propanone, 2-pentanone, 2-peptanone and 2-octanone which contribute to green odour notes 24 . Attending to OIR values (Table 3), acyclic ketones generated a great odour impact mainly in IG and AP. 2,3-Pentanedione, 2-heptanone, 6-methyl-, and 3,5-octadien-2-one (E,E)-presented the highest OIR values in IG while 2-heptanone, 6-methylwas also deemed important in AP. The latter compound has been included in the camphoreous odour family (Supplementary Table S1). Among cyclic ketones, β-ionone was the major odorant in freshwater strains, and one of the most abundant odorants in marine strains (Table 3). Previous studies reported β-ionone as a potent odorant in some microalgae such as Scenedesmus sp. and CV and also in macroalgae 25,26 . In addition, this compound has been related to the characteristic odour of microalgae 10,25 together with β-ionone, 5,6-epoxy-.
Alcohols. Alcohols in microalgae are mainly formed as secondary decomposition of hydroperoxides of fatty acids, with the exception of branched alcohols that may also derive from carbohydrates via glycolysis or from amino acids through the Ehrlich pathway 27 . In the present study, marine eukaryotes IG, NG, and TS were the strains presenting higher diversity of acyclic alcohols with 9, 8, and 8 individual compounds detected respectively. The number of cyclic alcohols found in these species was lower and similar between them (Table 2). Similarly, the percentage of relative abundance of acyclic alcohols was, in general, significantly higher in marine (11.2-18.6%) than in freshwater strains (7.2-15.6%), except for CV that presented a high proportion of acyclic alcohols (15.6%). Overall, the abundance in cyanobacteria was significantly smaller (0.65-5.55%) when compared with microalgal strains. The diversity of cyclic alcohols was similar among the studied strains and TS presented the highest proportion (P ≤ 0.05) of cyclic alcohols (7.48%) ( Table 2). Regarding individual compounds, 1-penten-3-ol, 2-penten-1-ol, (Z)-, and 1-octen-3-ol were major compounds in all eukaryote strains while very different alcohol composition was observed between prokaryotes. Briefly, SY showed negligible amounts of both acyclic and cyclic alcohols in comparison to the other strains (P ≤ 0.05) whereas major alcohols in AP were 1-hexanol, 2-hexen-1-ol, (Z)-and cyclohexanol, 2,4-dimethyl-. It is remarkable the absence of geosmin and isoborneol, 2-methyl-in freshwater microalgae and specially in cyanobacterial strains since these two compounds have been previously reported as odorant compounds related to earthy-muddy odour notes in cyanobacterial 24 . Commercial high purity standards of both compounds were analysed in order to ensure their absence in cyanobacteria strains. These results were comparable to those of previous reports in which geosmin and isoborneol, 2-methyl-were no detected in AP 16 nor in other freshwater species 9 . Results reported herein were in line with previous research on VOCs content in microalgae, where TS presented lower content of alcohols when compared to CV and NG (Table 1) 9 . In contrast, Van Durme et al. 9 found low amounts of 1-octen-3-ol while in the present study this acyclic alcohol was major in most of the studied strains as in dehydrated edible seaweed 19 . Although alcohols have relatively high OT values, some unsaturated alcohols may exert an important impact in odour 20 . In the present work, 1-octen-3-ol (earthy, green, oily, fungal, grassy, and fatty) was deemed as one of the compounds with higher OIR in most species (Table 3).
Nitrogen and sulfur containing compounds. Nitrogen-and sulfur-containing compounds were also detected in the headspace of the selected strains. Sulfur compounds were deemed as major compounds (P ≤ 0.05) (in terms of relative abundance; www.nature.com/scientificreports/ sented 29.6 and 37.5% of total VOCs, respectively (Table 2). In turn, sulfur compounds were not detected, or detected in negligible abundance, in cyanobacteria and freshwater strains. The most abundant sulfur compound in microalgae was methyl sulfide, although the abundance of dimethyl sulfoxide was also important in some species such as IG (Table 1). Nitrogen compounds abundance was lower than that of other chemical families and only in IG, TS, SA and AP reached percentages between 3 and 5% of total VOCs abundance ( Table 2). The formation of sulfur compounds has been previously related with catabolism of free, peptide and protein sulfur-containing amino acids 28 , while in marine algae the presence of dimethyl sulfide was related to the degradation of sulfonio propionate, dimethyl-18 . On the other hand, nitrogen compounds are commonly described as derived from Maillard reactions due to high temperatures applied to food or biological matrices 27,28 , and pyrazines and pyridines have been reported as abundant in dried seaweeds 29 . In the present work, the microalgal strains were dehydrated by freeze-drying and the temperature used for volatile extraction was low (30 ºC). Therefore, it is unlikely that those procedures could originate thermal-derived nitrogen compounds. Probably, the pyrazines found in the headspace of microalgae and cyanobacteria could be originated from trimethylamine N-oxide degradation during storage 15 .
Both nitrogen-and sulfur-containing compounds have been related to fish odours 10,13 . Sulfur compounds were associated with the characteristic aroma of marine crustaceous 13 while alkylpyrazines generally deliver roasty odour notes 28 and nutty odour such as the case of pyrazine, methyl-(Supplementary Table S1). In fish, this compound has been related with fishy and ammonia odours. Nitrogen compounds seem to contribute little to the final odour due to their low OIR values (lower than 100) in most species (Table 3). On the contrary, the low OT values for methyl sulfide together with its high abundance in marine microalgae strains make methyl sulfide the most important odorant in these species (Table 3). In this regard, previous studies have reported that sulfur compounds were responsible for characteristic odours of marine microalgae such as cooked shrimp/ cooked seafood and marine and fishy odours 9 .
Esters, furans and other compounds. The RA of the esters in the strains was small and each individual compound seemed to be characteristic of each specie. Ethyl and methyl acetate were previously reported in microalgae biomass 10 . Diethyl phthalate was found in all the studied strains, being phthalate products regarded as toxic pollutants 18 .
Furans were found as constant compound present in microalgae VOC composition. The number of furan compounds was very similar in both marine and freshwater strains although a significantly (P ≤ 0.05) higher percentage of RA was observed in CV (9.00%) in comparison with the rest of the species ( Table 2). The abundance of furan 2-methyl-, furan, 2-pentyl-, and furan, 2-ethyl-was particularly considerable in IG, TS, and CV (Table 1). Moreover, furans have been reported previously as microalgae VOCs 9,16 . Furans can be formed by Amadori pathways from the oxidation of fatty acids or by glucose pyrolysis 28 . In general, furans have been identified as off-flavours of fat and oils imparting a beany, grassy liquorice, and tobacco odour notes 27 .

Materials and methods
Microalgae and cyanobacteria strains selection and production. Selected  Daily maximum, minimum, and average temperature inside the greenhouse were 27.0 ± 2.2, 11.7 ± 1.7 and 18.4 ± 1.7 °C, respectively. Average irradiance during the approximately 12 h of sunlight was 600.2 ± 72.2 µE/ m 2 ·s with peaks of 1500-1600 µE/m 2 ·s at midday. The pH of all the strains except for AP was controlled by ondemand injection of carbon dioxide at 8.0. Culture media used for the production of CV and SA was the Arnon medium 30 for AP the Arnon medium supplemented with sodium bicarbonate (16.8 g/L; pH 9.5 ± 0.2) and for NG, TS, IG and SY the Algal medium 31 . Once the biomass concentration reached approximately 1.5 g/L, the biomass was harvested and concentrated by centrifugation using a Sigma 3-18 KS centrifuge (Sigma Laborzentrifugen, Osterode am Harz, Germany) operating at 8000g for 10 min. The concentrated biomass with a concentration of approximately 20 g/L was immediately frozen at −80 ºC and freeze-dried using a Crydos-50 freeze-dryer (Telstar, Barcelona, Spain). The obtained dried powder was stored in a sealed plastic container at room temperature until further analysis.
Solid-phase microextraction of volatile compounds. Detailed description of the chemicals and suppliers used in the present experiment are described in the Supplementary Data.
Freeze dried microalgal/cyanobacterial biomass was weighted (0.300 ± 0.001 g) in triplicate in 10 mL amber vials (Agilent Technologies, Madrid, Spain) and 10 μL internal standard (IS) were added (0.1 mg/mL of cyclohexanone in hexane solution). Vials were subsequently sealed with PTFE septa and a steel magnetic cap (18 mm PTFE/SIL, Agilent Technologies), vortexed for 15 s, and left in a chilled room (4 ± 1 ºC) for 24 h prior to analysis.
Volatile compounds trapped onto the fiber were desorbed in the front injection port of the GC equipment for 15 min at 240 °C in splitless mode (split valve was opened at 200 mL/min after 10 min of the injection) using www.nature.com/scientificreports/ the autosampler device. After thermal desorption, the fiber was directly cleaned in the back injection port for 30 min at 270 °C. The working routine of the automatic sampler was to perform a blank (empty 10 mL amber vial) every three sample analyses. All the microalgal and cyanobacterial samples were analysed on the same day, and the samples were randomly located in the autosampler tray.
Gas chromatography-mass spectrometry analysis. Volatile compounds were analysed using a 7820A gas chromatograph (Agilent Technologies) equipped with two split/splitless injectors and coupled to a 5975 series mass spectrometry detector (Agilent Technologies). The volatile compounds were separated in a Supelcowax-10 (Supelco) fused silica capillary column (60 m long, 0.25 mm i.d., 25 µm film thickness) as described in Moran et al. 12 .The mass spectrometer consisted in a single quadrupole operating in full scan mode (1.4 scans/s, m/z range 26-350) at 230 ºC with a total ion current of 70 eV.
Chromatographic data were analysed with MSD ChemStation Data Analysis (version 5.52, Agilent Technologies). The limit of detection (LOD) was calculated from the noise obtained in the analysis of ten blanks. LOD was set as twice the average noise for each chromatographic zone. Mean linear retention index (LRI) values were calculated using the average real retention time of three replicates of each compound and the retention time of the standard saturated alkanes certified reference material. LRI values showed a variation coefficient less than 0.15% for all individual volatile compounds.
Tentative identification of volatile compounds was performed by comparing their mass spectra (matching factor > 800) with those of the National Institute of Standards and Technology (NIST version 2.0, Gaithersburg, USA). Additionally, peak identifications were confirmed by comparison of experimental LRI values with those previously published for volatile compounds analysed under similar chromatographic conditions when available. Positive identification was performed by comparison of the experimental LRI and mass spectra with those of commercial standards. Chromatographic peak areas were measured using selective integration for the four more abundant m/z ions of each target compound according to NIST mass spectra.
Peak areas (> LOD) of individual volatile compounds detected in at least two of the three replicates were used to calculate mean abundances in each sample. The volatile compound content of the samples was expressed as relative abundance (RA, arbitrary area units) to the area of IS according to the following equation: Peak areas were multiplied by 10 -5 for easier comprehension and three significant figures were used to express the RA of volatile compounds in the samples.
Odour impact ratio of volatile compounds. The odour intensity of the different volatile compounds identified was estimated by means of the odour impact ratio (OIR). Briefly, available odour threshold (OT) values measured in water were collected from available databases [32][33][34] , and the OIR for the individual volatile compounds was calculated as follows: Additionally, odour notes for volatile compounds were described according to The Good Scents Company database 35 and Giri et al. 27,28 and are provided in Supplementary Table S1. Statistical analysis. Statistical analysis of data was performed using IBM-SPSS version 25.0 (IBM, Armonk, USA). One-way analysis of variance (ANOVA) was applied to determine the statistical significance of the differences in the volatile composition of microalgae and cyanobacteria individual species. Levene's test was used to verify data homoscedasticity. Tukey's test was used for pairwise comparison among individual species. When a variable was not homoscedastic, the robust Welch test was applied, and Games-Howell test was used for pairwise comparisons. In case of lack of normality of the variables for both individual RA and percentage of RA of chemical families, the non-parametric Kruskal-Wallis H test was applied. Statistical significance was declared at P ≤ 0.05.

Conclusions
Microalgae and cyanobacteria are rich in VOCs and their characteristic volatile profile is strongly strain-dependent. Prokaryotic cyanobacteria, generally included within the term microalgae, are less prone to generate acyclic diketones which have been related to fishy odour notes when compared with eukaryotic microalgae. However, the influence of the individual species on the volatile profile was very significant since AP and SY cyanobacteria species showed completely different volatile profiles. Sulfur compounds can be considered as characteristic volatile compounds in marine microalgae, whereas some freshwater species such as AP and SA are rich in branched hydrocarbons. Results presented herein indicate that the volatile profile of microalgae should be individually evaluated since this profile is strongly strain-dependent. Further research is needed to relate the volatile profile with the biochemical and other compositional characteristics of each specie. Assessing the organoleptic attributes of microalgae (and cyanobacteria) is important when used for food applications, as the marine flavour and odour attributed to many microalgae strains could be used as a strategy to potentiate culinary preparations or develop novel innovative foods.