Supplement of Concerted measurements of lipids in seawater and on submicrometer aerosol particles at the Cabo Verde islands: biogenic sources, selective transfer and high enrichments

Synechococcus-like cells and Nanoeucaryotes


between (a) the concentration of dissolved PE and TCN in the ULW samples, between (b) the concentration of particulate PE and TCN in the ULW (black) and SML (blue) samples and between (c) the Lipolysis index of the total particulate lipids (LI∑PL) and the TCN concentration in the ULW (black) and SML (blue) samples
In addition to SIMPOL.1 (Pankow and Asher, 2008), the EVAPORATION model (Compernolle et al., 2011)

Pigments, nutrients and microbiological investigations in seawater
The pigment measurements of bulk surface water (Table S2) indicated temporal changes in the composition of the community with total pigment 5 concentrations between 0.4 µg L -1 and 1.5 µg L -1 during the campaign and an increasing trend in pigment concentration towards the end of the campaign. A nutrient limitation (Table S1), especially for phosphate (averaged concentration of 0.09 µmol L -1 in each the ULW and the SML) could explain the low total abundance of autotrophic organisms, which is also reflected in low chl-a concentrations. Table S3. The mean TCN value was 1.2•10 6 ±1.3•10 5 cells mL -1 in the ULW and 1.3•10 6 ±1.9•10 5 cells mL -1 in the SML samples, which is consistent with previous reports for surface water of subtropical regions (Zäncker et 10 al., 2018). The abundance of Synechococcus-like cells was 1.9•10 4 ±1.2•10 4 cells mL -1 (ULW), 1.9•10 4 ±9.1•10 3 cells mL -1 (SML) and for Nanoeucaryotes 1.5•10 3 ±8.2•10 2 cells mL -1 (ULW), 1.5•10 3 ±9.6•10 2 cells mL -1 (SML). Duhamel et al. (2019) reported a Synechococcus cell abundance of 2.5-9.2•10 3 cells mL -1 for seawater samples (20 m depth) taken in the western subtropical Atlantic Ocean. Although comparable data are lacking, the low cell abundances in the present study are indicative of an oligotrophic system. Synechococcus-like cells and Nanoeucaryotes showed a similar trend as chl-a (two slight increases, followed by depression), indicating that autotrophic organisms followed a similar temporal pattern. 5 10 15 20    XLogP3-AA value of compound by using the known log KOW of a reference compound as a starting point (Cheng et al., 2007). For each compound we also used the PubChem database (https://pubchem.ncbi.nlm.nih.gov/), an open chemistry database at the National Institutes 5 of Health (NIH), to extract chemical and physical properties.

Surfactant activity of investigated individual lipid classes:
To estimate the surface activity of the individual lipid classes based on their physico-chemical properties, individual parameters, namely the density, the partitioning coefficient between octanol and water (log KOW) and the topological polar 10 surface area were considered. One physical parameter which shows a superficial correlation with bubble and/or currenttransport susceptibility is specific gravity, also called relative density. An inverse relationship of this function with increased rates of surface accumulation by current seems logical since substances with lower densities would probably resist remixing into the water column once they have been dissolved and concentrated at the air-water interface (Brown et al., 1992). Based on available density data from the literature presented within Table S7, we may roughly estimate that more nonpolar lipids such as FFA and ALC should have higher susceptibility for the air-water surface (surface accumulation) in comparison to GL and PP. In fact, FFA and ALC have lower molecular masses and densities compared to GL and PP. The octanol-water partition ratio, in turn, is the most common way of expressing the lipophilicity of compounds in logarithmic form. Referred to as log KOW or log P it is obtained either through experimental procedures (Rothwell et al., 2005) or prediction approaches (Mannhold et al., 2009). In general, the positive values for log KOW indicate some hydrophobic character, whereas larger values lead to an 5 increased hydrophobicity. Molecules with low or negative values for KOW, however, are often defined as polar, although no direct link between KOW and the charge distribution in the molecule exists. Based on the ranking of XLogP3-AA data presented in Table S7, the most hydrophobic lipid would be TG. Another important factor which affects the adsorption is the solubility.
According to the Lundelius' rule (Lundelius, 1920), the extent of the adsorption of a surfactant could be assumed to be inversely proportional to its solubility in water. An increase of polar moiety contribution in a molecule heightens its hydrophilicity which 10 leads to an enhanced solubility in water. As the topological polar surface area (TPSA) of a molecule is defined as the surface sum over all polar atoms or molecules, primarily oxygen and nitrogen, also including their attached hydrogen atoms, it can serve as suitable measure to get a rough estimate of the magnitude of the surfactant activity of investigated lipids as illustrated in Table S5. It becomes apparent that the lower the TSPA, the higher is the surfactant activity. Consequently, the higher polar surface area of GL and PP shows a larger hydrophilicity in comparison with FFA and ALC. Since PP is indeed more soluble 15 than FFA or ALC, we expect less surface activity.  Considering the sample preparation, the averaged EFaer(∑lipids) dissolved (3•10 5 ) is considered in the following, since the filtration (sample preparation, section 2.2.1) of the particles in the dissolved fraction of seawater samples (≤0.7 µm) are in the size-range of PM1 aerosol particles (≤1 µm). The particulate fraction in seawater covers the particle size-range 0.7-200 µm. The averaged EFaer(∑lipids) particulate with 2•10 5 is similar to the EFaer(∑lipids) dissolved (3•10 5 ). For the calculation of EFaer, Eq. (3) was used. For the analysis of sodium in the SML n=5 samples were investigated. In the SML, the sodium concentration was 12.53 ± 0.53 g L -1 . Due to small relative standard deviation (4.2 % for SML), the mean value of sodium concentration in 5 SML samples (12.53 g L -1 ) was used for the calculation of EFaer. For the sodium concentration on PM1 aerosol particles, the measured atmospheric concentrations, listed in Table S8, were used. Moreover, for ( ) in Eq.
(3) the measured concentration of the respective lipid class of the dissolved fraction in the SML was used for the calculation as shown in Table S8. PP* -Phospholipids, including PE, PG and PC GL* -Glycolipids, including MGDG, DGDG, SQDG  (2015) c Henry's Law Constants were calculated based on the method by Meylan and Howard (1991) as mentioned in Sander (2015) 2 The adsorption coefficient in water (Kaq) was calculated using equation ( (1 • 10 −6 ), (saturation vapor pressure at 298.15 K in Pa=N m -2 = kg m -1 s -2 ), the gas constant (8.314 kg m 2 s -2 mol -1 K -1 ) and the temperature (298.15 K).

Adsorption of the individual lipid classes at the bubble air-water interface:
To estimate the adsorption of the individual lipid classes at the air-water interface, both the adsorption coefficient in water (Kaq) and the one in air (Ka) were calculated, as shown in Table S10. Considering the concentration of the adsorbed solute at the air-water interface as well as the equilibrium concentration of the solute, the principle of Ka was based on the approach of Kelly et al. (2004). According to it, the concentration in the SML and the saturation vapor pressure (p) of the analyte describe 15 the distribution of the analyte between the interface and the air, namely Ka. Ka expresses the maximum gas-phase concentration of the analyte before the condensation on the surface occurs. Also, another new adsorption coefficient, Kaq, was introduced in this context. It takes into account the concentration of the adsorbed solute at the air-water interface, but uses the saturation concentration of the solute in water instead of air. Kaq expresses the maximum amount of the analyte that can be dissolved. If this value is exceeded (Ka>Kaq), enrichment takes place in this medium. The saturation concentration of the solute in water 20 was calculated by multiplying p with the Henry's Law (HA) constant of the analyte. As for most analytes no HA constants has been determined, however, which is also the case for p, estimation programs had been applied to calculate these values shown by Table S10. The parameters p and HA for the standards of the individual lipid classes were calculated. A comparison of p by using different models (SIMPOL.1 vs. EVAPORATION) is further discussed in Fig. S16.
Overall, the results in Table S10 help evaluating the possible adsorption of the individual lipid classes at the bubble air-water 25 interface. Assuming that the differences between both adsorption coefficients, Kaq and Ka, were between 10 1 and 10 2 , Kaq~Ka was defined. For example, this was applied to TG with Kaq(TG):2.96•10 17 , Ka(TG):5.14•10 18 and ALC with Kaq(ALC):1.81•10 -2 , Ka(ALC):1.75•10 0 . Based on the ratio of the two adsorption coefficients to each other (Kaq~ Ka), we conclude that the lipid classes TG and ALC are preferably distributed at the interface, the bubble surface. As regards the EFaer (Table S9), TG and ALC showed the comparatively highest EFaer with 3•10 6 and with 1•10 6 , respectively. In contrast, if we look at the lipid class which 30 had the comparatively lowest EFaer (4•10 4 ), the ratio of the adsorption coefficients was Ka>>Kaq for MGDG, meaning that it was preferably distributed into water.