Biogeosciences Low temporal variation in the intact polar lipid composition of North Sea coastal marine water reveals limited chemotaxonomic value

Temporal variations in the abundance and com­ position of intact polar lipids (IPLs) in North Sea coastal marine water were assessed over a one-year seasonal cy­ cle, and compared with environmental parameters and the microbial community composition. Sulfoquinovosyldiacylglycerol (SQDG) was the most abundant IPL class, followed by phosphatidylcholine (PC), phosphatidylglycerol (PC) and diacylglyceryl-(lV, A, lV)-trimethylhomoserine (DGTS) in roughly equal concentrations, and smaller amounts of phosphatidylethanolamine (PE). Although the total concen­ trations of these IPL classes varied substantially throughout the year, the composition of the IPL pool remained remark­ ably constant. Statistical analysis yielded negative correla­ tions between IPL concentrations and dissolved inorganic nutrient concentrations, but no changes in the overall plank­ tonic IPL composition due to nutrient limitation were ob­ served. Significant correlations between SQDG, PC, PC and DGTS concentrations and chlorophyll a concentrations and algal abundances indicated that eukaryotic primary produc­ ers, in particular Phaeocystis globosa, were the predominant source of IPLs at this site. However, while IPL concen­ trations in the water were closely tied to total algal abun­ dances, the rapid succession of different algal groups bloom­ ing throughout the year resulted in only minor shifts in the IPL composition. Principal component analysis showed that the IPLs were derived from multiple sources, and that no IPL species could be exclusively assigned to a particular algal taxa or (cyano)bacteria. Thus, the most commonly occur­ ring IPLs appear to have limited chemotaxonomic potential, highlighting the need to use targeted assays of more specific biomarker IPLs.


Introduction
Intact polar lipids (IPLs) and their derived polar lipid fatty acids (PLFAs) are widely used in ecological and biogeo chemical studies as biomarkers to determine the abundance and composition of extant microbial communities.These lipid molecules are mostly glycerol-based with a hydrophilic (polar) head group attached to the sn-3 position and a wide variety of fatty acid chains at the s ; j -1 and sn-2 positions (see Fahy et al., 2005Fahy et al., , 2009 for an overview and classifi cation).As basic building blocks of cell membranes, lipids comprise 11-23 % of the organic carbon in marine plankton (Wakeham et al., 1997), and they often contain key elements such as nitrogen, phosphorous or sulfur.The characterization of the lipid content of marine microbes has shown that spe cific types of IPLs or PLFAs are synthesized predominately, or sometimes exclusively, by specific microbial groups.For example, the sulfur-bearing glycerolipid sulfoquinovosyldiacylglycerol (SQDG) is only found in thylakoid membranes of photosynthetic organisms (Benning, 1988;Frenzten, 2004), while long-chain polyunsaturated fatty acids (PUFAs) are typical of marine microalgae (Volkman et al. 1998;Guschina and Flarwood 2006).Although this primarily culture-based chemotaxonomic record is still far from comprehensive, spe cific IPLs or PLFAs may be used as biomarkers for the pres ence of their source organisms in different environments, with IPLs containing more structural information than their derived PLFAs (e.g.Shaw 1974;Lechevalier and Lechevalier 1989;Sturt et al. 2004).Moreover, IPLs are thought to be exclusively derived from living microbes, due to their comparatively rapid degradation upon cell death (White et al., 1979;Harvey et al., 1986), and IPL abundances are con sequently used as a proxy for the extant microbial biomass in environmental samples (e.g.Petsch et al., 2001;Lipp et al., 2008;Zink et al., 2008).Finally, microbes have the ability to adjust the IPL composition of their membranes in response to changes in their environment, such as temperature or nu trient availability (e.g.Van Mooy et al., 2009), although such adaptations have mostly been studied in cultures maintained under controlled conditions (Minnikin et al., 1974;Benning et al., 1995;Pernet et al. 2003;Martin et al., 2010).
At present the number of studies into IPL dynamics in the marine water column is still limited.This is partly due to the comparatively recent development of suitable instrumentation for IPL analysis, using multistage mass spectrometry coupled to high performance liquid chromatog raphy by electrospray ionization interface (HPLC/ESI-MS11; Brrigger et al., 1997;Fang and Barcelona 1998).Thus far, IPL compositions in marine waters have been de termined in the Black Sea (Schubotz et al., 2009), the Sargasso Sea and Pacific Ocean (Van Mooy et al., 2006, 2009;Van Mooy and Fredricks, 2010), the western North Atlantic (Popendorf et al., 2011a), the Mediterranean Sea (Popendorf et al., 2011b), and the North Sea (Brandsma et al., 2012).At all of these sites the IPL composition is dominated by a relatively small number of IPL classes, which are the glycerolipids sulfoquinovosyldiacylglycerol (SQDG) and mono-and digalactosyldiacylglycerol (MGDG and DGDG), the glycerophospholipids phos phatidylcholine (PC), phosphatidylglycerol (PG) and phosphatidylethanolamine (PE), and the betaine lipids diacylglyceryl-(A, A , A)-trimethylhomoserine (DGTS), diacylglyceryl-hydroxymethyl-(A, A , A)-trimethylalanine (DGTA) and diacylglyceiyl-carboxyhydroxymethylcholine (DGCC).Comparisons with other parameters measured in the same waters yielded tentative relationships between the IPL composition and the in situ microbial community composition (Van Mooy and Fredricks, 2010;Popendorf et al., 2011a;Brandsma et al., 2012), as well as the nutrient availability (Van Mooy et al., 2006, 2009;Popendorf, 2011b).
However, each of these studies presents a snapshot anal ysis, as all the data were collected within short amounts of time (several weeks at most), and thus the temporal variability of IPLs in marine waters has not yet been resolved in any detail.For this study we monitored the IPL abundance and composition of coastal North Sea surface water during a one-year seasonal cycle.We compare this IPL time series with the microbial abundances, community composition and environmental conditions at the same site and time interval, in order to determine how these are reflected in the IPL com position and abundances.

Study site and time series
From 1974 onwards, bucket water samples for environmen tal and microbial analyses have been collected from the NIOZ sampling jetty (53°00/06"N , 4°47/21"E) at the en trance of the Marsdiep tidal inlet, which connects the North Sea and the westernmost basin of the Dutch Wadden Sea (Fig. 1).Sampling is performed at high tide, to assure that the water originates from the southeastern coastal North Sea (Alderkamp et al., 2006), and includes measurements of salinity, water temperature and dissolved nutrients, as well as chlorophyll a concentrations, phytoplankton and bacterial abundances, marine algal species composition, and primary production.The sampling frequency is 40 to 60 times per year, varying from once or twice a month in winter up to twice a week during phytoplankton spring blooms (Cadée and Hegeman, 2002).The current study was synchronized with this long-term time series and ran over a one-year time period, from March 2007 to March 2008, comprising 28 sam pling dates.

Microbial analyses
Chlorophyll a concentrations were assessed from 0.5-1.01water samples (filtered over MgCCH-coated filters, as per Cadée and Hegeman, 2002), and calculated from non acidified values of chlorophyll a according to Philippart et al. (2010).Primary production was measured in an incu bator, kept at in situ temperature and constant light condi tions, using the 14C method of Cadée and Hegeman (1974), and including actual daily irradiation in the estimation model (Philippart et al., 2007).Phytoplankton samples were pre served with acid Lugol's iodine, and cells were counted with a Zeiss inverted microscope using 5 ml counting chambers.Most algae were identified to species level, but some were clustered into taxonomic and size groups (Philippart et al., 2000).Analysis of changes in the phytoplankton species composition covered the nine most numerous marine algal taxa, which together comprised more than 85 % of the total numbers of marine algae in the Marsdiep during the study period.
Samples for (cyano)bacterial abundances were preserved with formalin (final concentration 1.5%) and snap-frozen in liquid nitrogen before storage at -80 °C.After thawing, the microbial community composition was analyzed with a bench-top flow cytometer (Beckman Coulter XL-MCL) with reduced sheath-flow to enhance the sensitivity of the instru ment.Chlorophyll fluorescence (>630 nm) and phycoerythrin fluorescence (5 7 5 ± 2 0 n m ) of the cyanobacteria were collected in separate photomultipliers (Veldhuis and Kraay, 2004), and used as the primary selection criteria for the pres ence of cyanobacterial cells.Total bacterial numbers were determined by flow cytometry after staining the cells with the green nuclear stain PicoGreen (MP, P-7581), according to Veldhuis et al. (1997).Briefly, 10 pi of a working solution PicoGreen (100 times diluted in TBS buffer) was added to 100 pi of sample and incubated for 15-30 min prior to anal ysis.Green fluorescence of the stained DNA (525 ± 20 nm) was as used as the primary selection criterion for the pres ence of bacterial cells.

Intact polar lipid analysis
Surface water samples for IPL analysis (~2 0 1) were taken with an acid-rinsed Nalgene bottle from a depth of less than 1 m.The water was filtered through pre-combusted 0.7 pm GF/F filters (142 mm diameter; Whatman, Clifton, NJ, USA), using a table-top filtration unit.All filters were then freeze-dried and extracted using a modified Bligh-Dyer procedure (Bligh and Dyer, 1959;Brandsma et al., 2012).IPL analysis of the extracts was performed by high performance liquid chromatography electrospray ion ization tandem mass spectrometry (HPLC/ESI-MS2), us ing the chromatographic conditions described by Jaeschke et al. (2009) and source and fragmentation parameters de scribed by Boumann et al. (2006) and Brandsma et al. (2012).Initially, the extracts were analyzed in positive and nega tive ion mode (two separate runs) using a data dependent M S2 routine in which a full scan {m/z 300-1000) was fol lowed by fragmentation of the base peak of the resulting mass spectrum.Identification of the major IPL classes was based on diagnostic fragmentation patterns in the M S2 mass spectra (Kato et al., 1996;Brügger et al., 1997;Keusgen et al., 1997;Fang and Barcelona, 1998).Subsequently, tar geted mass spectrometric experiments were used to elucidate the structural diversity within each of the identified IPL classes, and for quantification of the IPL classes and their constituent species.IPLs with a phosphatidylcholine (PC) or diacylglyceryl-trimethylhomoserine (DGTS) head group were measured in positive ion mode by parent ion scanning {m/z 300-1000) of fragment ions diagnostic for their polar head groups (i.e.m/z 184 and m/z 236, respectively).IPLs with a phosphatidylglycerol (PG), phosphatidylethanolamine (PE) or sulfoquinovosyldiacylglycerol (SQDG) head group were measured by neutral loss scanning (m /z 300-1000) for losses of 189 Da, 141 Da and 261 Da, respectively.The car bon number and degree of unsaturation of the fatty acid moi eties of the various IPLs were calculated using the m/z of the molecular species, and these are denoted as such below (i.e.C32:2 PC refers to an IPL with a phosphatidylcholine head group and two fatty acids that contain a total of 32 carbon atoms and two double bond equivalents; note that this does not include the glycerol moiety).Information on individual fatty acid compositions of the predominant IPL species were based on fragment ions or neutral losses diag nostic for fatty acids, obtained in the data dependent M S2 experiments (Brügger et al., 1997).
For quantification of the PCs, PCs, PEs, SQDGs and DGTSs, the peak areas of each IPL class (total ion cur rent) and their constituent IPL species (extracted ion chro matogram) were compared with the respective peak ar eas of known quantities of authentic standards.The stan dards used in this study were: Ci6:o/Ci6:0 PC, C i6:o/Ci6:0 PG and Ci6:o/Ci6:0 PE (all Avanti Polar Lipids, Alabaster, AL, USA), a mixture of SQDGs containing predominately C i6 :i/C i8:2 SQDG (~60 %), but also small amounts of SQDGs with Ci6:0-i6:t> Ci8:o-18:1 and C20:5 fatty acid com binations (Lipid Products, Redhill, Surrey, UK), and a stan dard of C 14 ;o/C i8:i DGTS, which was purified from IPL ex tracts of Isochrysis galbana (CCMP 1323) as described by Brandsma et al. (2012).Limits of detection were 50-100 pg on column for the glycerophospholipids, 100 pg on column for the DGTSs and 1 ng on column for the SQDGs.All IPL quantifications were reproducible within a 10 % error be tween duplicate runs, and the instrument response was mon itored by repeated analysis of blanks and quantitative stan dards every 10 samples.

Statistical analyses
Relationships between the various datasets (IPL concen trations, environmental parameters, microbial abundances) were tested statistically in Systat 13 (Systat Software, San Jose, CA).The measures of association between different variables were determined by calculating their Spearman's rank correlation coefficients (p).This test was chosen as many of the variables showed a highly skewed distribution.Only variable dependencies having corrected probability val ues (p) of less than 0.05 were considered significant and are reported here.In addition, principal component analysis (PCA) was used to extract principal components that could explain the variance in the IPL dataset, both for the total IPL classes and the main IPL species (130 individual com ponents), plus the environmental parameters and microbial groups.

Temporal variability o f environmental parameters
During the time series the sea surface temperature in the Marsdiep varied from around 6 °C in winter to almost 19 °C in summer (Fig. 2a).Salinity was fairly stable at 26-31, although lower values (down to 23) were measured in De cember 2007 and in early spring.Levels of dissolved in organic nutrients (P, N and Si) were highest at the end of winter, then decreased sharply at the onset of spring and re mained low throughout most of the summer, before grad ually increasing again through fall and winter (Fig. 2b).Dissolved inorganic phosphate (DIP) levels ranged from a maximum of 1.15pm oll-1 to a minimum of 7 0 n m o ll-1 , while silicate (DISi) concentrations ranged from 42 pinoi I-1 to 42 0 n m o ll-1 , and nitrogen (DIN) concentrations ranged from 85 pinoi I-1 to 2.4 pinoi I-1 .NOj(" was the most abun dant DIN species in winter (> 9 0 % of the DIN pool), but comprised only 30-50 % in spring and summer, concurrent with strong increases in NO(7 (4-8 %) and NH)j" (30-65 %).
The N:P ratio of dissolved inorganic nutrients was highest at the end of winter and in spring (generally around 80, but with brief maxima up to 722), and lowest in summer (gener ally around 30, with a minimum of 13).

Microbial abundances and community composition
Primary production and chlorophyll a concentrations varied strongly throughout the year in response to the environmen tal conditions (Fig. 2c).In winter the primary production was low at 3-6 pgC I-1 h-1 , but increased to 1 7 2 p g C l_ 1 h_1 during the spring bloom.The same pattern was observed for the chlorophyll a concentrations, which increased from 2 .7 p g l-1 to 55 pg 1-1.After the spring bloom the primary production and chlorophyll a concentrations remained fairly high throughout the summer and fall, before decreasing to their low winter values.Within the eukaryotic algae, a sequence of blooms was observed at various times in the year, with total cell num bers of the nine most numerous taxa reaching 1.0 x IO5 cells ml-1 between mid-March and mid-May (Fig. 2d).The first and by far the most pronounced algal bloom occurred in spring and was formed by the prymnesiophyte Phaeocystis globosa, with the colonial form predominating during the first part of the bloom and the solitary form during the second part (Fig. 3).Concurrently, blooms of the diatoms Chaeto ceros socialis, Skeletonema costatum and Pseudonitzschia delicatissima, as well as other prymnesiophytes and var ious unidentified flagellate algae were observed.A sec ond and more moderate algal bloom occurred between mid-May and June and was formed by the diatoms Thalassiosira spp.and Chaetoceros socialis, together with the ciyptophyte Plagioselmis spp.and various unidentified flagellate algae (Fig. 3).Finally, the third and least pronounced algal bloom occurred during summer (July and October) and was again formed by the diatoms Thalassiosira spp.and Chaetoceros socialis, together with the cryptophytes Plagioselmis spp.and Hemiselmis spp.(Fig. 3), as well as cyanobacteria (up to 3.2 x IO5 cells ml-1 : Fig. 2d).Bacterial numbers were fairly constant throughout the year (3-5 x IO6 cells m l-1 : Fig. 2d), but were lowest at during the algal spring bloom (1.5 x IO6 cells m l-1 ) and highest around its end (6.1 x IO6 cells m l-1 ).

IPL composition and abundances
Five major IPL classes were detected in the surface waters of the Marsdiep (Fig. DGDG, and a number of unidentified compounds were de tected in some of the samples as well.Each of the identi fied IPL classes comprised a wide range of IPL species with differing fatty acid combinations (Table SI).The least vari ety was observed in the SQDGs and PGs (around 50 species each), followed by the PEs and DGTSs (around 90 species each), while the PCs were the most varied class (more than 120 species).Fatty acid chain lengths generally ranged from C i2 to C2 2 , although C 14 to Cis fatty acids predominated (Table SI).W hile the majority of the fatty acids in each of the IPL classes had even chain lengths, some odd-carbon number fatty acids (C13 to C 1 9 ) were also detected, in par ticular in the PEs and PGs.Long-chain polyunsaturated fatty acids (PUFAs) were common in the PCs, but rare in the other glycerophospholipids and DGTSs, and not detected in the SQDGs.The average fatty acid chain length and degree of unsaturation within each of the IPL classes remained stable throughout the year.Average fatty acid chain lengths were highest in the PCs (34.8 ± 0.7 carbon atoms), followed by the PEs (33.3 ± 0 .7 ), PGs (33.2 ± 0 .2 ) and DGTSs (33.1 ± 0 .7 ), and were lowest in the SQDGs (3 1 .0± 0 .5 ).Similarly, the average degrees of unsaturation were highest in the PCs (2 .5 ± 0 .4double bond equivalents), followed by the PGs, PEs and DGTSs (each 1.7 ± 0 .2), and lowest in the SQDGs (0.8 ± 0.2).
The glycerophospholipids (i.e.PC, PG and PE) detected in the coastal North Sea waters were always present in lower concentrations that the SQDGs.Summed glycerophospholipid concentrations ranged from 0 .6p g l-1 in winter to 9 .6p g l-1 at the peak of the spring bloom.PGs and PCs were present in more or less equal amounts, with concen trations ranging from 0.3 to 4 .8p g l-1 and 0.1 to 3.9pg 1-1, respectively (Fig. 2F).The PEs were the least abundant of the quantified IPLs, with concentrations ranging from less than lO n g l-1 to l.O p g l-1 .
In summary, the IPL pool in the surface waters of the Marsdiep contained a large number of IPL species (at least 400, but likely more), almost all of which could be de tected throughout the year, but with only a limited number of species (less than 40) making up the largest part of the total IPL pool.These predominant species showed little tempo ral variation, constituting fairly constant fractions of their re spective classes over time.However, during the spring bloom and in summer a number of IPL species were present in ele vated abundances (typically around 5 %, rather than <1 %), and the IPL pool appeared to be somewhat less diverse than at other times.

Statistics relationships
The measures of dependence between each of the measured variables (Spearman's p, n = 30) are given in Table S3.
Significant positive correlations were found between total SQDG, PC, PG, and DGTS concentrations, and chlorophyll a concentrations {p > 0.68), primary production {p > 0.68) and algal abundances {p > 0.53), and the four classes were also strongly inter-correlated {p > 0.75).Scatter plots of the log-transformed data revealed the relationship between these IPL classes and the algal abundances to be linear, with R2 values ranging from 0.45 for DGTS to 0.71 for SQDG (« = 28; Fig. 5).Furthermore, significant negative relation ships were observed with the dissolved nutrient concentra tions in the water {p < -0.51;Table S3).PE was the only IPL class that was not correlated with any environmental or microbial parameter measured here, or any other IPL class (Fig. 5; Table S3).The concentrations of the predominant individual IPL species were in general positively correlated with the total concentrations of their respective classes.
Principal component analysis (PCA) of the concentrations of the five IPL classes and microbial parameters yielded three principal components, explaining 86 % of the total variance (56 %, 17 % and 13 % for principal components 1, 2 and 3, respectively) in the dataset (Fig. 6 upper panel).PC, PG, SQDG and DGTS concentrations were positively loaded on the first axis, together with the algal abundances, chloro phyll a concentrations and primary production, while PE was positively loaded on the second axis.The bacterial abundances were positively loaded on the third axis, while the cyanobacterial abundances were negatively loaded on the second axis, but positively on the third axis.A further PCA of the main IPL species in each class (130 in total) and the algal community data and cyanobacterial and bac terial abundances, yielded only two principal components, which explained 61 % of the variance, although it should be noted that the number of variables far exceeded the number of sample points in this analysis.In this PCA, single-celled Phaeocystis globosa, Pseudonitzschia delicatissima and the unidentified flagellates group were strongly positively loaded on the first axis (Table S4), while the remaining algal taxa, cyanobacteria and other bacteria all showed minor loading factors (between 0.24 and -0.10).On the second axis, both forms of Phaeocystis globosa, the Prymnesiales and the unidentified flagellates group were negatively loaded (be tween -0.64 and -0.44;Table S4), while Plagioselmis spp., Hemiselmis spp., cyanobacteria and bacteria were positively loaded (between 0.35 and 0.70).The remaining taxa showed only minor loading factors between -0.28 and 0.04.The IPL species were all positively loaded on the first axis, with only 8 out of 130 species having a factor loading of less than 0.40.On the second axis, the factor loadings of the IPL species were fairly uniformly distributed between values of -0.64 and 0.76.
Finally, PCA of the IPL classes and environmental param eters yielded three principal components, explaining 88 % of the variance (Fig. 6, lower panel).PC, PG, SQDG and DGTS concentrations were again positively loaded on the first axis (44 %), with salinity and temperature positively loaded on the second axis (32 %) and PE positively loaded on the third axis (12%).The dissolved nutrient concentrations (DIP, DIN, DISi) were all negatively loaded on both the first and the sec ond axis.

IPL diversity in the coastal North Sea
The predominant IPL classes observed in the coastal North Sea waters were the glycerolipid SQDG, three glycerophos pholipids (PG, PC and PE) and the betaine lipid DGTS.The same classes have so far been found to dominate the IPL composition in a range of marine waters, from the Pacific Ocean and Sargasso Sea (Van Mooy et al., 2006, 2009;Van Mooy and Fredricks, 2010) to the Black Sea (Schubotz et al., 2009), the western North Atlantic Ocean (Popendorf et al., 2011a), the Mediterranean Sea (Popendorf et al., 2011b) and the North Sea and English Channel (Brandsma et al., 2012).W hile the glycerolipids MGDG and DGDG are often present in substantial quantities in some of these waters as well, they were only detected in trace amounts in the M ars diep samples, similar to the observations made by Brandsma et al. (2012)  Principal com ponent 1 (56%) Principal component 2 (32%) Principal component 1 (44%) studies, the structural diversity in IPLs was large, comprising at least 400 different IPL species, but of these only a limited number made up the bulk of the total IPL pool.Furthermore, despite the substantial changes in environ mental conditions and microbial community composition (Figs. 2 and 3), the temporal variations in the IPL pool ob served in the coastal North Sea waters were mostly quantita tive and not qualitative.In other words, while the abundances of the IPL pool varied greatly throughout the year, its internal composition showed relatively little change, and was mostly limited to an increased contribution during the spring and summer blooms of several IPL species (e.g.C4o:io PC, C4 2 :11 PC, C34:4 PG, C35:0 PC, C30:0 PE, C38:6 PE and C40:6 PE) that were otherwise present in low concentrations.The principal component analyses and Spearman results both indicated a high degree of covariance between the SQDGs, PCs, PGs and DGTSs (p > 0.77), while the PEs were unrelated (Fig. 6 and Table S3).The cause for the different statistical behavior of the PEs compared to the other IPL classes lies predom inately in its behavior during the spring bloom.W hile the IPLs in general increased in concentration from mid-March onward, PE concentrations remained at low values through out this period (Fig. 2f).However, all IPLs reached maxi mum concentrations at the start of May, and PE concentra tions behaved in much the same way as those of the other IPLs throughout the rest of the year.With this one signif icant exception, the general IPL composition in the coastal North Sea thus remained fairly stable throughout the year, unlike the variable environmental conditions and microbial community composition.

Relationship o f IPLs with environmental parameters
The IPL concentrations were statistically compared with the environmental data, in order to determine the influence of ex ternal parameters, such as temperature or nutrient concentra tions.The results from the statistical tests all showed either a negative or no relationship between the IPL abundances and environmental parameters, with the exception of temperature (Fig. 6 lower panel and Table S3).Concentrations of SQDG, PG, PC and DGTS were all negatively correlated with the nu trient concentrations in the water, and positively with temper ature.These are likely indirect relationships, with nutrients being incorporated into microbial biomass during the spring and summer blooms, which are triggered by rising tempera tures and light availability.The strong decrease in DIP con centration during the spring bloom resulted in high N:P ra tios of dissolved inorganic nutrients (up to 722), especially in early April and late May.This may have led to phospho rous limitation of the phytoplankton community, although it should be noted that nitrogen rather than phosphorous could still have been the limiting nutrient during the spring bloom, due to the comparatively more rapid and complete recycling of the latter (e.g.Dodds, 2003).Culture and environmental studies have shown that marine phytoplankton can rapidly substitute glycerophospholipids with non-phosphorous IPLs (i.e.SQDG and betaine lipids) when phosphate is scarce (Benning et al., 1995;Van Mooy et al., 2009;Martin et al., 2010;Van Mooy and Fredricks, 2010).However, in the coastal North Sea the ratios of SQDG to PG and DGTS to PC (as proposed by Van Mooy et al., 2009) remained fairly stable throughout the year (around 7.7 and 1.6, respec tively) .This implies that the phytoplankton community was not sufficiently limited in nutrients to necessitate substantial IPL substitution.Indeed, it was proposed by Van Mooy and Fredricks (2010) that this process only occurs at DIP con centrations below 30 nmol I-1 , whereas in the Marsdiep this value did not decrease below 70 nm oll-1 (Fig. 2b).

Sources of IPLs
The IPL concentrations were also statistically compared with chlorophyll a concentrations, primary productivity and the microbial abundances and community composition to inves tigate their sources.The significant correlations between SQDG, PG, PC and DGTS concentrations with the primary production rate, chlorophyll a concentrations and total al gal abundances imply that the majority of the IPLs in the coastal North Sea were related to the biomass of the eukary otic primary producers (Table S3).This was also reflected in the first PCA where these four IPL classes grouped together with these parameters (Fig. 6 upper panel).Scatter plots of the log-transformed IPL concentration and algal abundance data showed that the relationship was linear and strongest for SQDG (Fig. 5), in agreement with its role as the main an ionic IPL in thylakoid membranes of photosynthetic organ isms (Benning, 1988;Janero and Barrnett, 1982;Frentzen, 2004).
Like the total concentrations of their classes, the con centrations of the predominant SQDG, PC, PG and DGTS species could be related to the total algal abundances, as well as to the abundances of most individual algal taxa occurring at this site (Table S3).Indeed, studies of cul tured Thalassiosira (Zhukova, 2004;Martin et al., 2010), Chaetoceros (Servel et al., 1993;Zhukova and Aizdaicher, 2001), Skeletonema (Berge et al., 1995), cryptophytes such as Hemiselmis (Chuecas and Riley, 1969) and prymnesiophytes such as Phaeocystis (Al-Hasan et al., 1990;Hamm and Rousseau, 2003), have shown that each of the differ ent algal groups occurring in the coastal North Sea predom inately synthesize PC, PG, SQDG and betaine lipids (Sato, 1992;Dembitsky, 1996;Kato et al., 1996), containing com binations of C 14:0, C 16:4-16:0.Ci8:5-18:0, C20:5 and C22:6 fatty acids.A PCA of the main IPL species of each class together with the microbial community composition data was performed to determine if IPL species could be related more specifically to specific algal taxa (Table S4).The re sults clearly show that the IPL pool in the Marsdiep water has mixed origins, as no IPL species could be exclusively W hat can be inferred more is that single-celled Phaeocystis globosa, Pseudonitzschia delicatissima and the unidentified flagellates group were the dominant source of IPLs at this site, as almost all IPL species were positively loaded on the same axis as these algae.However, the same IPL species were simultaneously loaded on the second axis, indicat ing that they were also partly derived from colony-forming Phaeocystis globosa and Prymnesiales (negatively loaded; Table S4), and partly from Plagioselmis spp., Hemiselmis spp., and (cyano)bacteria as well (positively loaded; Ta ble S4).The few IPL species that were weakly loaded on both axes were possibly derived from the other algal taxa in vestigated here (i.e.Chaetoceros socialis, Thalassiosira spp. or Skeletonema costatum diatoms).
Interestingly, cyanobacterial cell numbers did not corre late significantly with total concentrations of any of the IPL classes, and in the PCA results plotted on different axes than the IPLs (Fig. 6 and Table S3).Combined with the low abun dances of MGDG and DGDG, which are common IPLs in cyanobacterial membranes (e.g.Murata and Nishida, 1987;Harwood and Jones, 1989), this suggests that cyanobacteria did not contribute substantially to the total IPL pool in the coastal North Sea.A likely reason for this is the small cell size of cyanobacteria compared to eukaryotes, which trans lates into a much lower total amount of IPLs per cell (see Veldhuis and Kraay, 2004 for a comparable argument on cell size and chlorophyll a content).
The PE concentrations could not be related to any of the measured microbial abundances, despite the fact that PE is presumed to be the main glycerophospholipid in bacte rial membranes (Shaw, 1974;Lechevalier and Lechevalier, 1989).The sharp increase in PE concentrations at the end of the spring bloom would point to a bacterial source, as max imum bacterial production rates in the Marsdiep are known to coincide with the collapse of the bloom (Van Boekei et al., 1992).However, bacterial abundances in the Marsdiep are strongly suppressed by heterotrophic nanoflagellate grazing (Brussaard et al., 1995), and it is therefore possible that this led to a mismatch between bacterial numbers and bacteriallyproduced IPLs (including PEs), or that the bacterial IPLs were rapidly transferred to higher trophic levels.Addition ally, concentrations of two PE species containing the PUFA C22:6 (be.C38:6 PE and C40:6 PE) were not related with the total PE concentrations, but rather with the concentra tions of the other glycerophospholipids and DGTS.As those IPL classes and long-chain PUFAs are normally associated with eukaryotic algae (Gushina and Harwood, 2006), a nonbacterial origin for those two PE species is likely.

Implications for IPL chemotaxonomy
Despite the large number of IPL species quantified, gen eral IPL analysis as performed in this study appears to lack the chemotaxonomic resolution to accurately differentiate within the microbial community, beyond the level of "ma rine algae", "phototrophs", or "(cyano)bacteria".The com munity composition analysis showed a rapid succession of algal species, with subsequent bloom periods throughout the year, which were only partly reflected in changes in the over all IPL composition.However, the rapid fluctuations in IPL abundances were closely linked to changes in the total algal counts, showing that in this type of environment IPLs pro vide a good biomarker for living microbial biomass.The lack of large temporal variations in the IPL composition and the fact that no unique source could be identified for any of the main IPL species, suggests that the IPL contents of the different algal groups occurring in the coastal North Sea must have been relatively similar.Indeed, studies of the main algal taxa occurring in this region show that each predomi nately synthesizes PC, PG, SQDG and betaine lipids, which contain combinations of C i4:o, Ci6:4-16:0, Cl8:5-18:0, C20:5 and C22:6 fatty acids.The prevalence of these IPLs across a wide range of algal groups and throughout the w orld's oceans (e.g.Schubotz et al., 2009;Van Mooy and Fredricks, 2010;Popendorf et al., 2011a, b;Brandsma et al., 2012) further suggests that general IPL screening of marine waters may yield little chemotaxonomic information.In future studies it will therefore be necessary to target more specific biomarker IPLs, such as anammox bacterial ladderanes (Jaeschke et al., 2009;Brandsma et al., 2011) or cyanobacterial glycerolipids (Bauersachs et al., 2009), in order to accurately track the presence of specific microbial populations in the environ ment.

Conclusions
The coastal marine waters of the Marsdiep tidal inlet con tain a wide range of IPLs, whose composition is comparable to that of the adjacent southern North Sea.Despite substan tial variations in their abundances, the IPLs showed relatively little compositional changes over the year.Concentrations of SQDGs, PGs, PCs and DGTSs mostly co-varied, and their abundances were linked to the total algal biomass in the w a ter.The origin of the PEs at this site remains unclear, al though they may have been related to bacterial production at the end of the algal spring bloom.Intriguingly, the overall IPL species distribution through time did not reflect the suc cession of algal groups, implying that their IPL compositions are similar.Finally, no direct influence of environmental con ditions on the IPL composition was observed.

Fig. 1 .
Fig. 1.Map o f the southeastern North Sea and Wadden Sea; the arrow marks the sampling site at the entrance o f the Marsdiep tidal inlet.

□Fig. 3 .
Fig. 3. Abundances o f the different algal groups in the Marsdiep.The upper graph shows the absolute abundances, while the lower graph shows relative abundances (normalized to the total counts).

Fig. 4 .
Fig. 4. Partial base peak chromatogram (positive ion -Gaussian smoothed) showing the IPL classes identified in Marsdiep water during the phytoplankton spring bloom in late April.Unidenti fied peaks are indicated with a question mark.Example structures are given for each of the quantified IPL classes: diacylglyceryltrimethylhomoserine (DGTS).phosphatidylglycerol (PG).phosphatidylethanolamine (PE), phosphatidylcholine (PC) and sulfoquinovosyldiacylglycerol (SQDG).Each peak comprises a wide range o f IPLs with the same head group, but different fatty acids at the s n -l and sn-2 positions (R; and R/; in the example structures).Due to differences in response factors between the IPL classes, their relative abundances in the base peak chromatogram are not neces sarily indicative o f their respective absolute abundances.

Fig. 6 .
Fig. 6.Principal component analysis (PCA) plots for the total concentrations o f the IPL classes with microbial abundances, chlorophyll a concentrations and primary productivity (upper panel).and with environmental parameters (lower panel).