Lipid biomarkers for algal resistance to viral infection in the ocean

Significance

Algal blooms, rapid proliferation events of marine microalgae, are of great importance to the marine food web and to the biogeochemical cycles of nutrients, such as carbon and sulfur. The extent and duration of these blooms are controlled by multiple environmental factors. Blooms of the coccolithophore Emiliania huxleyi are frequently terminated following viral infection. The annual formation of E. huxleyi blooms indicates the existence of resistant cells that survive viral infection and form the seed population for subsequent blooms, nevertheless, their occurrence has not been reported in natural blooms to date. Here, we identified lipid biomarkers for resistant E. huxleyi cells and applied them to detect resistant cells in an open-ocean bloom of E. huxleyi.

Abstract

Marine viruses play a key role in regulating phytoplankton populations, greatly affecting the biogeochemical cycling of major nutrients in the ocean. Resistance to viral infection has been reported for various phytoplankton species under laboratory conditions. Nevertheless, the occurrence of resistant cells in natural populations is underexplored due to the lack of sensitive tools to detect these rare phenotypes. Consequently, our current understanding of the ecological importance of resistance and its underlying mechanisms is limited. Here, we sought to identify lipid biomarkers for the resistance of the bloom-forming alga Emiliania huxleyi to its specific virus, E. huxleyi virus (EhV). By applying an untargeted lipidomics approach, we identified a group of glycosphingolipid (GSL) biomarkers that characterize resistant E. huxleyi strains and were thus termed resistance-specific GSLs (resGSLs). Further, we detected these lipid biomarkers in E. huxleyi isolates collected from induced E. huxleyi blooms and in samples collected during an open-ocean E. huxleyi bloom, indicating that resistant cells predominantly occur during the demise phase of the bloom. Last, we show that the GSL composition of E. huxleyi cultures that recover following infection and gain resistance to the virus resembles that of resistant strains. These findings highlight the metabolic plasticity and coevolution of the GSL biosynthetic pathway and underscore its central part in this host–virus arms race.

Viruses are the most abundant biological entities in the marine environment and serve as major evolutionary and biogeochemical drivers in the oceans (1–4). Algae-infecting viruses are estimated to turn over a substantial portion of the photosynthetically fixed carbon, thus fueling microbial food webs, short-circuiting carbon transfer to higher trophic levels and promoting its export to the deep sea (5, 6). Recent advancements in single-cell technologies allow to better quantify infected cells in the natural environment (7–10), yet studying host–virus dynamics in natural populations (11) remains a major challenge, which limits our understanding of the possible phenotypic outcomes of viral infection.

The ongoing evolutionary arms race between algae and their viruses leads to diverse defense strategies, supported by continuous genetic and phenotypic adaptations of the algal cells (12–14). Resistance to viral infection has been reported for several algal species, both as isolates from natural populations and as sub-populations that emerge following infection under laboratory conditions (13, 15, 16). Nevertheless, the prevalence of resistant phenotypes in nature is currently unknown, as we lack sensitive tools to detect resistant cells in mixed populations, hindering our understanding of their ecological importance.

The cosmopolitan alga Emiliania huxleyi and its specific virus, E. huxleyi virus (EhV), are an attractive system to study host–virus interactions. E. huxleyi forms vast annual blooms in the ocean that play an important role in regulating the global biogeochemical cycling of carbon and sulfur (17–19) and are frequently infected and terminated by EhV (20–23). Viral infection leads to profound rewiring of the E. huxleyi metabolism, including changes in glycolysis, elevated fatty acid (FA) synthesis, and alterations in lipid composition (24–27). Particularly, EhV is the only virus known to date to encode almost a complete pathway for sphingolipid (SL) biosynthesis, resulting in the production of structurally distinct virus-derived glycosphingolipids (vGSLs) by infected cells (28–30). vGSLs were found to trigger host programmed cell death and are central components of the EhV membranes (30, 31). In addition, E. huxleyi cells produce host-derived GSLs (hGSLs), which are found in all E. huxleyi strains and serve as a proxy for healthy cells (31, 32), and sialic acid GSLs (sGSLs), which characterize susceptible E. huxleyi strains and were suggested to be involved in viral attachment and entry (31). Given their structural variability and diverse roles, SLs are key players in the arms race between E. huxleyi and its virus.

Resistance to viral infection has been described in several E. huxleyi strains and was previously attributed to ploidy level, genome and transcriptome variations between the strains (13, 33), to expression and activity of specific enzymes, such as DMSP-lyase, and to metacaspase expression (34, 35). Resistant cells were also identified in low numbers (<1%) in infected E. huxleyi cultures (36), revealing that resistance can also be triggered by viral infection. These resistant cells were found to be morphologically distinct from their susceptible progenitors, indicating the involvement of a life-phase transition and highlighting the phenotypic plasticity within E. huxleyi populations during infection (13, 36). Nevertheless, single-cell isolation coupled with infectivity assays remain the only method to identify resistant E. huxleyi cells. Moreover, the metabolic basis of E. huxleyi’s resistance to viral infection is unknown, as is the prevalence of resistant E. huxleyi cells in natural populations. In this study, we aimed at addressing this conundrum by identifying specific lipid biomarkers for resistant E. huxleyi cells and applying them to natural mixed populations.

Results

Untargeted Lipidomics Profiling of Virus-Resistant and Susceptible E. huxleyi Strains.

To identify lipids that are characteristic of resistant strains, we compared the lipidome of four E. huxleyi strains that differ in their susceptibility to viral infection by EhV201: the resistant E. huxleyi strains CCMP373 and CCMP379 and the susceptible E. huxleyi strains CCMP2090 and CCMP374 (hereinafter, E. huxleyi strains 373, 379, 2090, and 374, respectively) (31, 33, 34). Previous studies reported that following infection of E. huxleyi cultures by EhV, a small proportion of the population (< 1%) survives and acquires resistance to the virus (13, 36). We were therefore interested to delineate possible correlations between the lipid profile of resistant strains and the evolving resistant population within infected susceptible cultures.

The lipidome of the resistant and susceptible strains in the presence and absence of the lytic virus EhV201 (hereinafter, EhV) was compared over a three-day time course using liquid chromatography–high-resolution mass spectrometry (LC-HRMS)-based untargeted lipidomics. All untreated cultures grew throughout the experiment, reaching 1.0 to 2.4 × 10⁶ cells per mL (Fig. 1A). The resistant E. huxleyi strains 373 and 379 grew throughout the experiment regardless of the presence of EhV and with no accumulation of virions in the media (Fig. 1B). In contrast, upon addition of EhV, the susceptible E. huxleyi strains 2090 and 374 showed growth arrest one day post infection (dpi) and were subsequently lysed (Fig. 1B). Concomitantly, accumulation of virions was detected in the media starting from 1 dpi. In all cultures, cells were harvested at four different time points (0, 1, 2, and 3 d) for lipid extraction and untargeted lipidomics analysis.

Fig. 1.

Untargeted LC-HRMS-based lipidomics analysis reveals differences between virus-resistant and susceptible E. huxleyi strains. (A) Cell abundance during growth of E. huxleyi strains that differ in their susceptibility to viral infection: the resistant (R) E. huxleyi strains 373 and 379 and the susceptible (S) E. huxleyi strains 2090 and 374. (B) Cell abundance (black lines) and production of virions (gray lines) following addition of EhV. Values for A and B are presented as mean ± SD (n = 3). (C) Clustering of resistant and susceptible E. huxleyi strains based on untargeted lipidomics (12,190 mass features) and k-means clustering (k = 4, SI Appendix, Fig. S1A), as visualized by PCA. (D) Clustering of resistant and susceptible E. huxleyi strains in the presence and absence of EhV based on untargeted lipidomics (12,190 mass features) and k-means clustering (k = 4, SI Appendix, Fig. S1B), as visualized by PCA. Percentage of explained variance is stated in parentheses. Each cluster (CL) is surrounded by an ellipse, with the mean marked by “×.”

First, we compared the lipidome of the four strains in the absence of EhV. Unsupervised k-means clustering of the extracted data (n = 48; 12,190 mass features, k = 4, SI Appendix, Fig. S1A), visualized by principal component analysis (PCA), separated the strains into four distinct clusters (clusters 1 to 4, Fig. 1C). The first PC axis (31.8%) revealed a clear separation between the susceptible and resistant strains (clusters 1 and 2 vs. clusters 3 and 4, respectively), and the second PC axis (16.7%) highlighted further differences between the strains. Next, we applied k-means clustering to the whole dataset of cultures with and without addition of EhV (n = 96; 12,190 mass features, k = 4, SI Appendix, Fig. S1B), which showed a clear separation between susceptible and resistant strains (clusters 5 and 6 vs. clusters 7 and 8, respectively) along the first PC axis (40.1%, Fig. 1D). The second PC axis (15.5%) further separated the susceptible strains at late infection stages (2 and 3 dpi; cluster 5) from early infection stages (0 and 1 dpi) and the uninfected cultures (cluster 6).

Next, we focused on mass features that were differential between the strains in the absence of EhV, as the strains were separated to distinct clusters (Fig. 1C). We used a comparative analysis (one-way repeated measures ANOVA, P < 0.01), which reduced the data to 231 differential mass features. Following feature deconvolution and manual curation, these mass features were grouped into 54 putative lipid species (SI Appendix, Table S1). Interestingly, we putatively annotated five of these lipid species as GSLs (based on characteristic neutral losses and fragments of long-chain bases (LCBs) and amino FAs visible in the MS/MS spectra), of which four have not been described before in the E. huxleyi-EhV system (2, 9, 12, and 14, see Table 1 and SI Appendix, Figs. S2–S6 and Table S1). Since GSLs are known to play various roles in the E. huxleyi-EhV system, this encouraged us to manually search for additional GSL species that are differential between the strains and might have been filtered out in the initial analysis. Indeed, we identified ten additional GSL species that had differential intensity between the strains (1, 3–8, 10–11, 13, see Table 1 and SI Appendix, Figs. S2 and S7–S16 and Table S2).

Table 1.

Putative annotation of GSL species that differ between resistant and susceptible strains and are previously undescribed in the E. huxleyi-EhV system

Group	#	GSL species LCB/FA	RT (min)	Measured m/z ([M+H]+)	Predicted formula
A. Higher in resistant strains	1	d18:3/h22:1	13.12	794.6119	C46H83NO9
	2	d18:3/h22:2	12.47	792.5980	C46H81NO9
	3	d19:3/h21:1	13.03	794.6107	C46H83NO9
	4	d19:3/h23:2	13.18	820.6278	C48H85NO9
B. Only in resistant strains and during infection	5	d18:0/h22:0*	14.44	802.6722	C46H91NO9
	6	d18:0/h22:1†	14.22	800.6600	C46H89NO9
	7	d18:1/h22:1*	14.01	798.6440	C46H87NO9
	8	t18:0/h22:0*	14.00	818.6702	C46H91NO10
	9	t18:0/h22:1	13.77	816.6531	C46H89NO10
	10	t18:0/h22:2‡	13.18	814.6346	C46H87NO10
C. Only in resistant strains	11	d19:4/h22:1 (resGSL)	12.92	806.6127	C47H83NO9
	12	d19:4/h22:2 (resGSL)	12.25	804.5975	C47H81NO9
D. Only in the susceptible E. huxleyi strain 374	13	d19:3/h22:2 (374-GSL)§	12.98	806.6143	C47H83NO9
	14	d19:3/h22:3 (374-GSL)§	12.34	804.5981	C47H81NO9

Differences in the abundance profiles were tested by a one-way ANOVA, accounting for the strain and addition of EhV, followed by Tukey’s post-hoc test, P < 0.01 (SI Appendix, Tables S11 and S12).

^*Ceramides d18:0/h22:0 and t18:0/h22:0 were previously found to increase during infection (29). Ceramide d18:1/h22:1 was previously found to increase during infection and in resistant haploid cells (37).

^†GSL d18:0/h22:1 (6) presence in infected cells could not be verified using MS/MS due to low intensity.

^‡GSL t18:0/h22:2 (10) was detected in infected cells based on MS/MS analysis.

^§374-GSL d19:3/h22:2 (13) has the same fragmentation pattern as hGSL d19:3/h22:2, yet appears at a slightly later retention time (SI Appendix, Fig. S29). 374-GSL d19:3/h22:3 (14) was previously described as a hGSL species (31); however, it was not detected in E. huxleyi 373, 379, and 2090 in this study. GSL species were identified as “Level 2 – putatively annotated compounds” according to the Metabolomics Standards Initiative (38). LCB, long-chain base; FA, fatty acid; RT, Retention time.

We then applied two-dimensional hierarchical clustering to the subset of differential lipid species using the whole dataset (that is, with and without addition of EhV; Fig. 2A). This subset of lipid species recapitulated the previously observed separation (Fig. 1D) between the resistant and susceptible strains and between each pair of strains. Similarly, while there was no clear separation between resistant strains in the presence and absence of EhV, the susceptible strains infected with EhV were clustered separately from the uninfected cultures as early as 1 dpi. The putative lipid species were grouped into two main clusters, each further divided into two sub-clusters (Fig. 2A): (i) lipids with higher intensity in the resistant strains (especially E. huxleyi 379), of which most had higher intensity also in infected E. huxleyi 2090 cultures; (ii) lipids with higher intensity in both resistant strains; (iii) lipids with higher intensity in E. huxleyi strain 374 or in both susceptible strains; and (iv) lipids with higher intensity in the resistant E. huxleyi strain 373 and the susceptible E. huxleyi strain 374. Thirty-one putative lipid species were higher in one or both resistant strains (sub-clusters i and ii). Some of these species were elevated in the resistant E. huxleyi strain 379 compared to E. huxleyi strain 373, shedding light on possible metabolic differences between these two resistant strains. Seventeen putative lipid species were higher in the susceptible E. huxleyi strains 374 and 2090 (sub-cluster iii), one of which was identified as the known sGSL d18:2/c22:0 (31), and eight of which were higher in E. huxleyi 374 compared to E. huxleyi 2090.

Fig. 2.

Putative lipid biomarkers for E. huxleyi strains differing in their susceptibility to viral infection. (A) Two-dimensional hierarchical clustering of 64 putative lipid species (SI Appendix, Table S1) in four E. huxleyi strains in the presence and absence of EhV throughout a time course of four days (n = 32). Clustering was performed on log-transformed and standardized mean peak areas (n = 3) of the adduct ion with the highest intensity (see SI Appendix, Fig. S28 for the nonaveraged data). Samples are grouped into two main clusters that separate the resistant (R) strains from the susceptible (S) ones. Each cluster forms two sub-clusters that further separate the strains. The putative lipid species are divided into four sub-clusters (i to iv). Fourteen identified GSL species are marked by numbers (Table 1). *sGSL d18:2/c22:0. (B) Putative structures of four previously undescribed GSL species in the E. huxleyi-EhV system (2, 9, 12, 14), which are differential between the resistant and susceptible E. huxleyi strains. See SI Appendix, Fig. S2 for putative structures of all GSL species identified in this study. The structures, including LCB and FA composition, were determined based on LC-MS/MS analysis (SI Appendix, Figs. S3–S6). The positions of the double bonds and functional groups were assigned based on the most common structures in the LMSD (39).

The identified GSL species varied in their LCB composition, including dihydroxylated LCBs d18:0, d18:3, d19:3, and d19:4, and the trihydroxylated LCB t18:0 (Fig. 2B). We classified these GSL species into four groups based on their abundance in the different strains (Table 1 and SI Appendix, Figs. S17 and S18): (A) GSL species that are highly abundant in the resistant strains compared to susceptible strains (1–4, difference of >1 order of magnitude). These GSL species contain LCB d18:3 and d19:3; (B) GSL species that are found in resistant strains and in infected susceptible strains (5–10). These contain LCB d18:0, d18:1, and t18:0. (C) GSL species that are found only in the two resistant strains, with higher abundance in E. huxleyi 379 compared to E. huxleyi 373 (11–12, SI Appendix, Fig. S18). These contain LCB d19:4 and were termed resistance-specific GSLs (resGSLs) due to their detection in resistant strains and their absence in susceptible strains (of which, resGSL d19:4/h22:2 (12) appears in higher abundance). (D) GSL species that are found only in E. huxleyi 374 (13–14). These contain LCB d19:3 and were termed E. huxleyi 374-specific GSLs (374-GSLs). We analyzed the GSL composition of three additional strains, one susceptible (E. huxleyi RCC1216) and two resistant (E. huxleyi RCC1217 and CCMP2090-Rec-C2) (SI Appendix, Fig. S19). resGSL d19:4/h22:2 (12) was detected in the two resistant strains but not in the susceptible strain (SI Appendix, Fig. S19A). Interestingly, the resistant strain CCMP2090-Rec-C2, which was isolated following recovery of an infected E. huxleyi 2090 culture, contained low amounts of sGSL d18:2/c22:0 (SI Appendix, Fig. S19B), which was previously detected only in susceptible E. huxleyi strains (31).

GSL species containing LCBs d18:1, d18:3 and d19:4 were not detected thus far in the E. huxleyi-EhV system. LCBs d18:0, d19:3, and t18:0 were previously reported in the E. huxleyi-EhV system: LCB d19:3 in hGSL species and LCB d18:0 and t18:0 in infection-derived GSL and ceramide species (SI Appendix, Table S3) (29, 40). Intriguingly, GSL species containing LCB t18:0 (8–10), which were detected in resistant strains and in infected cultures (SI Appendix, Fig. S17), varied in their FA composition: resistant strains produce GSL species with a preference for monounsaturated and diunsaturated FAs over saturated ones (h22:1 and h22:2 vs. h22:0, SI Appendix, Fig. S17). Infected cultures, on the other hand, produce GSL species with saturated and monounsaturated FAs, as was previously described for t17:0-based vGSL species (29, 30). Importantly, trihydroxylated LCBs were previously found only in vGSL species and were considered a unique attribute of viral infection, derived from the virus-encoded biosynthetic pathway. The tetraunsaturated LCB d19:4, on the other hand, appears only in resGSLs found resistant strains, and therefore, we suggest that these unique resGSLs can be used as a biomarker for resistant cells in natural populations. Detection of GSL species with tetraunsaturated LCB and trihydroxylated LCB in resistant strains suggests the involvement of specific modifying enzymes in these strains.

Potential Enzymes Involved in Modulating GSL Composition in Resistant Strains.

The detection of resGSL species with LCB d19:4 (11–12), which contain an additional double bond compared to the LCB d19:3 found in hGSL species (SI Appendix, Table S3), indicates the involvement of an additional sphingolipid desaturase (SLD) in resistant strains. A gene encoding a putative SLD was previously identified in E. huxleyi (sld2) (24, 41), and we identified four additional genes based on the E. huxleyi genome and expressed sequences (sld1, sld3-sld5, SI Appendix, Table S4). Phylogenetic analysis of the conserved domain of the SLD proteins revealed three distinct clades (I-III, SI Appendix, Fig. S20A and Table S5), each consisting of diverse taxonomic groups. We further examined the expression of these genes using previous transcriptomics experiments with E. huxleyi strains 373, 379, 2090, and 374 (33, 42). Out of the five putative E. huxleyi sld genes, sld1 and sld4 were expressed in the resistant E. huxleyi strains 373 and 379 and not in the susceptible strains (SI Appendix, Figs. S20B and S21A). The other genes (sld2, sld3 and sld5) were expressed in all strains (SI Appendix, Figs. S21A and S22A). While SLD1 clustered together with the viral SLD (EhV201 SLD, AET97947.1, clade I), SLD4 fell into a different clade (clade III). Since infected cells do not produce GSL species with LCB d19:4, we suggest that SLD4 might be responsible for the formation of the fourth double bond in resGSLs (11–12). SLD1, on the other hand, might share a similar role to that of the viral SLD in the GSL biosynthetic pathway. Further biochemical analyses are required to validate the roles of the specific family members.

We were further intrigued to identify possible similarities between the viral and the host biosynthetic pathways that are responsible for the production of GSL species with trihydroxylated LCBs in infected and resistant cells. Previous studies suggested that the characteristic trihydroxylation of the LCB in infection-derived vGSL species is facilitated by a viral sphingoid base hydroxylase (EhV201 SBH, AET97919.1) (29, 43), which is highly expressed at early stages of infection (SI Appendix, Fig. S23). LCB t17:0 is the major LCB in vGSL species, while LCB t16:0 and t18:0 are found in lower abundances (29). In GSL species of resistant strains (8–10), on the other hand, only LCB t18:0 was detected. A gene encoding a putative SBH was previously identified in E. huxleyi (sbh1) (24, 33), and we identified six additional genes based on the E. huxleyi genome and expressed sequences (sbh2-sbh7, SI Appendix, Table S4). Phylogenetic analysis of the conserved domain of the SBH proteins revealed that the E. huxleyi SBHs do not form a clade together but rather show similarities to diverse phyla, indicating different evolutionary origins (SI Appendix, Fig. S20C and Table S6). Interestingly, SBH4 and SBH5 clustered together with the viral SBH, indicating a possible host–virus coevolution. Out of the seven SBHs, sbh4 and sbh5 were highly expressed in the resistant E. huxleyi strains 373 and 379 and not in the susceptible E. huxleyi strains (SI Appendix, Figs. S20D and S21B). Concomitantly, sbh2 was differentially expressed in the susceptible strains, while sbh1 and sbh6 were expressed in all four strains. sbh7 was detected in all four strains, with higher expression in infected E. huxleyi 2090 cultures. The expression of sbh3 was not detected in all strains and conditions tested (SI Appendix, Figs. S21B and S22B). Future functional analysis of these SLDs and SBHs will allow to determine their role in the biosynthetic pathway of GSL species in different E. huxleyi strains and during viral infection.

Detection of Resistant Algal Cells in an Open Ocean Bloom Using a Lipid Biomarker.

Since little is known about resistance to viral infection in algal blooms, we sought to utilize our resistant metabolic biomarker (resGSL) to assess the occurrence of resistant cells in an oceanic E. huxleyi bloom. To that end, biomass samples for lipidomics analysis were collected during the “Tara Breizh Bloom” cruise in the Celtic Sea, capturing the demise phase of an E. huxleyi bloom (Fig. 3A) (44). The occurrence of hGSL species (Fig. 3B), which are known lipid biomarkers for E. huxleyi and are present in all strains (30, 32), confirmed the presence of E. huxleyi cells, as was also visible using scanning electron microscopy (44). sGSL species, which characterize susceptible strains (31), were also detected (Fig. 3C), indicating the presence of virus-susceptible E. huxleyi cells in the water. We could also detect 374-GSL species (group D, 13-14) at a similar intensity as the hGSL species (Fig. 3D), indicating that some E. huxleyi cells share similarity to the susceptible E. huxleyi strain 374. Importantly, we detected resGSL d19:4/h22:2 (12) in four out of five days of sampling (Fig. 3E), revealing the presence of resistant E. huxleyi cells during bloom succession of E. huxleyi. The occurrence of hGSL, sGSL, 374-GSL, and resGSL species during the demise phase of the bloom suggests a complex population composition toward the end of the bloom.

Fig. 3.

Detection of resGSL in an open-ocean E. huxleyi bloom. (A) Satellite ocean true-color image from the Visible Infrared Imaging Radiometer Suite (VIIRS) onboard the Suomi National Polar-orbiting Partnership (SNPP) depicting the bloom area on May 21, 2019 (marked by a rectangle, source: https://www.star.nesdis.noaa.gov/sod/mecb/color/ocview/ocview.html). (Scale bar, 50 km.) Relative intensity of (B) hGSL (all E. huxleyi cells), (C) sGSL (susceptible E. huxleyi cells) and (D) 374-GSL (susceptible, 374-like E. huxleyi cells, 13–14) species during five days of sampling. (E) Relative intensity of resGSL d19:4/h22:2 (resistant E. huxleyi cells, 12) analyzed using high-sensitivity multiple reaction monitoring (MRM) mode during five days of sampling.

Targeted GSL Analysis during E. huxleyi Bloom Succession and Virus-Induced Demise.

Sampling open-ocean bloom provides only a snapshot of the bloom dynamics. Therefore, we sought to gain a detailed temporal resolution for our suite of biomarkers to assess the various phenotypes that occur during E. huxleyi bloom succession. We conducted an in situ mesocosm experiment in the coastal waters of southern Norway (45, 46), where annual blooms and viral infection of E. huxleyi occur naturally (20). Briefly, the experiment included seven mesocosm bags that were filled with natural marine microbial communities and monitored daily over 24 d. Four bags (bags 1to 4) were sampled for lipidomics analysis and are discussed hereinafter. All bags were supplemented with nutrients at a nitrogen to phosphorous ratio of 16:1 to favor the growth and induce a bloom of E. huxleyi (47). E. huxleyi blooms were observed starting from day 10 in all bags, reaching a concentration of up to 8 × 10⁷ cells per L at day 17, followed by bloom demise starting from day 18 (SI Appendix, Fig. S24A). Viral infection varied between the bags, as was visible by measurement of biomass-associated EhV by quantitative PCR (qPCR) using the major capsid protein (mcp) gene. Bag 4 showed the strongest increase in EhV starting from day 17, followed by bag 2 and bag 1. No viral proliferation was observed in bag 3 (SI Appendix, Fig. S24B). Concomitantly, the extent of bloom demise also varied between the bags, reaching the lowest cell abundance in bag 4 (SI Appendix, Fig. S24A).

We followed changes in the GSL composition of the particulate fraction (1.6 to 25 µm) during the bloom and demise of E. huxleyi (days 10 to 23) by LC-HRMS. hGSL species, present in all E. huxleyi strains (30, 32), correlated with E. huxleyi abundance (SI Appendix, Fig. S24C, Pearson correlation, r = 0.66 to 0.73, SI Appendix, Table S7). sGSL species, which characterize susceptible strains (31), also correlated with E. huxleyi abundance, primarily in the bloom phase (SI Appendix, Fig. S24D, r = 0.58 to 0.72, SI Appendix, Table S7). vGSL species, which indicate active viral infection of E. huxleyi (29, 30), positively correlated with the degree of infection in the bags (SI Appendix, Fig. S24E, r = 0.87 to 0.95, SI Appendix, Table S7).

Interestingly, we could not detect resGSL species (group C, 11–12), which are characteristic of resistant cells, suggesting that the abundance of resistant E. huxleyi cells throughout the bloom and demise phases (and within our sampling period) was below level of detection. GSL species of group A (1–4), which are found in higher intensity in resistant strains compared to susceptible strains (SI Appendix, Fig. S18), appeared from the beginning of the bloom and were highly correlated to hGSL species and to a lesser extent to E. huxleyi abundance and sGSL species (SI Appendix, Fig. S24 F, r = 0.76 to 0.96, 0.53 to 0.75, and 0.70 to 0.81, respectively, SI Appendix, Table S7). Accordingly, the detection of these GSL species is most probably derived from susceptible cells that dominated the bloom rather than rare resistant cells. 374-GSL species (group D, 13–14) appeared in a similar pattern to group A (SI Appendix, Fig. S24G and Table S7), indicating a high abundance of susceptible cells that share some similarity to E. huxleyi strain 374. GSL species of group B (5–6, 8–10), which are found in resistant strains and infected susceptible strains, appeared mostly from day 17 onward and were highly correlated to the abundance of EhV and the main vGSL species (vGSL t17:0/h22:0, SI Appendix, Fig. S24H, r = 0.69 to 0.99, SI Appendix, Table S7), as was also observed in the laboratory (SI Appendix, Fig. S17). Nevertheless, the infection-related occurrence of these GSL species might suggest an infection-derived initiation of cellular processes that eventually lead to resistance.

Remodeling of GSL Composition and Induction of Resistance in Infected Cultures.

To assess whether resistant E. huxleyi cells appear in low numbers during bloom succession, as detected in the open-ocean bloom (Fig. 3E), we isolated numerous E. huxleyi clones during the mesocosm experiment and determined their susceptibility to infection by EhV strain M1 (EhVM1), which was isolated during the same mesocosm experiment (48). Most E. huxleyi isolates were found to be susceptible to EhVM1, among them isolates RCC6918, RCC6936, and RCC6912 (Fig. 4 A and B and SI Appendix, Fig. S25 A and B), which were isolated during the bloom phase of E. huxleyi (46). We also isolated a few resistant E. huxleyi clones, among them E. huxleyi isolate RCC6961 (SI Appendix, Fig. S25C), which was collected during the virus-induced demise phase of the E. huxleyi bloom (46). Interestingly, some of the isolated susceptible strains showed rapid recovery 1 to 2 wk after viral infection (RCC6918 and RCC6912, Fig. 4B and SI Appendix, Fig. S25B, respectively). The recovered populations (named RCC6918-Rec and RCC6912-Rec, respectively) were resistant to the virus, as was validated by re-exposing the cultures to viral infection (SI Appendix, Fig. S26A). To examine whether the identified GSL markers for resistant cells can differentiate between the different phenotypes, we compared the GSL composition of the isolates and the recovered cultures (Fig. 4 C and D and SI Appendix, Fig. S26B). All isolates had similar amounts of hGSL species. The susceptible mesocosm isolates RCC6936, RCC6918, and RCC6912 had a similar GSL composition to E. huxleyi 374, having high intensity of sGSL and 374-GSL species and a lower intensity of group A GSL species (Fig. 4 C and D and SI Appendix, Fig. S26B). GSL species from groups B and C were not detected in these susceptible isolates. The resistant isolate RCC6961 had a similar GSL composition to the resistant strains 373 and 379: higher intensity of group A GSL species (compared to the susceptible isolates) and presence of GSL species from groups B and resGSL species (Fig. 4C). The distinct occurrence of resGSL species in this resistant isolate further supports its use as a biomarker for resistant cells. Remarkably, the cultures that recovered following infection of isolates RCC6918 and RCC6912 and acquired resistance to the virus had a similar GSL composition to the resistant isolate RCC6961 and the resistant strains 373 and 379, including the presence of resGSLs (Fig. 4D and SI Appendix, Fig. S26B).

Fig. 4.

Plasticity in the GSL composition of E. huxleyi cultures that recover following viral infection. (A) Cell abundance in cultures of the susceptible mesocosm isolate RCC6918. (B) Cell abundance (black) and production of virions (gray) following addition of EhVM1 to E. huxleyi isolate RCC6918. A recovered resistant population emerged a week after infection. Values of A and B are presented as mean ± SD (n = 2). The black arrow indicates the addition of EhVM1 to the cultures. (C) GSL composition of the susceptible mesocosm isolate RCC6936 and the resistant mesocosm isolate RCC6961 (n = 3). (D) GSL composition of the susceptible mesocosm isolate RCC6918 and of the culture that recovered following infection and was resistant to the virus (RCC6918-Rec, n = 3). Values for each lipid species (row) in (C and D) are shown after normalization. GSL species are grouped and numbered based on Table 1.

We analyzed the GSL composition of two additional recovered resistant cultures and eight additional mesocosm isolates, of which five are susceptible and three are resistant (SI Appendix, Fig. S19). As before, resGSL d19:4/h22:2 (12) was detected in the recovered cultures and resistant isolates but not in the susceptible isolates (SI Appendix, Fig. S19A). Additionally, sGSL d18:2/c22:0 was detected in all the susceptible isolates and in most of the resistant isolates and recovered cultures, as was observed in the recovered E. huxleyi strain CCMP2090-Rec-C2 (SI Appendix, Fig. S19B). These results indicate a metabolic plasticity in GSL metabolism, which corresponds to the change in phenotype from susceptibility to resistance toward viral infection. Furthermore, the detection of resGSL species (group C) in resistant isolates from the mesocosm and recovered resistant cultures suggests that these GSL species might have been produced during the mesocosm experiment by these rare populations, albeit in concentrations below our detection limit.

Discussion

Resistance to viral infection has been described in various phytoplankton cultures under laboratory conditions (12, 13, 49, 50). Nevertheless, the extent of resistance in natural algal populations is unknown as we lack the tools to detect resistant cells, hindering our ability to understand the metabolic basis of resistance to viral infection and its ecological significance.

Susceptibility to viral infection in the E. huxleyi-EhV system is a complex, nonbinary trait that depends on both the alga and the virus. Some E. huxleyi strains are resistant to all known EhV strains, while others resist only some. On the other hand, some EhV strains are generalists, infecting all known E. huxleyi strains, while others are specialists, infecting only specific strains (35, 51). The difference in susceptibility of E. huxleyi strains to viral infection has been previously associated to ploidy level during life cycle changes, as well as to genome and transcriptome variations between the strains (13, 33, 34, 36, 52). Resistant cells were also identified as a small sub-population in infected cultures (36). Yet, to date there exists no specific metabolic biomarker for algal resistance to viral infection, and the mechanisms underlying resistance are largely unknown.

Proposed Functional Role of Resistance-Specific LCBs.

resGSL species found in resistant cells are characterized by an uncommon tetraunsaturated LCB 19:4, which has been previously identified only in a few dinoflagellates and other haptophytes (e.g. GSL d19:4/h22:1, which was detected in Isochrysis galbana) (53). This LCB has an additional double bond compared to the LCB d19:3, which is found in GSL species in E. huxleyi (hGSL, group A and 374-GSL species, SI Appendix, Table S3), other haptophytes and dinoflagellates, and in SLs of fungi and marine invertebrates (32, 53–56). Interestingly, resistant E. huxleyi strains are also characterized by GSL species containing the trihydroxylated LCB t18:0 (Fig. 2B and Table 1 and SI Appendix, Fig. S2), which is highly abundant in plants and fungi (57) and was thus far found only in vGSL species produced by infected cells (in addition to t16:0- and t17:0-based vGSL species, SI Appendix, Table S3) (29, 31).

Both LCB unsaturation and hydroxylation were found to affect the biophysical properties of membranes: LCB unsaturation hinders the ability of SLs to form ordered domains within lipid bilayers, known as “lipid rafts” (58), while additional hydroxyl groups facilitate the formation of more hydrogen bonds, leading to an increased stability and decreased permeability of the membrane and to lateral diffusion of membrane proteins (57). Such changes in SL composition can also initiate signal transduction within the cells, as was found in plants, yeast, and mammals (57, 59). Subsequently, they allow organisms to cope with environmental stress, such as low temperature (60), and can alter the susceptibility of cells to viral infection (61–63). Specifically, GSL-rich lipid rafts in host cells can serve as cellular entry or egress points in diverse systems (64), suggesting that membrane lipids are under strong selection pressure during host–virus coevolution, possibly driving the plasticity in lipid composition.

In the E. huxleyi-EhV system, the lipid envelope of EhV and the plasma membrane of E. huxleyi seem to play an important role at the onset of the infection process, mediating the entry of the virus to E. huxleyi cells by endocytosis or membrane fusion mechanism (65). It was previously suggested that sGSLs, which characterize susceptible cells, mediate viral adsorption to host cells (31). The occurrence of resGSLs and t18:0-based GSLs in resistant cells might therefore hinder viral adsorption to the cells by impeding membrane fusion. Nevertheless, further structural and biochemical analyses are needed to determine the role of resGSLs and t18:0-based GSLs in modulating E. huxleyi resistance to viral infection.

Plasticity in GSL Composition during E. huxleyi-EhV Interactions.

The lipidome of E. huxleyi has been identified as a sensitive metabolic indicator for environmental stress conditions, such as nutrient limitation and viral infection, reflecting the physiological state of the cells (25, 37). In particular, GSLs were found to play a distinct role in the E. huxleyi-EhV system due to their involvement in cell signaling during infection and in viral assembly and egress (29–32). The identification of resGSLs and other GSL species characteristic of resistant E. huxleyi cells (Fig. 2B and Table 1 and SI Appendix, Fig. S2) broadens our view of the GSL diversity in the E. huxleyi-EhV system and adds valuable biomarkers that were thus far missing (Fig. 5A and SI Appendix, Table S3). While hGSL and group A GSL species are shared among all E. huxleyi cell types (that is, the different strains and phenotypes), most GSL species are produced only by some: sGSL and 374-GSL species mostly by susceptible cells; vGSL species by infected cells; group B GSL species by both infected and resistant cells; and resGSL species by resistant cells (SI Appendix, Table S3). A recent study further found that resistant E. huxleyi strains have a more diverse GSL composition than susceptible ones under nutrient-replete conditions (66).

Fig. 5.

The GSL-based arms race between E. huxleyi and EhV. (A) The GSL composition of susceptible, infected, and resistant E. huxleyi cells. Each GSL group is marked with a different color and consists of different LCBs. Infection by EhV leads to the production of vGSL and group B GSL species, while recovered cells and resistant strains present a unique GSL composition, consisting of group B and resGSL species. Scheme created with BioRender.com. (B) LCB composition of GSLs in the E. huxleyi-EhV system. Presented are the structure of the different LCBs (Left) and the LCB profile of susceptible (S), infected (I) and resistant (R) cells (Right). Infected cells produce trihydroxylated LCBs (found in vGSL and group B GSL species), while resistant cells produce both trihydroxylated and tetra-unsaturated LCBs (found in group B and resGSL species, respectively). Colors mark the GSL group in which the LCB is found, as in A. The position of the double bonds and functional groups were assigned based on the most common structure in the LMSD (39). (C) Expression pattern of sld and sbh genes which are differential between susceptible (S), infected (I) and resistant (R) E. huxleyi strains. sld and sbh genes are involved in LCB modification as part of the GSL biosynthetic pathway. EhV sld and EhV sbh are encoded by the EhV genome.

GSL species vary in their sugar headgroup, FA and LCB (67). In E. huxleyi, except for sGSL species that contain a sialic acid headgroup (31), all other known GSL species contain a hexose-based sugar headgroup (SI Appendix, Table S3). Additionally, most species have a highly similar FA composition (with the hydroxylated h22:0, h22:1, and h22:2 FAs being the most common), except for vGSL species that contain also longer FAs of 23-24 carbons, group A GSL species that contain FAs with 21 and 23 carbons (Table 1), and sGSL species that contain nonhydroxylated FAs (c22:0, c22:1, see SI Appendix, Table S3) (31). LCB composition, on the other hand, seems to be the main factor that differentiates between the various GSL groups and, consequently, between the cell types, thus driving the phenotypic plasticity in the E. huxleyi-EhV system (Fig. 5 A and B). Some LCBs are shared among several GSL species and cell types (LCB d18:1, d18:3 and d19:3), while others appear only in specific GSL species and cell types (LCB d18:2 in sGSLs of susceptible cells, LCB d19:4 in resGSLs of resistant cells, LCB d18:0 and t18:0 in group B GSLs of resistant and infected cells, and LCB t16:0 and t17:0 in vGSLs of infected cells), leading to a unique LCB profile for each cell type (Fig. 5B). Biosynthetic genes at various steps of the GSL pathway determine LCB composition, from the formation of the LCB to its hydroxylation and unsaturation (68). The presence of these genes and their differential expression under various biotic and abiotic conditions determine the GSL composition of the cells (69). In infected E. huxleyi cells, virus-derived SL biosynthetic enzymes lead to the production of vGSLs with a trihydroxylated LCB (predominantly t17:0) (29). In resistant E. huxleyi strains, our results suggest that the differential expression of specific sld and sbh genes (Fig. 5C and SI Appendix, Fig. S20 C and D) accounts for the biosynthesis of d19:4-based resGSL and t18:0-based GSL species. Remarkably, resistant E. huxleyi cells that emerge from infected susceptible cultures as early as 1 wk post infection (Fig. 4B and SI Appendix, Fig. S25B) produce resGSL and group B GSL species that are characteristic of resistant strains, consisting of LCBs that are not found in the parent susceptible strains (Fig. 4D and SI Appendix, Fig. S26B). This striking difference between the parent cells and the derived resistant cultures delineates the plasticity of the E. huxleyi lipidome. Such a rapid modulation of GSL composition following viral infection is therefore not restricted only to infected cells but might occur also in cells that evade infection or survive and become resistant to the virus. If so, viral infection might directly induce changes in host LCB biosynthesis and lead to the formation of GSL species that facilitate resistance. Alternatively, resistant cells may already exist as a rare sub-population in cultures of susceptible strains. Such cultures can recover from infection following the death of susceptible cells due to viral infection, which allows the resistant cells to proliferate. The phylogenetic similarity between the enzymes expressed by resistant strains (SLD1, SBH4 and SBH5) and their viral analogues (EhV201 SLD and EhV201 SBH, Fig. 5C and SI Appendix, Fig. S20 A and C) may further indicate competing biosynthetic pathways that are coexpressed during infection and affect its outcome, shedding light on the ongoing coevolution between E. huxleyi and its virus.

Detecting Resistant E. huxleyi Cells in Natural Populations.

Resistance to viral infection has been long studied under laboratory settings, describing a wide array of E. huxleyi strains that vary in their susceptibility to viral infection, and of EhV strains that vary in their level of infectivity (34, 36, 51). E. huxleyi strains can recover from lytic infection and gain resistance to the virus, highlighting their phenotypic plasticity and the rapid change in the dominating phenotypic state in the host cell population (36, 70). Nevertheless, although we are able to detect susceptible and infected E. huxleyi cells in natural samples using GSL biomarkers (SI Appendix, Fig. S24 D and E) (31, 32), it remains challenging to detect resistant cells in nature and to monitor their dynamics in natural heterogeneous populations (71).

In this study, we were able to detect a resGSL species during the demise of an open-ocean E. huxleyi bloom (Fig. 3E), indicating the occurrence of virus-resistant E. huxleyi cells in natural populations. The absence of resGSLs in samples from the mesocosm experiment stresses the scarcity of resistant E. huxleyi cells during the bloom phase and the early phase of the virus-induced bloom demise. This is further supported by the detection of resGSLs in resistant E. huxleyi isolates that originate from the mesocosm experiment. Additionally, the emergence of resistant cells containing resGSLs in recovered cultures (Fig. 4B and SI Appendix, Fig. S19A) suggests that these cells can be detected during late and postbloom phases in nature, as observed in samples from the oceanic boom (Fig. 3E). Thus, the sampling time of the mesocosm experiment might not have been long enough to see such an emergence of resistant sub-populations.

Importantly, the variation in E. huxleyi strain susceptibility to viral infection and the occurrence of generalist and specialist EhV strains (35, 51) could indicate multiple modes of defense employed by E. huxleyi to resist viral infection, as was reported in diverse bacteria–phage interactions (50, 72). The presence of resGSL, which was detected in all resistant strains analyzed in this study (SI Appendix, Fig. S19A), could represent one mode of defense. The variation in cellular quotas of resGSL and sGSL between the strains (SI Appendix, Fig. S19) and the levels of additional GSL species might further determine the extent of susceptibility of each E. huxleyi strain to different EhV strains. Having a broader view of the GSL diversity in the E. huxleyi-EhV system and the variability in GSL composition between different strains and phenotypic states might enable us in the future to assess population composition in natural blooms. Coupling the GSL biomarkers with advanced methods, such as single-cell lipid profiling and single-cell RNA sequencing (10, 73, 74), could further allow us to untangle the metabolic and phenotypic outcome of viral infection in complex microbial consortia in the marine environment. Studying and identifying the various cell types that constitute algal blooms and the metabolites they use to communicate will provide valuable insights into the host–virus arms race during bloom succession.

Materials and Methods

Strains of E. huxleyi and EhV Used in This Study.

Four E. huxleyi strains were used for untargeted lipidomics profiling: CCMP2090, CCMP373, CCMP374, and CCMP379, all noncalcifying. E. huxleyi 2090 and 374 are susceptible to viral infection, e.g., by EhV201, while E. huxleyi 373 and 379 are resistant. Transcriptomics data of all four strains are publicly available (33, 42, 75). The E. huxleyi cultures were supplemented with the lytic virus EhV201 (34), whose genome data are publicly available (76). Eighteen additional strains were used for targeted GSL analysis and for absolute quantification of resGSL d19:4/h22:2 (12) and sGSL d18:2/c22:0. Information regarding all E. huxleyi strains used in this study is presented in SI Appendix, Table S8.

To isolate E. huxleyi strains from the mesocosm experiment, water samples were collected during the bloom phase and during the virus-induced demise of E. huxleyi (46), and single E. huxleyi cells were sorted within 2 wk of collection at the Roscoff Culture Collection (RCC) laboratories. The susceptibility of these strains to viral infection was determined using the lytic virus EhVM1, which was isolated during the same mesocosm experiment and its genome data are publicly available (48).

Untargeted Lipid Profiling of E. huxleyi Cultures Using UPLC-HRMS.

Cultures of E. huxleyi strains 2090, 373, 374, and 379 with and without addition of EhV201 were analyzed for cellular lipid composition at days 0, 1, 2, and 3 of the experiment in three biological replicates. At day 0, samples were collected 4 h after the addition of EhV201. The samples (30 to 150 mL of each culture, equivalent to ~5 × 10⁷ cells per sample) were collected by vacuum filtration onto glass microfiber filters (grade GF/C, 47 mm in diameter, precombusted at 460 °C for >8 h, GE Healthcare Whatman, Buckinghamshire, UK), immediately plunged into liquid nitrogen, and stored at −80 °C until extraction (25). In total, 96 biological samples were collected.

Lipid extraction was performed as previously described (77) with slight modifications. Briefly, biological triplicates were divided into three batches, with 32 samples in each batch. Filters were placed in 15-mL glass tubes and extracted with 3 mL of a precooled (−20 °C) methanol: methyl tert-butyl ether (MTBE) (1:3, ν:ν) solution containing sphingolipid standard mixture (~150 nM of each species). The samples were shaken for 30 min at 4 °C and sonicated for 30 min. The samples were then supplemented with 1.5 mL water:methanol (3:1, ν:ν) solution, vortexed for 1 min, and centrifuged for 10 min at 3,200 × g and 4 °C. The upper organic phase (1.5 mL) was transferred to 2-mL centrifuge tubes and dried under a flow of nitrogen (TurboVap LV, Biotage, Uppsala, Sweden). The polar phase was reextracted with 1.5 mL of MTBE. The upper organic phase (2.25 mL) was combined with the organic phase from the first extraction and dried under a flow of nitrogen. The samples were stored at −80 °C until UPLC-HRMS analysis. Two extraction blanks were collected following the same procedure using blank filters.

Per batch, samples were thawed, redissolved in 300 μL acetonitrile:isopropanol (7:3 ν:ν) with 1% 1 M ammonium acetate and 0.1% acetic acid, vortexed, sonicated for 10 min and centrifuged at 20,800 × g for 10 min at 10 °C. The supernatants were transferred to 200-μL glass inserts in autosampler vials and directly used for LC-MS analysis. A pooled quality control (QC) sample was generated by combining aliquots of 10 μL from all biological samples and was injected at the beginning, middle, and end of each batch. An aliquot of 1 µL was analyzed using UPLC coupled to a photodiode detector (ACQUITY UPLC I-Class, Waters, Milford, MA) and a quadrupole time-of-flight (QToF) mass spectrometer (SYNAPT G2 HDMS, Waters), as described previously (77) with slight modifications (SI Appendix).

An additional analysis was performed to quantify several GSL species with higher sensitivity (SI Appendix, Fig. S18). The samples (250 mL of each culture at the exponential phase, 1 to 1.5 × 10⁶ cells per mL, equivalent to ~4 × 10⁸ cells per sample) were extracted as described above.

Comparative Analysis of Untargeted Lipid Profiling Data.

Raw LC-MS files were converted from the vendor’s format to the open-format “netCDF” using a “DataBridge” (MassLynx version 4.1). Preprocessing of the CDF files was done using the R (78) packages “xcms” (79) and “CAMERA” (80) obtained from the Bioconductor repository. This yielded a matrix of 12,190 aligned mass features across samples with corresponding peak intensity values. Parameters for mass feature detection, smoothing, alignment, binning, and filtering are specified in SI Appendix, Tables S9 and S10. The feature matrix was normalized to the total ion current (TIC, per sample) and standardized. The elbow method was applied to determine the number of clusters for a subset of samples (without addition of EhV, 48 samples, k = 4, SI Appendix, Fig. S1A) and for the whole dataset (with and without addition of EhV, 96 samples, k = 4, SI Appendix, Fig. S1B), followed by k-Means clustering and PCA analysis using the R package “factoextra” (81) for both the subset and the whole dataset. k-Means clustering (k = 5) and PCA were performed also for the whole dataset with the pooled QC samples, resulting in the same separation to clusters, while the pooled QC samples were grouped together in a fifth cluster (SI Appendix, Fig. S27).

Comparative analysis between clusters 1 to 4 (each containing a distinct E. huxleyi strain) in the subset without addition of EhV (Fig. 1C) was performed by one-way repeated measures ANOVA with “time” as a within-subject factor and “strain” as between-subject factor, P < 0.01, following log-transformation using the R package “rstatix” (82), reducing the data to 10,848 mass features. The mean intensity of each mass feature was then calculated for all clusters, followed by calculation of the fold change between the cluster with the maximum mean intensity and the other clusters. A fold change of ≥2 (in log scale) between the first and third highest clusters was selected, yielding a list of 231 differential mass features, which underwent further manual annotation to obtain a smaller number of feature groups. The peak shape of the extracted ion chromatograms (EICs) from coeluting mass features was compared using MassLynx (Version 4.1, Waters), and isotopes, adducts, and apparent neutral losses (e.g., of water) were annotated, grouping the mass features into 54 feature groups (SI Appendix, Table S1). Out of the 54 feature groups, four were putatively identified as GSL species following manual annotation, based on the accurate mass, adducts, and apparent in-source fragments (2, 9, 12, and 14, see SI Appendix, Figs. S2–S6 and Table S1). Ten additional GSL species, which were filtered out in the initial analysis due to a fold change <2 (in log scale, ranging from 1.2 to 2), were manually identified (1, 3–8, 10–11, 13, see SI Appendix, Figs. S2 and S7–S16). Next, the adduct ion with the highest intensity was selected for each feature group, and the corresponding peak area was extracted using QuanLynx (Version 4.1, Waters) across samples in the whole dataset. Peak areas above a signal-to-noise threshold of 10 (limit of quantification) were normalized to the TIC. Per feature, zero values were replaced with half of the minimal value. Hierarchical cluster analysis was then applied to the whole dataset (SI Appendix, Fig. S28) and to the dataset after averaging the peak areas of the biological replicates (Fig. 2A) using the extended number of putative lipid species (64 feature groups) following log-transformation using Matlab R2021a, with row-wise (per feature) scaling, row- and column-wise clustering using the default “Eucledian” method and the “redblue” color panel.

Putative Annotation and Phenotypic Grouping of GSL Species.

The annotation of GSL species that were previously undescribed in the E. huxleyi-EhV system (listed in Table 1) was based on LC-MS/MS analysis and the Lipid Maps computationally generated database of lipid classes and the Lipid Maps Structure Database (LMSD) (39), and carried out according to the Metabolomics Standards Initiative, “Level 2—putatively annotated compounds” (38). The annotation of previously described GSL species was performed according to the accurate mass and LC-MS/MS fragmentation pattern (29, 31, 37, 45). LC-MS/MS analyses were performed in positive ionization mode for the protonated molecules using a collision energy ramp of 10 to 45 eV and a scan time of 0.5 s. Analyses were performed on samples with high intensities, with injection volumes of 3 to 5 µL. The data were analyzed and processed with MassLynx (version 4.1, Waters). For MS/MS spectra and a list of fragments of the GSL species that were previously undescribed in the E. huxleyi-EhV system, see SI Appendix, Figs. S3–S16.

Quantification and phenotypic grouping of the GSL species (Table 1) was based on their abundance profiles in the different E. huxleyi strains in the presence and absence of EhV (SI Appendix, Figs. S17 and S18). The abundance profiles were generated by normalizing the peak area of each GSL species (extracted as described above) to the extraction standard (glucosylceramide d18:1/c12:0) and to the total number of extracted cells. Differences in GSL abundance were tested for day 2 of the experiment by a one-way ANOVA followed by Tukey’s post hoc test, P < 0.01 (SI Appendix, Tables S11 and S12). Day 2 was chosen since it was the first time point in which infected samples appeared as a separate cluster in the k-means clustering analysis (Fig. 1D).

Sampling of Natural E. huxleyi Blooms for Targeted Analysis of GSLs.

Water samples of an open-ocean E. huxleyi bloom were collected during the “Tara Breizh Bloom” cruise in the Celtic Sea from May 29 to June 2, 2019 (44). Water samples of 50 L were first filtered through a 20-μm nylon net to remove large particles. Cells were then collected by vacuum filtration onto glass microfiber filters (grade GF/C, 125 mm in diameter, precombusted at 460 °C for > 5 h, GE Healthcare Whatman). The filters were transferred to 50-mL centrifuge tubes and immediately plunged into liquid nitrogen. The filters were kept at −80 °C, freeze-dried (Gamma 2–16 LSCplus, Martin Christ, Osterode am Harz, Germany) within 6 mo after collection, and stored at −80 °C until further processing. Lipid extraction was performed as described above, using different solution volumes: 20 mL of the precooled (−20 °C) methanol:MTBE (1:3, ν:ν) solution containing sphingolipid standard mixture (~150 nM of each species), 15 mL of water:methanol (3:1, ν:ν) solution, and 11 mL of MTBE for reextraction. The upper organic phase (11 mL for the first extraction, 15 mL for the second extraction) was dried under a flow of nitrogen. An extraction blank was collected following the same procedure using a blank filter.

Water samples of an induced E. huxleyi bloom were collected during a mesocosm experiment (AQUACOSM VIMS-Ehux) that was carried out over 24 d (May 24 to June 17, 2018) in Raunefjorden at the University of Bergen’s Marine Biological Station Espegrend, Norway (60.38° N; 5.28° E), as previously described (45, 46). Water samples for cellular lipidomics analysis were collected daily from bags 1 to 4, as described previously (45). Lipid extraction was performed for bag samples in days 10 to 23 (bloom and demise of E. huxleyi, 56 biological samples in total) and for the extraction blank as described for the laboratory samples, without the addition of a sphingolipid internal standard mixture and including two solvent blanks.

Targeted analysis of GSL species was performed using UPLC-HRMS (SI Appendix).

Targeted GSL Analysis of Mesocosm-Derived E. huxleyi Isolates and Recovered Cultures.

E. huxleyi cultures of the mesocosm isolates (RCC6912, RCC6918, RCC6936 RCC6961) and recovered strains (RCC6912-Rec and RCC6918-Rec) were analyzed for cellular lipid content at the exponential phase in three biological replicates (1 × 10⁶ to 2.5 × 10⁶ cells per mL, see SI Appendix, Fig. S26C). To generate recovered resistant strains, EhVM1 was added to cultures of E. huxleyi isolates RCC6912 and RCC6918, RCC6936 at a ratio of 1:1 viral particles to E. huxleyi cells. Of the three susceptible isolates, isolates RCC6912 and RCC6918 recovered about a week following infection. The recovered resistant cultures were continuously refreshed until no EhV was detected using flow cytometry, yielding strains RCC6912-Rec and RCC6918-Rec, respectively (Fig. 4B and SI Appendix, Fig. S25B). Samples (100 to 150 mL of each culture, equivalent to ~2 × 10⁸ cells per sample) were collected by gentle vacuum filtration onto glass microfiber filters (grade GF/A, 47 mm in diameter, precombusted at 460 °C for > 5 h, GE Healthcare Whatman), immediately plunged into liquid nitrogen, and stored at −80 °C until extraction. In total, 18 biological samples were collected. Lipid extraction was performed as described for the laboratory strains, including three extractions blanks. Targeted analysis of GSL species was performed using UPLC-HRMS (SI Appendix).

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Data supporting the findings of this study are available in the paper and its Supplementary Information. Flow cytometry, qPCR, nutrient, and temperature data from the mesocosm experiment are available in Dryad (https://doi.org/10.5061/dryad.q573n5tfr) (83). Mass spectral raw data (full MS and MS/MS) were deposited in the EMBL-EBI MetaboLights repository with the identifier MTBLS3323 (www.ebi.ac.uk/metabolights/MTBLS3323) (84). Nucleotide sequences were deposited in GenBank and given accession numbers: MZ152812 to MZ152827 (https://www.ncbi.nlm.nih.gov/nuccore?term=MZ152812:MZ152827[accn]). Full sequences, domain sequences, and alignments used for the phylogenetic analysis are available on Figshare: https://doi.org/10.6084/m9.figshare.20448579 (85).