Environment-specific virocell metabolic reprogramming

Abstract Viruses impact microbial systems through killing hosts, horizontal gene transfer, and altering cellular metabolism, consequently impacting nutrient cycles. A virus-infected cell, a “virocell,” is distinct from its uninfected sister cell as the virus commandeers cellular machinery to produce viruses rather than replicate cells. Problematically, virocell responses to the nutrient-limited conditions that abound in nature are poorly understood. Here we used a systems biology approach to investigate virocell metabolic reprogramming under nutrient limitation. Using transcriptomics, proteomics, lipidomics, and endo- and exo-metabolomics, we assessed how low phosphate (low-P) conditions impacted virocells of a marine Pseudoalteromonas host when independently infected by two unrelated phages (HP1 and HS2). With the combined stresses of infection and nutrient limitation, a set of nested responses were observed. First, low-P imposed common cellular responses on all cells (virocells and uninfected cells), including activating the canonical P-stress response, and decreasing transcription, translation, and extracellular organic matter consumption. Second, low-P imposed infection-specific responses (for both virocells), including enhancing nitrogen assimilation and fatty acid degradation, and decreasing extracellular lipid relative abundance. Third, low-P suggested virocell-specific strategies. Specifically, HS2-virocells regulated gene expression by increasing transcription and ribosomal protein production, whereas HP1-virocells accumulated host proteins, decreased extracellular peptide relative abundance, and invested in broader energy and resource acquisition. These results suggest that although environmental conditions shape metabolism in common ways regardless of infection, virocell-specific strategies exist to support viral replication during nutrient limitation, and a framework now exists for identifying metabolic strategies of nutrient-limited virocells in nature.


Introduction
Microbes are central to ecosystem health and function, and microbial diversity and abundance is shaped by bottom-up (nutrients) and top-down (predation) forces.Among predators, viruses inf luence host mortality, metabolism, and horizontal gene transfer [1][2][3].During infection, viruses transform their microbial hosts into new entities termed "virocells" [4] that are metabolically and ecologically distinct from uninfected cells [5][6][7][8].Specifically, in virocells, host machinery is redirected toward obtaining energy and intra-and extra-cellular resources for viral infection [6,9] in ways that can alter cells and the environment.However, most virocell studies are done in nutrientreplete conditions that do not represent the resource limitation that virocells likely encounter in nature, e.g. the generally low and f luctuating oceanic nutrient levels [10][11][12][13].This leaves the interplay between bottom-up and top-down processes largely unstudied.
Among nutrients, phosphate (P) is essential for diverse biomolecules (e.g.DNA, RNA, modified-proteins, ATP) and is disproportionately higher in viruses than hosts [14].In the oceans, P is often co-limiting with nitrogen [11,13,15,16] and shapes both viral gene content [17,18] and infection mechanisms [9,[19][20][21].During P-starved Prochlorococcus infections by a phage encoding P-acquisition auxiliary metabolic genes, host and phage P-acquisition marker genes were expressed [22], orchestrated by the host's two-component regulatory system, phoR/phoB [17].In the unicellular eukaryotic algae Micromonas pusilla, transcriptional responses to low-P were stronger than to viral infection [23].Although these studies show that P-status strongly inf luences transcripts during infections, little is known about how P-status impacts other biomolecules or virocell metabolic reprogramming in other taxa, nor how P-stressed virocells interact with the environment.
Here we studied the effects of low-P on two Pseudoalteromonas virocells obtained from independent infections with dsDNA phages, siphovirus HS2 and podovirus HP1.We follow the host and both phages via time-resolved multi-omics (transcriptomics, proteomics, lipidomics, and endo-and exo-metabolomics) to compare low-P responses (current study) against our prior findings in high-P conditions [7] that are augmented here with lipidomics and metabolomics.Our prior work revealed that metabolic reprogramming was phage-specific in nutrient-rich conditions and likely driven by phage-host genomic complementarity [7].This current study asks: what are the intra-and extra-cellular impacts of low-P on cells and virocells across integrated omics datatypes?

Bacterial growth and phage infections
Growth and infections were conducted as described previously [7,24,25] with modifications for phosphate (P) limitation (low-P herein).Brief ly, Pseudoalteromonas sp.[13][14][15] were grown shaking at 150 rpm at 21 • C in either 1% Z + CNP medium for high-P or 1% Z + CN medium for low-P conditions.These media consisted of 1% Zobell (26 g sea salts, 1 g yeast extract, and 5 g proteose peptone per L), 8.3 mM ammonium sulfate, 0.15 mM phosphoric acid to high-P only, and 11 mM glucose added after autoclaving.The high-P medium had 55 μM of total organic phosphate (TOP) and 48 μM of inorganic phosphate (PO 4 ), whereas the low-P medium had 5 μM TOP and 0.6 μM PO 4 (Supplementary Table S1).
For growth curves, one colony was inoculated into 10 mL of either high-P or low-P media and grown for 12 h while measuring optical density at 600 nm (OD 600 ).This starter culture was transferred 1:20 in triplicate into 200 mL of the same medium in 1 L f lasks, from which colony forming units (CFUs) and OD 600 were sampled to obtain OD-CFU regression equations used to estimate cell density for phage infections.Growth rate (h −1 ), calculated during exponential growth as ln(OD 2 /OD 1 )/(time 2 -time 1 ), was obtained using R's "growthcurver" package [26].Growth was also assessed in this manner in the 1% Zobell medium only (without added C, N, or P).
For phage infections, 5 × 10 8 cells from the starter culture were transferred to 200 mL in 1 L f lasks and grown to mid-to lateexponential phase.Then, ∼1 × 10 8 cells were transferred in triplicate to a 1.5 mL tube and the volume was adjusted to 1 mL.Phages (either PSA-HP1 or PSA-HS2) were then added at a multiplicity of infection (MOI) of 0.1 for initial characterizations, or at MOI ∼3 for the time-resolved 'omics experiments.During infections, phages were allowed to adsorb for 15 min, and then they were diluted 100-fold in a 250 mL f lask, or for 10-fold for time-resolved 'omics sampling, in a 1 L bottle with the same type of medium used for infections, following previously described marine phagehost transcriptomics [7,27,28].Cells, free phages, and total phages were sampled periodically.Cells were spread on Zobell plates and incubated for 2 days at room temperature (RT).Free phages had cells removed with 0.2 μm syringe filters.Free and total phages were enumerated by top-agar plating technique (plaque assay; [29] previously used [7,24]).Brief ly, phages were diluted in 26 psu artificial seawater (26 g sea salts per L), mixed with 0.4 mL of bacterial overnight culture grown in Zobell and 3.5 mL molten soft agar (Zobell medium containing 0.6% low melting point agarose), and dispersed on 20% Zobell agar plates.Plates were incubated at RT and plaques were visible after 1-2 days.
A detailed account of all 'omics measurements and statistical analyses is provided in the Supplementary Materials and summarized below.

Transcriptomics
From diluted samples, 15-25 mL were collected in biological triplicates from 0, 30, 60, 80, and 100 min (low-P HP1-virocells and uninfected) or 0, 30, 60, 80, 100, 120, and 140 min (low-P HS2virocells and uninfected), and pelleted for 11 min at 20000 g.The supernatant was discarded before f lash-freezing in liquid N 2 .RNA was extracted using the Zymo Quick RNA Mini kit (R1054).RNA concentration and integrity were assessed using the Agilent 2100 Bioanalyzer RNA 6000 Pico assay with the prokaryote protocol.Glow cell sequencing was performed on the HiSeq2500 sequencer using HiSeq TruSeq SBS sequencing kits, v4, following a 2 × 100 indexed run (Illumina, San Diego, CA).Raw gene counts were generated with FeatureCounts whereby both reads from a pair-end run were aligned to the same feature in the reference genome.Read normalization and differential expression (DE) analyses were performed following previously published scripts and procedures [7,27,28].DE analyses were performed between host-infected and uninfected samples at every time point and under the same medium using the statistical package edgeR [30] whereby genes with a false discovery rate and P values < 0.05 were considered DE.Metabolic pathway reconstruction was performed with KEGG [31] and Ecocyc [32].

Proteomics
From diluted samples, 80 mL were collected in biological triplicates from 0, 20, 40, 60, 80, and 100 min (low-P samples) and pelleted for 8-11 min at 20000 g.The supernatant was discarded prior to f lash-freezing with liquid N 2 .After protein extraction (see supplementary methods), mass-spectrometry (MS) analysis was performed using a Q-Exactive Plus mass spectrometer (Thermo Scientific, San Jose, CA) outfitted with a home-made nanoelectrospray ionization interface.MS-GF+ version v2017.01.13 was used to identify peptides from the LC-MS/MS spectra.Search parameters included a +/−20 ppm parent mass tolerance, partial trypsin rules, and dynamic oxidized methionine residues.The candidate protein list was assembled by combining the Pseudoalteromonas sp.13-15 proteome with the 6-frame translation of both phages and a collection of 195 contaminant proteins.The spectra were filtered with an MSGF E-value score of <1 × 10 −9 and 2 or more unique peptide sequences were required to consider a protein identified.Passing spectra were counted and used as a relative abundance value for comparing across datasets.
Proteomics data are available at MassIVE and the ProteomeXchange repositories with accession numbers MSV000083626 and PXD013204, respectively.

Endometabolomics
From the diluted samples, 80-90 mL from biological triplicates from 0, 10, 20, 30, and 40 min (high-P samples) or 0, 20, 40, 60, 80, and 100 min (low-P samples) was deposited by vacuum filtration onto a 47 mm 0.4 μm polycarbonate filter and washed with an equal volume of PBS.Three media-only blanks were included.The extraction was the same as for the lipid samples described above.Analyses were done as reported previously [34].MS data files were processed using Metabolite Detector [35].Peaks were matched to PNNL augmented version of Agilent metabolomics database and additionally cross-checked with Wiley Registry 11th Edition and NIST17 GC-MS spectral databases.Identified metabolites were validated manually.Peak area values of detected metabolites were log-transformed for further analyses.

Exometabolomics
From the diluted samples, 90 mL collected at 0, 30, and 60 min in triplicates for all samples were spun down 5000 g for 5 min to pellet cells, the supernatant was filtered through a 0.2 μm filter, and f lash-frozen.Three respective media-only blanks were included.After removing salts (see supplementary methods), high resolution mass spectra of the filtrate and media-only blanks were collected by direct injection using a Bruker 9.4-Tesla Fourier transform ion cyclotron resonance mass spectrometer (University of Arizona).Putative chemical formulas were assigned using Formultitude (previously named Formularity) software [36].Gibbs free energy (GFE) was calculated as described previously [37], with peaks that were assigned a putative molecular formula in all samples.Organic matter (OM) transformation analysis for each individual sample replicate was done via network analysis using MetaNetter [38].

Linear mixed effects models
Separate linear mixed effect models were fitted for host proteins, host transcripts, phage proteins, phage transcripts, phage structural proteins, endometabolites, lipidomic data, and exometabolomic data.These models include a random intercept term to account for longitudinal correlation for genes.Gaussian family distribution is used for the transcript data (FPKM counts).Poisson family distribution is used for protein data (MS counts).Count was predicted by the main effects which are infection status, media type, and time, plus interaction terms between each pair of predictor variables.Type II Wald chi-square test ANOVA was performed to determine whether each term is statistically significant in predicting counts and the relative amount of variance explained by each term.

Overview of infection, transcription, and translation
Uninfected cells, HP1-virocells, and HS2-virocells were grown in low phosphate (low-P herein), profiled for multi-omics (transcriptomics, proteomics, endometabolomics, exometabolomics, and lipidomics), and compared with our previously published high-P data [7].The high-P medium had 11× more total organic P and 77× more inorganic P (P i ) than the low-P medium (Supplementary Table S1), and low-P significantly (P value < 0.05) impacted the composition of each omics data type (Supplementary Fig. S1; Supplementary Table S2).To evaluate the impact of low-P on virocell metabolic reprogramming and ecosystem footprint, virocell transcriptomes, proteomes, endo-metabolomes, endo-lipidomes, and exo-metabolomes were collected (Fig. 1A).To describe these biomolecular changes, we used the following language: (i) transcripts were over-expressed (OE), under-expressed (UE), or not differentially expressed (not DE) when their fold-change was >1, <1, or 1, respectively, in infected relative to the uninfected cells in the same medium and time point; (ii) proteins were enriched, depleted, or not different from the mean when their z-scores from comparing peptides levels across all proteins, treatments, replicates, and time points were >0, <0, or =0, respectively, and (iii) metabolites and lipids were enriched or depleted when their fold-change in infected relative to uninfected under the same condition and time point was >1 or <1, respectively.
Phage gene expression dynamics, previously measured in high-P conditions [7], were largely maintained under low-P conditions (Fig. 1D and E; Supplementary Fig. S3A and B), consistent with multiple studies showing phage transcriptional programs are relatively invariant [8,27,28,41,42].However, ∼60% of phage genes had lower expression levels in low-P than in high-P conditions (Supplementary Fig. S3C and D), suggesting low-P impacted phage transcript levels.Because phage transcript versus protein levels during nutrient limitation has not been explored, we assessed RNA-to-protein ratios (R/P), which in uninfected bacteria is both positively correlated with fitness (i.e. higher R/P leads to higher growth rates) and nutrient-sensitive [43,44].Phage R/Ps fell above (for HS2) and below (for HP1) the 1:1 line, for both non-structural (Spearman correlation, rho = 0.95 for HS2 and rho = 0.88 for HP1, P value < 0.001) and structural (Spearman correlation, rho = 0.64 for HS2 and rho = 0.89 for HP1, P value < 0.001) genes (Fig. 1F and G).This suggests opposing fitness strategies, whereby HS2 makes more RNA per protein than HP1 in low-P.

Low-P reduces transcription and translation in all cells, with virocell-specific differences
To further evaluate phage fitness strategies, and cell/virocell transcription and translation across environments, we conducted linear mixed effects models (LMEMs) on host and phage transcripts and proteins.Pseudoalteromonas (strain PSA 13-15) was independently infected with two different phages (PSA-HS2 and PSA-HP1) and sampled throughout the latent period ( boxed) for time-resolved multi-omics measurements in either low-P or high-P conditions and compared with uninfected control cells in the same condition.Transcriptomics, proteomics, and exo-metabolomics were all used as primary data, and endometabolomics and lipidomics as supporting data.The manuscript presents the story from three angles: (1) environment-specific effects (high-P versus low-P), (2) infection-specific effects (infected versus uninfected cells), and (3) phage-specific effects (HS2-virocells versus HP1-virocells) in the same environment.(B) Infection dynamics of phage PSA-HS2 in both media.(C) Infection dynamics of phage PSA-HP1 in both media.For both, (B) and (C), phage abundance, measured as plaque-forming units (pfus) per mL for free phages are represented over time, and 0 min represents 15 min after diluting the infection.The average of three biological replicates is plotted along with the error.All data points have been normalized against the first time point (0 min).Estimates of latent period are represented with rectangle boxes under each graph.(D) Gene expression dynamics of phage PSA-HS2 in both media.(E) Gene expression dynamics of phage PSA-HP1 in both media.For both, (D) and (E), relative transcript abundance for all genes is plotted as standardized log 2 FPKM (using all time points and media) over infection time, in high-P (dashed) and low-P (solid) media and colored according to earlier (blue) or later (black) temporal expression.(F, G) Scatter plot of the RNA-to-protein ratio (RNA/protein) for PSA-HS2 (F) and PSA-HP1 (G) phage genes in high-P (x-axis) and low-P (y-axis) conditions.Phage structural genes were determined based on annotation and are colored blue (HS2) or yellow (HP1); non-structural genes displayed in gray.The RNA/protein was calculated by dividing the average normalized transcript abundance (FPKM) by the average normalized protein abundance, both averaged across three replicates.(H) Bar plot representing the fraction of RPs that are enriched (z-score > 0), depleted (z-score < 0), or not detected, in every treatment and condition, including the uninfected control, HS2-virocells (HS2viro), and HP1-virocells (HP1viro), obtained from proteomics.Ribosome recycling proteins are not included.
Across environments there were significantly (P value < 0.0001) fewer host and phage transcripts and proteins in both virocells and uninfected cells in low-P than in high-P conditions (Supplementary Tables S3 and S4).This suggests environmentspecific effects where low-P reduced transcription and translation in cells and virocells.
We applied LMEMs to compare the virocells to each other in low-P (Supplementary Tables S3 and S4).For host-derived biomolecules, HS2-virocells had significantly (P value < 0.001) more transcripts, but fewer proteins than HP1-virocells.These host biomolecule differences could result from virocell differences in either degradation (i.e.host takeover, where one virocell degrades more transcripts/proteins than the other) or production (i.e.resource allocation, where one virocell has a transcript/protein surplus relative to the other).To help discern between the two, we examined ribosomal proteins (RPs).Uninfected cells and HS2-virocells had ≥79% of RPs enriched, whereas HP1-virocells only 17% (Fig. 1H, Supplementary Fig. S4).Additionally, the ribosome recycling factor protein, which reuses ribosomes for translating different mRNAs [45], was only enriched in HP1-virocells (Supplementary Fig. S4).Although there may also be differences in host takeover, these results suggest differences in resource allocation.HS2-virocells likely invest in new protein synthesis, which is possible due to the high codon and amino acid similarity between HS2 and this host [7].Contrastingly, HP1 is less well matched with this host's translational machinery [7] and, instead, HP1-virocells likely recycle ribosomes and proteins-which are energetically expensive [46]-as the strategy for responding to f luctuating environments, similar to nutrient-limited uninfected bacteria [47].As growth (fitness) is proportional to RPs in uninfected bacteria [48], we posit that increased RP production in HS2-virocells contributes to phage HS2's greater relative fitness [7].
The LMEMs were next applied to phage-derived biomolecules in low-P.This revealed that HS2-virocells had significantly fewer transcripts and proteins than HP1-virocells, but HS2-virocells had a significantly faster modeled rate of protein production than HP1-virocells (both P values < 0.001; Supplementary Tables S3 and  S4).This suggests that, like host proteins in HS2-virocells, HS2 synthesizes phage proteins faster than HP1, likely due to high codon and amino acid similarity between HS2-host [7].
In summary, low-P reduces transcription and translation in all cells and triggers different virocell strategies for transcript/protein resource allocation.

Virocell metabolic reprogramming strategies in low-P
Given the environment-dependent and virocell-specific biomolecule differences observed above, we next evaluated virocell metabolic reprogramming in low-P.

Proteins and transcripts inform the P status differently
In environments where P i is low, bacteria commonly display the canonical P i -stress response, which can be identified by marker genes alkaline phosphatase (phoA), phosphate ABC transporter 2C permease protein (pstA), phosphate ABC transporter ATP binding protein (pstB), phosphate ABC transporter permease (pstC), phosphate ABC transporter substrate binding protein (pstS), phosphate regulon sensor histidine kinase (phoR), and DNA binding response regulator (phoB) [49] (Fig. 2A).To ensure that our cells would display the P i -stress response, our low-P medium had an order of magnitude less P i than the 0.6 μM P i needed to activate Escherichia coli's P i -stress response [50] (Fig. 1A, Supplementary Table S1).
We confirmed that low-P uninfected Pseudoalteromonas and both virocells displayed the canonical P i -stress response because ≥92% of the P i -stress proteins were only enriched in low-P and were either depleted or not detected in high-P conditions (Fig. 2B and C; Supplementary Fig. S5A).Then, we examined the transcripts to ask whether infection enhanced or depleted the P i -stress response observed in uninfected cells.This revealed that ≤42% of the transcripts were OE in both virocells regardless of the media (Fig. 2D, Supplementary Fig. S5B).These results suggest that phage infection generally elicits a P i -stress response even in high-P conditions, which was also observed in cyanovirocells [8], with a relatively small additional P i -stress response in low-P conditions.

Both low-P virocells favor a specific pathway for nitrogen assimilation
In bacteria, P i -stress impacts nitrogen (N) assimilation, through crosstalk mediated by PhoB [49].Hypothesizing that both virocells would display this crosstalk, we examined transcripts and proteins from N metabolism in low-P, which include (i) ammonium transport (AmtB), (ii) regulation by the regulatory proteins PII-1 (GlnB) and PII-2 (GlnK), the two-component system (NtrB and NtrC), the uridylyltransferase (GlnD), and glutamine synthetase adenylyltransferase (GlnE), and (iii) assimilation by glutamine synthetase (GlnA), glutamate synthase (GltB and GltD), and glutamate dehydrogenase (GdhA) (Fig. 3A).AmtB was not detected in either virocell, but HS2-virocells did OE the transcript (Fig. 3B and C).AmtB is the protein that cells use for active N import into the cell to capture scarce ammonium [51], and in the oceans has been found highly expressed in low-N environments [52].
Of the six N regulation genes, HS2-virocells enriched fewer proteins than HP1-virocells (1 vs 3, respectively), but OE more genes (4 vs 0, respectively; Fig. 3B and C; Supplementary Fig. S6)-consistent with the global trends described above using LMEM, whereby overall HP1-virocells have more host proteins and HS2-virocells more host transcripts relative to each other.These included the kinase ntrB and transcriptional regulator ntrC that controls transcription of N assimilation genes, and their expression is regulated by PhoB [49].These co-regulated genes, ntrBC, and phoB, were only OE in low-P HS2-virocells (Fig. 3C), suggesting a HS2-virocell-specific transcriptional control of the known crosstalk between P i -stress and regulation of N assimilation.
Both virocells favored the glutamine oxoglutarate aminotransferase/glutamate synthase (GOGAT) over the glutamate dehydrogenase (GDH) cycle.Specifically, glnA from GOGAT was OE in both virocells and its protein enriched, whereas gdhA from GDH was UE and its protein depleted (Fig. 3B and C; Supplementary Fig. S6).In uninfected bacteria, GOGAT is the cell's ATP-dependent pathway for N assimilation that is preferred when P and N levels are low, but when energy is sufficient; whereas GDH utilizes TCA cycle intermediates in an ATP-independent manner and is preferred when energy is low and P and N levels are high [53] (Fig. 3A).This resource trade-off has not been previously described in virocells, but here it suggests virocells are limited for TCA cycle intermediates in addition to P and N.

Virocell-specific central carbon metabolism reprogramming toward energy and resource production in low-P
As viruses need energy and resources for replication [46], and central carbon metabolism (CCM) is a source of both [54], we Figure 2. The canonical inorganic phosphate (P i )-stress response in high-P and low-P conditions.(A) Cartoon representation of the canonical inorganic phosphate (P i )-stress response in a bacterial cell.(B) Bar plot representing the fraction of proteins that are enriched (z-score > 0), depleted (z-score < 0), or not detected, in every treatment and condition, obtained from proteomics.(C) Heatmap of the relative protein abundance (z-score across all samples and conditions) in high-P and low-P conditions, obtained from proteomics.Proteins are considered enriched if their z-score > 0, and depleted if their z-score < 0. (D) Bar plot representing the fraction of genes OE, UE, or not DE in the virocells relative to uninfected control cells in its respective condition, obtained from transcriptomics.asked whether CCM would be impacted by infection in low-P ( Supplementary Figs S7 and S8).In low-P relative to high-P conditions, the fraction of enriched proteins increased by 21% in uninfected cells and decreased by 52 and 17% in HS2-virocells and HP1-virocells, respectively (Fig. 4A and B; Supplementary Fig. S8A).This suggests environment-specific effects on infection whereby virocells reduce CCM activity in low-P conditions.
To contextualize these findings, prior work revealed that HS2 shares higher genomic similarity with the host than HP1, and that in high-P conditions, HP1-virocells have greater energy and resource demands [7,24].Applied to our current findings, it appears that HP1-virocells during nutrient limitation continue to have greater energy and resource needs relative to those of HS2-virocells, which may be alleviated through increasing CCM activity.
To test the hypotheses that low-P cells degraded FA, and HP1virocells did so the most, we characterized the intracellular lipidome.An LMEM revealed that all detected lipids had significantly (P value < 0.0001) lower abundance in low-P than in high-P conditions for both virocells, but not for uninfected cells (Supplementary Table S5).Additionally, lipid depletion was phagespecific, as HP1-virocells had significantly (P value < 0.0001) fewer lipids than HS2-virocells in low-P (Supplementary Table S5).These data suggest that lipid depletion is environment-, infection-and phage-specific, as both virocells, but not uninfected cells, deplete lipids in response to low-P conditions, and HP1-virocells more so.

Virocell ecosystem footprint
Viruses impact nutrient availability and cycling, especially carbon, in myriad ways [55][56][57][58][59][60][61][62][63].Yet, the extent to which virocells impact extracellular dissolved organic carbon (DOC) composition during infection is poorly understood, particularly in contrasting environmental conditions relevant at the micron-scale [64].To address this unknown, we tracked DOC chemical composition via high resolution MS as a diagnostic of environmental transformation, asking: what are the relative impacts of P availability and infection on cells, virocells, and their ecosystem footprints?

Low-P has a greater impact on DOC than phage infection
We evaluated which factors contributed to the observed semiquantitative variation in DOC chemical composition between samples.From a principal component analysis, in which 49.7% of the observed variation was captured in the first two principal components (Supplementary Fig. S1E), low-P was the only factor that significantly correlated with the observed variation in the DOC molecule types detected (linear regression, P value < 0.001, R 2 = 0.0004; time, P value = 0.06, and infection status, P value = 0.39, were not significant).Group-aggregated analyses also indicated that low-P was more significant (ANOVA, F = 26.9,P value < 0.001) than time (ANOVA, F = 2.7, P value < 0.005), and infection was not significant (ANOVA, F = 0.9, P value > 0.05; Supplementary Table S6).Together, these data suggest that, even though marine viruses are known to impact OM [65,66], low-P conditions more strongly shape DOC chemical composition than phage during infection.

Infection enhances extracellular lipid and peptide utilization in low-P
We investigated the chemical composition of the DOC across all samples and time points.Over 7000 DOC compounds were detected ranging in mass from ∼154-1195 Da, with 53.4% not assignable to molecular formulas or biogeochemical classes, and the rest consisting mainly of peptide-, lipid-, and polyphenol-like compounds (Supplementary materials).
How did low-P and infection impact these DOC classes (Supplementary Table S6)?First, in low-P versus high-P conditions, both virocells and uninfected cells had significantly (P value < 0.05) fewer lipid-and peptide-like compounds (Fig. 5A and B), suggesting environment-specific effects on these exometabolites.Second, in low-P, both HS2-and HP1-virocells had significantly (P value < 0.05) fewer lipids than uninfected cells (Fig. 5A), suggesting infection-specific effects on extracellular lipid abundance.However, there were no significant differences (P value > 0.05) between virocells to each other (Fig. 5A), suggesting extracellular lipids in low-P were not impacted by phage (i.e.phage-independent).For peptides, HP1-virocells had significantly (P value < 0.05) fewer extracellular peptides relative to both uninfected cells and HS2-virocells (Fig. 5B), suggesting phagespecific effects.Coupled with our intracellular observations (Fig. 3B and C; Fig. 4A-E), peptides (for N) and FAs (for energy and P) may have been partially derived extracellularly and contributed to these trends.Extracellular resource sourcing has been observed in other non-nutrient-limited virocells such as from Synechococcus [62,67] and Sulfitobacter [55], and likely helps close the elemental composition gap that exists between viruses and their hosts [14].

DOC transformation decreases in low-P and is virocell-specific
Given ∼50% of the DOC species were unclassified, we calculated the functional diversity of the exometabolome using Rao's quadratic entropy (RQE) [68][69][70].RQE measures differences between DOC pools by considering both compound abundances and differences in chemical properties, such as elemental composition and reactivity [37].This leverages a prior hypothesis that similarities in chemical properties translate to similarities in ecological function [71][72][73][74].First, in low-P versus high-P conditions, all cells had significantly (P value < 0.05) lower DOC RQE values, i.e. lower functional diversity, in terms of both elemental composition and reactivity (Fig. 5C and D; Supplementary Table S6), suggesting environment-specific effects.Second, in low-P, only HP1-virocells were significantly different (P value < 0.05), to both HS2-virocells and uninfected cells (Fig. 5C and D), suggesting virocell-specific effects whereby HP1-virocells likely transformed or exuded more diverse compounds during infection than the other cells.
Another way of understanding these DOC transformations is through their GFE and network heterogeneity (NH).GFE is routinely used in soil ecology to understand microbial OM transformation, as it represents the maximum work that can be performed in a thermodynamic system [37,[75][76][77].NH values denote the degree of substrate and end-product transformations in an entire system [78].Higher NH and GFE values denote microbial systems that are actively transforming metabolites to more recalcitrant states.First, in low-P versus high-P conditions, all cells had significantly (P value < 0.05) lower GFE (Fig. 5E), with no significant differences (P value > 0.05) between virocells, between virocells and uninfected cells 5E; Supplementary Table S6).Furthermore, GFE distribution showed a more diverse range of labile and recalcitrant metabolites in high-P than in low-P conditions (Supplementary Fig. S9), suggesting DOC transformation was environment-specific. Contrastingly, NH was significantly (P value < 0.05) lower in low-P than in high-P conditions for both virocells, but not for uninfected cells (P value > 0.05; Fig. 5F).Second, in low-P there were no significant differences (P value > 0.05) between virocells or between virocells and uninfected cells for either GFE or NH (Fig. 5F, Supplementary Table S6), suggesting an infection-dependent metabolic transformation of OM due to the environment.Together, these pre-lysis observations suggest that (i) low-P reduces DOC transformation (i.e.lower GFE) in all cells, and (ii) infection in low-P results in fewer metabolic transformations  (A and B) Boxplots representing the range of percentages of total exometabolites identified in each treatment that were classified as (A) lipid-like and (B) peptide-like compounds for uninfected cells, HS2-virocells, and HP1-virocells.(C) Elemental composition, (D) reactivity, (E) GFE of exometabolites, and (F) network heterogeneity in uninfected control cells, HP1-and HS2-virocells.Asterisks denote the P value resulting from a Wilcoxon test comparing virocells to uninfected cells, or between high-P and low-P conditions for each cell (uninfected, HS2-virocell or HP1-virocell), with " * " denoting significant comparisons (P value < 0.05) and "ns" denoting "not significant" comparisons (P value ≥ 0.05).9) invest in P to make more transcripts and, using the host's translational machinery, (10) rapidly synthesize host proteins when needed.Contrastingly, HP1-virocells struggle to make (11) transcripts and ( 12) RPs, and are less well-matched with the host translational machinery, thus resulting in (13) slower phage protein translation than HS2-virocells.Therefore, instead of relying on gene expression (transcript-to-protein) as HS2-virocells, HP1-virocell's strategy may be to (14) accumulate existing host proteins (recycling instead of synthesis) to be able to respond to f luctuating environments.The trade-off, however, is that host proteins occupy precious cellular space needed to make phage proteins, (15) resulting perhaps in fewer HP1 virions compared with HS2.Other virocell-specific phenomena are that (16) HS2-virocells likely use active transport to import ammonium, (3) HP1-virocells have fewer extracellular peptides, and HP1-virocells need more energy and resources, as shown by (17) the activation of a wider range of energy pathways in central C metabolism and ( 18) the increase in FA degradation relative to HS2-virocells.Cells and virocells are not drawn to scale and molecules do not ref lect quantity.
of DOC (i.e.lower NH), but these transformations did not significantly change the bioavailability of the system.

Virocell metabolic reprogramming and ecosystem footprint in low-P conditions
Here we report virocell intracellular metabolic reprogrammingfollowing transcripts, proteins, metabolites, and lipids-and extracellular DOC changes in low-P, relative to previously published high-P conditions assayed via transcripts and proteins only [7].We find three tiers of nested responses to the dual stresses of low-P and viral infection: (i) shared between all cells regardless of infection (i.e.environment-specific effects, but infection-independent), (ii) shared between virocells (i.e.infection-specific, but virocellindependent effects), and (iii) unique to each virocell due to the infecting phage (i.e.virocell-or phage-specific effects) (Fig. 6).
For environment-specific effects, in low-P all cells (uninfected, HP1-virocells, and HS2-virocells) reduced overall transcription and translation, activated the canonical P i -stress response, had fewer extracellular peptides and lipids, had a less diverse DOC pool (i.e.lower elemental composition and reactivity), and decreased overall DOC transformation (i.e.lower GFE), relative to high-P conditions (Fig. 6A).
For infection-specific effects, both low-P virocells had fewer extracellular lipids than uninfected cells, lower metabolic transformations (i.e.NH) than in high-P conditions, and reprogrammed intracellular metabolic pathways of N assimilation and energy (Fig. 6B).
For virocell-specific effects in low-P, extracellularly HP1-virocells had fewer peptides and higher DOC diversity (i.e.RQE) than HS2-virocells (Fig. 6C).Intracellularly, the virocells appeared to adopt different strategies for managing resource needs: gene expression regulation (HS2-virocells) versus resource accumulation and recycling (HP1-virocells).Based on nutrient-limited uninfected bacteria [79], there appears to be a trade-off between transcriptionally regulating proteins to save energy (HS2-virocells) versus recycling proteins (i.e.resource accumulation) to quickly respond to environmental changes (HP1-virocells).For example, HS2-virocells, using similar codons as the host [7], likely maintain low protein abundance and invest in transcriptional regulation for responsiveness (Fig. 6C).Contrastingly, HP1-virocells, being more codon-mismatched with the host, likely require more resources and energy (inferred from [7] and our CCM, and FA degradation results here), produce fewer RPs, recycle existing ribosomes, and maintain high host protein abundance to be metabolically ready.Based on prior work, this would come at the expense of utilizing valuable cellular space or other limiting resources for virions [80].Therefore, nutrient limitation triggers common virocell responses as well as nuanced metabolic reprogramming strategies imposed by resource and energy needs that may ultimately narrow the nutrient limitation gap and enable phages to reproduce equally well across environments (as seen from the similar phage titers obtained in our closed system, Fig. 1B and C).Finally, we posit that the nature and magnitude of the ecosystem footprint left by a virocell is not static, but rather depends on the environment of such infection.
This study provides a framework for developing empirical datasets needed to bridge the gap between laboratory phage-host experimental multi-omics datasets across taxa and conditions, and those in nature.Beyond experimental measurements, helping empiricists refine which data are most impactful in metabolic models, and myriad advances, such as in genome-scale and f lux balance analysis modeling [81], will be critical to place observational data into the mechanistic frameworks needed to predict dynamic outcomes.These steps are critical toward establishing the rules of life that govern the virocells that transform planet Earth [82,83].

Figure 1 .
Figure 1.Experimental design, phage properties, and host RPs.(A) Experimental design.Pseudoalteromonas (strain PSA[13][14][15] was independently infected with two different phages (PSA-HS2 and PSA-HP1) and sampled throughout the latent period ( boxed) for time-resolved multi-omics measurements in either low-P or high-P conditions and compared with uninfected control cells in the same condition.Transcriptomics, proteomics, and exo-metabolomics were all used as primary data, and endometabolomics and lipidomics as supporting data.The manuscript presents the story from three angles: (1) environment-specific effects (high-P versus low-P), (2) infection-specific effects (infected versus uninfected cells), and (3) phage-specific effects (HS2-virocells versus HP1-virocells) in the same environment.(B) Infection dynamics of phage PSA-HS2 in both media.(C) Infection dynamics of phage PSA-HP1 in both media.For both, (B) and (C), phage abundance, measured as plaque-forming units (pfus) per mL for free phages are represented over time, and 0 min represents 15 min after diluting the infection.The average of three biological replicates is plotted along with the error.All data points have been normalized against the first time point (0 min).Estimates of latent period are represented with rectangle boxes under each graph.(D) Gene expression dynamics of phage PSA-HS2 in both media.(E) Gene expression dynamics of phage PSA-HP1 in both media.For both, (D) and (E), relative transcript abundance for all genes is plotted as standardized log 2 FPKM (using all time points and media) over infection time, in high-P (dashed) and low-P (solid) media and colored according to earlier (blue) or later (black) temporal expression.(F, G) Scatter plot of the RNA-to-protein ratio (RNA/protein) for PSA-HS2 (F) and PSA-HP1 (G) phage genes in high-P (x-axis) and low-P (y-axis) conditions.Phage structural genes were determined based on annotation and are colored blue (HS2) or yellow (HP1); non-structural genes displayed in gray.The RNA/protein was calculated by dividing the average normalized transcript abundance (FPKM) by the average normalized protein abundance, both averaged across three replicates.(H) Bar plot representing the fraction of RPs that are enriched (z-score > 0), depleted (z-score < 0), or not detected, in every treatment and condition, including the uninfected control, HS2-virocells (HS2viro), and HP1-virocells (HP1viro), obtained from proteomics.Ribosome recycling proteins are not included.

Figure 3 .
Figure 3. Nitrogen assimilation during low-P conditions.(A) Cartoon summarizing nitrogen (N) transport, regulation, and assimilation in a cell.(B)Heatmap of the relative protein abundance (z-score across all samples and conditions) in low-P, obtained from proteomics.Proteins are considered enriched if their z-score > 0, and depleted if their z-score < 0. (C) Heatmap with relative gene expression (log 2 FC relative to uninfected control cells in the same condition and time point) of the N-metabolism genes (N transport, N regulation, and N assimilation) in low-P, obtained from transcriptomics.Genes are OE if log 2 FC > 0, UE if log 2 FC < 0, and not DE if log 2 FC = 0.The gene, ID, and functional category assignment are the same for both heatmaps.

Figure 5 .
Figure 5. Ecosystem footprint inferred from DOC composition from the exometabolome collected during infection.Exometabolome samples were collected for virocells and uninfected controls during the latent period, at 0, 30, and 60 min post-dilution of each infection in both environments.(Aand B) Boxplots representing the range of percentages of total exometabolites identified in each treatment that were classified as (A) lipid-like and (B) peptide-like compounds for uninfected cells, HS2-virocells, and HP1-virocells.(C) Elemental composition, (D) reactivity, (E) GFE of exometabolites, and (F) network heterogeneity in uninfected control cells, HP1-and HS2-virocells.Asterisks denote the P value resulting from a Wilcoxon test comparing virocells to uninfected cells, or between high-P and low-P conditions for each cell (uninfected, HS2-virocell or HP1-virocell), with " * " denoting significant comparisons (P value < 0.05) and "ns" denoting "not significant" comparisons (P value ≥ 0.05).

Figure 6 .
Figure 6.Virocell metabolic reprogramming and ecosystem footprint of cells and virocells under varying phosphate conditions.(A) Environmentspecific effects (common to all cells and virocells as a response to low-P): decrease in (1) intracellular transcripts and proteins, (2) extracellular DOM transformation and diversity, (3) extracellular peptides, (4) extracellular lipids, and (5) the activation of the canonical P i -stress response in low-P.Asterisks indicate that extracellular peptides, lipids, and DOC diversity are also affected in B and/or C. (B) Infection-specific effects (common to both virocells, but not uninfected control cells, as a response to low-P): (5) enhancing the P i -stress response of the uninfected host, (6) assimilating N via the GOGAT pathway, (7) activating central C metabolism, (4b) importing and (8) degrading lipids, which likely results in decreasing the extracellular lipid pool, and fewer metabolic transformations.(C) Virocell-specific effects (specific to each phage infection in response to low-P): HS2-virocells (9) invest in P to make more transcripts and, using the host's translational machinery,(10) rapidly synthesize host proteins when needed.Contrastingly, HP1-virocells struggle to make(11) transcripts and (12) RPs, and are less well-matched with the host translational machinery, thus resulting in(13) slower phage protein translation than HS2-virocells.Therefore, instead of relying on gene expression (transcript-to-protein) as HS2-virocells, HP1-virocell's strategy may be to(14) accumulate existing host proteins (recycling instead of synthesis) to be able to respond to f luctuating environments.The trade-off, however, is that host proteins occupy precious cellular space needed to make phage proteins,(15) resulting perhaps in fewer HP1 virions compared with HS2.Other virocell-specific phenomena are that (16) HS2-virocells likely use active transport to import ammonium, (3) HP1-virocells have fewer extracellular peptides, and HP1-virocells need more energy and resources, as shown by(17) the activation of a wider range of energy pathways in central C metabolism and (18) the increase in FA degradation relative to HS2-virocells.Cells and virocells are not drawn to scale and molecules do not ref lect quantity.