Strains and culture conditions

The C. reinhardtii strain closely related to strain S24-, which is often studied with M. japonicum, was used in the first experiment (Sample Set 1) under continuous light [26]. The C. reinhardtii cell-wall reduced strain CC-5390 was used in the experiments comparing continuous and diurnal light (Sample Set 2). This strain is amenable to synchronization under diurnal cycles, yielding populations of cells exhibiting nearly the same physiological state and gene expression, which provide a system with exceptional signal-to-noise for resolving diurnal patterns [41]. M. japonicum strain MAFF303099 was used in all experiments.

Cultures were grown in amended High Salt minimal medium (HSM) [26,42], which contains Kropat’s trace element solution [43], plus 2 mM MgSO₄·7H₂O, 10 µg/l biotin, and 10 µM CoCl₂ to support bacterial growth, but lacks a reduced carbon source. Cultures were grown in 250 ml Erlenmeyer flasks containing 100 ml medium and bubbled with filter-sterilized air (provided by an aquarium pump). Algal cultures and cocultures were incubated at 28°C, agitated at 110–130 rpm, and illuminated with 200 µmol photons/m²/s cool white light (Sylvania Dulux L55WFT55DL/841) continuously or under diurnal 12-h-light/12-h-dark cycles.

C. reinhardtii cells were precultivated axenically in photoautotrophic conditions. For the experiment comparing continuous and diurnal light, algal precultivation occurred in 400 ml flat-panel photobioreactors (Photon System Instruments, Drásov, Czechia) as previously described under the respective light regime to achieve adequate synchrony of the diurnal populations [41]. Bacterial cells for all experiments were precultivated in 14 ml polystyrene round-bottom tubes with 3 ml amended HSM containing 0.2% sucrose, agitated at 200 rpm, and incubated at 28°C. Bacterial cells were collected by centrifugation at 8000 xg and washed three times in 0.85% NaCl to remove sucrose before use as inoculum. Optical density at 600 nm (OD₆₀₀) was used to estimate the cell density of washed cell suspensions used for inoculation. Biological replicates refer to replicate flasks inoculated and incubated together in an experiment.

Estimation of cell density

C. reinhardtii cell density was determined using a Z2 Coulter Particle Count and Size Analyzer (Beckman Coulter, CA, USA). M. japonicum cell density was estimated as CFU determined from serial dilutions of culture samples on Tryptone Yeast (TY) agar plates after 4 d of incubation at 30°C in the dark.

In silico assessment of protein physicochemical characteristics

The physicochemical properties of proteins encoded by C. reinhardtii and M. japonicum were assessed according to Shi et al. 2024 [36]. The C. reinhardtii reference genome assembly and annotations v6.1 protein FASTA file from was downloaded from https://phytozome-next.jgi.doe.gov [44]. The M. japonicum MAFF303099 RefSeq protein FASTA file was downloaded from https://www.ncbi.nlm.nih.gov/data-hub/taxonomy/266835/. FASTA files were read into R using the “seqinr” package (v4.2-36). The isoelectric point (pI) of each protein was estimated using the pI function in the R package “Peptides” (v2.4.6) using pKscale = “EMBOSS”. The grand average of hydropathy (GRAVY) hydrophobicity score of each protein was estimated using the hydrophobicity function from the R package “Peptides” (v2.4.6) using scale = “KyteDoolittle”. The molecular weight of each protein was determined using the mw function from the R package “Peptides” (v2.4.6) using monoisotopic = FALSE, avgScale = “expasy”, label = “none”, and aaShift = NULL.

Proteomic sample collection and digestion

40 ml of culture was collected either 24, 36, or 48 h after inoculation. Cell pellets were collected by centrifugation at 8000 xg for 2 min at 4°C and were resuspended in 10 mM Na-PO₄ buffer (pH 7.0). Cell pellets were again collected by centrifugation at 8000 xg for 1 min at 4°C, snap frozen in liquid N₂, and stored at –80°C until further processing.

Cell pellets were diluted in 300 µl of 8 M urea and cells were lysed by bead beating in Micro-Organism Lysing Mix glass bead tubes for 45 s at 5.5 m/s using a Bead Ruptor Elite bead mill homogenizer (OMNI International, Kennesaw, GA). Lysate was collected by centrifugation at 1,000 xg for 10 min at 4°C. Protein concentration was determined by bicinchoninic acid (BCA) assay (Thermo Scientific, Waltham, MA USA). Then, 10 mM dithiothreitol (DTT) was added and samples were incubated at 60°C for 30 min with constant shaking at 800 rpm. 40 mM iodoacetamide (IAA) was then added and samples were incubated at ambient temperature in the dark for 30 min.

Protein samples were then digested with sequencing grade trypsin (United States Biological, Salem, MA) at a 1:50 (w/w) trypsin-to-protein ratio for 3 h at 37°C in the presence of 100 mM NH₄HCO₃ and 1 mM CaCl₂. Digested samples were desalted using a 4-probe positive pressure Gilson GX-274 ASPEC™ system (Gilson Inc., Middleton, WI) with Discovery C18 50 mg/1 ml solid phase extraction tubes (Supelco, St. Louis, MO), using the following protocol: 3 ml of methanol was added for conditioning followed by 3 ml of 0.1% trifluoroacetic acid (TFA) in H₂O. The samples were then loaded onto the column followed by 4 ml of 95:5 H₂O:acetonitrile (ACN), 0.1% TFA. Samples were eluted with 1 ml 80:20 ACN:H₂O, 0.1% TFA. The solvent was completely removed in a vacuum concentrator (Labconco, Kansas City, MO) and the peptides subsequently reconstituted in 100 µl pure water. Peptide concentration of the final mixture was determined using a BCA assay, and samples were diluted to 0.10 µg/µl with pure water for global proteomics LC-MS/MS analysis.

Reversed phase fractionation of peptides

Peptide mixtures were separated into 6, 12, or 24 fractions by high-resolution reversed phase UPLC using a nanoACQUITY UPLC® system (Waters Corporation, Milford, MA) equipped with an autosampler. Capillary columns, 200 µm i.d. x 65 cm long, were packed with Jupiter 3.0 µm C18 300 Å bonded particles (Phenomenex, Torrance, CA). Separations were performed at a flow rate of 2.2 µl/min on binary pump systems, using 10 mM ammonium formate (pH 7.5) as mobile phase A and 100% ACN as mobile phase B. 48 µl of the peptide mixtures (0.25 µg/µl) were loaded onto the column and separated using a binary gradient of 1% B for 35 min, 1–10% B in 2 min, 10–15% B in 5 min, and 15–25% B in 35 min, 25–32% B in 25 min, 32–40% B in 13 min, 40–90% B in 43 min, held at 90% B for 2 min (column washed and equilibrated from 90–50% B in 2 min, 50–95% B in 2 min, held at 95% B for 2 min and 95–1% in 4 min). The capillary column eluent was automatically deposited every minute into 12 mm x 32 mm polypropylene vials (Waters Corporation, Milford, MA) starting at minute 60 and ending at minute 170, over the course of the 180 min LC run. Fractions were concatenated into 6, 12, or 24 vials by collecting fractions from vial 1 to vial 6–24 and then returning to vial 1 (and so on). Prior to peptide fraction collection, 100 µl of water was added to each vial to facilitate the recovery of small droplets added into the vial. Each vial was completely dried in a vacuum concentrator (Labconco, Kansas City, MO) and reconstituted in 20 µl 2% ACN, 0.1% formic acid.

LC-MS/MS proteomics acquisition

A nanoACQUITY dual pumping UPLC system (Waters Corporation, Milford, MA) was configured for on-line trapping of a 5 µl injection at 5 µl/min for 10 min followed by reverse-flow elution onto the analytical column at 200 nl/min. The trapping column was slurry packed in-house using 360 µm o.d. x 150 µm i.d. fused silica (Polymicro Technologies Inc., Phoenix, AZ) Jupiter 5µm C18 media (Phenomenex, Torrence, CA) with 2 mm sol-gel frits on either end. The analytical column was slurry packed in-house using Waters BEH 1.7 µm particles packed into a 35 cm long, 360 µm o.d. x 75 µm i.d. column with an integrated emitter (New Objective, Inc., Littleton, MA). Mobile phases consisted of (A) 0.1% formic acid in water and (B) 0.1% formic acid in ACN with the following gradient profile (min, %B): 0, 1; 10, 8; 105, 25; 115, 35; 120, 75; 123, 95; 129, 95; 130, 50; 132, 95; 138, 95; 140, 1.

MS analysis was performed using a Thermo Eclipse mass spectrometer (Thermo Scientific, San Jose, CA). The ion transfer tube temperature and spray voltage were 300 ºC and 2.4 kV, respectively. Data were collected for 120 min following a 27 min delay from when the trapping column was switched in-line with the analytical column. FAIMS was used at CVs –40 V, –60 V, and –80 V. FT-MS spectra were acquired from 300–1800 m/z at a resolution of 120 k (AGC target 4e5) and while the top 12 FT-HCD-MS/MS spectra were acquired in data-dependent mode with an isolation window of 0.7 m/z and at an orbitrap resolution of 50 k (AGC target 5e4) using a fixed collision energy (HCD) of 32 and a 30 s exclusion time.

LC-MS/MS proteomics data processing

All LC-MS/MS datafiles were converted to mzML using MSConvert with all peaks kept during centroiding [45]. MZRefinery was used to calibrate search mass values from the MS2 data, which were then input to MS-GF+ [46]. All data were searched against a single concatenated FASTA file containing amino acid sequences from both the C. reinhardtii reference genome assembly and annotations v6.1 (https://phytozome-next.jgi.doe.gov/info/CreinhardtiiCC_4532_v6_1) (32,670 protein entries, 17,691 protein-coding genes) [44], the M. japonicum MAFF303099 genome assembly and annotations (https://www.ncbi.nlm.nih.gov/datasets/genome/GCF_000009625.1/) (7,105 protein entries, 7,162 protein-coding genes), and commonly observed contaminant proteins (16 protein entries). We employed a target/decoy approach with +/–20 ppm parent mass tolerance, dynamic modification of oxidized methionine, and allowance of partially tryptic peptide candidates [47]. MS abundances were extracted using StatsMomentsArea via the MASIC software to provide area-under-the-elution-curve values (i.e., extracted ion chromatograms) [48]. Data were collated using the MAGE software suite (https://github.com/PNNL-Comp-Mass-Spec/Mage), imported into SQL Server, and cross-checked for completeness. FDR was controlled to 1% (q-value ≤ 0.010297 gave 16,354 decoy identifications out of 1,635,330 total for Sample Set 1; q-value ≤ 0.0103619 gave 31,592 decoy identifications out of 3,134,149 total for Sample Set 2). Peptide-to-protein relationships were subjected to parsimony such that a protein was assigned if a) the peptide was uniquely mapped to that protein, b) was non-uniquely mapped to only one protein with a different uniquely mapping peptide, c) was non-uniquely mapped to a protein more other non-uniquely mapping peptides than any others, or d) was mapped to the first occurrence of the peptide in the search FASTA file when more than one protein had similar maximal peptide counts.

Data from samples that had been fractionated were then processed separately from data from samples that had not been fractionated. To infer protein abundance from peptide abundance, the simple roll-up method was applied, in which peptide abundances were log₂ transformed, mean central tendency normalized, then grouped by their gene ID and summed per sample, resulting in protein abundances as log₂-transformed MASIC values. This common approach was preferred here over other existing approaches for label-free protein quantification (e.g., MaxLFQ [49]), as it is computationally simple, uses all peptide information available, inherently weights signals based on their abundances (i.e., more abundant peptides influence protein abundance determination more than less abundant peptides whose signal may be noisier), and is suitable for LC-MS-level quantitation [50].

Only proteins with at least two unique peptides for quantitation that were detected in at least two of the three biological replicates in any condition were analyzed. In the unfractionated pilot experiment (Sample Set 1), this included 9871 proteins (6113 C. reinhardtii proteins, 3758 M. japonicum proteins). In the unfractionated mono-mix experiment (Sample Set 2), this included 9784 proteins (7039 C. reinhardtii proteins, 2745 M. japonicum proteins). In the mono-mix experiment comparing protein detection before and after peptide sample fractionation (Sample Set 2, fractionated), this included 15,464 proteins (10,858 C. reinhardtii proteins, 4606 M. japonicum proteins). Missing values were handled by exclusion rather than imputation.

Log₂-transformed protein MASIC values were quantile normalized using the normalize.quantiles function from the R package preprocessCore (v1.66.0) [51,52]. Quantile normalization was performed independently for each organism: the distributions of C. reinhardtii protein abundances across the relevant samples were quantile normalized together, and the distributions of M. japonicum protein abundances across the relevant samples were quantile normalized together. Normalized and unnormalized abundances for proteins analyzed in the unfractionated pilot experiment (Sample Set 1) are provided in S2 Table. Normalized and unnormalized abundances for the proteins analyzed in the unfractionated mono-mix experiment (Sample Set 2) are provided in S3 Table. Normalized and unnormalized abundances for the proteins analyzed in the mono-mix experiment before and after peptide sample fractionation (Sample Set 2, fractionated) are provided in S5 Table.

The average protein abundances were calculated after excluding missing values and the log₂-fold change in protein abundance was calculated as (mean abundance in coculture) – (mean abundance in monoculture). To calculate the Z-score abundance for each protein in a given experiment, missing values were excluded, then the mean and standard deviation of quantile-normalized MASIC values for each protein across the experimental conditions was calculated, and the Z-score was calculated as (mean norm. MASIC – (mean of all mean norm. MASIC))/(sd of all mean norm. MASIC).

Clustering analyses

k-means clustering was performed on the Z-score of quantile-normalized mean protein abundances in the unfractionated mono-mix experiment (Sample Set 2), separately for C. reinhardtii proteins and M. japonicum proteins that were detected in at least two biological replicates in all conditions (4737 C. reinhardtii proteins, 828 M. japonicum proteins). The fviz_nbclust function from the R package factoextract (v1.0.7) was used to generate elbow plots to determine the optimal k for clustering. The kmeans function from the R package stats (v4.4.0) was then applied using iter.max = 1000, centers = 6 for C. reinhardtii proteins or centers = 3 for M. japonicum proteins. Cluster assignment is provided in S3B Table. Heatmaps were generated using the pheatmap function from the R package pheatmap (v1.0.12).

Principal component analysis (PCA) was performed on quantile-normalized protein abundances of individual replicate samples in the unfractionated mono-mix experiment. This was done separately for C. reinhardtii proteins and M. japonicum proteins that were detected above the limit of quantitation (two standard deviations below the mean MASIC value in the experiment) in all samples (all replicates in all conditions: 3817 C. reinhardtii proteins, 569 M. japonicum proteins). The pca function was applied from the R package PCAtools (v2.16.0) and the results were visualized using the biplot function.

Statistical tests

Significant differences in culture densities and the number of proteins detected in different sample groups were assessed by two-tailed Student’s t-tests. Significant differences in the distribution of protein abundances (only for those proteins that were detected in at least two of the three biological replicates in both the monoculture control condition and the respective coculture condition) were assessed by Wilcoxon rank-sum tests (Mann-Whitney tests) using the wilcox.test function in the R package stats (v4.4.1).

Significant differences in the abundance of a given protein in coculture relative to monoculture were assessed using the t.test function in the R package stats (v4.4.0). Only proteins that were detected in at least two of the three biological replicates in both the monoculture control condition and the respective coculture condition were analyzed. In the unfractionated pilot experiment (Sample Set 1), this included 5368 C. reinhardtii proteins and 1737 M. japonicum proteins. In the unfractionated mono-mix experiment (Sample Set 2), it included 6246 C. reinhardtii proteins and 2128 M. japonicum proteins. In the fractionated mono-mix experiment (Sample Set 2, fractionated), it included 9934 C. reinhardtii proteins and 4179 M. japonicum proteins). Missing values were handled by exclusion rather than imputation. p values were adjusted for the number of t-tests conducted per species in each experiment using a Benjamini-Hochberg correction by applying the p.adjust function with method = “fdr”. Significant differences in protein abundance were defined as those where |log₂-fold change| > 1, p-adj. < 0.05, and the mean log₂-transformed MASIC value was greater than the limit of quantitation in both the coculture and the monoculture control.

Gene ontology (GO) enrichment analysis

GO term assignments for the C. reinhardtii v6.1 proteome were downloaded from Phytozome (https://phytozome-next.jgi.doe.gov/). GO terms were assigned de novo to the M. japonicum MAFF303099 proteome (available from the NCBI with accession GCF_000009625.1) using InterProScan (v5.59-91.0) with --iprlookup –goterms [53]. Then, the lists of proteins with significant changes in their abundance in coculture relative to monoculture were tested for GO term enrichment using the enricher function in the R package clusterProfiler (v3.0.4). p values were corrected using a Bonferroni-Holm adjustment, and enriched GO terms where p-adj. < 0.05 were considered significant.

Accession numbers

Raw proteomics data have been deposited at http://massive-ftp.ucsd.edu/ under the accession MSV000098360 and are available as of the date of publication.

Bacterial protein detection by DDA LC-MS/MS is impaired in the presence of an algal partner

To determine whether label-free DDA LC-MS/MS could be used to monitor changes to a bacterium’s proteome during coculture with an alga, we grew the model alga C. reinhardtii and the bacterium M. japonicum under four conditions and performed proteomics: algal monoculture, bacterial monoculture supplemented with sucrose, coculture, and coculture supplemented with sucrose (S1 and S2 Tables). We found that the presence of C. reinhardtii reduced detection of bacterial proteins substantially (Fig 1). In the absence of the alga, nearly 4000 unique bacterial proteins were detected, but in coculture, the number of unique bacterial proteins detected decreased as the relative abundance of the alga increased, such that <500 bacterial proteins could be detected in the coculture in the absence of sucrose (Fig 1A and 1B). For bacterial proteins detected in all three sample types, the measured abundance was also significantly decreased proteome-wide in coculture samples relative to monoculture samples (Fig 1C, Wilcoxon rank-sum tests, p < 0.05). Whereas the histogram of algal protein abundances in the monoculture samples overlapped with the histograms of algal protein abundances in coculture samples (muddied color from overlapping purple, turquoise, and orange histograms), the histograms of bacterial protein abundances were shifted to lower abundances as the relative abundance of the alga increased (Fig 1C, Wilcoxon rank-sum tests, p < 0.05). These results suggest a systematic decrease in the selection of bacterial peptide precursor ions in the presence of relatively abundant algal peptides in coculture samples.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Detection of bacterial proteins decreases in coculture samples.

Triplicate continuous-light-grown monocultures and cocultures with and without sucrose were collected for LC-MS/MS proteomics. See also S1 and S2 Figs. (A) Density of C. reinhardtii cells (left) and M. japonicum CFU (right) in triplicate monocultures (purple), cocultures (orange), and cocultures with sucrose (turquoise) upon collection for proteomics. The bars represent the mean of three replicate cultures, and error bars represent the standard deviation from the mean. Asterisks indicate significant differences by two-tailed Student’s t-tests (p < 0.05). (B) Number of unique C. reinhardtii proteins (left) and M. japonicum proteins (right) detected in samples from (A) by LC-MS/MS proteomics. Asterisks indicate significant differences as in (A). (C) Distribution of algal (left) and bacterial protein abundances (right) in cocultures (transparent orange), cocultures with sucrose (transparent turquoise), and monocultures (purple) for proteins detected in all three conditions. When distributions overlap, the color is muddied. The unnormalized MASIC values were averaged across the three biological replicates. Median values are shown above the dashed lines. Asterisks next to the medians indicate significant differences between the indicated distributions by a Wilcoxon rank-sum test (p < 0.05). (D) Distribution of algal (left) and bacterial protein abundances (right) in cocultures (transparent orange), cocultures with sucrose (transparent turquoise), and monocultures (purple) after quantile normalization, presented as in (C). (E) Changes in quantile-normalized protein abundances in coculture relative to monoculture for C. reinhardtii (left) and M. japonicum (right). Significant differences (triangles) were defined as those where |log₂-fold change| > 1, p-adj. < 0.05 from a Welch two sample t-test of the quantile-normalized MASIC values, and the mean MASIC value was greater than the limit of quantitation in both the coculture and the monoculture.

https://doi.org/10.1371/journal.pone.0340253.g001

Similar observations were apparent upon reanalysis of proteomics data from a published study on the interaction between C. reinhardtii and the actinomycete Arthrobacter strain P2b [21]. Although bacterial density was not impacted by the presence of the alga, the number of bacterial proteins detected by label-free DDA LC-MS/MS decreased in coculture relative to monoculture (S1A Fig), and the measured abundance of bacterial proteins that were detected in both conditions was decreased globally in samples containing C. reinhardtii proteins (S1B Fig, Wilcoxon rank-sum tests, p < 0.05) [21]. We wondered if decreased detection of bacterial proteins in coculture samples was unique to those containing eukaryotic partners with larger cells, or rather generalizable to any sample with another microbe. Therefore, we also reanalyzed label-free DDA LC-MS/MS proteomics data from syntrophic bacterial cultures of Dehalococcoides ethenogenes strain 195 and Desulfovibrio vulgaris Hildenborough [54]. In this case, the two organisms are closer in both cell size and the number of protein-coding genes. In addition, coculture increased growth of D. ethenogenes, and D. ethenogenes cells made up roughly 80% of the biomass in the cocultures analyzed by proteomics [54]. Nevertheless, detection of D. ethenogenes proteins was markedly decreased in cocultures relative to monocultures (S1C and S1D Fig). Furthermore, a similar observation was reported from proteomic analysis of the archaeon Nanoarchaeum equitans in the presence and absence of another archaeon, the epibiont Ignicoccus hospitalis [34]. These data on diverse symbiotic interactions suggest that increased complexity of coculture samples decreases microbial protein detection regardless of the identity or relative abundance of species present.

We sought to assess how differences in detection influence differential expression analysis of algal and bacterial proteins. We used standard criteria for defining significant changes in protein abundance: |log₂-fold change| > 1 and p-adj. < 0.05 from a Welch two sample t-test of the protein’s abundance (MASIC values) in replicate coculture and monoculture samples. We found that no C. reinhardtii proteins showed significant changes in abundance in coculture relative to monoculture (S2A Fig). This was not surprising, as we previously showed that wild-type C. reinhardtii growth and physiology are not significantly impacted by M. japonicum [26]. When we compared the abundance of the 1737 M. japonicum proteins detected in bacterial monocultures and cocultures with sucrose, we found that 1350 proteins (78%) were significantly decreased in abundance in the coculture samples, whereas only 12 (0.7%) were increased (S2B Fig). Downregulation of such a large proportion of the bacterium’s proteome was not expected, particularly since M. japonicum’s preferred carbon source, sucrose, was provided in both conditions. Given the systematic depression of bacterial protein abundance observed in Fig 1C, it was not possible to distinguish whether these differences in protein abundance were biological or methodological.

Data normalization has been shown to be important for differential expression analysis using label-free DDA proteomics [55–58], and it has been employed in order to perform differential expression analysis on data collected from microbial cocultures and monocultures [34,36,59]. Therefore, we tested whether data normalization could allow for appropriate comparisons of the abundance of bacterial proteins in monoculture and algal coculture. We applied quantile normalization, a complete normalization strategy that equalizes sample distributions by ranking the values in each sample and substituting the values by the rank-wise means [51]. This strategy has been shown to optimally reduce technical variability in shotgun LC-MS/MS datasets [56]. Since the distributions of algal protein abundances were markedly different from the distributions of bacterial protein abundances (Fig 1C), proteomics data from each organism were quantile normalized independently: the abundances of each organism’s proteins in each replicate sample were quantile normalized relative to the abundances of that organism’s proteins in all relevant samples. For example, the abundances of M. japonicum proteins in Replicate A of the monoculture condition were quantile normalized together with the abundances of M. japonicum proteins in the other monoculture replicate samples, in the replicate samples from cocultures, and in the replicate samples from cocultures with sucrose, but not with the abundances of proteins in the C. reinhardtii monoculture samples. We found that this per-organism quantile normalization strategy led to greater overlap in the distributions of protein abundances for both species (Fig 1D, a greater proportion of the histograms are overlapping and their color is muddied; S2B Table). However, significant differences in the distributions remained (Fig 1D, Wilcoxon rank-sum tests, p < 0.05). Furthermore, differential expression analysis using the quantile-normalized values still suggested that most bacterial proteins were significantly decreased in abundance in coculture relative to monoculture in the presence of sucrose (Fig 1E). In this case, 949 bacterial proteins (55%) decreased in abundance in coculture samples, whereas only 28 bacterial proteins (1.6%) increased. Thus, comparison of protein abundance determined by shotgun LC-MS/MS analysis of monoculture and algal coculture samples is not an appropriate way to discern symbiosis-induced changes in the bacterial proteome, even after data normalization.

The “mono-mix” strategy allows comparable detection of bacterial proteins from a coculture and a monoculture control by DDA proteomics

In order to enable differential expression analysis of the bacterial proteome in cocultures relative to monocultures from label-free DDA, we sought to develop a more appropriate monoculture control that could emulate the analyte complexity of the coculture condition. In a previous study investigating the interaction between the archaeon Methanococcus maripaludis and the bacterium D. vulgaris, protein abundances in coculture were compared to abundances in a “synthetic blend,” where equal amounts of total protein from each organism had been mixed together prior to analysis [60]. Then, the fold change in each protein’s abundance in cocultures relative to the synthetic blend was multiplied by a constant correction factor according to the estimated relative abundance of the organisms in coculture. This correction factor approach assumes that differences in organism relative abundance would impact detection equally for all peptides regardless of their physicochemical properties or their abundances. Proteins encoded by C. reinhardtii and M. japonicum exhibit a wide range of isoelectric points, hydropathies, and molecular weights, which can influence their detection (S3 Fig).

We developed a modified experimental design that we term the “mono-mix” strategy, in which DDA peptide detection in control samples may more closely emulate detection in coculture samples. We grew C. reinhardtii monocultures and M. japonicum monocultures supplemented with sucrose in parallel to cocultures in minimal medium. Then, immediately prior to collecting cell pellets for proteomics, we combined the two monocultures to approximate the bacteria-to-algae ratio expected in the cocultures (Fig 2A). We call this combined monoculture the “mono-mix” control.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. The mono-mix strategy enables comparable DDA detection of bacterial proteins in the coculture and the monoculture control when a similar bacteria-to-algae ratio is achieved.

To assess whether C. reinhardtii’s diurnal rhythms influence M. japonicum, cultures were grown in either continuous or diurnal light. See also S3 and S4 Figs. (A) Schematic of experiment. Triplicate cocultures of C. reinhardtii (green) and M. japonicum (pink) were grown in parallel with triplicate monocultures of C. reinhardtii and M. japonicum with 150 µg/ml sucrose. Then, the M. japonicum monocultures were added to the C. reinhardtii monocultures to achieve a “mono-mix” control (purple) with a similar bacteria-to-algae ratio as the coculture (orange). Cultures were grown in either continuous (sun icon) or diurnal light (eclipsed sun icon). Continuous light cultures were collected 36 h after inoculation, and the diurnal light cultures were collected at the end of the night (36 h after inoculation) and the end of the day (48 h after inoculation). (B) Density of C. reinhardtii cells and M. japonicum CFU in triplicate cocultures (orange) and mono-mix controls (purple) grown in continuous light (sun icon) or diurnal light (eclipsed sun icon) upon collection for proteomics at the end of the dark or light phases (black or white bars, respectively). The bars represent the mean of three replicate cultures, and error bars represent the standard deviation from the mean. Asterisks indicate significant differences by two-tailed Student’s t-tests (p < 0.05). (C) Number of unique algal (left) and bacterial proteins (right) detected in the samples from (B). Asterisks indicate significant differences as in (B). (D) Distribution of normalized abundances of algal proteins in cocultures (transparent orange) and mono-mix controls (purple) grown in continuous light (sun icon) or diurnal light (eclipsed sun icon) collected at the end of the dark or light phases (black or white bars, respectively). When distributions overlap, the color is muddied. The quantile-normalized MASIC values were averaged across the three biological replicates. Median values are shown above the dashed lines. Asterisks next to the medians indicate significant differences between the indicated distributions by a Wilcoxon rank-sum test (p < 0.05). (E) Distribution of normalized abundances of bacterial proteins presented as in (D).

https://doi.org/10.1371/journal.pone.0340253.g002

Like other algae, C. reinhardtii coordinates its metabolism and gene expression with the time of day under diurnal light. The alga performs photosynthesis, cell growth, and starch synthesis during the day period, initiates replication at dusk, then degrades starch and induces fermentative metabolism during the night in the G0 phase of the cell cycle [41,61]. To assess whether these diurnal rhythms influence M. japonicum in coculture, we applied the “mono-mix” strategy to cultures grown under continuous light and diurnal light. C. reinhardtii strain CC-5390 was used for this experiment, as this strain is amenable to synchronization under diurnal cycles. This yields populations of cells exhibiting nearly the same physiological state and gene expression, which provide exceptional signal-to-noise for resolving diurnal patterns [41]. We sampled cultures grown in diurnal light at both the end of the night (36 h after inoculation) and the end of the day (48 h after inoculation), and we compared them to cultures grown in continuous light sampled 36 h after inoculation (Fig 2A, S1 and S3 Tables).

Preparing a mono-mix control with the same bacteria-to-algae ratio as the parallel coculture was challenging. We determine the number of bacterial cells in a culture as CFU on TY agar plates, which is assessed only after 4 d of incubation. Thus, only a rough estimate of the bacterial cell density is available at the time samples are collected for proteomics. Therefore, there was variation in how comparable the cell densities were in the mono-mix controls and their parallel cocultures (Fig 2B). We found that although the number of bacterial proteins detected in the cocultures (roughly 800–2200) was more comparable to that of the mono-mix controls (roughly 1600–2300), it was still significantly lower in the samples collected from the continuous light cultures and the diurnal light cultures at the end of the day (Fig 2C, Student’s t-test, p < 0.05). The mono-mix control was most comparable to the coculture in the diurnal night samples, both in terms of organism relative abundance and of the number of bacterial proteins observed (Fig 2B and 2C). In this condition, there was no significant difference in the total number of bacterial proteins detected (Fig 2C, Student’s t-test, p = 0.48) nor in the distribution of bacterial protein abundances in the coculture relative to the mono-mix control (Fig 2E, Wilcoxon rank-sum test, p = 0.30). This was true whether data were quantile normalized or not (Fig 2E and S4B Fig, respectively).

Data normalization and comparison to mono-mix controls reveal that light regime may influence the M. japonicum proteome during coculture with C. reinhardtii

Having achieved more comparable detection of the bacterial proteome in cocultures and mono-mix controls, we next sought to determine whether these data could reveal the impact of light regime and time of day on the organisms during their interaction. We performed k-means clustering and principal component analysis (PCA) on the quantile-normalized abundances of proteins that were detected in all conditions (the cocultures and mono-mix controls in continuous light, diurnal night, and diurnal day). The C. reinhardtii proteome could be separated into 6 clusters that exhibited distinct patterns with respect to light regime and the presence of the bacterium (Fig 3A, S3B Table). For example, Cluster Cr1 proteins were more abundant in coculture under all light regimes, and Cluster Cr5 proteins were more abundant in samples collected during the light phase (either continuous or diurnal) regardless of the presence of the bacterium during growth.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Data normalization and comparison to mono-mix controls reveals that light regime may influence the M. japonicum proteome during coculture with C. reinhardtii.

(A) k-means clustering (k = 6) of the Z-score of quantile-normalized mean algal protein abundances in the mono-mix experiment. 4737 C. reinhardtii proteins that were detected in at least two biological replicates in all conditions were included in the analysis. (B) k-means clustering (k = 3) of the Z-score of quantile-normalized mean bacterial protein abundances in the mono-mix experiment. 828 M. japonicum proteins that were detected in at least two biological replicates in all conditions were included in the analysis. (C) PCA of quantile-normalized protein abundances for the 3817 C. reinhardtii proteins that were detected in all samples: three biological replicates of the mono-mix (circles) and coculture samples (triangles) under continuous light (white), diurnal night (dark grey), and diurnal day (light grey). (D) PCA of quantile-normalized protein abundances for the 569 M. japonicum proteins that were detected in all samples, presented as in (C). (E) Intersections and differences in the bacterial proteins that were significantly increased (yellow) and significantly decreased (blue) in samples collected from continuous light, diurnal night, and diurnal day. Intersection sizes (indicated by lines connecting colored dots) and differences (indicated by single colored dots) are shown as the vertical bars, while the total number of bacterial proteins that are significantly increased and decreased are shown as the horizontal bars. Significant differences were defined as those where |log₂-fold change| > 1, p-adj. < 0.05 from a Welch two sample t-test of the quantile-normalized MASIC values, and the mean MASIC value was greater than the limit of quantitation in both the coculture and the monoculture. See also S5 Fig.

https://doi.org/10.1371/journal.pone.0340253.g003

On the other hand, the M. japonicum proteome could only be separated into 2-to-3 distinct clusters (Fig 3B, S3B Table). There was only a subtle difference in the pattern of expression between bacterial proteins in Clusters Mj1 and Mj2 across the different light and culture conditions. Most bacterial proteins appeared to have lower abundance in the continuous light coculture samples (for which the distribution of protein abundances was significantly lower, Fig 2E, Wilcoxon rank-sum test, p < 2x10^–16) and high abundance in the continuous light mono-mix controls and the diurnal light cultures collected at the end of the night. Bacterial proteins in Cluster Mj3 appeared higher in cocultures than in monocultures, particularly in the diurnal night samples. These results suggest that complexity-related differences in detection of bacterial peptides in the various samples still influence the data.

PCA revealed that 27% of the variation in C. reinhardtii’s proteome could be explained by the presence or absence of light, and 13% of the variation was driven by the interaction with the bacterium (Fig 3C). PCA of M. japonicum’s proteome showed that the continuous light cocultures (white triangles) were most distinct from their respective mono-mix controls (white circles) (Fig 3D). This was also evident in the k-means clustered heatmap, where the majority of bacterial proteins detected appeared less abundant in the cocultures grown in continuous light than in the mono-mix controls grown in continuous light (Fig 3B). As discussed, bacterial protein detection was significantly lower in the continuous light cocultures relative to their mono-mix controls (Fig 2C and 2E). In addition, differential expression analysis suggested that in continuous light, many more proteins were significantly decreased in abundance in coculture than were increased (S5 Fig, S4 Table). Therefore, the patterns observed by clustering analyses likely reflect differences in organism relative abundance and consequent differences in peptide detection in the coculture and mono-mix samples collected from continuous light.

Interestingly, however, PCA of M. japonicum’s proteome also revealed that while the mono-mix controls (circles) clustered relatively close to one another on both the first and second principal components, the cocultures (triangles) were quite separated from one another by the first two principal components (Fig 3D). This suggested that light regime had a greater influence on the bacterial proteome in coculture than it did in monoculture.

Comparative analysis of the significant changes in bacterial protein abundance revealed that the majority of changes occurred uniquely in either continuous light, diurnal night, or diurnal day (Fig 3E, S5 Fig, and S4 Table). Two bacterial proteins were significantly increased regardless of light regime: WP_010914712.1, annotated as an extracellular solute-binding protein, and WP_010912842.1, annotated as a sugar-binding protein. Light-insensitive changes in proteins with such functions could be expected in coculture, where the bacterium relies on C. reinhardtii for reduced carbon substrates, compared to monocultures grown with sucrose, M. japonicum’s preferred carbon source.

Taken together, these results suggest that although complexity-related differences in peptide detection still influence proteomic comparisons between cocultures and monocultures, the mono-mix strategy together with data normalization may allow one to capture more meaningful changes in the proteome resulting from an algal-bacterial interaction by label-free DDA.

Fractionation improves detection and differential expression analysis of sparse proteins in coculture

Sample fractionation is a strategy that physically separates analytes in complex samples (e.g., by liquid chromatography (LC)) to reduce the diversity of ions entering the mass spectrometer at any given moment [62–66]. This simplification enables the instrument to perform additional fragmentation scans to identify and quantify detectable compounds, decreasing the impact of complexity, particularly for low-abundance analytes.

To test whether fractionation could improve detection of bacterial proteins from coculture by label-free DDA, we analyzed the proteome of a coculture sample after separation into 1, 6, 12, or 24 fractions. Each level of fractionation substantially increased the number of proteins that could be detected in coculture, both for M. japonicum as well as for C. reinhardtii (Fig 4A). We next sought to combine fractionation with the mono-mix strategy to determine changes in the M. japonicum proteome upon interaction with the alga. Since the diurnal night samples had the most comparable cocultures and mono-mix controls (Fig 2), we chose these samples as our test-case and analyzed them each as a single fraction or as 12 fractions by label-free DDA (S1 and S5 Tables). This level of fractionation resulted in a 1.8-fold increase in the number of algal proteins detected and a 1.9-fold increase in bacterial proteins detected (Fig 4B). In addition, the fractionated samples showed no significant difference in the distribution of protein abundances between the coculture and the respective mono-mix controls for either organism (Fig 4C bottom, Wilcoxon rank-sum tests, p > 0.05).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Fractionation improves detection of sparse bacterial proteins in coculture.

(A) The number of unique algal (left) and bacterial proteins (right) detected when a coculture sample was analyzed as 1, 6, 12, or 24 fractions. (B) The number of unique algal (left) and bacterial proteins (right) detected from cocultures (orange) and mono-mix controls (purple) collected at the end of the night from diurnal light-grown cultures when analyzed as 1 or 12 fractions. Bars represent the mean of three replicate cultures, and error bars represent the standard deviation from the mean. Asterisks indicate significant differences by two-tailed Student’s t-tests (p < 0.05). (C) Distribution of normalized abundances of algal (left) and bacterial proteins (right) in cocultures (transparent orange) and mono-mix controls (purple) collected at the end of the night when analyzed as 1 (top) or 12 fractions (bottom). When distributions overlap, the color is muddied. The quantile-normalized MASIC values were averaged across the three biological replicates. Median values are shown above the dashed lines. Asterisks next to the medians indicate significant differences between the indicated distributions by a Wilcoxon rank-sum test (p < 0.05).

https://doi.org/10.1371/journal.pone.0340253.g004

We next determined how combining fractionation with the mono-mix strategy influenced differential expression analysis of cocultures relative to monocultures. We found that running samples as 12 fractions nearly doubled the number of proteins that met the detection criteria for differential expression analysis. Over 4000 more algal proteins and 2000 more bacterial proteins could be assessed for changes in coculture relative to mono-mix controls when samples were fractionated (Fig 5A and 5B). Most of these newly captured proteins were of lower abundance on average.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Sample fractionation improves differential expression analysis of proteins in cocultures relative to mono-mix controls.

See also S6 Fig. (A) Changes in quantile-normalized C. reinhardtii protein abundance in coculture relative to mono-mix controls at the end of the night when samples were analyzed as 1 (left) or 12 fractions (right). Points are colored based on their detection criteria: lighter blue indicates that the protein only met the criteria when the sample was analyzed as either 1 or 12 fractions, navy blue indicates that the protein met the criteria when the sample was analyzed both as 1 and as 12 fractions. Significant differences (triangles) were defined as those where |log₂-fold change| > 1, p-adj. < 0.05 from a Welch two sample t-test of the quantile-normalized MASIC values, and the mean MASIC value was greater than the limit of quantitation in both the coculture and the mono-mix. Yellow and blue boxes indicate the number of significantly increased and decreased proteins, respectively. Histograms show the distribution of mean protein abundances (x axes) and of log₂-fold changes (y axes) for proteins that only met the detection criteria when the sample was analyzed as either 1 or 12 fractions (lighter blue) and for proteins that met the detection criteria when the sample was analyzed both as 1 and as 12 fractions (navy blue); median values are shown above the dashed lines. (B) Changes in quantile-normalized M. japonicum protein abundance in coculture relative to mono-mix controls at the end of the night when samples were analyzed as 1 (left) or 12 fractions (right) presented as in (A). (C) Intersections and differences in the bacterial proteins that were significantly increased (yellow) and significantly decreased (blue) in coculture when samples were analyzed as 1 or 12 fractions. Intersection sizes (indicated by lines connecting colored dots) and differences (indicated by single colored dots) are shown as the vertical bars, while the total number of bacterial proteins that are significantly increased and decreased are shown as the horizontal bars. (D) GO term enrichment in bacterial proteins that were significantly increased in quantile-normalized abundance in coculture when samples were analyzed as 1 or 12 fractions. The number of proteins with the corresponding GO term in each comparison is indicated by dot size and the p-adj. by the shading. No GO terms were significantly enriched in the bacterial proteins that were significantly decreased in quantile-normalized abundance in coculture, regardless of sample fractionation.

https://doi.org/10.1371/journal.pone.0340253.g005

While sample fractionation did not reveal any additional significant changes in the algal proteome (Fig 5A), nearly 4X the number of significant changes in bacterial protein abundance could be observed if samples were fractionated (Fig 5B and S6 Fig, S6 Table). Of the 55 M. japonicum proteins that exhibited a significant change in abundance in coculture relative to the mono-mix controls, 15 did not meet our detection criteria without sample fractionation. The remaining 40 proteins were detected in samples regardless of fractionation, but a significant change (|log₂-fold change| > 1, p-adj. < 0.05) was only observable if the samples were fractionated. Thus, sample fractionation improves differential expression analysis of bacterial proteins in coculture not only by increasing the detection of lowly abundant proteins, but also by increasing the signal-to-noise, such that changes in abundance can be observed.

Of the 36 significant increases in protein abundance in coculture observed if samples were analyzed as 12 fractions, 34 of them were missed if samples were analyzed as 1 fraction (Fig 5C). Gene ontology (GO) enrichment analysis revealed that in addition to “outer membrane-bounded periplasmic space” proteins, M. japonicum increased the abundance of groups of proteins significantly enriched for transport, thiamine-binding, and carbohydrate-binding activity (Fig 5D). These additional changes detectable upon sample fractionation may provide clues about the kinds of metabolites that are exchanged between C. reinhardtii and the bacterium.