Identifying Aspects of the Post-Transcriptional Program Governing the Proteome of the Green Alga Micromonas pusilla

Abstract

Micromonas is a unicellular motile alga within the Prasinophyceae, a green algal group that is related to land plants. This picoeukaryote (<2 μm diameter) is widespread in the marine environment but is not well understood at the cellular level. Here, we examine shifts in mRNA and protein expression over the course of the day-night cycle using triplicated mid-exponential, nutrient replete cultures of Micromonas pusilla CCMP1545. Samples were collected at key transition points during the diel cycle for evaluation using high-throughput LC-MS proteomics. In conjunction, matched mRNA samples from the same time points were sequenced using pair-ended directional Illumina RNA-Seq to investigate the dynamics and relationship between the mRNA and protein expression programs of M. pusilla. Similar to a prior study of the marine cyanobacterium Prochlorococcus, we found significant divergence in the mRNA and proteomics expression dynamics in response to the light:dark cycle. Additionally, expressional responses of genes and the proteins they encoded could also be variable within the same metabolic pathway, such as we observed in the oxygenic photosynthesis pathway. A regression framework was used to predict protein levels from both mRNA expression and gene-specific sequence-based features. Several features in the genome sequence were found to influence protein abundance including codon usage as well as 3' UTR length and structure. Collectively, our studies provide insights into the regulation of the proteome over a diel cycle as well as the relationships between transcriptional and translational programs in the widespread marine green alga Micromonas.

Fig 1. Experimental design and global dynamics of the Micromonas pusilla proteome.

(A) Cartoon modeling cell growth and size during the diel experiment based on details of cell growth provided in [3]. Sampling points are marked with arrows. (B) Distribution across the sampled time points of proteins from the nuclear genome. Observation within at least 50% of proteomics datasets (technical and biological replicates), and identification by ≥2 unique peptides (i.e., matching only one protein in the genome) were required for categorization as “present” and inclusion here. A unique peptide is one that matches only one protein in the genome. A substantial number of proteins were observed across all time points. (C) Relative quantification comparison of global proteome expression over the course of the diel cycle. A progressive shift in the number of proteins exhibiting differential abundance (p<0.05) was observed when progressing from dark to light. The largest increased proteome response was observed at the introduction of light (T4), and this response decreased with time until relative abundance favored the dark condition. (D) KOG class assignments for proteins that exhibited relative abundance differences (p<0.05). For three KOG classes, 100% of proteins increased in abundance when comparing T4 to T1, suggesting that these classes are strongly influenced by circadian regulation in M. pusilla.

Fig 2. Temporal dynamics observed in mRNA and protein expression using the high-confidence set.

(A) Comparison of absolute abundance values for the mRNA and protein expression data (all time points combined) indicates a moderate correlation between the data types (Spearman’s correlation coefficient, R_S = 0.428, p<0.0001). (B) Comparison of the log-ratios relative to T3 indicates a slightly negative correlation (R_S = -0.168, p<0.0001). These contrasting results suggest that while a relationship exists between mRNA and protein expression, there are considerable temporal differences between the respective expression programs.

Fig 3. Comparison of protein and mRNA expression patterns across the time course.

(A) Comparison of the degree of the correlation (Pearson, R_P) between the mRNA and protein expression profiles, per gene (Z-transformed). Less than 10% of the genes considered were correlated over the course of the experiment (using a threshold of 0.75); while 26% were delayed by 1 time point (1 TPT) and 9% by 2 time points (2 TPTs). (B) Concordance of Gene Set Enrichment Analysis (GSEA) of pairwise correlation (as measured by CS_p; see Methods) indicates there is considerable concordance between the expression programs of several key metabolic pathways, such as the Oxygenic Photosynthesis and TCA pathways. Note this is limited to those pathways that are concordant. Concordant pathways from a similar analysis of log-ratios include many of the same critical pathways (Fig M in S1 File). Complete representations of all pathways from the analysis of abundances and log-ratios are also provided (Figs N and O in S1 File). (C) A global comparison of the expression dynamics observed in the mRNA and protein expression programs.

Fig 4. Coverage of the Oxygenic Photosynthesis (OP) Pathway by the joint expression clusters.

(A) Cartoon of the OP pathway, with selected interactions color-coded to indicate cluster membership. Light-dependent reactions are indicated by yellow background. Of the 17 interactions in the pathway, 15 were mediated by genes in the high confidence data set (Tables E-G in S1 File). Note that within the light-independent reactions of the Calvin-Bensen-Bassham Cycle we identified two fructose-bisphosphate aldolase (FBA) proteins, wlab.223910 (Class I, Cluster 6) and wlab.149815 (Class II, Cluster 3), both with predicted chloroplast transit peptides. The Class II FBA of the cyanobacterium Synechococcus shows higher reactivity for sedoheptulose-1,7-bisphosphate than for fructose-1,6-bisphosphate than its Class I FBA [94] and thus, although they have not been experimentally characterized, the M. pusilla FBAs depicted here may also partition within the pathway. (B) Joint mRNA and protein expression profiles of the clusters enriched with OP pathway genes (Clusters 6, 7, & 15). Cluster 6 displays considerable correlation (R² = 0.643) between the mRNA and protein expression patterns, while Cluster 7 and 15 display either marginal (R = 0.161) or inverse (-0.446) correlation. Profiles for Clusters 2 and 3 are show in Figure P in S1 File.

Fig 5. Testing the effectiveness of sequence features as proxies for post-transcriptional control.

(A) Partial correlation matrix of the most correlated and anti-correlated features that were significantly partially correlated with protein expression in all time points, as indicated by bootstrap testing. The left four columns indicate the Spearman correlation between mRNA expression and the features per time point; the right four columns show the partial correlation of the features with the protein expression, when accounting for their correlation with the mRNA expression. As expected, features such as CDS sequence length were anti-correlated with mRNA and protein expression. An anti-correlation observed with the minimum free energy (MFE) in 3’ UTRs was notable, indicating that greater 3’ UTR structure is correlated with protein expression. (B) Comparison of the real versus predicted protein abundances for each sample. The blue dotted line indicates the slope and intercept for a perfect correlation (y = x); while the red dotted lines indicate the 5% and 95% quartiles for the residuals from the predicted protein expression abundances. (C) Categories of sequence features used in two or more of the MARS models. While features from the CDS, mRNA and UTR sequences comprised roughly 38% of the selected features, the majority were proportions of amino acid and amino acid classes in the protein sequences.