High-Throughput Metabolic Profiling for Model Refinements of Microalgae

Phenotype Microarray screening of model alga Chlamydomonas reinhardtii
The PM assays test the ability of the alga to utilize various sources of carbon, nitrogen, sulfur, and phosphorus in a minimal medium. In this methods description, we demonstrated how PM assays were used to identify carbon and nitrogen metabolism. Carbon and nitrogen utilization kinetics were measured with a microplate reader. Data were analyzed using PMI software. The summary kinetics of selected PM assay plates (PM01 and PM03) are shown in Figure 1. The "xy plots" display the respiration measurements over time plotted for the 96-well plates' assays, where the y-axis and x-axis represent the values of raw measurements and time, respectively. The data was converted to a heat-map pattern to comparatively analyze the assembly of the kinetic data.

The pipeline of refining genome-scale metabolic network using PM data (Figure 2) illustrates the integration of the high-throughput PM assays with experimental evidence provided by genomic searches can expand a metabolic network model.

To determine the reproducibility of the PM data obtained from PM01 – 04 and PM10 plates, a linear regression was analyzed to plot the data from two independent replicate experiments against each another (Figure 3). Figure 3 shows that the majority of the data were almost similar as they fall on the 45° line, with only a few outliers being present, and their coefficient of determination R² was 0.9. The consistency and reproducibility of the experiments for the alga are verified by this plot.

Identification of new metabolites
The PM assay identified 662 metabolites in seven plates; PM01-PM04 and PM06-PM08, while Gas Chromatography Time-Of-Flight (GC-TOF) had identified 77 metabolites³² (Figure 4). When comparing these two sets with the 1068 metabolites accounted in the iRC1080, only six metabolites overlapped between the three sets, and 149 overlapped between the iRC1080 and the PM. This result demonstrates that the metabolic profiling platform can be a significant source of new metabolic information.

Acetic acid was the only carbon source detected in plate PM01 as a supporting carbon after subtracting the background signal. This finding is consistent with the literature³³ and shows the specificity of the PM assays. The PM assays revealed new sulfur, phosphorus, and nitrogen sources that C. reinhardtii can utilize for growth. The sulfur metabolites were sulfate, thiosulfate, tetrathionate, and DL-Lipoamide. The phosphorus sources were thiophosphate, dithiophosphate, D-3-phospho-glyceric acid, and cysteamine-S-phosphate. The nitrogen source metabolites were L-amino and D-amino acids, including less common amino acids; L-homoserine, L-pyroglutamic, methylamine, ethylamine, ethanolamine, and D,L-α-amino-butyric, and 108 Di-peptides and five Tri-peptides (Table 1). All the 128 newly identified metabolites were searched in KEGG and MetaCyc for their associated reactions, EC numbers, and pathways.

The new 128 metabolites were associated with 49 unique EC numbers. Of these, 15 ECs were linked to their genomic evidence using five sources including; Phytozome Version 10.0.2³⁴ JGI Version 4³⁵, AUGUSTUS 5.0, and 5.2¹⁰ annotations from Manichaikul et al.³⁶ and KEGG¹³. Metabolites without genomic evidence were entered into the Universal Protein Resource website (UniProt, http://www.uniprot.org/)^37,38 where their related sequences were found in other organisms. Homologous sequences in C. reinhardtii were identified by running Position-Specific Iterated BLAST (PSI-BLAST, https://blast.ncbi.nlm.nih.gov/Blast.cgi) from the NCBI website considering only sequences that produced significant alignments (E-value <0.005).

Model refinement
Reactions associated with the new 128 metabolites, along with their encoded genes, were added to the iRC1080 model, expanding the network. The resulting model iBD1106, accounts for 2,444 reactions, 1,959 metabolites, and 1,106 genes (Table 2). The new 254 added reactions were 20 amino acid oxidation reactions, 108 di-peptide hydrolysis reactions, five tripeptide hydrolysis reactions, and 120 transport reactions, encoded by four genes (Cre02.g096350.t1.3, au.g14655_t1, e_gwW.1.243.1, Cre12.g486350.t1.3).

A total of 113 added new reactions account for the hydrolysis of di-peptides and tri-peptides. The hydrolysis of di-peptides and tri-peptides are associated with two genes, one for di-peptides (Cre02.g078650.t1.3) and one for tri-peptides (Cre16.g675350.t1.3).

Concerning sources of phosphorus, a reaction for hydrolysis of cysteamine-S-phosphate into cysteamine and phosphate was added associated with the gene JLM_162926.

The WoLF PSORT tool³⁹ (http://www.genscript.com/psort/wolf_psort.html) and results reported by Ghamsari et al.³⁵ were applied to obtain the specification of the cellular compartments where the new reactions take place. By analyzing protein sequences associated with the new reactions, WoLF PSORT predicted cytosol as the cellular compartment for the reactions.

A generated metabolic model may contain gaps when the biochemical information is incomplete. In such cases, gapFind, A COBRA command, is used. It lists root gaps and allows the identification of new gaps introduction in the new model. The metabolites that cannot be produced in a metabolic model are referred to as root gaps^40,41. Analyzing the root gap indicated that both iRC1080 and iBD1106 models contain the same 91 gaps. This shows that adding the new metabolites and their associated reactions did not introduce any additional root gaps. It should be noted that the phenotyping method used in this protocol does not close root gaps, because the original root gap metabolites lack transport or production mechanisms, which were not addressed in the phenotyping assays. Flux balance analysis was carried out to test the metabolic behavior of iBD1106 under light and dark conditions; (no acetate) and (with acetate), respectively. The algorithm maximizes the biomass precursor reactions for an objective function (biomass growth). To evaluate the involvement of each metabolite to the set objective function, "shadow prices" for all metabolites were calculated. The change in the objective function concerning flux changes of the metabolite defines the shadow price of a metabolite^30,42. The indication of whether a metabolite is in "excess" or is "limiting" the objective function can be determined by shadow price analysis, e.g., biomass production. Negative or positive shadow price values reveal metabolites that, upon addition, will decrease or increase the objective function. Zero values of shadow prices reveal metabolites that will not affect the objective function. The comparison of shadow prices between iBD1106 and iRC1080 in Figure 5 shows that, for most metabolites, a significant change is not observed; though, differences are found in 105 and 70 cases under light and dark growth conditions, respectively. Table 4 includes examples of such metabolites.

Figure 1: Phenotypic microarray profiling of C. reinhardtii. Respiration XY-plots and level plots of the PM01 (Carbon sources; A, C) and PM03 (Nitrogen sources; B, D) assay plates are shown. The figure is an 8×12 array where each cell represents a well-plate and, thus, a given metabolite or growth environment. Within each cell or well representation, curves represent dye conversion by reduction (y-axis) as a function of time (x-axis). PM respiration curves from the CC-503 and blank wells are shown in each cell and are indicated by color (teal color represents blank wells and purple color represents CC-503). The level-plot represents each respiration curve as a thin horizontal line changing color (or remaining unchanged) over time. Heatmap color changes are from light yellow (little to no respiration has taken place) to dark orange or brown (significant respiration has taken place). Metabolites utilized by C. reinhardtii (CC-503) and the blank plates are shown. This figure is from a previously published work by Chaiboonchoe et al.¹² Please click here to view a larger version of this figure.

Figure 2: Genome-scale metabolic network refinement pipeline using PM data. After a new compound tests positive in a PM assay, its Enzyme Commission number (EC), reaction, and pathway are identified from available databases, e.g., KEGG and MetaCyc. Genomic evidence is then extracted from genomic and annotation resources when available and constitutes a link between genotype and phenotype. When direct genomic evidence is unavailable, the protein sequence is identified from the EC numbers, and genetic evidence is identified via PSI-Blast. The reconstructed metabolic network is then refined based on newly identified compounds, but only after a quality control step that entails querying the protein domains using relevant databases. This figure has been modified from previously published work by Chaiboonchoe et al.¹² Please click here to view a larger version of this figure.

Figure 3: Reproducibility of PM tests. The PMI values were collected over a 168 hours period, and the maximum PMI values were plotted for two replicate studies. Each axis represents the maximum PMI values for each study (the x-axis being one replicate study and the y-axis another). Reproduced values are equidistant from each axis. Each point represents a single maximum value. The linear regression was performed by a spreadsheet software, and the resulting coefficient of determination (R²) is shown. This figure has been modified from previously published work by Chaiboonchoe et al.¹² Please click here to view a larger version of this figure.

Figure 4: Venn diagram of metabolites. The Venn diagram enumerates metabolites identified by PM plates, the iRC1080 metabolic model, and Gas Chromatography Time of Flight (GC-TOF) experiments. Each circle indicates the total number of metabolites that exist in each respective method of study. At the same time, the overlapping regions represent the number of metabolites shared between those methods. The iRC1080 metabolic model contains a total of 1,068 unique metabolites. The GC-TOF identified a total of 77 metabolites³², while there are a total of 662 metabolites identified using the PM plates. This figure is from previously published work by Chaiboonchoe et al. ¹² Please click here to view a larger version of this figure.

Figure 5: Shadow prices of metabolites in iRC1080 and iBD1106 under different conditions for biomass maximization. Each circle on the "radar plots" corresponds to a shadow price value, while each line extending from the center of a plot indicates a metabolite. (A) Shadow prices and metabolic behaviors of iRC1080 and iBD1106 under a light growth condition; (B), different metabolic behaviors of iRC1080 and iBD1106 under a dark growth condition. This figure is from previously published work by Chaiboonchoe et al. ¹² Please click here to view a larger version of this figure.

Table 1: List of identified positive substrate utilization metabolites (C, P, S, N) not present in the iRC1080 metabolic model.  *Reaction was not included if no gene was identified.  ¹Phytozome version 10.0.2 (http://phytozome.jgi.doe.gov/pz/portal.html#!info?alias=Org_Creinhardtii).   ²JGI version 4 ³⁵.  ³Augustus version 5¹⁰.  ⁴KEGG (http://www.genome.jp/kegg/kegg1.html).  ⁵JGI version 3.1³⁶.  This table is from previously published work by Chaiboonchoe et al. ¹²

Table 2: Contents of iRC1080 and iBD1106.  This table is from previously published work by Chaiboonchoe et al. ¹²

Table 3: Summary of new reactions in iBD1106. This table is from previously published work by Chaiboonchoe et al. ¹²

Table 4: Example of significant shadow prices for iRC1080 and iBD1106. This table is from previously published work by Chaiboonchoe et al. ¹²