Genome-Scale Metabolic Reconstruction and in Silico Perturbation Analysis of the Polar Diatom Fragilariopsis cylindrus Predicts High Metabolic Robustness

Abstract

Diatoms are major primary producers in polar environments where they can actively grow under extremely variable conditions. Integrative modeling using a genome-scale model (GSM) is a powerful approach to decipher the complex interactions between components of diatom metabolism and can provide insights into metabolic mechanisms underlying their evolutionary success in polar ecosystems. We developed the first GSM for a polar diatom, Fragilariopsis cylindrus, which enabled us to study its metabolic robustness using sensitivity analysis. We find that the predicted growth rate was robust to changes in all model parameters (i.e., cell biochemical composition) except the carbon uptake rate. Constraints on total cellular carbon buffer the effect of changes in the input parameters on reaction fluxes and growth rate. We also show that single reaction deletion of 20% to 32% of active (nonzero flux) reactions and single gene deletion of 44% to 55% of genes associated with active reactions affected the growth rate, as well as the production fluxes of total protein, lipid, carbohydrate, DNA, RNA, and pigments by less than 1%, which was due to the activation of compensatory reactions (e.g., analogous enzymes and alternative pathways) with more highly connected metabolites involved in the reactions that were robust to deletion. Interestingly, including highly divergent alleles unique for F. cylindrus increased its metabolic robustness to cellular perturbations even more. Overall, our results underscore the high robustness of metabolism in F. cylindrus, a feature that likely helps to maintain cell homeostasis under polar conditions.

Keywords: arctic; flux balance analysis; gene deletion; metabolic network; reaction deletion; systems biology.

Figure 1

Local sensitivity analysis of the model output (growth rate) to an increase (red bars) or decrease (blue bars) by an arbitrary factor of 40% for each of the 69 model parameters taken independently. The percentage of variation of the growth rate is given on the y-axis (on a log scale) as a function of changes in parameter values (see Table 1 and Section SI.1 for details and units of parameters). (1) chla: chlorophyll a; (2) chlc: chlorophyll c; (3) Fucoxan: fucoxanthin; (4) BCarot: beta-carotene; (5) DD: diadinoxanthin; (6) GCprop: proportion of GC in DNA; (7) Genomesize: total genome size; (8) RNADNA: RNA:DNA ratio; (9–27) lip1 to lip19 refers to lipids tabulated in supplementary excel Table Supp6; (28) ala: alanine; (29) arg: arginine; (30) asn: asparagine; (31) asp: aspartate; (32) cys: cysteine; (33) gln: glutamine; (34) glu: glutamate; (35) gly: glycine; (36) hist: histidine; (37) ile: isoleucine; (38) leu: leucine; (39) lys: lysine; (40) met: methionine; (41) phe: phenylalanine; (42) pro: proline; (43) ser: serine; (44) thr: threonine; (45) trp: tryptophane; (46) tyr: tyrosine; (47) val: valine; (48) totalProt: total protein; (49) totalCarb: total carbohydrate; (50) totalLip: total lipid; (51) totalDNA: total DNA; (52) totalRNA: total RNA; (53) totalPigm: total pigment; (54) PropGlu: proportion of glucan; (55) PropTGA: proportion of triacylglycerol; (56) glu: glucose; (57) galac: galactose; (58) mann: mannose; (59) xylu: xylulose; (60) arab: arabinose; (61) fuco: fucose; (62) rhamn: rhamnose; (63) glucur: glucuronate; (64) mannSul: mannose-sulfate; (65) CN: total C:N ratio; (66) NP: total N:P ratio; (67) Ps: C uptake rate; (68) Ccell: carbon per cell; and (69) DWcell: dry weight per cell.

Figure 2

Global sensitivity analysis of the model output (growth rate) to a change by 40% of the 69 model parameters using the Morris method. The relative influence of each parameter on the model outcome (growth rate of F. cylindrus in the light) is quantified with a sensitivity index called “mean of elementary effect”, or mean EEs on the X-axis. The relative strength of interactions between parameters can also be visualized with the standard deviation of elementary effects (standard deviations of EEs) on the Y-axis. All points are numbered following the order of appearance of parameters in Figure 1 from left to right. Note that the parameters #7, #30, and #33 are not shown, since they have standard deviations of EEs and mean EEs smaller than 10⁻¹¹.

Figure 3

Cumulative percentage of active reactions, for which deletion of each reaction (taken independently) decreased the growth rate by a proportion equal to the value on the Y-axis or lower. Calculation of intracellular flux distribution and growth rate after single reaction deletion were performed using flux balance analyses (FBA, blue line) or quadratic minimization of metabolic adjustments (MOMA, red line). For instance, the curve describing the reaction deletion with FBA means that a single reaction deletion for around 15% of active reactions does not inhibit growth rate. For around 32% of all active reactions, single reaction deletion decreases growth rate by less than 1%. For around 68% (100−32) of all active reactions, single reaction deletion abolishes growth rate, so the cumulative sum on the X-axis gives 100. The other curves can be interpreted analogously.

Figure 4

Relative frequency histogram of total degree (A), in-degree (B), and out-degree (C) for the metabolites involved in sensitive reactions (sensitive metabolites) or robust reactions (robust metabolites). “Total degree” refers to the total number of reactions connected per metabolite, “in-degree” refers to the number of reactions producing a metabolite, and “out-degree” is the number of reactions using a metabolite as the substrate. The X-axis is set to values smaller than 20. Full data are available in Supporting Information.

Figure 5

Cumulative percentage of active genes, for which deletion of each reaction (taken independently), decrease the growth rate by a proportion equal to the value on the Y-axis or lower. Calculation of intracellular flux distribution and growth rate after single gene deletion were done using flux balance analyses (FBA, blue line) or quadratic minimization of metabolic adjustment (MOMA, red line). The gene deletion analysis was performed without (A) or with (B) 53 allelic variants.

Figure 5

Cumulative percentage of active genes, for which deletion of each reaction (taken independently), decrease the growth rate by a proportion equal to the value on the Y-axis or lower. Calculation of intracellular flux distribution and growth rate after single gene deletion were done using flux balance analyses (FBA, blue line) or quadratic minimization of metabolic adjustment (MOMA, red line). The gene deletion analysis was performed without (A) or with (B) 53 allelic variants.