Model simulates algae cell behavior under heterotrophic condition
The final calibrated model adequately simulates the experimental data (Fig. 2). Cell growth, as well as extracellular metabolites such as glucose and glycine are closely simulated. More importantly, total lipids and starch as the main products were also simulated adequately. The model simulation also followed closely the dynamics of intracellular metabolites, which distributed in glycolysis, PPP pathway, TCA cycle as well as energy metabolism. These results thus confirm the model structure as well as its calibrated kinetic parameters to simulate algae cells metabolism and products accumulation dynamics. Indeed, in this work, both the experimental data and model simulations show glycolysis and PPP pathways being more affected by glucose supply while TCA metabolism, which is fed by both carbon and nitrogen metabolisms, seems more robust to perturbations such as extracellular glucose depletion. The intracellular nutrition storage pool in the form of TCA cycle metabolites seemed to maintain biomass in the later growth phase. From both simulation and experimental data, algae biomass still accumulates while glycine and other amino acids pool (AA) reached values under the detection limits. However, this phenomenon was only observed for nitrogen sources since cell biomass growth stopped simultaneously to glucose depletion. This intracellular nutrients management phenomenon has also been modeled and proved in phytoplankton and plant cells [23]. Where the model premises were based on observations that cell growth continued after the exhaustion of external nitrogen pool, being then supported by the consumption of intracellular nitrogen pools such as chlorophyll molecules.
Dynamic metabolic flux analysis reveals lipid and cellular metabolic behavior in Chlorella protothecoides
Considering all the above, it is thus clear that the model structure allows simulating heterotrophic Chlorella protothecoides cell behavior. The model was then taken as an in silico tool and perform a dynamic metabolic flux analysis estimating flux distribution. A dynamic metabolic flux analysis was performed from model simulation (Fig. 3). Looking at glucose flux (VHK), glycine flux (VGHMT) as well as cell specific growth rate (Vgrowth) (Fig. 3a), it is clear that cell growth proceeds simultaneously to carbon source uptake, but not proportionally to nitrogenous source uptake. Interestingly and as previously discussed for glycine concentration, glycine flux (VGHMT) ceased more than 24 h prior to growth cessation. Fluxes of PPP pathway and starch synthesis (Fig. 3c, d) originate from G6P are partially affected in some extent by glucose flux (Fig. 3a). For instance, model simulation VPGM flux showed being reversible from accumulation to decomposing at around day 3, where glucose reached depletion. This suggests that starch, which is an intracellular carbon storage pool, rapidly responds to a low carbon source level threshold, contributing providing continuous carbon flow feeding cell metabolism and maintenance. However, as an alternative carbon storage pool, net lipid flux shows a quasi-constant rate, composed of a synthesis flux (VFASN) that was slightly affected at glucose depletion and two catabolic fluxes (VLipase and VGPAT) which stayed quite constant (11.87–12.04 mmol gDW−1 day−1 and 0.05 mmol gDW−1 day−1 respectively) (Fig. 3e). Interestingly, TCA cycle fluxes (VISOD, VSDH) (Fig. 3f) exhibited a minimum value at glucose depletion, for increasing thereafter. As previously mentioned, the TCA cycle is closely related to carbon and nitrogen metabolism, it seems after carbon source depletion, some carbon and nitrogen dependent compounds (such as pigment) stopped synthesis, which squeezed the intracellular nitrogen source flux goes to TCA cycle. As in heterotrophic culture, C. protothecoides represents yellowish because of carotene content is higher than chlorophyll [24]. However, after glucose depletion, we found the color of culture turns from yellow to green, and the carotene gets to degrade. Moreover, CS flux dynamics closely follows the lipid synthesis flux although it’s quite low compared with lipid synthesis. As CS is competing the same substrate from FASN, the flux of these two enzymes are quite dependent on the concentration of AcCOA, which is in agreement from model prediction.
A closer view of flux rates were estimated at 48 h before glucose depletion in the exponential growth phase. For comparison purposes, all the flux values were normalized to an uptake flux of 100 mmol g−1DW h−1 glucose (Fig. 4). Flux results agree with that reported in the previous reports [13, 18]. For example, in [18], who performed a flux balance analysis at steady state for Chlorella sp. under heterotrophic condition, with a GPI flux of 66.28 mmol g−1 DW h−1 (leading to glycolysis), G6PDH of 12.15 (leading to PPP pathway) and PGM of 13.83 (leading to starch), compared to 49.8 mmol g−1 DW h−1, 32.04 and 17.3 respectively in our work. The total flux to G6P obtained from our model is of 92.26 mmol g−1 DW h−1 compared that of 99.22 in literature. The net flow From F6P to GD (PFK minus FBPase) was of 73.47 mmol g−1DW h−1 compared to 70.35 mmol g−1 DW h −1 (from F6P to GAP), and the flux from GD to PEP was of 150.51 mmol g−1 DW h−1 in our model versus 148.25 (from GAP to PEP) in literature. The fluxes of nucleotides synthesis (from RX to ADP) was of 1.36 mmol g−1DW h−1 compared to 0.67 mmol g−1DW h−1 (from PRPP to DNA and RNA). Biomass synthesis rate was of 9.19 compared to 7.36 in [18].
Furthermore, downstream fluxes to AcCOA, the sum of the downstream lipid and TCA cycle flux was of 74.01 mmol g−1DW h−1 compared to 86.14 g−1DW h−1 in literature. Although a similar total flux around the TCA cycle was obtained, with 73.93 mmol g−1DW h−1 at lipid branch and 0.08 at TCA branch, different results were reported in [18] with 81.21 mmol g−1DW h−1 at TCA cycle and 4.82 mmol g−1DW h−1 at lipid branch. This discrepancy may rely on a high lipid level (13.13% DW) in our cell culture compared to that in [18] (1% DW). Differences in culture conditions may be involved as well. Meanwhile, this difference of high lipid synthesis flux and low TCA cycle fluxes were also found in Wu’s work, where a 13C metabolic flux analysis was accompanied with flux balance analysis in Chlorella protothecoides [13]. Thus, from both the dynamic and steady state flux analysis, high lipid content in Chlorella is mainly due to low TCA split-flow. In our result, lipid was fully accumulated at 72 h, so the metabolic fluxes at 72 h was also extracted from the dynamic flux profile. Along with the glucose deleption, most of the metabolic fluxes decreased in different extent, some even decreased up to 78.87% (EMP pathway fluxes). However, the re-arrangement of flux distribution from starch and energy catabolism to lipid synthesis was obvious. At 48 h, starch and ATP were accumulating (VPGM is 13.83 mmol g−1DW h−1) along with glucose assimilation. However, after glucose deleption at 72 h, starch was catablising (VPGM is − 23.25 mmol g−1DW h−1) as a storage pool and ATP was also consuming (V PPRiBP is − 0.01 mmol g−1DW h−1) to maintain other metabolism. Lipid synthesis flux (VFASN), decreased a little bit from 73.93 to 63.89 mmol g−1DW h−1 (decreased 13.51%), which was less impacted by glucose deleption compared with other fluxes on EMP pathway. This may due to a constant feeding flux converting from starch and ATP catabolism. Therefore, the energy stored in starch and ATP seams to be converted to a more stable storage pool as lipid after glucose deleption.
We also looked at the major carbon distribution before glucose depletion. Model simulations show that the glucose uptake rate (VHK) and the glycolytic fluxes went down to a very low level after day 2.6, we have thus analyzed their related flux ratios only before glucose depletion (< 2.6 days). First, we evaluated that 6% of the glucose flux contributes to biomass synthesis and growth (Vgrowth-to-VPK ratio) (Fig. 5a), a value comparable to the literature with 3.9% [25]. Within the same range, 8% (8.323–8.325%) of the glucose flux feed lipid synthesis (VFASN–VLipase to VPDH ratio) (Fig. 5b). However, as a main product contributing to biomass, the lipid catabolism-to-biomass synthesis and growth ratio (VFASN –VLipase –VGPAT to Vgrowth) shows two successive constant values at around 60% increasing at 80% at mid-exponential growth phase (1.5 d) (Fig. 5e). Model simulations also suggest that around 1% of the glucose flux goes to starch synthesis (VPGM to VHK) (Fig. 5c), and that 15% to 7% of the glucose flux feed nucleotides synthesis (VPPRiBP to VG6PDH) (Fig. 5d). Concerning the PPP pathway activity, around 12% of the glucose uptake flux flow into the pentose phosphate pathway (VG6PDH to VHK) (Fig. 5f). Therefore, the dynamic metabolic flux analysis give our lights of the carbon distribution in Chlorella protothecoides.
Parameter sensitivity showed biological significance on metabolic kinetics
A sensitivity analysis on model parameters showed flux maximum rate constants (Vmax,i) to be more sensitive than affinity constants (Km,i). For the final calibrated model 21 parameters, 15 maximum flux rates and 6 enzyme affinity constant (Fig. 6), out of 77 revealed greater sensitivity, defined as affecting the objective function of more than 10% when applying a − 70% to + 150% parameter value change around its optimized value.
The most sensitive parameters are Vmax,HK and Vmax,GHMT, which are both at the entrance of the major carbon and nitrogen sources; Vmax,GPI, Vmax,PGM and Vmax,PDH, which refer to fluxes at the intersection of glycolysis, starch and lipid metabolisms are also highly sensitive, while PPP pathway (Vmax,TK) and TCA cycle parameters, showed a low sensitivity level. Since the major intersections from glycolysis including starch synthesis, PPP pathway which finally leads to nucleic acids, and TCA cycle which is related to protein synthesis, and the fatty acids synthesis which leads to lipids. The sensitive flux parameters on starch and lipid metabolism suggested that starch synthesis is the major competing pathway that affect lipid accumulation in C. protothecoides. It has also been reported in Chlamydomonas reinhardtii, when starch biosynthesis is blocked (sta6 mutant), the lipid content could be greatly boost, some can reach up to 30-fold [26]. Except for starch flux sensitivity, Vmax,FASN, Vmax,Lipase and Vmax,GPAT are also sensitive. As Vmax,FASN and Vmax,Lipase are related to lipid synthesis and degradation, they are responsible for the balance of cellular lipid pool. Meanwhile, Vmax,GPAT is in charge of providing glycerone-phosphate as the neutral lipid skeleton. The sensitivity of these fluxes gave us light on the genetic strategy for lipid yield promotion. Interestingly, there are two highly sensitive affinity constants (km,growth_lipid and km,Lipase_lipid), referring to the importance of lipid for cell biomass growth. Algae cell is a great platform accumulating lipids, some algae species could accumulate lipids up to 70% of their biomass. In C. protothecoides, the lipid content could reach 36% under heterotrophic condition, although in our work lipid content only reached 13% DCW, the sensitive of lipid affinity constant to growth suggest a huge potential to optimize the enzyme activity. Some reactions or pathways (i.e. their kinetic parameters) such as the maximum specific growth rate, PPP pathway and TCA cycle, showed a low sensitivity level, which suggest these are robust pathways. Final parameter values and the 95% confidence intervals for the sensitive parameters are shown in Table 4. They are all within ranges found in the BRENDA databank.