Serine synthesis through PHGDH coordinates nucleotide levels by maintaining central carbon metabolism

Phosphoglycerate dehydrogenase (PHGDH) catalyzes the committed step in de novo serine biosynthesis. Paradoxically, PHGDH and serine synthesis are required in the presence of abundant environmental serine even when serine uptake exceeds the requirements for nucleotide synthesis. Here, we establish a mechanism for how PHGDH maintains nucleotide metabolism. We show that inhibition of PHGDH induces alterations in nucleotide metabolism independent of serine utilization. These changes are not attributable to defects in serine-derived nucleotide synthesis and redox maintenance, another key aspect of serine metabolism, but result from disruption of mass balance within central carbon metabolism. Mechanistically, this leads to simultaneous alterations in both the pentose phosphate pathway and the tri-carboxylic acid cycle, as we demonstrate based on a quantitative model. These findings define a mechanism whereby disruption of one metabolic pathway induces toxicity by simultaneously affecting the activity of multiple related pathways.

For simplicity, the TCA cycle is represented as the PYR_TCA-1 and PYR_TCA-2 reactions, in which carbon atoms from pyruvate feed the TCA cycle through reactions catalyzed by PDH and PC. For the metabolites ribose-5-phosphate (R5P), serine (Ser), glycine (Gly), aspartate (Asp), inosine monophosphate (IMP) and uridine monophosphate (UMP), contribution from unlabeled sources such as the salvage pathways were also considered in this model. Coefficients in the biomass synthesis flux were evaluated from the molecular composition of dry cell weight from literature 1 ( Table 2).  R5P_input ( Table 2 that can be derived from this precursor. These coefficients are listed in Table 3. (3) The subscript of each EMU refers to the carbon atoms included in that EMU. For example, Asp is the EMU with the 2 nd , 3 rd and 4 th carbon atoms in aspartate. EMUs with the suffix 'input' refer to EMUs in unlabeled sources (e.g. salvage pathways) that also contribute to the production of the corresponding metabolite. denotes the metabolic flux through a reaction. Experimentally measured MIDs of IMP, UMP, G6P, R5P, PYR, OAA/Asp, 3PG, Ser and Gly were used in estimating the metabolic fluxes. CAP, CO2 and all 'input' EMUs were assumed to be unlabeled, in which each carbon atom has the probability 0.01109 to contain 13 C according to the natural isotope abundance. MIDs of Asp ,

G6P
, PYR , PYR , Asp , Ser and 3PG were estimated from the experimentally measured MIDs of the metabolites directly containing these EMUs based on the assumption that each carbon atom has the same probability to be labeled. The MID of 1CTHF was estimated from the MIDs of 3PG and Ser_input based on the same assumption.
Solving for the relative metabolic fluxes from the EMU model The EMU model was solved in two steps to determine the metabolic fluxes from mass isotopomer distributions (MIDs). In the first step, flux ratios at branch points were directly computed from MIDs of the EMUs linked to the respective branch point. The flux ratios were then used in combination with the stoichiometric matrix to solve all relative metabolic fluxes in the model.
Step 1. Calculate flux ratios at branch points: For a given branch point with two upstream fluxes 1 and 2 , suppose that 1 and 2 and are the MID vectors calculated from the convolution of MIDs of all EMUs directly converted to the related branch point metabolite by 1 and 2 , respectively, and 3 is MID of the metabolite at the branch point (i.e. 3 can be derived from either 1 or 2 ), then the isotopomer balance equations for this branch point are: 1 1 + 2 2 − ( 1 + 2 ) 3 = 0 (8) Let = 1 1 2 be the flux ratio at this branch point, we have: Which is identical to: ( 1 − 2 ) = 3 − 2 (10) Because the dimension of the vector ( + 1) is always larger than 1, the equations above are overdetermined. The equations were therefore solved using by least-squares minimization. This was implemented using the function numpy.linalg.lstsq in the Python package NumPy.
We note that most branch points in our model consist of one input flux from an unlabeled source and a flux synthesizing this metabolite from its precursor. Metabolites at these branch points include IMP, UMP, R5P, Asp, Ser and Gly. Thus, the equations for the corresponding flux ratios are: Step 2. Calculate relative fluxes from flux ratios and stoichiometric matrix: Given the flux ratios estimated in the previous step, all relative fluxes can therefore be determined by solving the equations below: