Assessment of flux through oleoresin biosynthesis in epithelial cells of loblolly pine resin ducts

Genome-scale models capture the remarkable specialization of epithelial cells for oleoresin biosynthesis in needles of loblolly pine (Pinus taeda L.).


Metabolic Reconstruction
The Pintae_Epi model was created to computationally predict flux distribution through metabolic pathways of epithelial cells of resin ducts in loblolly pine (Pinus taeda L.) needles. As a first step, primary metabolic reactions, including stoichiometry, were adopted from the well-curated Arabidopsis Core Model (Arnold and Nikoloski, 2014) (Scheme 1). Searches against several online databases pertaining to plant metabolism (MetaCyc Version 19.1, http://metacyc.org/ (Caspi et al., 2014) and AraCyc Version 15, https://www.plantcyc.org/databases/aracyc/15.0 (Lamesh et al., 2012)), augmented by traditional literature searches, were then conducted to assess the current knowledge regarding metabolic reactions in loblolly pine that differ from those known to occur in Arabidopsis (e.g., oleoresin biosynthesis). The E.C. numbers for the enzymes catalyzing the reactions represented in Pintae_Epi were transferred from the Arabidopsis Core Model or assigned manually based on the relevant literature on oleoresin biosynthesis. Each gene (and corresponding enzyme/reaction) was assigned to metabolic processes and pathways based on the Gene Ontology annotation of the most similar gene contained in the Arabidopsis genome (https://www.arabidopsis.org/tools/bulk/go/). The same process was employed to develop a metabolic reconstruction for mesophyll cells (Scheme 1).

Incorporation of Subcellular Compartmentation and Transport
Each of the reactions of the metabolic reconstructions was associated with a subcellular compartment (cytosol, plastid, mitochondrion, or peroxisome) based on the known or predicted localization of the enzyme encoded by the putative Arabidopsis ortholog (Hooper et al., 2014) (Scheme 1). The model allows for free exchange of O2 and water into and out of the cell, and contains a single light "import" reaction that is irreversible. Sucrose is the only nutrient that can freely enter into the cell (but not out; acconting for a proton symport mechanism), while others, such as NO3 -, PO4 3-, and SO4 2-, are imported at the expense of ATP and water (in stoichiometrically balanced reactions). All amino acids are generated de novo from intermediates of central carbon metabolism and nitrogen ultimately coming from imported NO3 -. Roughly 40% of all reactions are transport reactions, which are responsible for the exchange of metabolites between compartments, while the remaining 60% of reactions are involved in the biosynthesis or degradation of metabolites.

Refinements
Additional steps involved removing thermodynamically infeasible loops (which ensures a stoichiometrically balanced model) and eliminating reactions that cannot carry flux (orphans not connected to the remainder of metabolic network) (details in Johnson et al., 2017).  The contigs of the loblolly pine epithelial cell and mesophyll cell transcriptome assemblies were compared, using the Blastx algorithm, to the UniProt and TAIR sequences represented in the MetaCyc and AraCyc databases, respectively. Loblolly pine transcripts were associated with reactions from the Arabidopsis Core Model, AraCyc or MetaCyc (in this order of priority), based on global identity, and this information was transferred to our metabolic reconstructions. Reactions in the metabolic reconstruction with no associated transcripts in the appropriate data set were removed. These additional steps generated the metabolic models for loblolly pine epithelial cells (Pintae_Epi; 694 reactions) and mesophyll cells (Pintae_Meso; 722 reactions) (Scheme 1).

Determining resin duct volume
The oleoresin concentration had been determined for bulk needle tissue. For the Pintae_Epi model, it was important to calculate the production of oleoresin by individual resin ducts and their epithelial cells. We therefore determined the volume fraction of needles occupied by resin ducts. Details are described in the Materials and methods section of the main manuscript and a visual description is provided in Scheme 2.

Scheme 2.
Morphometric determination of fractional volumes of resin ducts. A, Use of the "Freehand" function of ImageJ to obtain cross-sectional areas of leaves. B, Tapering of leaves at the distal end. C, Modeling the shape of the leaves tip as elliptical cone. D, Use of the "Oval" function of ImageJ to obtain areas of resin ducts. E, Modeling the shape of a resin duct as cylinder.

Application of Flux Minimization Principle
Gene expression values (as Transcripts per Kilobase Million or TPM) reflecting averages of three replicate transcriptome data sets obtained with isolated epithelial cells of resin ducts and mesophyll cells (NCBI Sequence Read Archive, accession number SRP126587) were mapped onto each reaction of the Pintae_Epi and Pintae_Meso models, respectively. In instances where multiple genes code for an enzyme and/or multiple enzymes are associated with a reaction, the cumulative TPM for each reaction was calculated. To computationally predict intracellular flux distribution, a slight modification of the Expression data-guided Flux Minimization (E-Fmin) algorithm was applied (Song et al., 2014). The E-Fmin algorithm was chosen due its capability to analyze complex metabolic networks, and its ability to consistently predict flux distribution using a stoichiometric network and gene expression values for the enzymes involved in the network.