Constructing a Biosynthetic Pathway in Reverse
By: Carol A. Rouzer, VICB Communications
Published: March 28, 2014
A bioretrosynthetic approach, starting with the last step, yields an efficient biocatalytic pathway for the synthesis of didanosine.
Although biocatalytic processes are frequently used to synthesize complex molecules, it is rare that they are employed in multi-step pathways, particularly in the case of unnatural compounds. The difficulty of tailoring multiple enzymes to catalyze subsequent steps in a pathway involving artificial substrates can be daunting, and the lack of paradigms to guide the process magnifies the challenge. Now, inspired by the retrograde evolution hypothesis, Vanderbilt Institute of Chemical Biology (VICB) members Brian Bachmann and Tina Iverson propose the bioretrosynthetic approach to complex molecules. Successful application of this approach produced a highly efficient biocatalytic pathway for the synthesis of didanosine [W. R. Birmingham, et al. (2014) Nat. Chem. Biol., published online March 23, doi:10.1038/nchembio.1494].
Didanosine (2′,3′-dideoxyinosine) is an off-patent HIV-1 reverse transcriptase inhibitor. It serves as a prototype for a widely prescribed class of drugs for which manufacturing costs are substantial, so an efficient biocatalytic pathway could significantly reduce the cost of therapy. A hypothetical biosynthetic pathway, starting with 2,3-dideoxyribose, could readily be envisioned, based on the biosynthesis of inosine from ribose (Figure 1).
Figure 1. Bioretrosynthetic approach to a pathway for didanosine synthesis. (a) Biosynthesis of inosine from ribose, the prototype pathway. (b) Three proposed steps in the synthesis of didanosine from dideoxyribose. The enzymes are ribokinase RK, 1,5-phosphopentomutase (PPM), and purine nucleoside phosphorylase (PNP). In the retro-evolution approach, the steps are optimized in reverse order. Image reproduced by permission from Macmillan Publishers Ltd, from W. R. Birmingham, et al. (2014) Nat. Chem. Biol., published online March 23, doi:10.1038/nchembio.1494. Copyright 2014.
The retrograde evolution hypothesis, [N.H. Horowitz (1945) Proc. Natl. Acad. Sci. U.S.A., 31, 153] proposed that biosynthetic pathways evolved by gene duplication and optimization beginning with the final step and working forward. Bachmann and Iverson realized that designing a biosynthetic pathway in reverse order would offer a distinct advantage over the common approach of beginning optimization with the first step. Specifically, a single assay for formation of the final product could be used for development of every step. Key to success of this approach is the availability of a robust assay for final product formation that provides reliable data as stepwise progress is made towards the beginning of the pathway and the reaction mixture becomes increasingly complex.
The bioretrosynthetic approach required that the investigators begin with optimization of the final step in the proposed pathway, the conversion of 2,3-dideoxyribose 1-phosphate and hypoxanthine to didanosine through the action of purine nucleoside phosphorylase (PNP) (Figure 1), and that they develop a robust assay for this step of the pathway. This process was facilitated by prior work in collaboration with VICB member Jens Meiler that had optimized human PNP for the reverse reaction, phosphorolysis of didanosine. The researchers quickly tested the optimized enzyme, PNP46D6 (containing Y88F, M170T, G4E, Q172L, and T177A mutations), for its ability to catalyze didanosine formation. They found that, compared to wild-type PNP, PNP46D6 exhibited a 16-fold increase in the rate of didanosine formation accompanied by an 8.7-fold decrease in the rate of formation of inosine. This activity was strong enough that the PNP46D6 enzyme could be used to design a colorometric assay for didanosine synthesis, based on the consumption of hypoxanthine.
With the third and final step of the pathway secured, the investigators turned to the optimization of the second step, the conversion of 2,3-dideoxyribose 5-phosphate to 2,3-dideoxyribose 1-phosphate by 1,5-phosphopentomutase (PPM). Beginning with the PPM from Bacillus cereus, they confirmed a low level of didanosine formation in a tandem assay including PNP46D6. Co-crystal structures of PPM with ribose 5-phosphate and 2,3-dideoxyribose 5-phosphate revealed amino acids, Ser154, Val158, and Ile195, that were involved in substrate binding. The researchers targeted these residues in saturation mutation studies. The results yielded an S154G mutant enzyme with a 49-fold increase in selectivity for didanosine formation and a V158L mutant with an 880-fold redution in catalytic efficiency for ribose 5-phosphate. An enzyme bearing both mutations exhibited a marked reduction in kcat for both substrates, suggesting that the benefits of each mutation were mutually exclusive.
Random mutagenesis studies starting with either the S154G or the V158L mutants yielded four promising variants, which were then further modified by random recombination. The resulting variant with the best activity, PPM4H11, carried four mutations, V158L, T81I, V190K, and the silent mutation P361P. This enzyme exhibited 325% greater activity for didanosine production and retained 11% of the wild-type enzyme’s activity for formation of inosine. Although the investigators were unable to obtain co-crystal structure data for PPM4H11 with ribose 5-phosphate or 2,3-deoxyribose 5-phosphate, structural data for the unliganded enzyme indicated that the mutations caused a rotation of the cap domain, which contains substrate binding residues relative to the core domain, which contains the catalytic residues. Molecular modeling suggested that this rotation alters the orientation of the substrate with regard to the catalytic residues, possibly accounting for the change in substrate selectivity.
Having substantially improved the selectivity and catalytic efficiency of the second and third pathway enzymes for didanosine formation, the investigators now turned to the first step in the proposed pathway, the conversion of dideoxyribose to dideoxyribose 5-phosphate through the action of RK. However, addition of this step to their tandem assay quickly revealed a problem. The high concentrations of ATP required to sustain RK’s activity for an extended period of time were inhibitory to PPM. The researchers solved this problem by adding an ATP regenerating system, including phosphoenolpyruvate, pyruvate kinase (PK), and adenylate kinase (AK), to the assay reaction mixture. This allowed a lower concentration of ATP to sustain the reaction over an extended period of time. With the modified assay in hand, they began their work with the RK from Escherichia coli, which had rudimentary activity with dideoxyribose. A co-crystal structure of this enzyme with ribose indicated that Asp16 plays a key role in binding the 2- and 3-hydroxyl groups, both of which are absent in dideoxyribose (Figure 2). The investigators hypothesized that mutation of this residue would markedly reduce activity with ribose and possibly favor activity with dideoxyribose. This proved to be true, as a D16A mutant enzyme exhibited a 20-fold increase in didanosine production.
Figure 2. Co-crystal structure of E. coli RK with ribose showing the interaction of the substrate with Asp16 in the enzyme’s active site. Image reproduced by permission from Macmillan Publishers Ltd, from W. R. Birmingham, et al. (2014) Nat. Chem. Biol., published online March 23, doi:10.1038/nchembio.1494. Copyright 2014.
With enzyme variants exhibiting substantially increased activity for didanosine production now available to catalyze all three steps, the investigators assessed the efficiency of their catalytic pathway (Figure 3). Incorporation of RKD16A, PPM4H11, and PNP46D6, along with PK and AK, in the presence of dideoxyribose, phosphoenolpyruvate, ATP, and hypoxanthine led to a 50-fold increase in the production of didanosine and a 200-fold decrease in the production of inosine from ribose when compared to the comparable reaction mixture containing wild-type enzymes and ribose. This astounding 9,500-fold increase in selectivity for didanosine production was unexpected in light of the selectivity increases of the individual enzymes. Since the largest change in selectivity resulted from addition of RKD16A, the researchers investigated its activity more closely. They were surprised to discover that the D16A mutation had conveyed an unusual ribose 1-kinase activity to RK, allowing the enzyme to effectively catalyze the first two steps of the pathway, obviating the need for PPM4H11. Indeed, biocatalytic synthesis of didanosine using just RKD16A and PNP46D6, along with PK and AK, resulted in a 70-fold increase in the rate of didanosine formation and an 8,300-fold overall increase in selectivity for dideoxyribose over ribose when compared to wild-type enzymes.
Figure 3. Evaluation of the activity of wild-type versus optimized enzymes for the biosynthesis of inosine versus didanosine for the third pathway step (a), the second and third step (b) and the full pathway (c). Image reproduced by permission from Macmillan Publishers Ltd, from W. R. Birmingham, et al. (2014) Nat. Chem. Biol., published online March 23, doi:10.1038/nchembio.1494. Copyright 2014.
The investigators concluded that their bioretrosynthetic approach to the construction of new pathways for the synthesis of unnatural molecules is a new paradigm worthy of further exploration. They point out that the more conventional approach of investigating each enzyme individually may have failed to reveal the ribose 1-kinase activity of RKD164A because the alternative products, ribose 1-phosphate and ribose 5-phosphate would likely not be distinguished in a traditional kinase assay. They note that one disadvantage of their approach is that each step must be optimized individually and serially, while the conventional approach allows optimization of enzymes for multiple pathway steps to occur simultaneously. However, they suggest that the advantage of using a single enzyme assay that directly measures production of the desired final product outweighs this disadvantage.