Report

A Functional Protein Chip for Pathway Optimization and in Vitro Metabolic Engineering

See allHide authors and affiliations

Science  16 Apr 2004:
Vol. 304, Issue 5669, pp. 428-431
DOI: 10.1126/science.1096920

Abstract

Pathway optimization is difficult to achieve owing to complex, nonlinear, and largely unknown interactions of enzymes, regulators, and metabolites. We report a pathway reconstruction using RNA display–derived messenger RNA–enzyme fusion molecules. These chimeras are immobilized by hybridization of their messenger RNA end with homologous capture DNA spotted on a substrate surface. Enzymes thus immobilized retain activity proportional to the amount of capture DNA, allowing modulation of the relative activity of pathway enzymes. Entire pathways can thus be reconstructed and optimized in vitro from genomic information. We provide concept validation with the sequential reactions catalyzed by luciferase and nucleoside diphosphate kinase and further illustrate this method with the optimization of the five-step pathway for trehalose synthesis.

The properties of metabolic pathways uniquely depend on the relative activities of all the enzymes that they comprise. As such, metabolic pathways can be optimized by modulating the relative expression of the corresponding genes. Pathway optimization through the introduction of genetic controls is indeed a central tenet of metabolic engineering (1). However, despite recent advances in pathway optimization and metabolic engineering (29), this remains a demanding task owing to the complexity of metabolic pathways. Methods such as single-gene overexpression or overexpression of all pathway genes are optimal strategies only in special cases. Mathematical methods can aid in pathway optimization provided that satisfactory kinetic models are available. However, their use has been limited owing to the lack of reliable in vivo kinetic models. Hence, pathway optimization must rely on combinatorial experimental methods, whereby the relative amounts (or activities) of the pathway enzymes are altered and the effect on the pathway performance is measured. However, this is difficult to achieve because in general, all pathway enzymes are not available in pure and active form.

mRNA-protein fusions, originally developed for use in RNA display (10), are molecules linking covalently an expressed mRNA and its translated protein product. These fusion molecules (fig. S1A) comprise two distinct parts. The mRNA part carries genetic information that can be used as a specific tag to capture the fusion molecule by a homologous capture DNA, and the protein carries potential functionality. If the capture DNA is immobilized on a solid support, it could be used as an anchor for separating the entire chimeric molecule from a mixture of similar molecules and attaching it on the support (fig. S1B).

Weng et al. (11) demonstrated that mRNA-protein fusion molecules could be hybridized on a DNA microarray to build a protein microarray. Using three different types of proteins and their tagged antibodies, they showed that mRNA-protein fusion molecules could specifically bind to their corresponding capture DNA. We used this platform for catalytically active proteins in order to reconstruct metabolic pathways whose enzymes were not readily available. We generated all mRNA-enzyme fusion molecules in parallel, in a single in vitro translation step. The relative amounts of the different types of fusion molecules immobilized on the wells of a 96-well microplate were controlled by the amounts of unique capture DNA molecules designed for each target enzyme. All seven mRNA-enzyme fusions examined retained their catalytic activity when immobilized. This facilitated the experimental investigation of numerous combinations of relative enzyme activities and ultimately the optimization of the five-step pathway for trehalose synthesis.

We first examined the efficiency of cross-linking of DNA capture molecules on the support using fluorescein-labeled capture DNA. Fluorescein-labeled oligonucleotides coding luciferase capture DNA (synthesized from Integrated DNA Technologies, Coralville, IA) were loaded on wells with surfaces coated by polylysine, and cross-linking was carried out in the ultraviolet (UV) cross-linker (Stratagene, La Jolla, CA) (12). After cross-linking, wells were washed twice with 200 μl of 1× phosphate-buffered saline solution. Significant fluorescence was detected in the wash solution only when more than 100 μg of DNA was used, indicating that polylysine-coated microwells can immobilize up to 100 μg of DNA. Subsequently, fluorescein tags were released from the surface after treatment with deoxyribonuclease, and their fluorescence was measured. Almost all of the fluorescein-labeled capture DNA loaded was recovered (Fig. 1A), suggesting ∼100% cross-linking efficiency up to 100 μg of capture DNA loaded per well. Finally, the activity of the hybridized luciferase fusion molecule compared well with that of luciferase enzyme in solution, which shows that there is no significant activity loss by immobilization (Fig. 1B).

Fig. 1.

Efficiency of cross-linking and hybridization. (A) Cross-linking efficiency. Fluorescein-labeled luciferase capture DNA was cross-linked in UV cross-linker (Stratagene, La Jolla, CA). Fluorescence emitted by fluorescein in the well was measured by Fusion microplate reader (Perkin Elmer, Wellesley, MA). Cross-linking efficiency was estimated as the ratio of the fluorescence in the well after cross-linking to the fluorescence of total DNA added to the well. (B) Hybridization efficiency. Hybridization was carried out in wells with 5 μg of luciferase capture DNA each, as described in fig. S1. For enzyme in solution, fusion molecules were incubated for 30 min at room temperature in wells free of capture DNA. After hybridization, luciferase activities were measured (15).

Next, we investigated the sequential reaction catalyzed by nucleotide diphosphate kinase (NDK) and luciferase, as a model system (13): Math Math Math Math Math mRNA-enzyme fusion molecules were synthesized in the modified in vitro translation system (12) from rabbit reticulocyte lysate and were loaded in the wells of a 96-well microplate, each containing different amounts of DNA capture molecules for the two enzymes, NDK and luciferase (12).

Figure 2 summarizes the results obtained when different amounts of capture DNA in the wells are used with varying volumes of load solution obtained from the in vitro translation of individual fusion molecules (henceforth termed “solution”). For the same solution volume (i.e., same amount of chimeric molecules), Fig. 2A shows that the activity of each of the above reactions increases and then saturates as the amount of capture DNA spotted on the wells increases. Apparently, 0.3 μg of capture DNA is sufficient to hybridize all fusion molecules produced by in vitro translation of luciferase or NDK. When the amount of capture DNA in the wells is fixed at the maximum of 0.3 μg, the activity of each enzyme increases linearly (up to 40 μl of solution) with the amount of solution loaded on the well (Fig. 2B). To investigate the specificity of hybridization when fusion molecules of both enzymes were present in the solution, we loaded mixtures of fusion molecules (all translated in a single step) on wells containing different combinations of the two types of capture DNA. Figure 2C shows the activity of luciferase in wells each containing the same amount of solution but varying amounts of luciferase capture DNA molecule and the same amount (0.3 μg) of NDK capture DNA, and the activity of NDK in wells each containing varying amounts of NDK capture DNA and the same amount (0.3 μg) of luciferase capture DNA. The linear dependence seen in both cases is evidence of the specificity of hybridization between fusion molecules and capture DNA, as well as the activity of each fusion enzyme of the consecutive reaction system of Reactions 1 and 2. A replica set of reactions with ribonuclease-treated fusion molecules showed no activity, providing evidence of the critical role played by the mRNA portion of the fusion molecule in capturing the enzymatic activity (fig. S2).

Fig. 2.

Activity and specificity of the mRNA-enzyme chimeras immobilized by homologous recombination with capture DNA molecules attached on a microwell surface. The sequential reaction of luciferase and nucleoside diphosphate kinase (NDK) was used as a test system: (A) Effect of increasing amounts of capture DNAs on enzymatic activities. Solutions of fusion molecules (40 μl) were loaded in each well. (B) Effect of increasing amounts of fusion molecules on enzymatic activities with 0.3 μg of capture DNA molecules spotted on each well. (C) Specificity of hybridization between fusion molecules and capture DNA. Mixtures of fusion molecules (all translated in a single step) were loaded on wells containing different combinations of the two types of capture DNA. Shown are the activity of luciferase in wells containing varying amounts of luciferase capture DNA molecule and the same amount (0.3 μg) of NDK capture DNA and the activity of NDK in wells containing varying amounts of NDK capture DNA and the same amount (0.3 μg) of luciferase capture DNA. In both cases, the same amount of solution was used containing both types of fusion molecules. Closed and open symbols represent luciferase and NDK, respectively.

The above results demonstrate the two advantages of the described approach. First, only a single step of in vitro translation is required and reaction mixtures containing different types of chimeric molecules can be directly applied to the well without any purification. Second, processing time is minimized for immobilizing the enzymes on supports at numerous combinations suitable for high-throughput testing.

To investigate the applicability of the platform to a realistic pathway exhibiting the network and regulatory complexities present in real metabolic bioreaction networks, we selected the trehalose synthesis pathway from glucose (Fig. 3). Trehalose is a disaccharide with a large potential market as a multifunctional sweetener, moisture retainer in cosmetics, and preservative in pharmaceutical products and frozen foods (14). It is synthesized in yeast and has also been observed in bacterial fermentation products (15, 16). The pathway of trehalose synthesis harbors a branch point (glucose-6-phosphate) where flux distributions between two competing reactions can impact end-product accumulation. Additionally, very little is known about the kinetics and regulation of the five pathway reactions, only one of the enzymes is commercially available, and overabundance of any single enzyme or all five together does not yield an optimal system in terms of maximal trehalose accumulation rate. It is therefore of interest to determine the optimal relative enzymatic activities, or relative gene expression rates for an in vivo application, that would maximize the rate of this pathway.

Fig. 3.

Systematic optimization of the trehalose synthesis pathway using the functional protein chip of this study. Total trehalose was determined at 3 hours after initiation of reaction (21). For the sake of clarity, error bars have been omitted in (A) to (F); the error bars shown in (G) are representative of the accuracy of trehalose measurements. The effect of varying amounts of each pathway enzyme on trelahose synthesis was examined while keeping the other enzymes fixed at (A) 3 μg, (B) 4 μg, and (C) 5 μg of capture DNA molecules. Because the optimal amount of OtsA was determined to be 4 μg (C), further experiments were carried out with 4 μg of OtsA while the other enzymes were fixed at the levels of (D) 6 μg and (E) 7 μg of capture DNA molecules. (F) Effect of varying enzyme amounts on trehalose synthesis under the optimal condition (4 μg of OtsA and 6 μg of PGM) while the other enzymes were set at 8 μg of capture DNA. (G) Amount of trehalose product synthesized for varying amounts of total PGM and OtsA but with their ratio maintained at the optimal value of 3/2. Other enzymes were saturated at 8 μg of capture DNA in each reaction mixture. (H) Schematic of the trehalose synthesis pathway used in this study. All the genes were cloned from the Escherichia coli K12 genomic DNA except for hxk1 cloned from S. cerevisiae by polymerase chain reaction with the following primers: udpgp: Psense, 5′-ATGACGAAT T TAAAAGCAGT-3′; Pantisense, 5′-GAGCTCT TAT TCGCT TAACAGCT TCT-3′. hxk1: Psense, 5′-ATGGT TCAT T TAGGTCCAAA-3′; Pantisense, 5′-AGTACTAGCGCCAATGATACCAAGAG-3′. pgm: Psense, 5′-ATGGCAATCCACAATCGTGC-3′; Pantisense, 5′-GAGCTCCGCGT T T TTCAGAACTTCGC-3′. otsA: Psense, 5′-ATGAGTCGT T TAGTCGTAGT-3′; Pantisense, 5′-AGTACTCGCAAGCT T TGGAAAGGTAG-3′. otsB: Psense, 5′-GTGACAGAACCGT TAACCGA-3′; Pantisense, 5′-AGTACTATACTAACGACTAAACGACTC-3′.

We examined the effect of each of the five enzymes on trehalose synthesis by setting the amounts of all but one enzyme at a constant level and measuring the rate of trehalose synthesis as the amount of the enzyme of interest was increased from zero to the same level as that of the other four enzymes (17, 18). In each case, the entire process was repeated for a higher enzyme level (Fig. 3). Saturation profiles were obtained for the lower enzyme amounts (Fig. 3, A and B), but a trehalose synthesis rate maximum was observed (Fig. 3C) at higher enzyme levels. For trehalose-6-phosphate synthase (OtsA), the maximum trehalose synthesis rate occurred at 4 μg of OtsA capture DNA, and for the other four enzymes maximum synthesis was reached at 5 μg of capture DNA. Figure 3, D and E, show the product amounts obtained with OtsA set by 4 μg of its capture DNA and the amount of the other enzymes increased to 6 μg (Fig. 3D) and 7 μg (Fig. 3E) of capture DNA. Whereas a monotonic response was obtained for hexokinase (HXK1), UDP–glucose pyrophosphorylase (UDPGP), and trehalose-6-phosphate phosphatase (OtsB), the trehalose amount was maximized at 6 μg of capture DNA for phosphoglucomutase (PGM) (Fig. 3E). Each curve in Fig. 3F shows the trehalose product obtained when the amount of the corresponding enzyme is varied between 0 and 8 μg while the three enzymes HXK1, UDPGP, and OtsB are kept at the maximum of 8 μg of capture DNA, and OtsA and PGM are set at the optimal levels captured by 4 and 6 μg of spotted DNA, respectively. (For the two curves corresponding to OtsA and PGM, all enzymes are set to the maximum captured by 8 μg of spotted DNA, while the amounts of OtsA and PGM are varied between 0 and 8 μg of capture DNA.) In summary, for the enzyme levels examined, increasing the amounts of the three enzymes, e.g., HXK1, UDPGP, and OtsB, enhances trehalose synthesis, whereas PGM and OtsA must be maintained at optimal activities for maximum trehalose synthesis. The points of zero amount of capture DNA shown for each curve in Fig. 3, A to F, were obtained by spotting no capture DNA for the enzyme depicted by the corresponding curve but the maximum for all other enzymes. The fact that no trehalose accumulation was measured under these conditions suggests that cross-hybridization was kept to a minimum. This is due to the specific design of the capture DNA and confirms the results of Fig. 2.

PGM and OtsA are enzymes at the branch point of the pathway, and maintaining an optimal balance of their activities is key for optimal flux allocation. PGM activity that is too high will drain the pool of glucose-6-phosphate required in the condensation reaction catalyzed by OtsA. Similarly, OtsA that is too high cannot be supported by the available pool of UDP-glucose unless an optimal activity of PGM is also present. As shown in Fig. 3, A and B, no optimal amounts of PGM and OtsA are found, indicating that, at these levels, the activities of both enzymes were too low to deplete the pool of glucose-6-phosphate. Throughout all experiments conducted with various enzyme amounts (Fig. 3, C to F), trehalose synthesis is maximized when the three enzymes HXK1, UDPGP, and OtsB are maintained at maximum activity while the activity of PGM and OtsA is controlled at the optimal ratio of 3/2. This finding was tested in a subsequent experiment in which the total amount (and therefore activity) of PGM and OtsA was varied from 0 to 15 μg while maintaining the optimal ratio of 3/2 and with the other three enzymes saturated at 8 μg of capture DNA. As seen in Fig. 3G, the trehalose synthesis rate increases linearly with the total amount of pgm and otsA capture DNA (at the 3/2 ratio). Thus, maintaining an optimal profile of enzymatic activities eliminates pathway kinetic limitations.

Trehalose 6-phosphate is known to inhibit hexokinase in Saccharomyces cerevisiae (1921), and this effect could explain in part the observed results, although it cannot alone explain the observed effects of PGM on the flux. A complete explanation of the observed pathway kinetics should be sought in an overall flux distribution model derived from individual enzymatic kinetics and regulation. It is nevertheless of interest to observe that by fixing the relative enzyme activities around the branch point at an optimal ratio, the whole pathway is essentially reduced to a linear pathway whose flux increases linearly with the activity of the pathway enzymes according to Kacser's “Universal Theorem” (22).

Although the focus of this research is on metabolic pathway optimization, the chimeric molecules of RNA display can have much broader applications, such as microarrays for protein capture for analytical applications (protein microarrays) (8, 9), screening of protein libraries for binding to target molecules, and screening of peptides or natural products for inhibition of binding activity. We also envision applications for the production of high added value compounds by reconstructing entire pathways for the synthesis of such compounds. Examples include proteins with specific glycosylation patterns and other posttranslational modifications and reactions, in general, catalyzed by unavailable enzymes that can be produced in vitro from their genomic sequence and conveniently immobilized using their capture DNA homolog.

Supporting Online Material

www.sciencemag.org/cgi/content/full/304/5669/428/DC1

Materials and Methods

Tables S1 and S2

Figs. S1 and S2

References

References and Notes

View Abstract

Navigate This Article