Research Article pubs.acs.org/synthbio
Synthetic Protein Scaffolds for Biosynthetic Pathway Colocalization on Lipid Droplet Membranes Jyun-Liang Lin,†,‡ Jie Zhu,† and Ian Wheeldon* Department of Chemical and Environmental Engineering, University of California, Riverside, California 92521, United States
Downloaded via IOWA STATE UNIV on January 16, 2019 at 14:32:22 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.
S Supporting Information *
ABSTRACT: Eukaryotic biochemistry is organized throughout the cell in and on membrane-bound organelles. When engineering metabolic pathways this organization is often lost, resulting in flux imbalance and a loss of kinetic advantages from enzyme colocalization and substrate channeling. Here, we develop a protein-based scaffold for colocalizing multienzyme pathways on the membranes of intracellular lipid droplets. Scaffolds based on the plant lipid droplet protein oleosin and cohesindockerin interaction pairs recruited upstream enzymes in yeast ester biosynthesis to the native localization of the terminal reaction step, alcohol-O-acetyltransferase (Atf1). The native localization is necessary for high activity and pathway assembly in close proximity to Atf1 increased pathway flux. Screening a library of scaffold variants further showed that pathway structure can alter catalysis and revealed an optimized scaffold and pathway expression levels that produced ethyl acetate at a rate nearly 2-fold greater than unstructured pathways. This strategy should prove useful in spatially organizing other metabolic pathways with key lipid droplet-localized and membranebound reaction steps. KEYWORDS: biocatalysis, cascade catalysis, ester biosynthesis, membrane localization, metabolic engineering
A
bind pathway enzymes can create multienzyme aggregates with high enzyme concentrations, minimal distance between active sites, controlled ratios of enzymes, and optimized spatial organization.13 Using intracellular scaffolds of protein−protein interaction domains, this strategy has been used to colocalize pathways for mevalonate, glucaric acid, and butyrate biosynthesis in E. coli, resulting in 77-, 5-, and 3-fold improvements in product titers, respectively.11,14,15 Resveratrol and 1,2 propanediol titers have also been improved (5 and 4.5-fold, respectively) with plasmid-based DNA scaffolds that recruit and colocalize zinc-finger modified multienzyme pathways in E. coli.16 Controlled intracellular assembly of multistep pathways has also been achieved with RNA scaffolds. Modified enzymes that bind to aptamers within the scaffold were used to created two-, three-, and four-step pathways that enhanced product titers and in some cases overall production rates.17,18 These examples of multienzyme colocalization are evidence that pathway structure can improve metabolite titers, but the scaffolding technologies have been limited to cytosolic organization and are not accessible to membrane-bound pathways. The lack of membrane-based scaffolding technologies is beginning to be addressed with engineered synthetic lipid-containing scaffolds. Lipid/protein particles based on proteins from the lipid-encapsulated bacteriophage ϕ6 have recently been used to assemble a two-step biosynthetic pathway in E. coli.19 Enhanced product titers from engineered microbes with spatially organized pathways are not necessarily the result of
defining feature of eukaryotic biochemistry is the spatial organization of metabolic pathways in and on membranebound organelles. This organization provides potentially viable targets for improving the rates of microbial biosynthesis. Key steps in yeast long chain fatty acid synthesis and lipid storage are localized to the endoplasmic reticulum (ER),1,2 many plant secondary metabolite pathways that produce valuable pharmaceuticals are often organized on vacuole and ER membranes,3 and the focus of this study, yeast ester biosynthesis, localizes to lipid droplets (LDs).4,5 The intracellular organization of these and other membrane-bound pathways is catalytically useful. Native multienzyme structures with controlled interenzyme distances and ratios of active sites can concentrate metabolites and drive reactions counter to unfavorable thermodynamics, isolate toxic intermediates from the rest of the cell, and prevent the loss of substrates to alternative pathways.6,7 When engineering synthetic pathways, the benefits of spatial organization are often lost, as differences in eukaryote physiologies can disrupt localization and translating pathways to the common bacterial host E. coli or other prokaryotes results in a loss of native organelles. Metabolic engineering and synthetic biology use genome engineering and pathway refactoring tools to optimize flux along targeted biosynthetic pathways by disrupting nonessential competing pathways, increasing the activity of native and heterologous enzymes to alleviate kinetic bottlenecks, and modifying transcription to control pathway expression.8 Enzyme colocalization has emerged as an additional strategy to recreate some of the effects of natural spatial organization and introduce substrate (or metabolite) channeling in synthetic pathways.9−12 Molecular scaffolds with domains that selectively © 2017 American Chemical Society
Received: February 4, 2017 Published: May 12, 2017 1534
DOI: 10.1021/acssynbio.7b00041 ACS Synth. Biol. 2017, 6, 1534−1544
Research Article
ACS Synthetic Biology equivalent increases in biosynthetic rates. Analysis of in vivo scaffolded pathways points to a range of effects that can produce higher titers. Colocalization and controlled clustering of enzymes in engineered pathways can benefit microbial growth by limiting the concentrations of toxic pathway intermediates,11,20 improve the underlying kinetics of bottleneck enzymes,14 and minimize metabolic burden by enabling low pathway expression-levels.10,11 Overall reaction rates can also be increased: controlled spatial organization can limit parasitic side reactions, enhancing flux along the desired pathway;18 molecular scaffolds can increase local substrate concentrations driving higher rates at low substrate concentrations;21,22 and, enzyme proximity can enhance presteady state kinetics.22 The limits of rate acceleration have recently been modeled, suggesting that with optimized multienzyme cluster size, architecture, and number per cell, up to ∼6-fold enhancements in the rate of two-step pathways are possible.12 To take advantage of the benefits of spatial organization with membrane-bound pathways we engineered a protein scaffold in yeast to colocalize the ester biosynthesis pathway on the outer membrane surface of intracellular LDs. In yeast fermentations, low concentrations of short and medium chain volatile esters are synthesized through the condensation of an alcohol and acetyl-CoA by alcohol-O-acetyltransferase (AATase or Atf; Figure 1a). The produced esters (e.g., ethyl and isoamyl acetate,
We reasoned that the spatial organization of upstream enzymes around the native intracellular localization of the terminal reaction step, Atf1, would produce high local enzyme concentrations and limit crosstalk with other metabolic pathways, thus increasing pathway flux and the conversion of intracellular acetyl-CoA to ethyl acetate under ethanol producing fermentation conditions. The relocalization of enzymes from their native targets to LDs requires a localization tag that can be fused to target proteins. From a library of protein domains, we identified a terminal fusion that localizes to yeast LDs under fermentation conditions. The identified oleosin domain formed the basis of a synthetic protein scaffold that was able to colocalize native S. cereisiae Ald6 and Acs1 with Atf1 on LDs (Figure 1b). Enzyme colocalization was achieved using cohesion-dockerin protein− protein interactions pairs from Clostridia species: cohesion domains localized to LDs by fusion to the oleosin scaffold with the corresponding dockerin domains fused to Ald6 and Acs1. Fluorescence microscopy and Förster Resonance Energy Transfer (FRET) studies showed that the enzymes are tightly clustered on LDs and in vitro kinetic analysis revealed increased rates of ethyl acetate synthesis due to colocalization. To rapidly optimize scaffold and pathway expression and structure, we used a colorimetric assay for ethyl acetate biosynthesis and screened a combinatorial library of scaffolds with a series of different strength promoters to drive enzyme and scaffold expression. This approach enabled us to relocalize upstream pathway enzymes around the native LD-localization of Atf1 and demonstrate the effects of colocalization on pathway flux and ethyl acetate fermentation titers.
■
RESULTS Identification of a Yeast Fermentation Lipid DropletTargeting Domain. To identify a suitable LD-targeting domain, we screened a series of five protein domains known to localize to mammalian cell LDs and the plant LD-protein oleosin. The screened domains included terminal hydrophobic sections of a putative methyltransferase AAM-B,28 17βhydroxysteriod dehydrogenase (17βHSD),29 the antiviral protein viperin,30 and a hepatitis C virus core protein (HCVcp).31 The library also included an internal domain of a guanine nucleotide exchange factor GBF-132 and Zea mays oleosin.33 Fluoroescence microscopy of C-terminally modified cyan fluorescent protein (CFP) revealed that AAM-B-CFP and HCVcp-CFP colocalized with the ER protein marker Sec61 with C-terminally fused red fluorescent protein (Sec61-DsRed) during exponential growth. Trafficking to LDs during stationary phase was observed by colocalization with the LD protein marker Erg6 (Erg6-DsRed) after 24 h (Figure 2a). The LD localization domains of 17βHSD, viperin, and GBF-1 were found to be primarily cytosolic in S. cerevisiae and therefore not suitable as targeting domains (Figure S1). Fluorescence microscopy of cells overexpressing oleosin with C-terminally fused CFP (Ole-CFP) showed vesicle-like structures adjacent to Sec61-DsRed, suggesting oleosin localization to premature LDs. Similar to AAM-B-CFP and HCVcp-CFP, Ole-CFP colocalized with Erg6-DsRed during stationary phase, indicating LD targeting. Although Ole-CFP, AAM-B-CFP, and HCVcpCFP could localize to LDs under aerobic condition, only OleCFP maintain LD localization under fermentation conditions to 72 h (Figure 2b). In anaerobic cultures both AAM-B-CFP and HCVcp-CFP were dispersed in cytosol. As such, the plant
Figure 1. Pathway colocalization on intracellular LD membranes in S. cerevisiae. (a) Ethyl acetate biosynthesis in S. cerevisiae and the subcellular localization of aldehyde dehydrogenase (Ald6), acetyl-coA synthetase (Acs1), and alcohol-O-acetyltransferase (Atf1). Abbreviations: Cyt, cytosol; Mt, mitochondria; ER, endoplasmic reticulum; and, LD, lipid droplet. (b) A schematic diagram of the scaffolding strategy, colocalizing Ald6 and Acs1 with Atf1 on LDs via a membrane bound protein scaffold.
among others) are responsible for the distinctive smells of yeast fermentations and are industrially useful as solvents and flavor and fragrance compounds. The predominant AATase for short chain ester synthesis in S. cerevisiae is Atf1, with contributions from Atf2.23 Atf1 and -2 are membrane-associated proteins that localize to the ER and traffic to LDs during the stationary phase.4 Heterologous expression of Atf1 and -2 as well as AATases from other yeasts and fruit results in a loss of membrane localization and the formation of low activity aggregates in the cytosol.24 Overexpression in S. cerevisiae does not disrupt native localization and maintains high Atf activity. Upstream enzymes in the acetyl-CoA branch of yeast ester biosynthesis, aldehyde dehydrogenase (Ald6) and acetyl-CoA synthetase (Acs1), are naturally segregated from Atf1, localizing to the cytosol (Ald6 and Acs1) and mitochondria (Acs1).25−27 1535
DOI: 10.1021/acssynbio.7b00041 ACS Synth. Biol. 2017, 6, 1534−1544
Research Article
ACS Synthetic Biology
Figure 2. Identification of synthetic LD targeting domains in S. cerevisiae. Fluorescence microscopy analysis of subcellular localization of three LD tags including oleosin (Ole) and terminal domains of AAM-B(1−38) and hepatitis C virus core protein (HCVcp(118−164)) at 10 and 24 h (a) and under fermentation conditions (b). Sec61-DsRed and Erg6-DsRed are ER and LD protein markers, respectively. Cell morphology is shown by phase contrast (false colored blue). CFP: cyan fluorescent protein; and, DsRed: red fluorescent protein. Scale bar: 1 μm. (c) Fluorescence microscopy analysis of subcellular localization of Acs1-Ole and Ald6-Ole. Scale bar: 1 μm. (d) Enzyme activity of modified Acs1 and Ald6, and empty vector controls in units of μmol min−1 mg−1 of total lysate protein. Cells overexpressing fusion constructs were cultured to stationary phase and cell lysates were prepared for kinetic assays (n = 3). Schematic diagrams of LD localized Acs1-Ole-CFP and Ald6-Ole-CFP are shown above the activity data.
Functional Localization of Enzymes to Lipid Droplet Membranes. Our strategy to recover Acs1 and Ald6 function focused on an oleosin-based scaffold with protein−protein interaction domains to selectively recruit each enzyme to the outer surface of LDs. We selected cellulosomic cohesiondockerin interaction pairs because there are no natural homologues in S. cerevisiae and they have previously been used to selectively colocalize multienzyme structures in vitro35 and on yeast cell surfaces.36 The protein parts of the scaffold system are shown in Figure 3a including, oleosin (Ole) and two dockerin-cohesin pairs (D1-C1 from Clostridium perf ringens37 and D2-C2 from Clostridium thermocellum38). The scaffold and modified enzymes also contained a protein identification tag (e.g., HA, His, or V5) for Western blot analysis. The putative assembly of dockerin-modified Acs1 and Ald6 with an Ole-C1C2 LD-localized scaffold is schematically depicted in Figure 3b (left). To demonstrate pathway colocalization Ald6-YFP-D1, Acs1-CFP-D2, and Atf1-DsRed were coexpressed with the OleC1-C2 scaffold. S. cerevisiae harboring the pathway expression vector was cultured to stationary phase and imaged to confirmed pathway localization. In the absence of the scaffold, Ald6-YFP-D1 and Acs1-CFP-D2 were observed in the cytoplasm, while Atf1-DsRed localized to LDs (Figure 3b, right). Both Ald6-YFP-D1 and Acs1-CFP-D2 colocalized to LDs with expression of the scaffold. Control experiments of YFP tagged scaffold confirmed LD targeting by colocalization with the Erg6-DsRed LD marker (Figure S3) and culture conditions did not affect oleosin function (Figure S4). It is interesting to note that measurable levels of the dockerin modified enzymes
protein oleosin was used in subsequent experiments for ester pathway localization. Fluorescence microscopy of overexpressed Acs1 and Ald6 with C-terminal fusions of Ole-CFP (Acs1-Ole-CFP and Ald6Ole-CFP) showed that oleosin was able to relocalize both enzymes to LDs (Figure 2c). In the absence of oleosin, overexpressed Acs1-CFP and Ald6-CFP localized to the cytoplasm. Functional characterization was less successful as LD-targeted Acs1-Ole-CFP and Ald6-Ole-YFP exhibited whole cell lysate activities equal to the cell background (Figure 2d). Acs1-Ole-CFP produced acetyl-CoA from CoA-SH and acetate at a rate of 0.19 ± 0.01 μmol min−1 mg−1 of total protein, a value equal to the empty vector control (0.21 ± 0.02 μmol min−1 mg−1). Ald6-Ole-CFP converted acetaldehyde to acetate at a rate of 0.82 ± 0.04 μmol min−1 mg−1 of total protein, with the vector control limited to 0.75 ± 0.02 μmol min−1 mg−1. Nterminal fusions of oleosin were also found to have limited activity (Figure S2). Control experiments suggested that the low activities were not due to C-terminal fusions as cell lysates containing Acs1-CFP and Ald6-CFP exhibited activities significantly greater than the cell backgrounds (Acs1-CFP, 0.37 ± 0.01 μmol min−1 mg−1 of total protein, and Ald6-CFP, 1.32 ± 0.12 μmol min−1 mg−1). It is possible that relocalization to the membrane surface of LDs resulted in poor enzyme orientation restricting substrate access to the active sites. For example, the crystal structure of Acs1 shows a positively charged patch around the active site that could interact with the negatively charged phosphate end-groups of the LD membrane, thus limiting catalysis.34 1536
DOI: 10.1021/acssynbio.7b00041 ACS Synth. Biol. 2017, 6, 1534−1544
Research Article
ACS Synthetic Biology
Figure 3. Functional relocalization of enzymes to LDs via synthetic membrane scaffolds. (a) Protein parts of the synthetic membrane scaffold and abbreviations used in the figure. C1 and D1 are X82 (a cohesin-like protein) and dockerin from C. perf ringens, while C2 and D2 are cohesin and dockerin from C. thermocellum, respectively. (b) Schematic representation the LD assembled pathway (left). Fluorescence microscopy of the subcellular localization of Ald6-D1-YFP, Acs1-D2-CFP, and Atf1-DsRed in the presence and absence of the scaffold at stationary phase (right). Scale bar: 3 μm. (c) Enzyme activity of Acs1-D2 and Ald6-D1 in units of μmol min−1 mg−1 of total lysate protein. Cell lysates from overexpressed Ald6-D1, Acs1-D2, Atf1, and the scaffold Ole-C1-C2 were analyzed for Acs1 and Ald6 activity. The mean and standard deviation are shown (n = 3).
were not observed in the cytosol as judged by fluorescence microscopy (Figure 3). This suggests that the dockerin-cohesin interactions may occur early in the secretory pathway and/or there is a surplus of scaffold for enzyme attachment. Acs1 and Ald6 activities above cell lysate backgrounds were observed when enzymes were targeted to LDs by the synthetic protein scaffold. Localized Acs1-D2 produced acetyl-CoA at a rate of 0.19 ± 0.01 μmol min−1 mg−1 of total protein in cell lysate assays, while a vector control expressing Atf1 and the Ole-C1-C2 scaffold exhibited an activity of 0.16 ± 0.01 μmol min−1 mg−1 of total protein (Figure 3c). Functional localization of Ald6-D1 was also successful with measured activity reaching 0.82 ± 0.05 μmol min−1 mg−1 of total protein, a value greater than the background activity of 0.62 ± 0.06 μmol min−1 mg−1. On the basis of the kinetic results of overexpressed Ald6-D1 and Acs1-D2 along with our previous analysis of Atf1 kinetics,24 the order of activity of pathway enzymes was assumed to be Ald6 > Atf1 > Acs1. Of note is the reduced activities of LD-targeted Ald6-D1 and Acs1-D2 in comparison to cytosolically targeted enzymes (compare Figures 3c and 2d). Western blot analysis suggests that the reduction in measured activities was due to reduced expression levels (Figure S5). Spatial Organization of a Multienzyme Pathway Colocalized to LDs. Theoretical and experimental analysis of substrate channeling in coupled enzyme reactions suggests that the distance between enzyme anchor points should be less than ∼5 nm to produce significant enhancements in catalysis.39 The flexibility of the cohesion-based protein scaffold prevents accurate calculation of the distance between immobilized Ald6D1, Acs1-D2, and LD-localized Atf1, as such we quantified pathway organization by FRET microscopy (Figure 4). Acceptor photobleaching experiments showed that scaffold-
Figure 4. FRET microscopy and quantitative analysis of multienzyme colocalization. Stationary phase cells expressing scaffold-bound Ald6YFP-D1 and Acs1-CFP-D2, Ole-C1-C2-CFP and Atf1-YFP, Ole-C1-C2 with C-terminally fused CFP-YFP, and Ole-C1-C2-CFP with cytosolically expressed YFP were imaged and analyzed. Phase contrast micrographs and fluorescence images of the CFP donor, YFP acceptor, and raw FRET signal are shown. CFP signal is shown in green (Ex/ Em: 426−450 nm/502−538 nm), YFP signal is shown in yellow (Ex/ Em: 488−512 nm/532−554 nm), and the FRET signal is shown in blue (Ex/Em: 426−450 nm/532−554 nm). FRET efficiency was quantified using the Leica SP5 FRAP (fluorescence recovery after photobleaching) protocol with acceptor fluorescence photobleached to ∼10% of its initial value. Presented data represents the mean and standard deviation of five samples. The scale bar represents 1 μm.
1537
DOI: 10.1021/acssynbio.7b00041 ACS Synth. Biol. 2017, 6, 1534−1544
Research Article
ACS Synthetic Biology
prepared from stationary phase cultures were equilibrated with pathway substrates and cofactors (ATP, CoA-SH, and NADP+ for Ald6/Acs1; ethanol, ATP, and CoA-SH for Acs1/ Atf1) and reactions were initiated by the addition of acetaldehyde (Ald6/Acs1) and acetate (Acs1/Atf1). To confirm that LDs remained intact, cell lysates containing OleCFP were prepared in parallel and LD morphology was confirmed by fluorescence microscopy (Figure S7). S. cerevisiae with integrated ATF1 produced acetyl-CoA at a rate of 85 ± 20 nmol min−1 mg−1 of total protein from natively expressed Ald6 and Acs1 (Figure 5b). The rate of acetyl-CoA synthesis was 93 ± 15 nmol min−1 mg−1 of total protein when Ald6-D1 and Acs1-D2 were overexpressed. Colocalization of the pathway enzymes on LDs with the coexpression of the Ole-C1-C2 scaffold increased the rate of acetyl-CoA production to 159 ± 8 nmol min−1 mg−1 of total protein, 1.8-fold enhancement over the control strain and unassembled pathway. Similarly, the rate of the last two reaction steps of ethyl acetate biosynthesis, Acs1/Atf1, increased 2.9 and 1.9-fold over the control strain and unassembled pathway, respectively (Figure 5c; ATF1 integrated control, 14 ± 1 nmol min−1 mg−1 of total protein; unassembled control, 21 ± 0.4 nmol min−1 mg−1; and, colocalized pathway, 40 ± 2 nmol min−1 mg−1). Western blot analysis of the expressed pathway and scaffold revealed that enzyme levels were consistent across the tested strains (Figure S8). A traditional test for substrate channeling (i.e., the transfer of an intermediate from one active site to another without first diffusing to the bulk environment) is the competition for pathway intermediates by an orthogonal reaction.6,7 Competing side reactions reduce the overall pathway reaction rate by consuming pathway intermediates that diffuse to the bulk solution. The effect of substrate channeling is protection against competing reactions and an increase in the rate of the colocalized pathway. The cell lysate assays shown in Figure 5 approximates such a test because natively expressed enzymes (e.g., pyruvate dehydrogenase, citrate synthase, and acetyl-CoA synthetase-2, among others) compete for the acetate, CoA-SH, and acetyl-CoA intermediates of the Ald6/Acs1 and Acs1/Atf1 coupled reactions. The rates of the Ald6/Acs1 and Acs1/Atf1 reactions increased when colocalized on the LD-membrane while enzyme levels remained constant in comparison to the unassembled pathways, thus suggesting the possibility of substrate channeling along the structured ethyl acetate biosynthesis pathway. Additional channeling assays (e.g., controlled challenge assays or the tracking of radiolabeled intermediates; see ref 4) are necessary to draw a conclusion of substrate channeling. Optimizing Pathway Architecture and Expression Enhances Ester Biosynthesis. To optimize pathway flux during fermentation, we created a combinatorial library of the pathway by varying scaffold design and promoter strength. Kinetic analysis of Acs1-D2 and Ald6-D1 suggested that Acs1 activity was limiting (Figure 3). Western blot analysis of the expressed pathway supports this conclusion. Comparison of Ald6-D1, Acs1-D2, and Atf1 expression levels relative to a GFP standard revealed low Acs1-D1 expression (Figure S9). In light of these results, our optimization strategy focused on increasing Acs1-D2 expression and varying the ratio of Acs1-D2 to Ald6-D1 on the LD-localization scaffold. A library of 60 pathway variants was created by homologous recombination in S. cerevisiae as previously described.43 The design of the genetic constructs and pathway variants are shown in Figure 6a. A second copy of
bound Ald6-YFP-D1 and Acs1-CFP-D2 maintained a FRET efficiency of 13 ± 4%, while a CFP-tagged scaffold and YFPtagged Atf1 had an efficiency of 15 ± 2% (
