· web viewsaccharomyces cerevisiae and violacein biosynthetic pathway (ukibe et al., 2009; lee et...
TRANSCRIPT
Artificial Protein Scaffold System (AProSS): An Efficient Method to
Optimize Exogenous Metabolic Pathways in Saccharomyces cerevisiae
Tianyi Lia,1, Xiuqi Chena,d,1, Yizhi Caib,c,*, Junbiao Daia,b,*
a Key Laboratory of Industrial Biocatalysis (Ministry of Education) and Center for Synthetic and
Systems Biology, School of Life Sciences, Tsinghua University, Beijing 100084, China
b Centre for Synthetic Genomics, Shenzhen Institutes of Advanced Technology, Chinese Academy
of Sciences, Shenzhen 518055, China
c Manchester Institute of Biotechnology, University of Manchester, 131 Princess Street, M1 7DN,
Manchester, UK
d Department of Biology, Johns Hopkins University, 3400 N. Charles Street, Baltimore, Maryland,
US
1 These authors contributed equally to this work
*Corresponding author:
Junbiao Dai
E-mail: [email protected]
Telephone: 86-755-86585244;
Yizhi Cai
E-mail: [email protected]
Telephone: +44- 0161 306 4214;
ABSTRACT
Scaffold proteins influence cellular signaling by orchestrating multiple enzymes, receptors or
ion channels, and could be tailored to enhance the efficiency of biochemical reactions by
positioning related enzymes physically together. However, the number of applicable domains
remains small, and the construction of scaffold proteins with optimal domain ratio could be tedious
and time-consuming. In this study, we outlined a modular design to quickly assemble scaffold
proteins using protein interaction domains, which have been constructed into a standardized vector.
We generated multiple protein interaction domains and ligands for making artificial scaffold
proteins. At the same time, we developed a robust Golden-Gate-based molecular toolkit for the
construction of artificial scaffold proteins, allowing a variance of domain types, number, and
positions. The synthesized domain-ligand interaction was verified by yeast two-hybrid and split-
GFP assays. We demonstrated an increase in the yield of the target product and a change of
metabolic flux using the synthetic scaffolds. Our system could be a useful tool for metabolic
engineering and beyond.
Keywords: Synthetic Biology; Metabolic Engineering; Artificial Scaffold Protein; Protein-Protein
Interaction Domain; Golden Gate Assembly
1.Introduction
Microbial production of natural products has been achieved by introducing product-specific
enzymes or the entire metabolic pathways into readily engineered organisms such as E. coli and
budding yeast, a process now we call metabolic engineering (Bailey, J. 1991). The range of
chemicals that can be produced from metabolic engineering has expanded substantially (Alonso-
Gutierrez et al., 2013; Jin et al., 2005; Chemler et al., 2010; Li et al., 2017; Cone et al., 2003; Meng
et al., 2017; Awan et al., 2017) since the term was coined around 20 years ago, in part due to
notable advances in fields related to metabolic engineering. However, introducing a metabolic
pathway into a heterologous host can cause problems. For instance, tThe synthetic metabolic
pathways lack endogenous regulatory components. Therefore, the introduction of such pathways
often leads to growth retardation and causes metabolic imbalance due to the accumulation of over-
expressed proteins, end products, and intermediates which are often toxic (Lee et al., 2012;Nielsen et al., 2017; Tan et al., 2016). Moreover, imbalances in a metabolic pathway often elicit a
stress response in central metabolism (Keasling et al., 2010). Currently, the regulatory landscape
within the cells is not well-understood and the tools to manipulate it is very few tools to alter the
regulatory landscape within the cells are neither available nor well-understood (Yadav et al., 2012).
This leaves pathway balancing as one of the most challenging issues in metabolic engineering that
require further investigation.
The principal objective of balancing a metabolic pathway is to produce more of a target product
by reducing potential flux imbalances and cellular burden in the host organism. This is mainly
accomplished by eliminating the production of excessive intermediate metabolites and precursors
which results in efficient conversion of intermediates, substrates, and co-factors to desired products
(Jones et al., 2015). Among the several successful pathway-balancing approaches, post-translational
modulation takes advantage of synthetic scaffolds to recruit metabolic enzymes of interest pathways
to form synthetic complexes and increase spatial orientation of substrates. This is achieved either by
protein fusions for enzyme cascades or synthetic scaffolds to dock the enzymes in close proximity
using DNA, RNA, or proteins (Sachdeva et al., 2014; Delebecque et al., 2011; Dueber et al. 2009).
Dueber et al. demonstrated that by introducing synthetic scaffold proteins, the production of an end
product, mevalonate could be improved significantly (Dueber et al., 2009). They assembled a
synthetic scaffold with three protein domains, GBD from rat N-WASP, SH3 from mouse Crk, and
PDZ from mouse α-syntrophin. Meanwhile, they fused the corresponding ligands on the three
enzymes in the mevalonate pathway, namely AtoB, HMGS, and HMGR. By varying the number of
each domain on the scaffold, they successfully found an optimal combination and improve the
mevalonate production by 77 folds.
Although introducing the synthetic scaffold proteins can be a very tempting approach for
metabolic engineers to optimize their pathways, two major problems lie ahead. The construction of
a synthetic scaffold protein with optimal configuration can be very time-consuming and laborious.
As indicated in the previous study, constructing the fusion proteins with multiple domains requires
a significant amount of repetitive lab work. One has to configure the scaffold protein specifically
for each metabolic pathway, which would be essential for the performance of the scaffolding
strategy. Having an only limited number of available protein domains is another obstacle for broad
application of this strategy. The most commonly used scaffold protein domains are limited to the
three candidates from Dueber’s study, whereas few tried to adapt new protein domains into the
scaffold (Kim et al., 2014) and none have built a scaffold protein with 4 or more protein domains.
However, there is need to co-localize more than three enzymes since most of the pathways we are
optimizing today involve three or more enzymes, such as astaxanthin synthesis in Saccharomyces
cerevisiae and violacein biosynthetic pathway (Ukibe et al., 2009; Lee et al., 2013). Meanwhile, the
scaffold systems engineered to recruit pathway enzymes are predominately in E. coli, while there is
an urgent demand for them in yeast and higher eukaryotes (Horn et al., 2015).
In this study, we designed a standardized assembly protocol and constructed a set of Golden-
Gate-based molecular parts, which we named Artificial Protein Scaffold System (AProSS). We
designed and constructed a hierarchical three-step assembly system to integrate any protein domain
with variable repeats into a scaffold protein. On the other hand, we searched through structural
studies and found three strong interaction protein domains with their corresponding ligands. These
three domains were named as FE, YAP, and BIR. FE, and YAP are two WW domains of human
FE65 (Meiyappan et al., 2007) and YAP65 protein (Macias et al., 1996), and BIR is a protein
domain from inhibitor-of-apoptosis protein(IAP) (Liu et al., 2000). We chose the interactional
center of these proteins and synthesized the gene de novo, with subsequently modularized the
protein domains into a standardized library together with the three previously verified domains,
namely, GBD, SH3 and PDZ. The domain-ligand interactions were verified by yeast two-hybrid
and split GFP assays. Constructing scaffold proteins with AProSS increased the production of
violacein and deoxyviolacein by 29% and 63% respectively, while violacein/deoxyviolacein ratio
increased by 18% with rewired flux.
2.Materials and Methods
2.1 Strains and cultivation.
E. coli DH5α was used for plasmid construction and amplification. Yeast strain JDY26 (MATa
ade2-101 trp1-901 leu2-3.112 his3∆200 ura3-52 gal4∆ gal80∆ SPAL::URA3 LYS2::GAL1-HIS
GAL2-ADE2 met2:GAL7-LacZ can1R) was used to host the constructs and subjected to yeast two-
hybrid analysis. Yeast strain BY4741 (MATa leu2Δ0 met15Δ0 ura3Δ0 his3Δ1) was used to host the
split GFP constructs and subjected to FACS analysis. Yeast strain JDY52-URR-His3 (MATa
his3Δ200 leu2Δ0 lys2Δ0 trp1Δ63 ura3Δ0 met15Δ0), a derivative of S288C with ISceI recognition
sites flanking the URR1-His3-URR2 fragment at HO locus (Guo et al., 2015) was used as a host for
violacein biosynthetic pathway. Yeast cells were cultured in YPD (10 g/L yeast extract, 20 g/L
peptone, 20 g/L glucose) and SC-Ura [6.7g/L YNB, 0.01μmol/L Fe(NH4)2(SO4)2, 20 g/L glucose;
The mixture of amino acids and other nutrients] medium.
2.2 Plasmids construction
For our designed plasmids, we took advantage of the pSMART®HCKan and
pSMART®LCAmp plasmids combined with the cloned RFP gene. We added two EarI restriction
endonuclease sites to both terminals of the linear DNA by PCR. The RFP gene was also constructed
by PCR with additional EarI on the primers, followed by EarI digestion and ligation. With the
KanR and AmpR genes already on the vectors, we inserted the BsaI and BsmBI onto the upstream
and downstream of the RFP site by mutagenic PCR. For pLV vectors, we separated the two
digestion sites with BsmBI inside and BsaI flanking outside around the RFP gene and inserted a 9-
amino acid of glycine-serine repeat between the BsaI site and BsmBI site upstream, in frame with
the RFP starting from pLV2.
2.3 Selection and synthesis of new protein interaction domains
The new protein interaction domains were cherry-picked by reading through structural studies
about protein-ligand interaction (Dueber et al., 2009). After we acquired the DNA sequence of the
protein domain, we removed the stop codon, optimized the codon composition using GeneDesign
(Richardson et al., 2006), mutated the BsaI, BsmBI and EarI digestion sites, and added the
additional sequences to the 5’ and 3’ ends:
Upstream: caggaaacagctatgaccGGTCTCaCTCGnnnnnnnnnnnnnnnnnnnn
Downstream: tgtaaaacgacggccagtGGTCTCgTTGAnnnnnnnnnnnnnnnnnnnn
“nnn” indicates the protein domain DNA region, the blue and red parts are the BsaI recognition
and digestion overhangs. The whole sequence could then be synthesized through primer alignment
(Richardson et al., 2006) or direct chemical synthesis.
2.4 One-pot AProSS assembly
One-pot Golden Gate assembly is described in many studies (Engler et al., 2008; Engler et al.,
2009;Yuan et al., 2017). In the AProSS method, constructing a scaffold ORF composes of three
steps. The first step, clone PCR product into a kanamycin resistant vector pMV, mixed with
reaction mix [1.5 μL 10× T4 DNA ligase reaction buffer (New England BioLabs, M0202), 0.15 μL
100× bovine serum albumin (BSA, New England BioLabs), 2.5 U T4 DNA ligase (Enzymatics,
Beverly, MA, L6030-HC-L), and 10 U of BsaI (New England BioLabs, Beverly, MA, R0535 or
R0580, respectively)] to a final volume of 15 μL. One-pot digestion-ligation assembly was
performed in a thermocycler as follows: 37 °C for 60 min, 50 °C for 15 min and 80 °C for 15 min.
Five microliters of each assembly reaction were transformed into 100 μL of competent DH5α E.
coli cells and plated on the appropriate selection media. The second step, subclone the fragment into
a ampicillin resistant vector pLV with location order, mixed with reaction mix [1.5 μL 10× T4 DNA
ligase reaction buffer (New England BioLabs, M0202), 0.15 μL 100× bovine serum albumin (BSA,
New England BioLabs), 2.5 U T4 DNA ligase (Enzymatics, Beverly, MA, L6030-HC-L), and 10 U
of BsmBI (New England BioLabs, Beverly, MA, R0535 or R0580, respectively)] to a final volume
of 15 μL. One-pot digestion-ligation assembly was performed in a thermocycler as follows: 55 °C
for 60 min and 25°C for 60 min with T4 DNA ligase was added as soon as the thermocycler cool to
25°C, followed 55 °C for 15 min and 80 °C for 15 min Product was transformed into competent
DH5α E. coli cells as the described in the first step. The third step, similar to the first step with a
series of pLVs and the corresponding pAV in a one-pot reaction. One-pot digestion-ligation
assembly was performed in a thermocycler as follows: 25 cycles of 37 °C for 2 min and 16 °C for 5
min, followed by 50 °C for 15 min and 80 °C for 15 min. If the number of domains in the scaffold
is under five, using 37 °C incubation for 60 min, followed by 50 °C for 15 min and 80 °C for 15
min also works. The assembly product was transformed into competent DH5α E. coli cells like
other steps.
2.5 Protein domain-ligand interaction verification
2.5.1 Yeast two-hybrid assay
We used PCR to amplify the domains and the ligands, and inserted them into yeast two-hybrid
vectors, pDB-Leu and pPC86 plasmids respectively. The protein domains with relatively larger size
were fused to the C terminus of the Gal4 binding domain, and the ligands were fused to the
activation domain in another vector. The co-transformed yeast strain was diluted from OD=1 to 5-5
on SC petri dish lacking uracil and incubated under 30 ℃.2.5.2 Split GFP assay
The frGFP sequence is from T. Magliery’s lab (Sarkar et al., 2008). We synthesized it from
oligos and attached the domains to the C terminus of NGFP and corresponding ligands to the N
terminus of CGFP by using pAV2 vector of AProSS system. Transcription units were constructed
into the yeast expression vector and co-transformed into yeast strain BY4741. The strain was
incubated for 48 hrs under 30 ℃ followed with observing and imaging by fluorescence
microscopy.
2.5.3 Flow cytometry
We cultured strains with split-GFP-fused ligands and domains and normalized them to the same
OD at 0.1. After culturing for 48 hrs, we measured the fluorescent intensity by flow cytometry. The
threshold was set by measuring the intensity of negative control, NGFP-GBDligand and SH3ligand-
CGFP.
2.6 Violacein expression and purification
We constructed yeast strain with violacein biosynthetic pathway genes (Lee et al., 2013). We
then transformed the yeast with 3 different scaffolds and incubated the yeast for 2 days. After
picking the single colonies and overnight culture, we diluted the culture OD=0.1. After 24 hrs, the
fermented culture was boiled at 80 ℃ for 30 min. We then centrifuged the culture and
discarded the supernatant. Afterwards, cells in the pellet were resuspended using 1ml ethanol,
sonicated by ultrasonic cleaner for 15 min and boiled at 94 ℃ for 15 min. The lysate was filled to 1 ml with ethanol. Finally, 500 μl lysate was sent to HPLC.
HPLC analysis condition: C-18 column (Inertsil ODS-SP 4.6×150mm). Solution A is ddH2O,
and solution B is 75% MeOH. Flow velocity is 1ml/min. The sample size is 10 μl. Detection
wavelength is 568 nm.
3. Results
3.1 Designing of AProSS
3.1.1 Design of an efficient assembly system
The artificial protein scaffolds are composed of several domains with variable repeats in a given
order. To construct any scaffold from protein domains according to the designer's specifications, it
is desirable to have a library of standard parts, which could be cherry-picked and quickly
assembled. Therefore, we adopted the Golden-Gate strategy (Engler et al., 2009; Yuan et al., 2017)
to design a series of vectors with sequentially compatible digestion overhangs. This helped us to
achieve seamless ligation of the DNA fragments with the designated order by assigning pre-
designed unique overhangs to both ends of the fragments. We categorized the assembly process into
three steps: interface standardization, positioning with linker insertion and domain assembly. For
each step, we designed a set of vectors, termed pMV, pLV and pAV.
Vector pMV is used to storage different protein domains and will provide a uniform and
standardized digestion interface for them. Vectors pLVs are the location vectors. This group of
vectors contains different pLVs with two universally compatible ends with pMV to receive protein
domains and different pre-designed downstream digestion overhangs to indicate positional
information of each received protein domain. Meanwhile, these vectors carry a 9-amino-acid
glycine-serine linker between each domain. Vectors pAVs are the assembly vectors which take and
put together all the protein domains from the pLV vectors with compatible digestion overhand at
both ends, providing a start codon at the 5’ terminus and a stop codon at the 3’ terminus. We
assigned specific sequences with two Type IIs restriction sites to generate a designed four-
nucleotide overhang in both upstream and downstream of the fragment. These four-nucleotide
overhangs sequentially connect each protein domain and each pair of upstream and downstream
overhangs defines a specific position of a protein domain. For example, the downstream joint of
pLV1 is complementary to the upstream joint of pLV2, the downstream joint of pLV2 is
complementary to the upstream joint of pLV3. For pAVs, the digestion overhangs are compatible
with pLV1 and the last pLV that we would need on the scaffold. To improve the efficiency of the
assembly, we chose 16 overhang combinations as the joints (Table 1) with at least two nucleotide
mismatches between any two sites.
We designed the recognition sites of BsaI and BsmBI type IIs endonucleases in the pMV, pLV,
and pAV vectors, respectively. To eliminate the template plasmid from previous assembly steps,
these three vector groups carry Kanamycin and Ampicillin resistance genes alternatively. An RFP
serves as an insertion marker, visible under natural light. (Figure 1A) To use YeastFab assembly
method (Guo et al., 2015; Yuan et al., 2017) to assemble transcription units, we design compatible
restriction four-nucleotide digestion overhang of pAVs to GATG and GCTA so that the correct
pAV-ScaffoldORF plasmids will have the same form as the basic part of YeastFab method.
3.1.2 The AProSS assembly workflow
To assemble artificial scaffold proteins with AProSS, we just need to obtain the protein-coding
sequences of a protein domain without the stop codon and flanked by two BsaI recognition sites by
either PCR or direct synthesis. This sequence is then cloned into the pMV vector. The protein
domains on pMV can subsequently be assigned to any designated position on the scaffold protein
and cloned into the corresponding pLV vectors. By putting the same domain into several successive
pLVs, we could achieve a variable number of domain repeats on the scaffold. The full-length open
reading frame of the scaffold is assembled by mixing all pLVs with one corresponding pAV vector
following the Golden gate assembly protocol. (Engler et al., 2009;Yuan et al., 2017) For yeast
expression, the assembled scaffold could be cloned into transcription units (TUs) by YeastFab
method with selected promoters and terminators (Guo et al., 2015; Yuan et al., 2017) to utilize in S.
cerevisiae. (Figure 1B)
3.2 Discovery and verification of applicable protein-protein interaction (PPI) domains
The candidate domains and ligands should possess high affinity, specificity and relatively small
sizes that won’t significantly perturb the conformations of the attached enzyme or affect normal cell
growth. With careful investigation and selection, we have successfully found three protein
interaction domains, namely, FE, YAP, and BIR. (Meiyappan et al., 2007; Macias et al., 1996; Liu
et al., 2000) Together with the other three domains, GBD, SH3, and PDZ, we synthesized the DNA
sequences of the six domains as well as the ligands, flanked the sequences with BsaI sites, and
cloned them into pMV vectors by BsaI enabled Golden Gate reaction. All of the six domains and
ligands are supposed to function as independent modules to interact with each other specifically.
After acquiring the pMV vectors containing domains or ligands, we use yeast two-hybrid assay
to verify the interaction of domains and ligands qualitatively (Figure 2A). We amplified the
sequences and inserted them into yeast two-hybrid vectors, with c-fos and c-jun transcription factor
as positive control. The yeast hybrid test successfully verified four out of six domain-ligand
interactions. We achieved a semi-quantitative measurement of the interaction between domain and
the ligand with a series dilution sets. PDZ and its ligand showed the strongest interaction, followed
by GBD, YAP, and FE (Figure 2B).
Compared with the negative control group, which contains plain Gal4 binding domain and
activation domain, the SH3 and BIR with their ligands are displaying near zero interaction in all
three repeats contradicting previous structural studies (Nguyen et al., 1998; Liu et al., 2000).
To further probe the functionality of these domains and ligands, we chose split GFP assay to
verify the protein-protein interaction quantitatively (Figure 2C) (Sarkar et al., 2008). We took
advantage of the AProSS system to fuse the NGFP with a protein domain on the C terminus and the
CGFP with the corresponding ligand on the N terminus by the pAV vector. With original frGFP as
positive control and none-interaction pairs fused with split GFP as negative control, we observed
fluorescence signals across all six pairs (Figure 2D, G) providing a semi-quantitative result to
compare the relative interaction strength of different domain-ligand groups. The cells with YAP
domain and its ligand display the highest fluorescence intensity which indicates the YAP domain
and its ligand might have the highest affinity.
To assess the interaction quantitatively, we used flow cytometry to measure the intensity of
different sets of domains and ligands. The results of YAP and SH3 groups show a higher interaction
between domains and their ligands, GBD and BIR show a middle interaction with their ligands,
PDZ and FE show a weaker interaction with their ligands (Figure 2E).
One advantage of AProSS system is that one module can be assembled to different positions of
the GFP halves easily. We went on the assay the positional effect of the domain-ligand pairs and the
split GFP parts. We used this quantitative method to detect the differences between the two domains
fusion types, fusion at N-terminus and fusion at C-terminus. It shows that the ligands fused at C-
terminus and domains fused at N-terminus have higher intensity the other way around (Figure 2F).
3.3 Assembly efficiency of artificial protein scaffolds.
To examine the assembly efficiency of AProSS system, we set up assembly reactions ranging
from three to seven parts, using white/red ratio as evaluation. We tested two scenarios:
heterogeneous and homogenous assembly. We sub-cloned seven identical domains from pMV to
seven successive pLV vectors and assembled these domains into scaffolds ranging from three units
to seven units. Five-unit assembly yields over 90% white colonies. Using the same method, we also
tested the efficiency of seven-unit heterogeneous assembly. Colony PCR bands pattern of selected
white colonies could be used to check the accuracy of scaffolds assembly. It shows that over 80%
selected white colonies have correct scaffold genes in heterogeneous assembly. The assembly
efficiencies were higher in homogenous assembly, but the accuracy decreased accordingly (Table
2).
The results suggest that the AProSS method have a high efficiency when the number of domains
of the target scaffold is below seven. The heterogeneous assembly accuracy is slightly higher than
the homogenous assembly with the efficiency decreasing as the number of the parts increases in
both situations.
3.4 Altering metabolic flux using artificial scaffolds.
Violacein is a very useful natural product to test the efficacy of the scaffold proteins as it
produces a visible purple pigment. It is synthesized through five enzymatic steps and one non-
enzymatic reaction from two molecules of tryptophan in yeast (Lee et al., 2013) (Figure 3A).
Meanwhile, the side chain product deoxyviolacein could also be used as an indicator. We fused the
ligands of GBD, SH3, and PDZ to VioC, VioD and VioE genes of the violacein biosynthetic
pathway respectively and assembled them with VioA and VioB genes into the full violacein
pathway and transformed it into yeast strain JDY52, enabling the yeast to synthesize violacein
(Figure 3B). The GBD ligand and PDZ ligand were attached at the C-terminus of VioC and VioE
gene, and the SH3 ligand at the N-terminus of VioD gene. If the tails of the enzyme polypeptide
chains are embedding in the protein’s interior, fusing with ligands may change the protein’s
conformation and inactivate the enzyme. Our preliminary tests showed that when fusing SH3 ligand
at the C-terminus of VioD gene or VioB, the enzymatic activity decreased significantly.
To test the function of the designed scaffolds, we constructed three different scaffolds, each
containing five domains, and transformed them into yeast strains with all genes required for
violacein synthesis respectively. At the same time, frGFP was served as negative control and
transformed into these strains (Table 3, line 1-3). The violacein produced by these strains was
purified and analyzed by HPLC. Compared with the control set, the strain with scaffold GBD-
double SH3-double PDZ (GSSPP) achieved 29% increase of violacein production (Figure 3C) and
GBD-SH3-triple PDZ (GSPPP) achieved 63% increase of deoxyviolacein production (Figure 3D).
The other scaffolds also increase the yield of violacein to a certain level. These results demonstrated
that the scaffold proteins constructed by AProSS are valid and efficient. Meanwhile, this experiment
also illustrated the rate-limiting step in the pathway, for the yields of the products are positively
related to the copy number of the PDZ domain, suggest that VioE may be a key enzyme in the
pathway.
Violacein biosynthetic pathway exhibited several common characteristics for regular metabolic
pathway engineering. One of them is that the branched pathway structure leading to off-target side
reactions, both enzymatic and spontaneous (Lee et al., 2013). Fortunately, the intermediates of the
pathway are all colored and detectable. Thus, it could be used to demonstrate the change in
metabolic flux by introducing protein scaffolds (Figure 3E). We assembled scaffold GBD-PDZ and
scaffold SH3-PDZ, (Table 3, line 4, 5) enabling the scaffolds recruiting VioC and VioE to
synthesize deoxyviolacein or recruit VioD and VioE to synthesize protoviolaceinic acid which is
the precursor of violacein (Figure 4). In the experiment, the strain with scaffold SH3-PDZ produced
a 118% higher violacein to deoxyviolacein ratio than the strain with scaffold GBD-PDZ (Figure
3F).
4. Discussion
Pathway balancing through scaffolds does not take advantage from reducing the production of
over-produced RNA or protein, but by improving the efficiency of substrate transfer from enzyme
to enzyme, minimizing diffusion, before the substrate reacts with the enzyme. Scaffold-based
optimization techniques benefit from the formation of micro-domains with extremely increased
metabolite concentrations in the cytosol (Shetty et al., 2008). This means that in pathways where
enzyme diffusion is one of the limiting steps for the overall production, the scaffold-based method
can be applied orthogonally to all pathways to provide a significant improvement on pathway
balance.
The AProSS system in this study shows much potential in building multiple component fusion
proteins. It can not only construct simple fusion proteins such as his-tag labeled enzymes but is also
capable of assembling a large scaffold protein with over ten different components while ensuring
the position of each part. This system will surely ease the application of artificial scaffold proteins
in many further studies. From now on, we are no longer limited to the existing scaffolds but are able
to design any appropriate scaffolds. Moreover, the available domains are expanded to six and
possibly beyond in the future to suit more complex situations. Meanwhile, the AProSS system could
be applied in any protein fusion situations, such as tagging, enzyme fusion, etc. It tremendously
simplifies the cloning process, so that researchers could get the target fusion proteins as early as in
one week.
Although it has already been verified by many researchers that scaffolds can increase the
pathway yields in exogenous expression systems, none of the studies can optimize domain
combination efficiently. AProSS provides an easy-to-use method to screen for the optimal protein
scaffolds in any metabolic pathway. Benefit from the standardization idea, the different domains
may get the same joints after type IIs restriction enzyme digestion. Thus, we could ignore the
differences among them and pool the different pLVs together with the corresponding pAV to
construct scaffolds randomlycombinatorially. Afterward, we would get a random scaffold pool
which contain a large number of different scaffolds with different domain varieties, repeats and
combinations. We could transform this scaffold pools to regulate the pathways in a single step. By
analysis the yield of target products, we may find the best performance strain. Then, we could
recycle recover the plasmid in this strain and sequence it to find the best domain combination
(Figure 4). Meanwhile, combined with a mathematical model, the optimal domain combination may
help us to understand balancing for heterologous pathways in normal expression systems.
With the domain library expansion, constructing artificial protein scaffolds to regulate metabolic
pathways could be applied at more and more complex situations both in prokaryotes and eukaryotes
systems thanks to the AProSS system. The standard, convenient assembly operations of this method
will reduce the bench work vastly for metabolic engineers. Meanwhile, the ease to assemble any
domain combination gives researchers a simple way to regulate the pathways at the post-
translational level which may also reduce the work for basic strain construction. Following the
potential optimal strain selection methods: scaffolds pool construction to selection, we may get
higher yields of products, at the same time understand the scientific principle of certain pathways.
These advantages will make AProSS a useful tool in metabolic pathways regulation in the future.
ACKNOWLEDGEMENT
This work was supported by the National Natural Science Foundation of China (31725002) and
by Bureau of International Cooperation, Chinese Academy of Sciences (172644KYSB20170042) to
JD. The work is also supported by UKRI BBSRC award (BB/P02114X/1) and EPSRC award
(EP/P017401/1) to YC.
REFERENCES
Alonso-gutierrez, J., Chan, R., Batth, T. S., Adams, P. D., Keasling, J. D., Petzold, C. J., & Soon, T. (2013). Metabolic engineering of Escherichia coli for limonene and perillyl alcohol production. Metabolic Engineering, 19, 33–41. https://doi.org/10.1016/j.ymben.2013.05.004
Awan, A. R., Blount, B. A., Bell, D. J., Shaw, W. M., Ho, J. C. H., McKiernan, R. M., & Ellis, T. (2017). Biosynthesis of the antibiotic nonribosomal peptide penicillin in baker’s yeast. Nature Communications, 8(May), 1–8. https://doi.org/10.1038/ncomms15202
Bailey, J. E. (1991). Toward a science of metabolic engineering. Science,252(5013), 1668-75.
Chemler, J. A., Fowler, Z. L., McHugh, K. P., & Koffas, M. A. G. (2010). Improving NADPH availability for natural product biosynthesis in Escherichia coli by metabolic engineering. Metabolic Engineering, 12(2), 96–104. https://doi.org/10.1016/j.ymben.2009.07.003
Cone, M. C., Yin, X., Grochowski, L. L., Parker, M. R., & Zabriskie, T. M. (2003). The blasticidin S biosynthesis gene cluster from Streptomyces griseochromogenes: Sequence analysis, organization, and initial characterization. ChemBioChem, 4(9), 821–828. https://doi.org/10.1002/cbic.200300583
Delebecque, C. J., Lindner, A. B., Silver, P. A., & Aldaye, F. A. (2011). Organization of Intracellular Reactions with Rationally Designed RNA Assemblies. Science, 333(6041), 470–474. http://doi.org/10.1126/science.1206938
Dueber, J. E., Wu, G. C., Malmirchegini, G. R., Moon, T. S., Petzold, C. J., Ullal, A. V, … Keasling, J. D. (2009). Synthetic protein scaffolds provide modular control over metabolic flux. Nature Biotechnology, 27(8), 753–759. http://doi.org/10.1038/nbt.1557
Engler, C., Kandzia, R., & Marillonnet, S. (2008). A one pot, one step, precision cloning method with high throughput capability. PloS One, 3(11), e3647. http://doi.org/10.1371/journal.pone.0003647
Engler, C., Gruetzner, R., Kandzia, R., & Marillonnet, S. (2009). Golden gate shuffling: a one-pot DNA shuffling method based on type IIs restriction enzymes. PloS One, 4(5), e5553. http://doi.org/10.1371/journal.pone.0005553
Guo, Y., Dong, J., Zhou, T., Auxillos, J., Li, T., Zhang, W., … Dai, J. (2015). YeastFab: The design and construction of standard biological parts for metabolic engineering in Saccharomyces cerevisiae. Nucleic Acids Research, 43(13), e88. http://doi.org/10.1093/nar/gkv464
Horn, A. H. C., & Sticht, H. (2015). Synthetic Protein Scaffolds Based on Peptide Motifs and Cognate Adaptor Domains for Improving Metabolic Productivity. Frontiers in Bioengineering and Biotechnology, 3. https://doi.org/10.3389/fbioe.2015.00191
Jin, Y., & Alper, H. (2005). Improvement of xylose uptake and ethanol production in recombinant Saccharomyces cerevisiae through an inverse metabolic engineering approach. Applied and Environmental Microbiology, 71(12), 8249–8256. https://doi.org/10.1128/AEM.71.12.8249
Jones, J. A., Toparlak, T. D., & Koffas, M. A. G. (2015). Metabolic pathway balancing and its role in the production of biofuels and chemicals. Current Opinion in Biotechnology, 33, 52–59. http://doi.org/10.1016/j.copbio.2014.11.013
Keasling, J. D. (2010). Manufacturing molecules through metabolic engineering. Science (New York, N.Y.), 330(6009), 1355–1358. http://doi.org/10.1126/science.1193990
Kim, S., & Hahn, J.-S. (2014). Synthetic scaffold based on a cohesin-dockerin interaction for improved production of 2,3-butanediol in Saccharomyces cerevisiae. Journal of Biotechnology, 192PA, 192–196. http://doi.org/10.1016/j.jbiotec.2014.10.015
Lee, J. W., Na, D., Park, J. M., Lee, J., Choi, S., & Lee, S. Y. (2012). Systems metabolic engineering of microorganisms for natural and non-natural chemicals. Nature Chemical Biology, 8(6), 536–546. http://doi.org/10.1038/nchembio.970
Lee, M. E., Aswani, A., Han, A. S., Tomlin, C. J., & Dueber, J. E. (2013). Expression-level optimization of a multi-enzyme pathway in the absence of a high-throughput assay, 41(22), 10668–10678. https://doi.org/10.1093/nar/gkt809
Li, Y., & Tan, H. (2017). Biosynthesis and molecular regulation of secondary metabolites in microorganisms, 60(9), 935–938. https://doi.org/10.1007/s11427-017-9115-x
Liu, Z., Sun, C., Olejniczak, E. T., Meadows, R. P., Betz, S. F., Oost, T., … Fesik, S. W. (2000). Structural basis for binding of Smac/DIABLO to the XIAP BIR3 domain. Nature, 408(6815), 1004–8. http://doi.org/10.1038/35050006
Macias, M. J., Hyvönen, M., Baraldi, E., Schultz, J., Sudol, M., Saraste, M., & Oschkinat, H. (1996). Structure of the WW domain of a kinase-associated protein complexed with a proline-rich peptide. Nature. http://doi.org/10.1038/382646a0
Meiyappan, M., Birrane, G., & Ladias, J. A. A. (2007). Structural Basis for Polyproline Recognition by the FE65 WW Domain. Journal of Molecular Biology, 372(4), 970–980. http://doi.org/10.1016/j.jmb.2007.06.064
Meng, X., Wang, W., Xie, Z., Li, P., Li, Y., Guo, Z., … Chen, Y. (2017). Neomycin biosynthesis is regulated positively by AfsA-g and NeoR in Streptomyces fradiae CGMCC 4 . 7387, 60(9), 980–991. https://doi.org/10.1007/s11427-017-9120-8
Nguyen, J. T., Turck, C. W., Cohen, F. E., Zuckermann, R. N., & Lim, W. A. (1998). Exploiting the basis of proline recognition by SH3 and WW domains: design of N-substituted inhibitors. Science (New York, N.Y.), 282(5396), 2088–92. http://doi.org/10.1126/science.282.5396.2088
Nielsen, J. C., & Nielsen, J. (2017). Development of fungal cell factories for the production of secondary metabolites: Linking genomics and metabolism. Synthetic and Systems Biotechnology, 2(1), 5–12. https://doi.org/10.1016/j.synbio.2017.02.002
Richardson, S. M., Wheelan, S. J., Yarrington, R. M., & Boeke, J. D. (2006). GeneDesign: Rapid , automated design of multikilobase synthetic genes GeneDesign : Rapid , automated design of multikilobase synthetic genes, 550–556. http://doi.org/10.1101/gr.4431306
Sachdeva, G., Garg, A., Godding, D., Way, J. C., & Silver, P. a. (2014). In vivo co-localization of enzymes on RNA scaffolds increases metabolic production in a geometrically dependent manner. Nucleic Acids Research, 42(14), 9493–503. http://doi.org/10.1093/nar/gku617
Sarkar, M., & Magliery, T. J. (2008). Re-engineering a split-GFP reassembly screen to examine RING-domain interactions between BARD1 and BRCA1 mutants observed in cancer patients. Molecular bioSystems, 4(6), 599–605. http://doi.org/10.1039/b802481b
Shetty, R. P., Endy, D., & Knight, T. F. (2008). Engineering BioBrick vectors from BioBrick parts. Journal of Biological Engineering, 2, 5. http://doi.org/10.1186/1754-1611-2-5
Tan, S. Z., Manchester, S., & Prather, K. L. J. (2016). Controlling Central Carbon Metabolism for Improved Pathway Yields in Saccharomyces cerevisiae. https://doi.org/10.1021/acssynbio.5b00164
Ukibe, K., Hashida, K., Yoshida, N., & Takagi, H. (2009). Metabolic engineering of Saccharomyces cerevisiae for astaxanthin production and oxidative stress tolerance. Applied and Environmental Microbiology, 75(22), 7205–7211. https://doi.org/10.1128/AEM.01249-09
Yadav, V. G., De Mey, M., Giaw Lim, C., Kumaran Ajikumar, P., & Stephanopoulos, G. (2012). The future of metabolic engineering and synthetic biology: Towards a systematic practice. Metabolic Engineering, 14(3), 233–241. http://doi.org/10.1016/j.ymben.2012.02.001
Yuan, T., Guo, Y., Dong, J., Li, T., Zhou, T., Sun, K., … Wu, Q. (2017). Construction, characterization and application of a genome- wide promoter library in Saccharomyces cerevisiae, 11(1), 107–108. https://doi.org/10.1007/s11705-017-1621-7
Table1. Standardized Prefix and Suffix Sequences for AProSS*
Plasmid name BsaI prefixes BsaI suffixes BsmBI prefixes BsmBI suffixes
pMV GGTCTCTCGAG GGTCTCATCAA CGTCTCGCTCG CGTCTCGTTGA
pLV
pLV1 GGTCTCTGGAA GGTCTCTCGCC
CGTCTCGCGAG CGTCTCGTCAA
pLV2 GGTCTCTGGCG GGTCTCTACCC
pLV3 GGTCTCTGGGT GGTCTCTCTGA
pLV4 GGTCTCTTCAG GGTCTCTTGGA
pLV5 GGTCTCTTCCA GGTCTCTTGCT
pLV6 GGTCTCTAGCA GGTCTCTAC GA
pLV7 GGTCTCTTCGT GGTCTCTCACT
pLV8 GGTCTCTAGTG GGTCTCTTAGA
pLV9 GGTCTCTTCTA GGTCTCTTACC
pLV10 GGTCTCTGGTA GGTCTCTCCTG
pLV11 GGTCTCTCAGG GGTCTCTCCTG
pLV12 GGTCTCTCGAC GGTCTCTGTCG
pLV13 GGTCTCTCTGA GGTCTCTTCAG
pLV14 GGTCTCTACTC GGTCTCTGGTT
pLV15 GGTCTCTAACC GGTCTCTGTTC
pAV pAV1-15 GGTCTCTTTCC Match to the corresponding pLVn** CGTCTCGGATG CGTCTCGGCTA
* Bold 6 bp sequences are recognition sites; underlined 4 base sequences are overhang sites. All sequences are written
5′ to 3′ on the top strand of the final part.
**pLVn tail = pLVn+1 head = pAVn tail. Inverse orientation.
Table2. The efficiency of AProSS assembly
Type Domain number Constitution Assembly
efficiency Accuracy
Different domains
3 (N)CGFP*-BIR-YAP(C) 98% 12/12
4 (N)CGFP-BIR-YAP-FE(C) 94% 12/12
5 (N)CGFP-BIR-YAP-FE-PDZ(C) 92% 11/12
6 (N)CGFP-BIR-YAP-FE-PDZ-SH3(C) 82% 10/12
7 (N)CGFP-BIR-YAP-FE-PDZ-SH3-GBD(C) 74% 10/12
Identical domains
3 (N)SH3-SH3-SH3(C) 100% 10/12
4 (N)SH3-SH3-SH3-SH3(C) 96% 11/12
5 (N)SH3-SH3-SH3-SH3-SH3(C) 95% 9/12
6 (N)SH3-SH3-SH3-SH3-SH3-SH3(C) 93% 6/12
7 (N)SH3-SH3-SH3-SH3-SH3-SH3-SH3(C) 80% 2/12
*We only have 6 different domains, so to test the assembly efficiency under 7 parts, we use CGFP as the first domain,
which has a similar size to scaffold domains.
Table3. The scaffolds to test the efficacy for recruitment enzymes in S. cerevisiae.
Name Code name Domain number ORF Constitution PRO TER
GSSPP S1 5 (N)GBD-SH3-SH3-PDZ-PDZ(C)
pTEF2 tCYC1
GSPPP S2 5 (N)GBD-SH3-PDZ-PDZ-PDZ(C)
GSSSP S3 5 (N)GBD-SH3-SH3-SH3-PDZ(C)
GP S4 2 (N)GBD-PDZ(C)
SP S5 2 (N)SH3-PDZ(C)
Figure 1 The three-vector design of AProSS and the system workflow. (A) The Schema graph of
the three group of vectors. (B) The workflow that using AProSS method to construct artificial
protein scaffolds to regulate metabolic pathways in S. cerevisea.
Figure 2 Domain-ligand interaction efficiency verification. (A) The model for yeast two-hybrid
assay to verify the interaction of domains and ligands qualitatively. (B) The results for yeast two-
hybrid assay. Show the growth conditions of different interaction groups on uracil lacking plates.
(C) The model for split GFP assay to verify the interaction of domains and ligands quantitatively.
(D) The results for split GFP assay. Show the florescence under microscope, 40× objective lens. (E)
The FACS analysis data of different domain-ligand group. The YAP group shows the highest
interaction intensity. (F) The FACS analysis data for different fusion type. When fusing ligand with
NGFP and domains with CGFP, the florescent intensity is higher than they fuse reverse. (G)
Florescence signal in yeast cells with different domain-ligand groups, 100× objective lens.
Figure 3 Protein Scaffolds to regulate violacein biosynthesis in S. cerevisiae. (A) The violacein
biosynthetic pathway. (B) The model for scaffolds recruiting enzymes and forming a reaction center
in cells. VioC VioD and VioE genes have been fused with different ligands and would be recruited
by the certain scaffolds. (C) The influence of adding scaffolds into yeast cells for violacein yielding.
(D) The influence of adding scaffolds into yeast cells for the by-product deoxyviolacein yielding.
(E) The model for scaffolds change the flux of metabolic pathway. (F) The relative concentration of
violacein and deoxyviolacein under the regulation of different scaffolds. The left scaffold leads to
produce deoxyviolacein, while the right scaffold leads to produce violacein.
Fig
u re
4
The workflow of pools method regulation.