CyanoConstruct: simple platform for cyanobacterial expression construct assembly and translational tuning

Abstract

Background

Despite the advances in synthetic biology, the construction of efficient and balanced artificial pathways remains challenging and a general bottleneck in bacterial strain development. Although translational tuning can be applied for balancing consecutive catalytic steps at protein expression level, the potential remains underexplored in cyanobacterial engineering. To complement existing modular cloning systems for this purpose the objective here was to simplify the construct assembly procedure to make translational tuning more accessible for expression optimization in cyanobacteria.

Results

This study describes the design and use of a one-pot DNA construct assembly system (CyanoConstruct) for the generation transformation-ready multi-gene expression plasmids in a single Golden Gate reaction. This approach allows the user to select the ribosome binding site element (RBS) for each target gene, thus serving as a tool for independently modulating the translation efficiency of the individual overexpressed enzymes. For easy adaptation, a custom online tool (www.cyanoconstruct.com) guides the sequence design of new compatible parts and the assembly of constructs from user-specified parts in silico. We demonstrate the use of the system by assembling different two-gene and three-gene expression constructs from parts selected specifically for optimal performance in Synechocystis sp PCC 6803; the constructs functioned as intended in vivo and showed different pathway fluxes construed by alternative RBS combinations. The efficiency and specificity of the assembly were shown to be high, enabling the generation of the final expression plasmids from the library parts in one assembly cycle.

Conclusions

CyanoConstruct offers a simple strategy for building bacterial operon-based expression constructs, specifically facilitating the use of modular cloning systems for RBS optimization in routine cyanobacterial engineering. With the help of the web-based design tool that also serves as a sequence repository, the part library described in this work (https://www.addgene.org/browse/article/28263931/) can be easily expanded with user-specified sequences. By increasing the throughput for generating pathway variants with different translational patterns, the system is expected to advance the design of more efficient strains with higher flux to the desired end-product, thereby contributing to the development of next-generation biotechnologies.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12896-025-01098-7.

Introduction

Construct assembly & different cloning methods in cyanobacterial engineering

Photosynthetic cyanobacteria are interesting engineering targets as potential next-generation biotechnology hosts, due to their ability to catalyze light-driven production of carbon based chemicals directly from CO₂ [1, 2]. Established synthetic biology tools and increasing knowhow on expression regulation in cyanobacteria enable more complex and fine-tuned genetic modifications to generate new and enhanced metabolic pathways with higher productivity [3–5]. Strain engineering typically involves the assembly of DNA constructs from a number of genetic elements, for which a wide selection of alternative molecular assembly tools have been developed. Modular BioBrick standards [6], that rely on repeated rounds of subcloning, have been frequently implemented for building constructs in cyanobacterial engineering [7–10]. Iterative rounds of enzymatic digestion, ligation, transformation and extraction, however, is laborious and time consuming for complex assemblies, which has prompted the use of methods that enable the combination of multiple DNA parts in a single reaction for higher preparative throughput. Gibson assembly [11] is one of the first modular systems that allows the combination of multiple DNA sequences in a single reaction used in cyanobacterial engineering [12–15]. The system is based on PCR to generate DNA parts with appropriate compatible overhangs, and provides position-independent assembly of multiple fragments using 5’ exonuclease, DNA polymerase and ligase in a single tube. While Gibson assembly is easily adaptable for different applications with relatively low user threshold, the use of PCR makes the system susceptible to spontaneous mutations during amplification, and requires sequencing which is not always practical. Golden Gate cloning [16, 17] is another extensively used strategy for one-pot multipart assembly. It is based on Type IIS restriction enzymes that cleave the DNA outside the restriction enzyme’s recognition sequence, and allow the generation of multiple position-specific sequence overhangs that can be ligated in a predetermined order in a single reaction. This has been used in the standardized modular system CyanoGate [18] adapted in cyanobacterial engineering based on the eukaryotic MoClo syntax [19, 20]. CyanoGate is highly versatile and applied for diverse uses including comparing alternative genetic control elements [21–23], generating markerless deletions using homologous recombination [24] or CRISPR/Cas9 mediated editing [23], and mapping neutral integration sites for new cyanobacterial hosts [23]. In addition, there are variations of the modular cloning tools such as BASIC [25–27], Start-Stop assembly [28] and AQUA Cloning [29, 30] that have been applied and offer alternatives for construct assembly in cyanobacterial engineering. Importantly, there is no single assembly standard that would be practical or optimal for all purposes, and the most convenient choice depends on the specific task, user preference, required throughput, and available resources including access to pre-existing DNA part libraries.

Need for translational tuning in metabolic engineering

In microbial metabolic engineering, the introduction of new cellular functions typically relies on the expression of exogenous genes that must operate in a new metabolic context, and under the regulation of non-native control circuits. As the artificial systems have not been optimized to function in concert through evolution, finding an optimal balance between the catalytic reactions and the surrounding cellular metabolic network presents a complex engineering challenge. In cyanobacterial engineering, our objective is to efficiently funnel fixed CO₂ towards specific target pathways and metabolic products, while maintaining control of the highly flexible metabolic environment that is well buffered against changes. Despite the accumulation of detailed species-specific molecular information, the improvement of computer-based modeling technologies, and the development of effective synthetic biology tools, the task is not trivial. Expression systems may suffer from either low catalytic activities, metabolic imbalance or toxicity effects that compromise the well-being of the host, all resulting in systematic underutilization of the cell’s biosynthetic resources which otherwise could be directed for productive use. While promoters are conventionally used as the master switches to control expression, the use of alternative ribosome binding site (RBS) elements provides a promising strategy to tune the expression of individual proteins at translational level in engineered cyanobacterial systems [22, 31, 32]. This type of translational modulation, based on the use of different RBSs with varying efficiencies, has been demonstrated to be a powerful approach in improving the metabolic flux through complex operons in Escherichia coli (E. coli from herein) [33–35]. The concept has been successfully applied also in improving the expression of independent key enzymes in cyanobacterial pathway engineering, including heterologous ethylene forming enzyme (EFE) in ethylene biosynthesis [31, 36], heterologous bisabolene synthase in alpha-bisabolene production [15], and native acetoacetyl-CoA reductase (PhaB) in 3-hydroxybutyrate production [37] in Synechocystis sp PCC 6803 (Synechocystis from herein). This has also been implemented in other cyanobacterial hosts as demonstrated by optimizing the expression of geranyl pyrophosphate synthase (GPPS) and limonene synthase (LS) in limonene biosynthesis in Synechococcus elongatus UTEX 2973 [38] and Synechococcus elongatus PCC 7942 [39], respectively. In addition, there are several combinatorial examples of balancing the translation of several pathway enzymes in concert, as exemplified by the expression of pyruvate decarboxylase (Pdc) and aldehyde reductase (Adh) in ethanol biosynthesis [26], and acetolactate synthase (AlsS), 2-acetolactate decarboxylase (AlsD) and Adh in 2,3-butanediol biosynthesis [32]. Although translation appears to have a limited role in regulating native protein levels in Synechocystis [40], the literature strongly suggests that translational tuning can offer a strategy for improving productivity in cyanobacterial systems, and could provide advantage in engineering multi-gene pathways as well as testing new cyanobacterial hosts.

Challenges in cyanobacterial translational tuning

In E. coli the translational control has been extensively studied, and the efficiency of a specific RBS can be predicted to a relatively reliable extent [41–43]. In cyanobacteria this is not as straightforward [44, 45], partially due to the lack of comparative information on the consensus sequences and associated sequence elements at a functional level. As demonstrated in several contexts, the translational efficiency is not determined by the 5’ UTR RBS region alone, but the outcome appears to be heavily dependent also on the downstream gene sequence as seen in Synechocystis [22, 31] and in Synechococcus [32]. Although such context dependence is a known phenomenon in biology [46], resulting at least in part from sequence-level interactions between the RBS region and the coding sequence, the extent to which this affects relative translation efficiencies in Synechocystis seems worth attention. The studies have demonstrated pronounced variance in translational efficiencies of different RBS-gene combinations, which frequently result in significant differences in expression levels from one target gene to another. Besides inconsistencies between the translation of yellow fluorescent protein (sYFP2), codon optimized green fluorescent protein (GFPmut3) and EFE [31], gene-specific variance has been reported in the expression levels of Pdc and eYFP expressed under the same RBS [26]. The effect has also been observed when expressing ketoreductase (LfSDR1M50), enoate reductase (YqjM) and Baeyer–Villiger monooxygenase (CHMO_mut) under multiple different RBSs in Synechocystis [22], and between heterologous acetolactate synthase (AlsS), 2-acetolactate decarboxylase (AlsD) and Adh expressed under alternative RBSs in Synechococcus [32]. From the viewpoint of rational engineering this is highly inconvenient, as the RBS prediction tools based on function in E. coli cannot be reliably used to estimate relative translational patterns in cyanobacteria [22, 31, 32] and the performance must be experimentally verified through trial and error for each new RBS-gene combination. In the case of artificial multi-gene pathways, the problem may become even more pronounced, as even small changes in expression efficiency of individual enzymes in a pathway may result in large changes in the flux through the successive reaction steps. One intuitive solution to this is to generate parallel variations of the same construct, using sets of different RBS elements in front of each target gene, followed by functional screening of the best-performing clones. We have previously adapted a modular subcloning system to achieve this [31], but the iterative assembly procedure is relatively labor-intensive for constructs composed of many individual elements as in multi-gene operons. The possibility to independently change the RBS regions has also been introduced in the multi-component assembly systems CyanoGate [22], BASIC [25] and Start-Stop Assembly [28]. In these hierarchical systems, higher level complexity can be obtained through intermediate assembly steps which depend on the specific assembly system and the number and type of elements that are being fused together. The independent genetic parts, including the RBS elements and the target genes, are typically combined into smaller functional units in the first reaction, followed by transformation and amplification in E. coli, and successive rounds of assembly to generate the final constructs.

Objectives of the study

With the realization that many DNA assembly standards already exist, the prime motivation here was to develop a complementary high-throughput system that could lower the user threshold for taking advantage of translational tuning in cyanobacteria. The goal was to design a simple strategy for compiling ready-to-use multi-gene constructs in a single one-pot reaction, while allowing the user to select the translation control elements independently for every target gene. The system takes advantage of the existing RBS library [31] that has been used in expressing a number of different proteins at varying levels in Synechocystis, but also gives the user full freedom to modify the 5’ UTR region as needed. With the support of an online tool that guides construct design and the domestication of new parts of interest, the setup and use of the system should be easily accessible, even without prior cloning experience.

Results

Design of the CyanoConstruct assembly platform

The cloning platform CyanoConstruct was designed for one-pot assembly of transformation-ready DNA constructs to facilitate protein expression, tagging and flexible translational tuning in cyanobacteria. The system uses Golden Gate cloning to generate circular plasmid constructs from up to nine DNA fragments in a single reaction (Fig. 1a). The design enables demonstration of one-step assembly of functional single-gene expression constructs (Fig. 1b) or operons for two (Fig. 1c) and three target proteins (Fig. 1d), with independently selected RBS elements for protein-specific translational control. The Type IIS restriction enzyme BbsI is used to generate the position-specific four-base overhangs (Fig. 2a; Supplementary Table S1) [47] that allow complementary DNA parts to assemble in the correct predetermined order (Fig. 2b). To accomplish this at high fidelity, the system takes advantage of optimized overhang pairs selected for highest accuracy with minimal mispairing [47]. The DNA parts were designed to assemble directly into the desired expression plasmid backbone carrying a short plasmid insertion domain sequence (Fig. 2c, Supplementary Figure S1a), thereby allowing immediate transformation of the reaction mix in E. coli for the selection and amplification of the final constructs. The RBS parts used for translational control were based on an element library compiled and characterized earlier for Synechocystis [31] (Fig. 2d). The CyanoConstruct assembly procedure, from construct design to the verification of the transformed cyanobacterial strains, takes place in five steps as described in Fig. 3a-e.

Fig. 1.

Simplified representation of the CyanoConstruct assembly system (see Fig. 2 for sequence details) a) CyanoConstruct is a Golden Gate -based assembly platform optimized for combining up to nine DNA fragments in a single one-pot reaction into transformation-ready plasmids that are directly compatible with Synechocystis. The system has been specifically designed for building expression constructs for b) a single target gene, c) operons composed of two target genes or d) operons composed of three target genes, while allowing easy exchange of the translational control elements (RBS) preceding each gene. The recipient backbone plasmid can be a replicative vector or an integrative construct, and can be easily exchanged for user-specific plasmids to be used in other cyanobacteria. The assembled fragments are maintained as a library of donor plasmids that can be amplified in E. coli, or they can be supplemented as linear PCR products. p: plasmid backbone; P: promoter; RBS: translational control element; GOI: gene of interest (target gene, indicated in red text); TT: transcription terminator

Fig. 2.

The compatible sequence overhangs and CyanoConstruct assembly. a) The library constructs contain specifically positioned Type IIS restriction enzyme BbsI recognition sequences 5’-GAAGAC-3’ (red text) that enable the generation of the four base pair single-stranded overhangs specific for the parts as shown here for P0-P8 in 9-fragment assembly. b) The digested fragments assemble in the correct order based on the compatible position-specific overhangs. These fragments no longer contain the BbsI recognition sequences, so once ligated, they are not cleaved again. The last fragment, the transcription terminator (TT; yellow highlight), links the operon to the backbone, and varies depending on the total number of parts in the assembly. c) The plasmid insertion domain sequence allows the fragments P1-P8 to be integrated into the recipient plasmid. This sequence element should be added to any user-specified vector that is to be used as the recipient backbone in the assembly. d) Alignment of the RBS, start codon and the coding sequence, as exemplified by the P4-P5 interphase; The translational control elements (RBSs at P2, P4 and P6) are composed of an invariable insulator sequence (grey highlight), the variable RBS (green highlight) and the start codon (underline). Consequently, the assembly overhangs are positioned at the beginning of the GOI open reading frame, and require two additional base pairs to prevent frameshift, thus adding two extra residues at the target protein N-terminus (Supplementary Figure S1)

Fig. 3.

Stepwise procedure of the CyanoConstruct assembly from sequence design to verified cyanobacterial expression strains a) The DNA parts for the assembly are designed by removing internal BbsI sites and addition of position-specific sequence overhangs for each fragment. The web-based design tool assists this process and compiles the output sequence file for verification. b) The DNA parts with the position-specific overhangs are ordered via commercial gene synthesis or generated by PCR. The parts are stored as inserts in high copy number plasmids, and maintained as transformed E. coli glycerol preps that allow easy amplification. c) The assembly reaction is carried out in a single reaction tube with all the combined DNA parts (i.e. plasmid minipreps) in temperature cycling with BbsI and T7 ligase (see Tables 2 and 3). d) The resulting reaction mix is transformed in E. coli, and the clones carrying the circularized plasmid are selected based on antibiotic resistance and verified by colony PCR. The final construct is stored in the freezer library as an E. coli glycerol prep and a plasmid miniprep. e) The extracted plasmid construct is transformed in Synechocystis, followed by selection of antibiotic-resistant colonies, verification by PCR, and storage at -80 °C. The procedure follows the same steps for assembling single-gene constructs or operons composed of two target genes (seven-fragment assembly) or three target genes (nine-fragment assembly)

Online design tool to guide user to set up the assembly

A web-based sequence tool (www.cyanoconstruct.com) was developed to facilitate the design of the CyanoConstruct assembly reaction for user-specified parts (Fig. 4). The tool was written in Python with Flask web server and SQLAlchemy database libraries to enable in silico design of the position-specific cloning motifs that allow the parts to be assembled in the intended order. The work flow is divided into two phases, design and assembly. In the design phase the user imports the desired parts as individual DNA sequences (plain text), specifies the type of the sequence (i.e. promoter, RBS, GOI or terminator), and names each part. Apart from the plasmid insertion domain (Fig. 2c, Supplementary Figure S1a), the tool does not accept sequences with BbsI recognition sequences (GAAGAC), which are prompted to be removed by the user. After the imported entries are saved in the tool’s sequence library, the desired position is specified for each part (i.e. P0-P8; see Fig. 2; Supplementary Table S1), assigning corresponding cloning motifs from a set of optimized overhangs [47]. In the assembly phase, the graphical component selection interface allows the user to pick compatible parts from the library for a desired construct. As the output, the tool provides annotated GenBank files for (i) each part with their cloning motifs, and (ii) the entire construct after assembly. In case new parts are generated by PCR from existing sequence templates, the tool also suggests overhang primer sequences for appending desired position-specific cloning motifs on the amplified fragments.

Fig. 4.

CyanoConstruct sequence design tool workflow. In the design phase (the top rank), the user inputs DNA part sequences and specifies the intended position of those parts in the final construct. In the assembly phase (the bottom rank), the user selects the parts in the assembly wizard and submits them to generate the output sequences for proofing. The dotted arrows indicate steps carried out by the user outside the tool. If the user has already designed parts they may proceed directly to the assembly phase. The tool is accessible at www.cyanoconstruct.com

Designing constructs for testing the assembly in practice

To evaluate the assembly system in practice, two different types of expression constructs were designed. All the DNA parts used for these constructs are listed in Table 1 with the corresponding sequence information in Supplementary Figures S1-S2. The first construct was composed of seven independent DNA fragments (i.e. 7-fragment assembly) (Fig. 1c) that together form a two-gene operon expressing sYFP2 and EFE (Fig. 5a-b; Table 1) enabling easy detection in vivo [31]. Two alternative plasmid backbones were used for the purpose, a pDF-based [31] replicative plasmid with a broad host-range replicon RSF1010 compatible with different cyanobacteria [12, 57–61] (Fig. 5a), and an integrative backbone (Fig. 5b) based on pSI1b [49] targeting slr1311 (psbA2) in the Synechocystis chromosome. The second construct was composed of nine independent DNA fragments (i.e. 9-fragment assembly) (Fig. 1d) constituting a three-gene ethanol operon expressing Adh, Pdc and pyruvate kinase (PK) (Fig. 5c; Table 1). These enzymes catalyze the last three consecutive steps in ethanol biosynthesis [62] that can be quantified from the culture medium. Apart from the backbone vectors, all the domesticated DNA fragments in the assemblies (Table 1) were ordered as ready-to-use synthetic inserts in high copy number carrier plasmids. These were stored as plasmid minipreps for the assembly, in parallel to corresponding E. coli glycerol freezer stocks used for plasmid amplification.

Table 1.

List of all DNA parts designed and used in construct assembly in this study (Fig. 5), indicating the relative part position in the final construct as defined by the position-specific overhangs, the description of part source and function, and the carrier plasmid backbone that allows part amplification in E. coli. The cloning motifs and overhangs for each part are described in Supplementary Table S1, and the full sequences in Supplementary Figs. 1–2

Position	Name	Part description	Carrier plasmid	Source
P0	pDFC	Replicative expression plasmid (shuttle vector) backbone	-	[31, 48]
	pSIC	Backbone for construct integration in Synechocystis locus slr1311	-	[49]
P1	PA1lacO−1	IPTG-inducible Plac promoter variant with lacIq repressor and RiboJ	pCloneEZ-NRS	[48]
P2	RBSS3	RBS from C-phycocyanin β gene cpcB; sll1577 (Synechocystis)	pCloneEZ-NRS	[31]
	RBSS4	RBS from photosystem II D1 gene psbA2; slr1311 (Synechocystis)	pCloneEZ-NRS	[31]
	RBSB	RBS designed in silico [50] for high expression in E. coli	pCloneEZ-NRS	[31, 33]
P3	syfp2	Gene encoding yellow fluorescent protein variant	pCloneEZ-NRS	[31, 51]
	adh	Gene encoding aldehyde reductase slr1192 (Synechocystis)	pCloneEZ-NRS	[52, 53]
P4	RBSS3	RBS from C-phycocyanin β gene cpcB; sll1577 (Synechocystis)	pCloneEZ-NRS	[31]
	RBSS4	RBS from photosystem II D1 gene psbA2; slr1311 (Synechocystis)	pCloneEZ-NRS	[31]
	RBSS5	RBS from RuBisCO large subunit gene rbcL; slr0009 (Synechocystis)	pCloneEZ-NRS	[31]
	RBSB	RBS designed in silico [50] for high expression in E. coli	pCloneEZ-NRS	[31, 33]
P5	efe	Gene encoding Ethylene Forming Enzyme (Pseudomonas syringae)	pCloneEZ-NRS	[31, 54]
	pdc	Gene encoding pyruvate decarboxylase (Zymomonas mobilis)	pCloneEZ-NRS	[55]
P6	RBSS3	RBS from C-phycocyanin β gene cpcB; sll1577 (Synechocystis)	pCloneEZ-NRS	[31]
	RBSS4	RBS from photosystem II D1 gene psbA2; slr1311 (Synechocystis)	pCloneEZ-NRS	[31]
	RBSB	RBS designed in silico [50] for high expression in E. coli	pCloneEZ-NRS	[31, 33]
	TT	rrnB T1 terminator from pTrc99A	pCloneEZ-NRS	[48]
P7	pk	Gene encoding pyruvate kinase (Enterococcus faecalis)	pCloneEZ-NRS	[56]
	pk*	Gene encoding pyruvate kinase (Enterococcus faecalis)	pUC57	[56]
P8	TT	rrnB T1 terminator derived from pTrc99A	pCloneEZ-NRS	[48]
	TT*	rrnB T1 terminator derived from pTrc99A	pUC57	[48]

* Tested redesigned overhang for P7-P8 junction

Fig. 5.

Representation of different types of reporter constructs assembled in this work. a) The seven-fragment reporter construct was composed of two genes syfp2 (encoding yellow fluorescent protein) and efe (encoding ethylene forming enzyme), the independently assigned RBS elements, the lac promoter variant P_A1lac0−1 (including also the repressor gene lacI), transcription terminator, and the replicative plasmid backbone pDFC. b) The corresponding integrative operon was identical but assembled in the pSIC backbone carrying upstream and downstream sequences of Synechocystis gene slr1311 for homologous recombination. c) The nine-fragment construct was as a) but composed of three reporter genes adh (encoding aldehyde reductase), pdc (encoding pyruvate decarboxylase) and pk (encoding pyruvate kinase) and their assigned RBS elements. See Supplementary Figures S1-S2 for sequence details

7-fragment assembly functions effectively with high specificity

The assembly conditions (Tables 2 and 3) were initially selected based on parameters described in literature [16] and several successive rounds of preliminary testing. In the default setup, the seven individual plasmids carrying the domesticated DNA elements (Table 1) were extracted, mixed together with the rest of the reaction components, and incubated in the thermal cycler to catalyze the assembly (Table 2). Each individual reaction mix was then directly transformed in E. coli DH5α and plated on antibiotic-containing media for selection and verification. Colony PCR analysis was routinely performed for 10–45 colonies per reaction plate to confirm correct assembly based on the amplification fragment size (Fig. 6a-b) using primers listed in Supplementary Table S2. The efficiency of the assembly was calculated based on the PCR as the percentage of the correct clones in the analyzed samples (Table 4). The reactions consistently resulted in hundreds of antibiotic-resistant colonies, with an average assembly efficiency of about 95% (Table 4). All the negative background corresponded to clones harboring only the empty backbone vector P0 (Fig. 2), and unspecific misassembly was never observed. Ultimately, extraction and transformation of the verified plasmids in Synechocystis resulted in strains that expressed both the target enzymes, sYFP2 (Fig. 7a) and EFE (Fig. 7b), which demonstrated conclusively that the assembly system for seven fragments functioned as designed.

Table 2.

Reaction composition and cycling parameters for the 7-fragment CyanoConstruct assembly. The reaction was carried out in 10 µl total volume, and the DNA components were added in the mix as plasmid minipreps (145 Ng per miniprep, except 290 Ng for the RBS plasmids in the reaction) in a maximum combined volume of 5 µl. The assembly reaction was followed by direct transformation in E. coli for amplification and screening

Reaction	Component (stock)	V per reaction	Final Concentration ~
Assembly Vtot ~ 10 µl 37 °C 5 min, 16 °C 5 min 50 cycles	BbsI (10 U/µl)	1 µl	1 U/µl
	NEBuffer 2.1 (10x)	1 µl	1x
	Dithiothreitol (10 mM)	1.3 µl	1.3 mM
	T7 Ligase (3000 U/µl)	0.5 µl	150 U/µl
	ATP (10 mM)	1.3 µl	1.3 mM
	DNA (plasmid MP)	5 µl in total	15–30 ng/µl per plasmid

Table 3.

Reaction composition and cycling parameters for the 9-fragment CyanoConstruct assembly. The reaction was carried out in 15 µl total volume, and the DNA components were added in the mix as plasmid minipreps (200 Ng – 800 Ng per miniprep in the reaction) in one or two steps (see Table 4) in a maximum combined volume of 6.5 µl. The additional exonuclease step at the end of the assembly reaction was necessary to reduce the empty plasmid background. The assembly reaction is followed by direct transformation in E. coli for amplification and screening

Reaction	Component (stock)	V per reaction ~	Concentration added~
Assembly Vtot ~ 15 µl 37 °C 5 min, 16 °C 5 min 30 + 50 cycles *	BbsI (20 U/µl)	2 µl	2.7 U/µl
	NEBuffer 2.1 (10x)	1.5 µl	1x
	Dithiothreitol (10 mM)	2 µl	1.3 mM
	T7 Ligase (3000 U/µl)	1 µl	200 U/µl
	ATP (10 mM)	2 µl	1.3 mM
	DNA (plasmid MP)	6.5 µl in total	~15–55 ng/µl per plasmid
[Optional] Empty vector removal Vtot ~ 25 µl 37 °C 30 min	Exonuclease V (RecBCD) (10 U/µl)	2.5 µl	1 U/µl
	NEBbuffer 4 (10x)	2.5 µl	1x
	ATP (25 mM)	2.5 µl	2.5 mM
	BbsI (20 U/µl)	2.5 µl	2 U/µl

* See Table 4 for details

Fig. 6.

E. coli colony PCR screen used for the identification of correct construct assembly based on amplification product size. The agarose gel image represents the a) 7-fragment assembly using pDFC replicative plasmid backbone, b) 7-fragment assembly using pSIC backbone for chromosomal integration, c) initial 9-fragment assembly using pDFC replicative plasmid backbone, d) 9-fragment assembly after the sequence overhang redesign. The correct amplification products that represent successful assembly (+ clone) are indicated by a green rectangle. The truncated products that represent misassembly (-clone) have been indicated by a red rectangle. The screening was carried out using antibiotic-resistant E. coli transformant colonies as PCR templates (see Fig. 3d), while correctly assembled plasmids and corresponding empty plasmid backbones served as the positive (+ ctrl) and the negative controls (-ctrl), respectively. All the different amplification products detected in the screening reactions have been shown in the image, and no other bands were observed. The information acquired from corresponding screening reactions has been used for compiling the quantitative assembly efficiency data shown in Table 4

Table 4.

Efficiency of the CyanoConstruct assembly system evaluated based on the relative number of positive E. coli clones obtained per reaction. List of assembly reactions for seven and nine fragment constructs carried out using different amounts of DNA and temperature cycles. the independent parts are specified in Tables 1, and the composition of the reaction mixtures and the reaction conditions are presented in Tables 2–3. #Col_tot.: the number of transformant colonies on selection plate; #Col./+: number of PCR-analyzed colonies / number of correct positive clones; Eff.%: the percentage of the correct positive clones identified in colony PCR

Assembled parts in circular plasmid	Relative DNA amount*	Parts included (#cycles)	#Coltot	#Col./+	Eff.%
7-fragment assembly
pDFC-PA1lacO−1-RBS03-syfp2-RBS05-efe-TT-	RBSs 2x; rest 1x	All (50)	742	45 / 42	93
pSIC-PA1lacO−1-RBS03-syfp2-RBS05-efe-TT-	RBSs 2x; rest 1x	All (50)	106	53 / 51	96
9-fragment assembly (initial rounds of testing; unoptimized)
pDFC-PA1lacO−1-RBSS4-adh-RBSS3-pdc-RBSS3-pk-TT-	RBSs 2x; TT 2x; rest 1x	All (50)	160	11/1	9
pDFC-PA1lacO−1-RBSS4-adh-RBSS3-pdc-RBSS3-pk-TT-	RBSs 2x; TT 2x; rest 1x	All (50)	516	18/3	17
pDFC-PA1lacO−1-RBSS4-adh-RBSS3-pdc -RBSS3-pk-TT-	RBSs 2x; TT 2x; rest 1x	All (50)	674	22/4	18
pDFC-PA1lacO−1-RBSB-adh-RBSB-pdc -RBSS3-pk-TT-	RBSs 2x; TT 2x; rest 1x	All (50)	316	32/3	9
pDFC-PA1lacO−1-RBSB-adh-RBSB-pdc -RBSS4-pk-TT-	RBSs 2x; TT 2x; rest 1x	All (50)	800	20/3	15
pDFC-PA1lacO−1-RBSB-adh-RBSS4-pdc-RBSB-pk-TT-	RBSs 2x; TT 2x; rest 1x	All (50)	480	24/4	14
pDFC-PA1lacO−1-RBSB-adh-RBSS4-pdc-RBSS3-pk-TT-	RBSs 2x; TT 2x; rest 1x	All (50)	208	22/5	23
pDFC-PA1lacO−1-RBSS4-adh-RBSS4-pdc-RBSB-pk-TT-	RBSs 2x; TT 2x; rest 1x	All (50)	180	22/1	5
9-fragment assembly (P7-P8 overhang redesign)
pDFC-PA1lacO−1-RBSB-adh-RBSS4-pdc-RBSS3-pk-TT-	RBSs 2x; TT 2x; rest 1x	All (50)	140	55/0	0
9-fragment assembly (reaction parameter optimization)
pDFC-PA1lacO−1-RBSB-adh-RBSS4-pdc-RBSS3-pk-TT-	RBSs 2x; rest 1x	w/o TT (30); All (50)	204	32/26	81
pDFC-PA1lacO−1-RBSB-adh-RBSS4-pdc-RBSS3-pk-TT-	All 1x	w/o TT (30); All (50)	248	32/27	84
pDFC-PA1lacO−1-RBSB-adh-RBSS4-pdc-RBSS3-pk-TT-	RBSs 2x; rest 1x	RBSB & adh (30); All (50)	216	32/22	69
pDFC-PA1lacO−1-RBSB-adh-RBSS4-pdc-RBSS3-pk-TT-	All 1x	RBSB & adh(30); All (50)	256	32/28	88
pDFC-PA1lacO−1-RBSB-adh-RBSS4-pdc-RBSS3-pk-TT	adh 4x; RBS 2x; rest 1x	All (80)	244	32/24	75
pDFC-PA1lacO−1-RBSB-adh-RBSS4-pdc-RBSS3-pk-TT	adh 4x; rest 1x	All (80)	160	32/18	56
pDFC-PA1lacO−1-RBSB-adh-RBSS4-pdc-RBSS3-pk-TT	All 1x	All (80)	352	32/27	81

* Based on the mass of the DNA plasmid carrying each specific part. 1x = 145 ng in 7-fragment assembly and 200ng in 9 fragment assembly

Fig. 7.

Verification of successful construct assembly by functional expression analysis in vivo. The assembled expression constructs were transformed in Synechocystis followed by the analysis of a) sYFP2 fluorescence and b) ethylene accumulation in culture vial headspace at two different time points after induction (7-fragment construct; see Fig. 5a), and c) ethanol concentration in culture medium using three different strains with alternative RBS element combinations (9-fragment construct; see Fig. 5c). The RBS elements in the constructs used were S3/S5 for sYFP2 and EFE (a-b), and S4/S3/S3 (#1), S3/S4/B (#2) and B/S4/S4 (#3) for Adh, Pdc and PK (c), respectively. The control strains used for comparison were the wild type Synechocystis with empty pDF plasmid (0 ref. in a-b) or the wild type background strain (WT in c). The average values and standard deviations in each case were calculated based on three parallel replicates (N = 3)

9-fragment assembly hindered by unspecific product formation

The established reaction setup was subsequently adapted for the 9-fragment ethanol construct assembly, which after slight modifications to the concentrations (Tables 3 and 4), produced hundreds of antibiotic-resistant E. coli colonies per reaction. Again, the constructs with the full-sized inserts (Fig. 6c) were functional in Synechocystis, and the transformed clones accumulated ethanol in the culture medium, showing varying production levels for parallel operons with different RBS combinations (Fig. 7c). Notably, however, the majority of all the analyzed E. coli colonies on the selection plates appeared to have a truncated insert (Fig. 6c), which reduced the total assembly efficiency to under 15% (Table 4). This amplification product was too short to represent the fully assembled operon, suggesting that some of the DNA elements ligated together unspecifically, forming incomplete misassembled circular constructs. As the incorrect PCR product was not observed in the 7-fragment assembly, the mismatch was expected to involve the position-specific overhangs of at least one of the additional fragments P7-P8 present only in the 9-fragment reaction (Fig. 2b; Table 1).

Identifying unspecific binding between P2 and P8 as the source of misassembly

To identify the source of the observed misassembly, four parallel truncated PCR products were extracted from the agarose gel and sequenced. The results revealed unspecific binding between the P2 RBS fragment right overhang (GTCT) and the P8 transcription terminator left overhang (CGGA) (see Fig. 2a), which was in perfect agreement with the size of the detected incorrect PCR product. Due to the 75% homology and the relative strength of the two GC pairs between mismatched sticky ends, the efficiency of the P2-P8 binding significantly hindered the correct assembly, leaving out the five elements P3-P7 in the majority of the assembled constructs.

Improving reaction specificity and assembly efficiency by reaction optimization

The low initial assembly efficiency of the nine fragments (Table 4) suggested that the overall performance could be improved simply by reducing the likelihood of the competing P2-P8 mispairing. Initially, this was achieved by excluding the P8 TT fragment at the first half of the reaction, thereby allowing the correct overhang pairs (specifically P2-P3) to ligate before the addition of the last P8 fragment. Several different setups were tested (Table 4) resulting in dramatic decrease in mispairing and improving the overall reaction efficiency to over 85%. Notably, up to ~ 80% efficiencies were also obtained by simply altering the reaction component concentrations, demonstrating that all the fragments could be mixed together at the same time, and still prevent the formation of the truncated product.

While testing the assembly, it became evident that the overall performance was largely determined by the relative empty vector background. Although the assembly itself was highly specific under the selected conditions (Table 4), the undigested (re-ligated) plasmid backbone easily resulted in false positive antibiotic-resistant colonies, and required excessive PCR screening of many transformant clones. To reduce this background, an exonuclease step was introduced at the end of the assembly reaction to chew the ends of any remaining linearized backbone fragments that could still ligate into circular plasmids [63]. Notably, as the correctly assembled complete constructs no longer contain BbsI recognition sequences and are thus not vulnerable to digestion, additional supplementation of BbsI combined with the exonuclease treatment appeared highly effective in removing remaining empty plasmids from the final reaction mix. Although not essential when combining only seven fragments, the treatment significantly increased the success rate for the 9-fragment assembly (Table 4).

Fragment overhang redesign did not improve reaction specificity

Besides optimizing the reaction setup to prevent the mispairing between P2 and P8 (Table 4), we also tested redesigning the part-specific overhangs to increase the annealing specificity. As the parts P0-P6 were shown to ligate as intended in the seven-fragment assembly, the redesign focused only on the P7-P8 overhang junction. To select alternative overhangs for the P8 cloning motif a Jupyter based NUPACK [64–66] simulation (Supplementary Figure S3) was used to calculate the free energy for duplex formation for every overhang pair in the system (Supplementary Table S3), as visualized in the heatmap (Supplementary Figure S4) using a custom R script (Supplementary Figure S5). From the unused overhangs that were predicted to promote correct assembly with minimal off-target affinity (Supplementary Figure S4) the sequence CTTC was selected to replace the original P8 transcription terminator left overhang CGGA. Repeated assembly reactions with the redesigned P7-P8 overhang pair, however, did not improve the efficiency but resulted in the appearance of second misassembly fragment (Fig. 6d) outcompeting the desired product. Due to the complexity of the overhang pairing and difficulty in making reliable mathematical predictions, this approach was not continued further.

The effect of the fragment sequences and relative concentrations

When evaluating the system performance, several alternative fragments were tested for specific positions, including three different RBS elements for P2, P4 and P6 in the 9-fragment assembly, and two alternative backbones (P0) in the 7-fragment assembly (Tables 1 and 4). The comparison showed no apparent difference between the use of the different backbones, and all the tested alternative fragment combinations assembled as designed. No obvious sequence-specific differences were observed in the efficiency when using alternative combinations, and some variation was always detected even between reactions with the same set of parts. The relative amounts of the DNA plasmid preps used, on the other hand, resulted in clear differences in the assembly efficiency, as seen between the different 9-fragment reactions (Table 4). Most notably, the observed unspecific binding between P2 and P8 that results in incorrect assembly was completely avoided simply by reducing the amount of the RBS and TT plasmids in the reaction.

Discussion

Justification for a new standard

Many different cloning systems and modular DNA assembly strategies using various standards have been described in literature. While the obvious lack of standardization hinders the development of universal modular language for genetic engineering [17], it offers alternatives for a wide range of purposes and specific user needs to choose from. As seen in the cyanobacterial field, there is no single platform applicable or most resource-efficient for all needs, and many laboratories still rely on conventional cloning methods. Setting up new systems or changing the cloning standards for routine preparative use may be surprisingly laborious, yet the ease and current low cost of DNA synthesis for generating ready-to-use DNA parts makes new options more accessible. CyanoConstruct described here is an application of Golden Gate assembly that offers a simple alternative for building multi-gene expression constructs for cyanobacteria, while providing access to flexible translational control with minimal cloning effort. We demonstrate the use of the system with several alternative reporter constructs (Figs. 5 and 7) which are essentially identical in architecture to constructs previously assembled by subcloning [31] that have been successfully used for expressing different proteins in vivo [49, 67, 68]. Although Synechocystis has been the primary expression host, the genetic elements used here (Table 1; Supplementary Figure S1) have been carefully selected so that they would be applicable also in other cyanobacterial species. Besides the prokaryotic operon-based design common in cyanobacteria, plasmids with the broad host-range replicon RSF1010 [12, 57, 59–61, 69], different versions of the rrnB terminator [21, 70], and P_lac-derived promoters [71, 72], have been widely used also in Synechococcus and Anabaena. To allow users to apply this system to custom expression targets and genetic elements, the functional parts described in this work in addition to an expanded set of RBSs for translational tuning are available as a plasmid library from Addgene (Supplementary Table S4) (https://www.addgene.org/browse/article/28263931/).

Number of assembly steps in comparison to other systems

We have validated CyanoConstruct for the assembly of transformation-ready expression constructs that carry up to three target genes with independently selected RBSs, using nine DNA parts (Fig. 5c). Compiling corresponding constructs with the modular subcloning system we have applied earlier [31, 49, 67, 68] takes six separate rounds of digestion, extraction, ligation, transformation, selection and plasmid isolation. With the established multi-part assemblies that rely on Type IIS enzymes this is typically accomplished in several hierarchical steps, often using two different enzymes and several vector backbones [18, 22, 73]. CyanoGate, for example, does not support polycistronic operons, but allows the sequential assembly of multi-gene constructs from independently constructed single-gene transcriptional units [22]. The parallel transcriptional units generated in the first phase are amplified in E. coli and extracted as plasmid minipreps, followed by a subsequent reaction to compose the final expression construct with multiple genes. Although the parts and the overall design vary between methods, the stepwise procedure is similar to those used in Start-Stop Assembly [73] and in BASIC [25] for larger constructs. In comparison, CyanoConstruct combines the library parts together in one reaction using a single Type IIS restriction enzyme; intermediate extraction steps or sequential assemblies are not required, allowing the reaction mix to be directly transformed in E. coli for selection and amplification (Fig. 3). If more than three independent proteins are to be expressed at the same time, the system can be used without modifications for six targets by using the two alternative backbones pDFC and pSIC in parallel.

Overhangs and assembly specificity

CyanoConstruct assembly uses a library of four-base overhangs that have been designed and validated for optimal combined ligation fidelity in vitro (Supplementary Table S1) [47]. While we have here used a maximum of nine parts per reaction (Fig. 5c) to ensure high assembly efficiency, accuracy and low negative background (Table 4), the predesigned overhangs in the library would allow the system to be expanded for more parts [47]. Although Golden Gate cloning has been used for joining together as many as 35 fragments in a single reaction [74], most applications limit to maximum of five [28] or seven fragments in one round [22, 26], and generate more complex constructs through iterative rounds of cloning. The assembly reactions can be performed using linear DNA, but in long-term use it is often most convenient to store the verified parts as plasmid minipreps that enable easy amplification in E. coli. The general advantage of Golden Gate is that the assembly reaction does not involve PCR, so amplification errors are not a concern, and the resulting constructs can be verified without sequencing based on the insert size alone. On the downside, existing parts with their cloning overhangs are fixed to a particular position in the assembly, which requires modification to alter the order. To aid the user in this, the online CyanoConstruct tool guides the position-specific part design, and provides part sequences for DNA synthesis, as well as primer sequences for overhang modification using PCR. Although the output sequences should also be checked using external cloning software to confirm correct assembly, the tool significantly reduces the workload at the design phase.

Cost of DNA synthesis

Based on the current average price of DNA synthesis of 0.08 USD per base pair, the calculated cost of the DNA parts for building a single three-gene construct presented here (Fig. 5c) would be about 500 USD without the plasmid backbone. The price for three alternative RBS parts (length ~ 70–80 bp) for each of the positions P2, P4, and P6, would be altogether about 50 USD. The resulting 27 (i.e. 3 × 3 × 3 permutations with a fixed gene order) construct variants with all possible RBS permutations would thus require 15 library plasmids in total, and cost around 20USD per construct in terms of gene synthesis. Therefore, although the synthesis of ready-to-use gene expression constructs may be feasible if a specific single plasmid is needed, it is significantly more economical to assemble the constructs from a library of parts if multiple variations are required for high throughput functional screening.

RBS parts for translational tuning and N-terminal tagging

CyanoConstruct has been specifically designed to facilitate the exchange of the RBS sequences upstream each target gene (Fig. 1d; parts P2, P4 and P6 in the 9-fragment assembly) to enable independent and flexible control over the translation of the expressed enzymes. We have previously used the strategy to optimize the expression of individual proteins including EFE and sYFP2 in Synechocystis [31], and now demonstrate the dynamic effect of translational tuning on ethanol production using different RBS combinations (Fig. 7c). While many engineering projects have used RBS calculators that are based on sequence information from E. coli [15, 32, 38, 39, 75], CyanoConstruct is founded on a set of sequences that derive primarily from cyanobacteria and have been characterized in several contexts in Synechocystis [22, 31]. The RBS parts are composed of an invariable 27 bp upstream insulator sequence separating the translational unit from the upstream sequence [33], and the ≤ 22 bp variable RBS-specific sequence composed of the core Shine-Dalgarno sequence and the flanking regions (Fig. 2d). This allows the user to apply and compare any 5’ UTRs as they appear in nature or in previous expression systems simply by exchanging the variable region of the RBS element (Fig. 2d; Supplementary Figure S1) with the desired sequence. In this design, the start codon is included in the RBS part, so that the overhang junction overlaps with the beginning of the coding region. Notably, to prevent frameshift, this requires the addition of two bases (Fig. 2d), which appends two extra amino acid residues at the N-terminal end of the expressed proteins (GR for P2; GL for P4 and GG for P6). With the realization that protein-specific effects of these N-terminal modifications may be difficult to predict, this was a deliberate compromise that enabled the use of high-fidelity 4 bp overhangs, while maintaining the upstream regulatory sequences untouched. This way the mRNA 5’ UTR sequence which is involved in ribosome interactions and could affect translation initiation, is independent of cloning. In the other systems that prioritize preserving the native N-terminus of the target protein, the overlap with the coding region is avoided by using shorter 3 bp overhangs that align with the start codon [73] or overlap at least partially with the upstream 5’ UTR [22, 25]. However, enzymes with N-terminal modifications, such as additional single residues from cloning scars, have been previously expressed in Synecocystis without any apparent issue [31, 49, 67, 68]. In addition, overexpressed proteins have been frequently fused with N-terminal tags in cyanobacteria, including His-tag [76–78], FLAG-tag [79–81] and Strep-tag [79], suggesting that the additional residues are not a critical inherent concern in the current design. Instead, together with the RBS of choice, this design now allows the introduction of an N-terminal tag sequence on any of the expressed target proteins without modifying the existing gene sequences in the library. The desired tag is ordered as part of the RBS fragments after the ATG, which is not feasible with standards that either use overhangs aligning with the start codon [73] or the end of the UTR [22, 25]. Besides purification tags, this can be also be applied for introducing tags to promote, for example, protein degradation [82], secretion [83, 84], or protein-protein interaction via proximity labeling [85].

Reaction optimization & troubleshooting

The specificity of the assembly was relatively high, and enabled the generation and screening of the correct constructs in each reaction round at over 85% average efficiency once the conditions had been optimized (Table 4). While we never observed misaligned products in the 7-fragment assembly when building two-gene constructs, the reaction with nine fragments initially resulted also in unspecific binding between the part overhangs (Fig. 6c) due to off-target ligation at the P2-P8 junction. These problems were solved with reaction condition optimization, simply by changing the relative concentrations of the DNA plasmids in the reaction mix (Table 4). In comparison, our attempt to avoid incorrect ligation at the P2-P8 junction by rational modification of the P7-P8 overhang proved unsuccessful. The rational substitution of the P7-P8 overhang rather than P2-P3 to improve the 9-fragment assembly specificity was justified by the efficacy of the 7-fragment assembly but not necessarily optimal. Considering an interaction matrix of the free energies of overhang duplex formation (Supplementary Figure S4), a Wilcoxon test later identified the P2-P3 junction as having the most differential off-diagonal negativity of the overhangs in the 9-fragment assembly. Although no overhang was calculated to significantly contribute to off-target ligation, the P2-P3 overhang appears energetically more problematic regarding incorrect assembly of P2-P8 than the P7-P8 overhang. The plasmid concentrations appeared to have a clear effect on the reaction outcome and would be the first optimization parameter to improve reaction specificity. As the assembly does not have a built-in screening system to distinguish assembled constructs from empty plasmid background, it is essential to ensure that the remaining intact backbone vector is removed before transformation in E. coli. Especially in the 9-fragment assembly which was shown to be more demanding, an additional supplementation of exonuclease that breaks down linear DNA was needed for efficient screening of the assembled plasmids. Together with E. coli transformant selection on antibiotic plates, colony PCR screening was a convenient and efficient method for assaying transformation, as the majority of colonies on the transformant plates carried the correct insert.

For what purpose is CyanoConstruct most useful?

The CyanoConstruct system is optimal for users who want to generate single transcription unit expression constructs for one, two or three target enzymes, without prior knowledge of the translational performance or the optimal relative levels of the co-expressed enzymes. The system provides an user-friendly platform for laboratories that need to build relatively simple constructs, for example, for i) testing alternative RBS combinations when optimizing the expression of a specific target protein, ii) balancing consecutive reaction steps catalyzed by several co-expressed enzymes, iii) comparing the translation of enzyme homologs from different sources, or iv) introducing various N-terminal tags to the target proteins. In addition, this can be a useful tool for v) expanding the host range to other cyanobacteria beyond Synechocystis, as the translation and optimal expression patterns may easily vary from one host to another, or for vi) studying ribosome association dynamics between the 5’-UTR and the start codon, which still remains poorly understood in cyanobacteria.

Summary

While commercial DNA synthesis has become more economical, modular cloning systems that allow synthetic genetic elements to be easily arranged in a defined order provide effective means for high-throughput assembly of numerous construct variants with relative ease. CyanoConstruct enables script-assisted design, and direct assembly of operon-based polycistronic expression constructs to produce transformation-ready plasmids compatible with cyanobacteria in a single one-pot reaction. The system prioritizes simplicity over different functionalities, provides an alternative to more complex systems with potentially higher starting threshold.

Materials and methods

Molecular biology reagents

All the molecular biology reagents and enzymes used in this work were purchased from New England BioLabs (USA) (see Tables 2 and 3 for the reaction components), except DreamTaq DNA Polymerase used in colony PCR which was from Thermo Fisher Scientific (USA). The Miniprep and Midiprep kits used for plasmid isolation were acquired from Qiagen (Germany).

DNA synthesis

All PCR primers (Supplementary Table S2) were ordered from Eurofins MWG Operon (Germany). Longer DNA sequences including all the parts used for the assembly (Table 1) were purchased as synthetic DNA fragments from GeneScript (USA), delivered as inserts in high copy number E. coli plasmid pCloneEZ or pUC57.

E. coli strains and culture conditions

Escherichia coli strain DH5α was used for all DNA cloning steps, including amplification of the plasmids carrying the designed DNA parts, and the selection and amplification of the final constructs after the assembly. The cells were cultured at 37 °C in Lysogeny Broth (LB) liquid medium with 150–200 rpm or on LB plates supplemented with 1.5% agar. The antibiotics used for transformant selection were 100 µg/ml ampicillin (pCloneEZ and pUC57), 50 µg/ml spectinomycin and 20 µg/ml streptomycin (pDFC) and 50 µg/ml kanamycin (pSIC). All E. coli strains were maintained in − 80 °C freezer library as 10% glycerol preparations.

Synechocystis strains and culture conditions

The glucose-tolerant substrain of Synechocystis sp. PCC 6803 [86] (originally from Prof. A. Kaplan Hebrew University of Jerusalem, IL) was used for all cyanobacterial experiments. The cells were cultured in BG-11 medium [87] buffered to pH 8.0 with TES-KOH or corresponding plates supplemented with 1.5% agar. The cultures were incubated under constant white light (50 µmol photons m^{− 2} s^{− 1}) in 1% CO₂ atmosphere in a growth chamber MLR-351 (Sanyo, Japan) or Algaetron 230 (Photon Systems Instruments, Czech Republic) at 30 °C, with 150 rpm agitation for liquid cultures. The antibiotics used on the BG-11 agar plates for transformant selection were 6 µg/ml spectinomycin and 3 µg/ml streptomycin (pDFC) or 6 µg/ml kanamycin (pSIC). The antibiotics used in liquid cultures were 12.5 µg/ml spectinomycin (pDFC) or 12.5 µg/ml kanamycin (pSIC) for the precultures, and 50 µg/ml for the corresponding main cultures.

Designing DNA parts for the assembly

Adaptation of DNA parts for the assembly (Table 1, Supplementary Figure S1) involved two main steps: (i) the inactivation of BbsI sites (GAAGAC) in the original sequences with single-base substitutions, and (ii) the addition of position-specific 12 bp overhang sequences for BbsI recognition and cleavage at each end (Fig. 2). For all gene sequences (i.e. GOI), the native start codons were omitted, and BbsI sites in the coding region inactivated using appropriate synonymous codons to avoid amino acid substitutions in the encoded protein. For the RBS elements, a translation start codon ATG was included immediately downstream the variable RBS region, followed by additional two guanidine bases to maintain the reading frame after ligation (Fig. 2d). After the design, all parts were subjected to in silico sequence analysis (Benchling Biology Software 2024 retrieved from https://benchling.com) to reaffirm assembly into circular constructs in all intended combinations, and to verify the reading frames against correct amino acid sequences. The designed parts were ordered as ready-to-use synthetic DNA inserts in high-copy E. coli plasmids for easy storage and amplification. Alternative to DNA synthesis, the position-specific DNA parts can be produced from available template sequences using PCR and the primers designed by the sequence tool.

Designing plasmid backbones for the assembly

Modification of existing plasmids to be used as assembly backbones involved three main steps: (i) the inactivation of interfering BbsI sites, (ii) removal of unwanted elements (e.g. promoters or associated repressors), and (iii) introduction of the 64 bp CyanoConstruct insertion domain sequence (Fig. 2c) to enable the ligation of the parts into the backbone (Fig. 2b). The replicative plasmid backbone pDFC (see Fig. 5a) that is compatible for amplification and selection both in Synechocystis and in E. coli was generated accordingly from pDF-lac2 [31]. This was accomplished by first replacing the NheI-PasI and ScaI-KasI fragments with corresponding synthetic sequences in which the unwanted BbsI sites had been inactivated by point mutations. The existing promoter P_A1lacO−1 and repressor lacIq were removed as an EcoRI-NsiI fragment, followed by Klenow blunting and backbone religation. Finally, the PsrGI-AclI fragment was replaced with a synthetic sequence to remove rrnB1 T1 terminator, and to add the plasmid insertion domain (Fig. 2c, Supplementary Figure S1a) in place. The alternative backbone for chromosomal integration pSIC (see Fig. 5b) was generated from pSI1B-slr1311-IS-syfp2-CmR [49] that allows homologous recombination at the psbA2 locus (slr1311) in Synechocystis. This was accomplished by replacing the XmaI-Sal fragment (containing the promoter P_A1lacO−1, lacI^q repressor and syfp2 gene) with the CyanoConstruct insertion domain. The resulting plasmids were verified with analytical restriction digestion and gel electrophoresis, and stored as minipreps and E. coli glycerol preps for subsequent use.

The assembly reaction

All DNA parts to be assembled were maintained as circular plasmids in E. coli (glycerol preps stored at -80 °C) and extracted as plasmid minipreps from 5 ml o/n liquid cultures. The concentrations of the minipreps were measured spectrophotometrically (BioDrop µLite), after which the plasmids were combined for the assembly in a PCR tube at final concentrations ~ 15–50 ng/µl each (see Tables 2, 3 and 4). After the supplementation of the remaining reaction components (BbsI, NEB buffer, Dithiothreitol, T7 Ligase, ATP) as specified for the 7-fragment assembly (Table 2) or the 9-fragment assembly (Table 3), the reactions were incubated in temperature cycling at 37 °C 5 min / 16 °C 5 min for 50–80 cycles using a PCR cycler (BIO-RAD T100 Thermal Cycler). To ensure the breakdown of remaining non-assembled plasmid backbone, and additional reagents (i.e. Exonuclease V (RecBCD), NEBuffer, ATP) were added in the same PCR tube followed by 30 min incubation at 37 °C. The resulting reaction mix was used directly for transformation in E. coli.

E. coli transformation for amplification and selection

The assembled constructs were amplified and selected in E. coli. For this, heat shock-competent E. coli DH5α cells were prepared using conventional rubidium chloride protocol [88] and stored at -80 °C until use. For the transformation, aliquots of the competent cells (50 µl) were thawed on ice, supplemented with 7.5–10 µl of the assembly reaction mix, incubated on ice for 20 min, and submerged in 42 °C water bath for 45 s. After 1 h incubation in 1 ml LB (37 °C at 150 rpm) all cells were plated on antibiotic-containing plates (50 µg/ml spectinomycin and 20 µg/ml streptomycin for pDFC; 50 µg/ml kanamycin for pSIC), and incubated o/n at 37 °C to obtain transformant colonies.

Assembly verification by E. coli colony PCR

After each transformation, typically 20–30 antibiotic-resistant colonies from each plate were analyzed by colony PCR [89] to confirm the presence of the correct sized insert. The PCR reactions were carried out with DreamTaq polymerase using primers listed in Supplementary Table S2 (0.4 µM each) and 1 µl of transformant cells suspended in MQ as template. The temperature cycling conditions were 95 °C / 10 s (denaturation), 60 °C for pDFC or 55 °C for pSIC (annealing), and 72 °C for 3 min 30 s (extension) for 35 cycles. The PCR product sizes were verified by gel electrophoresis (Mupid-One Electrophoresis System, Nippon Genetics Europe; Gel Doc EZ imager, Biorad) against known standards (NEB 1 kb+). Selected verified clones were stored as glycerol stocks in the strain library, and used for preparing the plasmid midipreps for cyanobacterial transformation.

Construct transformation and selection in Synechocystis

For transforming Synechocystis, cells from 5 ml of liquid culture at OD₇₅₀ = 0.8 were pelleted (3000 g / 5 min) in a 12 ml Falcon tube, and gently resuspended in 1 ml of fresh BG11. 15–20 µg/ml of extracted plasmid DNA was mixed with the cells. After incubation for 16 h in dark (150 rpm, 30 °C 1% CO₂) the cells were plated on antibiotic-containing BG11 and maintained under constant 50 µmol photons m^− 2 s^− 1 light 30 °C (1% CO₂). After the appearance of antibiotic resistant colonies in 10–30 days, the transformants were analysed by colony PCR. This was done as described for E. coli, with the exception that the template cells were subjected to three successive freeze-thaw cycles (5 min at – 80 °C /5 min at 60 °C, spin-down in between) to induce cell lysis. Verified strains were stored in the − 80 °C library (in triplicates) by pelleting cells (5 min at 4000 g) from 1 ml liquid culture at OD₇₅₀ = 1, followed by resuspension in 200 µl fresh BG11 with 7.5% DMSO and freezing in liquid N₂.

Construct functional verification in Synechocystis

For functional analysis, each generated Synechocystis strain and the control strains were cultured in three replicates in BG11 (V = 20 ml) w/o antibiotics in 50 ml Erlenmeyer flasks to OD₇₅₀ = 0.5, and induced by the addition of isopropyl-β-D-thiogalactopyranoside (IPTG) to the final concentration of 1 mM. To measure sYFP2 fluorescence, 150 µl of cell culture from each flask was transferred at different time points on 96-well black clear-bottom polystyrene plates (Costar Corning, USA) and analyzed using Tecan microplate reader (Tecan infinite 200 PRO) at 495 nm (ex) /535 nm (em) as described earlier [31]. To measure ethylene production in the efe-expressing strains, 1 ml of each cell culture was sealed in 10 ml air-tight serum bottles and incubated in gentle mixing under 50 µmol photons m^{− 2} s^{− 1} light until headspace gas analysis at given time points with GC-FID (Perkin Elmer AutoSystem with Varian CP-CarboBOND fused silica capillary column) as described before [31]. To analyze ethanol production, the induced cultures incubated for 18 h were centrifuged (5000 g / 10 min), supernatant filtered using RC 0.45 μm syringe filter, and analysed by GC-FID (Shimadzu Nexis GC-2030 with Hewlett-Packard HP-INNOWax column) as described earlier [90].

CyanoConstruct part library and complete sequences

All the functional DNA parts and plasmid backbones described in this study together with an extended set of RBSs (Supplementary Table S4) have been deposited with corresponding plasmid maps and complete sequences with Addgene (https://www.addgene.org/browse/article/28263931/).

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Design and conceptual planning by OMV, PK, PP, KPA. Experimental design by PK, RN. Experimental work by RN, AA, DN, OM, LK, ME, NK. Experimental supervision by RN. Initial web-tool script by AB. Web-tool second iteration by LT. Web-tool third iteration by AP. Data analysis and result interpretation by PK, RN, OMV, KPA. Manuscript preparation by PK, OMV, RN, KPA, EMA. Manuscript finalization by PK, OMV. Preparation of figures by PK, OV. Supervision and coordination by PK, KPA. All authors contributed to manuscript revision and approved the final version for publication.

Funding

This work was supported by the Sloan Foundation (G-2024-22710), NSF (2419641), the Academy of Finland (307335), NordForsk NCoE (82845) and the Jane and Aatos Erkko Foundation.

Data availability

All data generated or analyzed during this study are included in this article and its supplementary information files. The web-tool is found at http://www.cyanoconstruct.com.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.