A standardized genome architecture for bacterial synthetic biology (SEGA)

Philosophy and design of SEGA

In the Standardized Genome Architecture (SEGA), genomic integration of DNA fragments is enabled by λ-Red recombineering and landing pads that contain features that (i) enable insertion of additional genetic elements and (ii) provide well-characterized functional parts such as promoters and genes, and (iii) provides insulation against genome context-dependent effects. The SEGA landing pads allow for reusable homology regions and time-efficient construction of parallel genetic designs with a minimal number of reagents and handling steps. SEGA strains can be stored as ready-to-use competent cryostocks with the recombineering machinery activated. This heavily simplifies the genome engineering process for the end-user. SEGA bricks are integrated on the genome simply by combining the two reagents (i.e., competent cells and DNA), followed by incubation steps, and successful recombinants are identified by visual inspection on agar plates. Thus, DNA can be prepared by a simple PCR amplified directly from original source or even simpler by ordering of inexpensive synthetic DNA fragments. These steps do not rely on prior DNA assembly on plasmids, which may be problematic as some genes are toxic when handled on cloning vectors³³.

A typical engineered function in DNA includes a gene and genetic control elements that ensure the expression of this gene. We adopt the term cargo, defined as the DNA portion that bears the main functionality of the vector²⁵, but separate the control elements from the cargo: a SEGA landing pad typically hosts two major genetic control elements that influence gene expression on the transcriptional (C1), and translational (C2) level (Fig. 1). The SEGA strain collection presented here comprises different sets of experimentally validated promoters (with their regulators) as C1 and different translation initiation regions (TIRs) as C2 (Fig. 1). The TIR spans from the Shine-Dalgarno sequence to the 5th codon of the gene of interest³⁴.

We further introduced a number of gadgets (another concept adopted from SEVA) defined as dispensable DNA sequences that confer new utilities or properties to the basic frame of the vector²⁶. For example, to minimize polar effects on host genes and vice versa, insulating terminators flanking the landing pad were included from a comprehensive, experimentally characterized collection³⁵. Further, to aid the recombineering of parts, genetic markers for selection and counter-selection were added (Fig. 1a). A third gadget in between the control elements and the cargo provides additional features such as a bicistronic gene expression buffer³⁶, signal peptides, or peptide tags for, e.g., protein solubilization, detection, or purification. Since these gadgets contain a start codon and are closely interconnected with the C2 control elements, their sequences are key for translation initiation, and can standardize translation levels as outlined below.

Beyond the described functions, the gadgets facilitate modulization by constituting standard recombineering sites (RS1, RS2, RS3, Fig. 1; sequences can be found in Supplementary Table 4) that can be reused for recombination of different parts we term SEGA “bricks” (Fig. 1b).

SEGA enables highly simplistic genome engineering with green-white screening

An important function of the SEGA gadgets is to enable efficient and simplistic recombination of genetic parts, in addition to providing insulation from the genomic context with efficient terminators. The tetA gene is particularly suited for this task since it can be used for both positive selection, by conferring resistance to tetracycline, and counter-selection because the expression of tetA leads to sensitivity toward metal ions like Ni²⁺³⁷. Our standard SEGA landing pad contains, in addition to the control elements, a preliminary cargo encoding GFP, with the rationale that exchange of this cargo with any other DNA fragment would lead to the loss of GFP fluorescence enabling simple “green-white-screening” by visual inspection (Fig. 2a).

Fluorescence was exploited to test the performance of the downstream tetA gadget. A SEGA cargo brick, encoding the red-fluorescent protein mCherry, was integrated into the landing pad using homology regions RS2 and RS3 (Fig. 2b). Cryostocks of electrocompetent cells, with the λ-Red recombination machinery expressed¹⁴, were thawed and mixed with or without the mCherry fragment. After recovery, the cells were plated on M9 agar supplemented with or without NiCl₂ (Fig. 2c). As expected, a green-fluorescent bacterial lawn could be observed in the absence of negative selection. When adding NiCl₂, far fewer green cells could be observed and with the addition of the mCherry fragment, mainly red fluorescent colonies were present. The efficiency for integrating mCherry varied between 80 and 100% and was more efficient with a gel-purified PCR fragment, likely due to the absence of interfering PCR primers (Supplementary Fig. 1). This shows that DNA fragments can be integrated into SEGA landing pads with high efficiency by simply adding DNA to frozen SEGA competent cells, creating a final, antibiotic marker-free construct on the genome in a single step. More commonly, successful recombinants will be colorless instead of red. Therefore, we termed this method “green-white-screening” and validated the efficiency and easy identification by integration of another gene encoding a commercially interesting camelide-derived single domain antibody (Nanobody). We obtained between 81% and 94% white (positive) colonies across three replicates for the integration of the Nanobody gene (Supplementary Fig. 2).

While assembling SEGA landing pads and other complex designs, we encountered a technical challenge with tetA. Occasionally, we assembled different DNA fragments on a plasmid followed by sequence verification, PCR amplification, and genome integration. However, tetA gadgets optimized for the genome appeared toxic to bacterial cells when handled on plasmids, probably due to the higher gene copy number. To address this problem, we devised an alternative approach that includes only the 5′ end of tetA including the promoter. This truncation can be handled on low-copy plasmids and tetA is restored on the genome when the complementing 3′-end is present there. We found this approach highly useful and efficient for introducing landing pad bricks, including different control elements and cargos assembled on plasmids (Supplementary Figs. 3a, 6a). We further tested the truncated tetA approach for the integration of the 4542 bp operon crtEBIY from Pantoea ananatis encoding four genes necessary to produce the orange food coloring pigment β-carotene (Fig. 2d). Previously, our attempts to integrate this cargo brick with counter-selection against tetA were unsuccessful, probably due to the large size of the DNA fragment. We assembled the crtEBIY pathway together with the 5′tetA fragment and its accompanying control elements on a plasmid, amplified the whole construct by PCR, and subsequently integrated it in one step. The green-white screening showed no false positive (green fluorescent) clones while the orange color helped estimating that over 70% of the obtained colonies carried a functional crtEBIY operon (Fig. 2e, Supplementary Fig. 3d). Analysis of the remaining colorless clones showed sequence errors that presumably rendered the pathway non-functional.

The truncated tetA gadget increases the integration efficiency of large SEGA bricks, but also provides the basis for consecutively building larger and more complex synthetic biology designs on the genome, by cycling between tetA selection and counter-selection. As a simple demonstration of how the dual-selection marker can be used for sequential genome engineering, we chose to split the crtEBIY pathway into the four individual genes and integrated them in a stepwise manner. We cycled through four rounds of selection and counter-selection to introduce the complete operon (Supplementary Fig. 3b, c). In the first and third round of recombineering, tetA was truncated, whereas in the second and fourth round, functional tetA was reconstituted. This demonstrates successful recycling of a selection marker with minimal technical effort for building advanced synthetic biology designs in multiple rounds of genome engineering.

In E. coli, galK can be used as a positive selection marker on galactose as well as a counter-selectable marker by adding the toxic galactose analog 2-deoxy-galactose (DOG)³⁸. In the SEGA landing pad, a galK upstream gadget enables the exchange of control elements such as promoters. For this purpose, the native galK gene was deleted from the genomes of selected strains of the SEGA collection. To demonstrate the functionality of the galK gadget, we constructed a landing pad that contained galK and the rhamnose-inducible PrhaBAD promoter driving expression of gfp. This way, when the C1 element is exchanged for a constitutive promoter, recombinant genomes are identified as green in the absence of L-rhamnose (Fig. 2f), whereas recombinant non-constitutive promoters are identified as white in the presence of rhamnose (Supplementary Fig. 4). SEGA competent cells were electroporated with oligonucleotides encoding the constitutive J23100 promoter and plated on M63-glycerol agar with or without DOG and L-rhamnose. In the presence of DOG and the J23100 oligonucleotide and the absence of L-rhamnose, 95% of the colonies were fluorescent, indicating successful exchange of C1 (Fig. 2g). By omitting the selection pressure of DOG, a bacterial lawn could be observed that was fluorescent only when L-rhamnose was present. This shows that the galK gadget can be used to successfully exchange SEGA control elements with high efficiency by simple addition of oligonucleotides or PCR products and that different recombinant promoters can be identified by green-white screening.

Construction and validation of a SEGA strain collection with different control elements

We envision SEGA to be an extraordinary simple, but also comprehensive and versatile platform. To this end, we engineered a range of the most commonly used control elements into SEGA landing pads allowing streamlined construction of multiple variants encompassing all combinations of C1 and C2 by adding just a single cargo brick (see Figs. 3 and 4). The current SEGA collection of more than 100 strains is listed in Supplementary Table 1 and Supplementary Data 1. We included four promoters that have been characterized extensively before: the strong constitutive promoter J23100 from the Anderson collection and three inducible promoters: PT7, PrhaBAD, and Ptrc^39,40,41. These were all placed on the genome together with a GFP cargo and the tetA gadget. The performance of the four promoters in the landing pad context was tested by following growth and fluorescence in a microplate reader (Fig. 3a). Ptrc is highly inducible by IPTG, but also provides a basal level of (“leaky”) transcription in the absence of the inducer. PrhaBAD is completely off in the absence of the inducer, and the level of expression is titratable by the addition of L-rhamnose. Expression of gfp from PT7 is the strongest among the tested C1 elements, but it is tightly repressed in the absence of IPTG. For all tested promoters, the observed performance was in good agreement with their established performance in plasmid systems.

We additionally built landing pads with two inducible promoters from the “Marionette” collection as C1 elements: PSal and PTet⁴². This allowed a simple test of the effect of regulator position on the performance of inducible promoters in SEGA: either the regulator was arranged divergently to the promoter within the landing pad or it was present in a distant genome location in a regulator array as described by Meyer et al. The response function of PSal differed mostly in terms of maximum expression strength but not in sensitivity of the regulator between the two designs (Supplementary Fig. 5). Furthermore, the performance of Ptet in SEGA is severely affected by the presence of the tetracycline efflux pump expressed from the tetA gadget, but expression of gfp can be induced as long as the tetR regulator is integrated as part of the landing pad.

Besides transcription, translation is a major determinant of gene expression, and the ability to manage both provides highly stringent control of synthetic biology designs⁴³. To address this, we engineered C2 control elements—translation initiation regions (TIRs)—known to control the speed of translation initiation with a major impact on overall translational efficiency^44,45. In E. coli, the TIR covers both the Shine-Dalgarno sequence and the 5′end of the expressed gene^46,47,48,49. Thus, the TIR would typically be part of the incoming cargo, which precludes standardization of translation strength across different cargos. For this reason, we included a third gadget position between the control elements and the cargo that can serve multiple purposes: it provides a recombination site independent of C1 and C2, it encodes a standardized TIR independent from the cargo, and it can provide post-translational control and facilitate downstream applications. We picked a range of gadgets that offer useful features in the 5′end of various cargos: a bicistronic design (BCD) that will leave proteins of interest untagged, while still permitting normalization of expression levels³⁶; a ubiquitin-his tag for solubilizing proteins that at the same time enables histidine-tag mediated purification⁵⁰, a TEV protease site providing the opportunity to release a custom protein N-terminus⁵¹, a PROTi tag affecting the stability of proteins on demand⁵²; and two different secretion signals, encoding for a PelB signal peptide (PelBss) and YebF protein, serving the role to target proteins of interest to the periplasm or to secrete them outside of the cell⁵³.

To provide experimentally validated TIRs with defined translation levels for these different gadgets, we first constructed TIR sequence libraries, randomizing six nucleotides upstream from the start-codon as well as changing the first two codons after the ATG to all synonymous codons (Fig. 3b). This is an established and simple approach that can create variation in expression levels by more than 1000-fold⁴⁵. The TIR libraries were introduced into a landing pad with the constitutive J23100 promoter and the truncated 3′tetA gadget. The gadget variants were assembled on plasmids together with gfp and the 5′tetA fragment. The constructs were then amplified by PCR, using degenerate oligonucleotides to randomize the C2 elements, and integrated to complete the truncated tetA sequence on the genome (Supplementary Fig. 6a). The resulting TIR libraries were grown in liquid culture to mid-exponential phase and separated into three gfp expression levels—low, medium, and high—by fluorescence-activated cell sorting (FACS) (Fig. 3c, Supplementary Fig. 6b–g). Single clones were grown and selected from agar plates, purified by restreaking, and validated by sequencing (the TIR sequences can be found in Supplementary Table 5). The expression levels were subsequently verified by growth and fluorescence in a microplate reader and flow cytometer (Fig. 3d, Supplementary Fig. 7).

In our SEGA designs, the C1 and C2 sequences are always separated by an identical 23 bp sequence allowing the recombination of all different promoters and gadgets with different TIRs. This enabled the amplification of different C2 and gadgets together with gfp and 5′tetA from the FACS-sorted strains and integration into strains harboring the different C1 elements and the 3′tetA gadget (Supplementary Figs. 8a–10a). This way, we constructed all combinations of C1, C2, and gadget elements and verified the GFP expression levels for the resulting strains in a microplate reader and flow cytometer. Overall, we observed that the different C2-gadgets preserved the expression strength predictably at low, medium, and high levels when recombined with a different promoter (Fig. 3e, Supplementary Figs. 8–10). However, the total expression levels varied as expected depending on the promoter. One exception was the YebF gadget: the expected expression pattern was only observed in combination with PrhaBAD and we were unable to construct SEGA strains combining the J23100 promoter with YebF for extracellular secretion. The YebF strains were functional with Ptrc and PT7, but the expression strength deviated from the predicted low, medium, and high levels (Fig. 3h).

The main application of SEGA will probably be the simple exchange of cargos. Thus, it is relevant to test if different control elements and gadgets offer predicable expression of different cargos. To test this, gfp and tetA were replaced by rfp in all J23100 landing pads with different C2 gadgets (Fig. 3g) and performance of all constructs followed the same trend in expression levels as previously observed for gfp (Fig. 3e). This demonstrates an overall predicable performance of the SEGA modules with different cargos.

Multi-level control of a difficult-to-express gene is enabled by SEGA

The interplay of transcriptional and translational control is critical when it comes to gene expression burden or product toxicity⁴³. The standardized setup and the simplistic genome engineering workflow of SEGA with multi-level control might allow for optimizing genomic constructions with targets that show toxic effects. To explore this, we integrated the gene encoding the E. coli membrane protein YidC fused to GFP, which is known to be toxic when overexpressed⁵⁴. With the SEGA collection, only a single DNA fragment encoding yidC-gfp flanked by sequences homologous to RS2 and the truncated tetA gadget was required to construct six different production strains (Fig. 4a). The Ptrc, PT7, and PrhaBAD promoters were combined with the C2: low or C2: high translational control elements. We observed that integration into the Ptrc background yielded much fewer colonies, but positive recombinants were found for all combinations. Single clones were restreaked and validated by sequencing, but all sequenced PT7 constructs with a high TIR had a mutation in the T7 promoter region. Growth and expression were measured in a microplate reader. Based on fluorescence normalized to cell density, more YidC-GFP was produced under control of PT7 with the low translational control element than with the high TIR (Fig. 4b). We observed leaky production from Ptrc resulting in quite high protein titers without induction, but again the low TIR was performing better than the high TIR, when production was induced with IPTG (Fig. 4c). Moreover, induction of YidC-GFP expression from Ptrc and PT7 with IPTG resulted in arrested growth (Fig. 4d). However, the constructs under control of PrhaBAD produced the highest absolute amount of YidC-GFP (Fig. 4c) and there was no clear correlation between production levels and cell growth (Fig. 4d). This demonstrates that SEGA strains with different control elements can easily be screened simultaneously to identify the optimal production scenario for a challenging target.

Efficiency of integration of SEGA landing pads varies between different genomic locations

One of the first decisions to make when introducing new features on a genome is the target locus and this may affect both cellular fitness and the performance of the construct. To study SEGA landing pad integration in different locations, we chose 11 previously characterized loci evenly distributed across the E. coli K12 MG1655 genome^55,56,57 (Fig. 5b). The landing pad was typically placed directly downstream of a native gene, when possible in an intergenic region between two converging genes, to avoid disruption of promoters and other regulatory sequences (Supplementary Table 6).

We observed that some genomic locations seemed more difficult to target for manipulations than others. Similar observations were previously reported in a study targeting the whole genome with a transposon-based approach⁵⁷. Across 11 different loci, integration efficiencies ranged from 2 to 500 colony forming units (CFU) per µg DNA with the majority in the low-efficiency range (Fig. 5c). Possibly, different homology sequences used for targeting different loci could play a role in determining the integration efficiency. To control for this, we integrated the landing pads in different K12 sibling strains from the Keio collection⁵⁸ harboring identical kanR cassettes in different genomic locations equivalent to the loci targeted before. This enabled the use of identical homology sequences, i.e., the exact same DNA construct could be integrated in different loci (Fig. 5a). However, we still observed a wide distribution of integration efficiencies for different loci (Fig. 5c), and there was poor correlation between the integration efficiencies for the intergenic integration and the integration into the equivalent Keio strain. (Supplementary Fig. 12).

We then tested the integration efficiency of a SEGA cargo brick, in this case rfp, into the different genome locations harboring the SEGA landing pad (Fig. 5d). Selection for the correct integration was performed via NiCl₂ counter-selection and similar to what was observed previously, the integration efficiencies were distributed over a wide range, but the number of colonies obtained with the SEGA integration approach was sufficient to screen multiple clones in all tested loci (Fig. 5d). Due to the design of the SEGA landing pad, it is easy to differentiate false positive background colonies on the counter-selection plates by green-white-screening. We estimated the true positive rate of integration in 11 different genome locations by calculating the percentage of green and red fluorescent colonies across three replicates (Fig. 5e). The integration efficiencies into the SEGA landing pads, based on tetA counter-selection, were similar in most of the tested loci with true positive rates of above 80%. The only exceptions were the ycbX and wbbL loci where a high background of green colonies was observed. This provides important experience with engineering into different genomic loci and validates the functionality of SEGA landing pads across the E. coli genome.

Performance of SEGA modules in different genome locations

It is already well-established that the genome location affects the expression strength of native genes as well as heterologous constructs^56,57,59,60. We wanted to further investigate the effect of genome positioning on the performance of our SEGA landing pads and different modules such as C2 control elements. Non-disruptive integrations of the landing pads in 11 different genome locations were constructed as described above. For all integrations, a J23100 promoter landing pad brick was used, and for each locus two variants with low and high translation strength (C2) were constructed. The GFP expression of the 22 resulting strains was measured in late exponential phase in a microplate reader and the expression patterns across different loci followed the generally established notion that gene expression decreases with increasing distance from the origin of replication (Fig. 5f and g)^55,56,57,60. However, we observed outliers from this general trend: an example is the ykgH locus where gfp expression was 4-fold diminished compared to expression from the lacZ locus, which is located only 41 kb away. Importantly, the low and high expressing variants of the control element C2 showed the expected low- and high-range expression levels across the genome. The median expression from the high translation control element was 3.1-fold higher that expression from the low translation control element (Fig. 5h), emphasizing the standardized modular performance of SEGA. Comparing the expression from the landing pad in different loci yielded a median fold change of 1.3 and 2.1, respectively, for constructs with either the C2: low or C2: high translation initiation control element (Fig. 5h). Again, this demonstrates the standardized performance of different SEGA modules and provides clues to the relative importance of different factors such as promoters, TIRs, genomic locations etc. on the performance of synthetic biology designs.