A rhamnose-inducible system for precise and temporal control of gene expression in cyanobacteria

Cyanobacteria are important for fundamental studies of photosynthesis and have great biotechnological potential. In order to better study and fully exploit these organisms, the limited repertoire of genetic tools and parts must be expanded. A small number of inducible promoters have been used in cyanobacteria, allowing dynamic external control of gene expression through the addition of specific inducer molecules. However, the inducible promoters used to date suffer from various drawbacks including toxicity of inducers, leaky expression in the absence of inducer and inducer photolability, the latter being particularly relevant to cyanobacteria which, as photoautotrophs, are grown under light. Here we introduce the rhamnose-inducible rhaBAD promoter of Escherichia coli into the model freshwater cyanobacterium Synechocystis sp. PCC 6803 and demonstrate it has superior properties to previously reported cyanobacterial inducible promoter systems, such as a non-toxic, photostable, non-metabolizable inducer, a linear response to inducer concentration and crucially no basal transcription in the absence of inducer.


INTRODUCTION
Photoautotrophic microorganisms have great potential for the sustainable production of chemicals from carbon dioxide using energy absorbed from light. Cyanobacteria including Synechocystis sp. PCC 6803 ('Synechocystis' hereafter) and Synechococcus sp. PCC 7002 have been successfully engineered to produce 2,3-butanediol 1,2 , lactate 3 , isobutanol 4 , plant terpenoids 5 and ethanol [6][7][8][9] , and to allow the utilisation of xylose 10 . Cyanobacteria, particularly Synechocystis, are also used as model organisms for fundamental studies of important processes such as photosynthesis [11][12][13][14][15][16] , circadian rhythms [17][18][19] and carbon-concentrating mechanisms [20][21][22][23] . Due to specific challenges, genetic modification of cyanobacteria is more difficult than genetic modification of model heterotrophic microorganisms such as Escherichia coli and Saccharomyces cerevisiae. These challenges include polyploidy 24,25 , which makes the isolation of segregated recombinant strains slow and laborious; genetic instability of heterologous genes 26 and limited synthetic biology tools and parts such as promoters and expression systems. Improved synthetic biology capabilities for cyanobacteria would be useful for both fundamental and applied studies.
Inducible promoters are important tools which allow flexible control over gene expression, which is useful in many fundamental and applied studies. Unlike the limited number of constitutive promoters which have been shown to function well in cyanobacteria 6,10,[27][28][29], inducible promoters provide access to a wide, continuous range of gene expression levels using a single genetic construct, simply by varying inducer concentrations 30 . Furthermore, inducible promoters also allow control over the timing of expression of a gene of interest. An ideal inducible promoter system would have certain properties: Firstly, the promoter should not 'leak', that is, there should be no basal transcription in the absence of inducer, allowing very low expression levels to be used, and avoiding premature expression during strain construction and segregation, which can be associated with toxicity and mutation 26,31,32 . Secondly, the inducer molecule should be non-toxic, non-metabolisable, readily available and stable under experimental conditions (including under light in the case of photoautotrophic organisms), allowing sustained expression with no impact on growth caused as an artefact of the expression system itself. Thirdly, expression should demonstrate a near-linear response to inducer concentration over a wide range. Finally, expression should have a consistent unimodal distribution across a population of cells.
Several inducible promoter systems have been described in Synechocystis spp. and Synechococcus spp., but none are ideal. Metal ion-inducible promoters have been described which respond to nickel, copper, cadmium, arsenic and zinc [33][34][35][36] . Unfortunately these systems have disadvantages including the presence of many of the metals in standard growth media 37 , a narrow range of useful concentrations because the concentrations required for detectable and unimodal induction are close to toxic levels 34 , and some are 'leaky' in the absence of inducer.
The use of metals as inducers also has the potential to disrupt metal homeostasis, resulting in the sequestration of metals required as essential cofactors of many enzymes involved in photosynthesis and related metabolic pathways 11,[38][39][40] . Synthetic inducible promoters have also been constructed and used in cyanobacteria. Two promoter systems using the tetracyclineresponsive repressor TetR and its cognate operator sites have been engineered for use in cyanobacteria. The first example for use in Synechocystis resulted in a well-characterised, anhydrotetracyline (aTc)-responsive promoter with low leakiness and a good dynamic range 41 .
Unfortunately the inducer aTc is extremely sensitive to light and therefore induction from this promoter was transient and required high concentrations of aTc. The second example, in Synechococcus sp. PCC 7002, suffered the same issues with photolability of the inducer and low expression by comparison to a commonly-used strong constitutive promoter 42 . It is clear therefore that aTc-based inducible promoters are unsuitable for photoautotrophic growth conditions. The non-metabolisable analogue of lactose, isopropyl β-D-1-thiogalactopyranoside (IPTG) has also been tested for use as an external inducer of lac-based promoters in a variety of cyanobacterial strains 1,27,[43][44][45] , with mixed performance in terms of dynamic range and leakiness in absence of inducer. Finally, use of a green-light inducible promoter in Synechocystis sp. PCC 6803 has been reported 46 , but isolating the specific wavelengths required for induction from natural or white light used for growth is difficult, leading to unwanted induction.
To-date, heterologous promoters associated with the AraC/XylS family of positive transcriptional regulators have not been used in cyanobacteria. One promising candidate is the L-rhamnoseinducible rhaBAD promoter system of E. coli, which naturally has almost all of the ideal properties described above [47][48][49] . Recently this system was optimised in E. coli by the identification of L-mannose as a non-metabolisable inducer and constitutive expression of the activating transcription factor RhaS in order to make the system independent of the native regulatory cascade 50 .
Here we introduce the rhaBAD promoter of E. coli into Synechocystis sp. PCC 6803, characterise its behaviour, assess inducer stability and investigate the effects of modifying various promoter sequence elements and of varying expression of the transcription factor RhaS.
The result is an inducible expression system with several important advantages over expression systems previously characterised in cyanobacteria including precise control of the strength and timing of induction as well as sustained gene expression in the presence of light.
This system is likely to be very useful and widely applicable in Synechocystis and other cyanobacteria.

Analysis of the E. coli rhaBAD promoter in a heterologous Synechocystis context
In the heterologous context of a Synechocystis cell the rhaBAD promoter might be expected to perform differently than in the native E. coli host. To assess this, we considered the known The genome of Synechocystis encodes SYCRP1, a homolog of E. coli CRP 52,53 with 27% identity and 49% similarity to the E. coli protein. In E. coli, CRP binds to promoters containing specific binding sites when the concentration of cAMP is high, for example when glucose is scarce and other carbon sources must be metabolised for growth. The CRP-binding site in the rhaBAD promoter has been shown to be essential for this promoter to function fully in E. coli 47,54 .
In Synechocystis, SYCRP1 has been shown to positively and negatively regulate a number of promoters in response to changing cAMP concentrations 53,55 . The sequence of the CRP-binding site in Synechocystis (tgtgaNNNNNNtcaca) differs by only one nucleotide to the CRP-binding site sequence found in the E. coli rhaBAD promoter (tgtgaNNNNNNtcacg), which suggests SYCRP1 might bind to this heterologous promoter sequence 56,57 .
To the best of our knowledge, positively-regulated AraC/XylS-type expression systems like those in E. coli have not been reported in Synechocystis or in other cyanobacteria. In E. coli, the positive transcriptional regulator RhaS is essential for transcription from the rhaBAD promoter.
We used BLASTP 58 to search the genome of Synechocystis for a homolog of E. coli RhaS. No protein with significant similarity to E. coli RhaS was identified, suggesting that heterologous expression of the rhaS gene of E. coli which encodes this protein would be required for the heterologous rhaBAD promoter from E. coli to function in Synechocystis.
It has been hypothesised that differences in the RNA polymerase components between cyanobacteria and E. coli are one reason for E. coli promoters failing to function as expected when used in cyanobacteria 59  factor protein, corresponding to R412 and K406 respectively ( Figure 2). The above analysis suggested that the the E. coli rhaBAD promoter is likely to be functional in Synechocystis, and will probably require RhaS to be provided.

L-rhamnose is not metabolised by nor toxic to Synechocystis
Before testing whether the E. coli rhaBAD promoter is functional in Synechocystis, we first wanted to check if the natural sugar inducer L-rhamnose was metabolised by the cyanobacterium or if the use of a non-metabolisable analog of rhamnose would be required, as previously found in E. coli 50  Having confirmed that the rhaBAD promoter was functional in Synechocystis and demonstrated many of the desired properties of an ideal inducible promoter system, we next investigated if modifications to the promoter sequence itself or varying the concentration of RhaS in the cell affected the behaviour of the system. As the role of CRP is still poorly understood in Synechocystis and as the CRP-binding site is required for rhaBAD functioning in E. coli, we investigated the effect that deleting this operator sequence from the promoter would have on induction strength and/or kinetics. The reporter plasmids pCK305 and pCK306 were both modified through deletion of the CRP-binding operator sites, resulting in pCK313 and pCK314 respectively. Wild-type Synechocystis cells were transformed with either plasmid and integration and segregation confirmed as before. Transformants were then cultured in both photoautotrophic and mixotrophic growth conditions and the inducer-response and timecourse experiments repeated ( Figure 6). Results were very similar to those observed with pCK305 and pCK314, meaning the CRP-binding site is not required for the rhaBAD promoter to function in

Synechocystis.
Next we investigated whether increasing the cellular concentration of the transcriptional activator RhaS would change the response to inducer concentration, dynamic range or kinetics of rhaBAD promoter induction. The original rhaS RBS was predicted to have a low T.I.R. of just 72, so two synthetic RBSs were designed using the RBS Calculator 68 with much higher T.I.R. values of 5,000 and 18,000, and these new RBS sequences were inserted in place of the rhaS RBS used in pCK306, resulting in pCK320 and pCK321 respectively. These constructs were introduced into Synechocystis, integration and complete segregation was confirmed as before, then these transformants were used for inducer-response and timecourse experiments as before. The Synechocystis strains transformed with the new RBS variant plasmids pCK320 or pCK321 showed similar fluorescence response to inducer concentration and timecourses to cells transformed with pCK306 ( Figure S1).
Finally, we sought to directly compare all the functional rhaBAD expression system variants.
Absolute levels of fluorescence measured using flow cytometry cannot be directly compared between different days and experiments due to instrument variation. This is sometimes overcome in reporter studies by normalising to a reference promoter included in each separate experiment, allowing relative comparisons. Here, as we had a defined set of constructs to compare, we compared these directly in a single experiment. Synechocystis cells containing each of the rhaBAD-promoter reporter plasmids were cultured, both photoautotrophically and mixotrophically, in BG11 media supplemented with 1 mg/ml L-rhamnose, and the fluorescence intensity measured by flow cytometry after 191 h ( Figure S4). No statistically-significant difference was observed between cells containing constructs pCK306 (+rhaS), pCK314 (+rhaS, ΔCRP-binding site), pCK320 (+rhaS, T.I.R. of RBS of rhaS = 5,000) or pCK321 (+rhaS, T.I.R. of RBS of rhaS = 18,000). The inducible reporter constructs described above show non-zero levels of fluorescence in Synechocystis even in the complete absence of inducer, which could suggest that the promoter is 'leaky'. However, it was noted that even cells containing the non-functional promoter reporter constructs (such as pCK305) were slightly more fluorescent than wild type cells lacking any reporter plasmid ( Figure 7A). As these constructs are integrated into the Synechocystis genome, it was hypothesised that this basal fluorescence resulted from transcriptional read through from the chromosome rather than leaky expression from the rhaBAD promoter itself. To test this hypothesis, the rhaBAD promoter of pCK321 (one of the above-described derivatives of pCK306 which performs identically) was removed resulting in the promoterless plasmid The only observed flaw with this implementation of the rhaBAD promoter in Synechocystis was a low level of basal expression, which we found was independent of the rhaBAD promoter. The sll0410 insertion site adjacent to ndhB has been used previously, but seems to result in transcriptional read-through of inserts, presumably from the promoter found inside the ndhB ORF 69 . For most inducible expression studies, this observation will be unimportant and expression constructs reported here will be ideal, because in many cases the ability to specify extremely low expression levels is not required. Where extremely low or zero basal and induced expression is required, alternative integration sites or extrachromosomal plasmids may prove more suitable 70 .
We found that the rhaBAD promoter of E. coli was functional and inducible in Synechocystis without any modification of the promoter sequence itself. This was not obvious in advance given reports of difficulties in using E. coli promoters in cyanobacteria. In this case our analysis of the relevant transcription factor machinery and interacting residues successfully predicted function of this promoter in Synechocystis, so it is interesting to consider whether this promoter might function in other cyanobacteria such as Synechococcus sp. PCC 7002 or Arthrospira species.
For example, one of the sigma 70 factor residues important for interaction with RhaS, K593, is not found in the Synechococcus sp. PCC 7002 ortholog but is found in the Arthrospira plantensis ortholog. The residue found in the Synechococcus ortholog is an arginine, a similar basic amino acid, so may still interact appropriately with RhaS for function.
This study represents an important step towards addressing the shortage of reliable synthetic biology tools for the manipulation of cyanobacteria, both for fundamental and applied studies.
The characteristics of the rhamnose-inducible expression system shown in this work will allow greater control of gene expression in cyanobacteria than previously possible. Despite this progress, much work remains in the development and characterisation of other synthetic biology tools to address the unique challenge of engineering these important photoautotrophic organisms and realising their applied potential.

Plasmid Construction
A table of all plasmids and oligonucleotides (Table S1) is provided in the Supplementary Information. All plasmid construction was carried out using standard molecular cloning methods.
Full details are provided in the Supplementary Information.

Strain Construction
Wild-type Synechocystis cells were cultured in BG11 supplemented with 5 mM D-glucose to an optical density (measured at 750 nm) of 0.5 and 4 ml harvested by centrifugation at 3200 g for 15 mins. Pellets were resuspended in 100 μl BG11, 100 ng of plasmid DNA was added and the mixture was incubated at 100 μmol m −2 s −1 white light for 60 mins. Cells were spotted onto BG11 glucose plates and incubated at 100 μmol m −2 s −1 white light for 24 h at 30°C. Cells were collected and transferred onto BG11 glucose plates supplemented with 30 μg ml -1 kanamycin.
When single colonies appeared, transformants were segregated through passaging on selective plates and full segregation was confirmed by PCR.

Assays
After confirmation by PCR that Synechocystis transformants were fully segregated, cells were cultured to mid exponential phase before subculture to a final optical density (measured at 750 nm) of 0.1. Cultures were grown for 24 h and then L-rhamnose added to a variety of final concentrations. The optical density of cultures was monitored at 750 nm and high-resolution fluorescence intensity of each cell was performed using flow cytometry using an Attune NxT Flow Cytometer (ThermoFisher). Cells were gated using forward and side scatter, and GFP fluorescence (excitation and emission wavelengths: 488 and 525 nm [with 20 nm bandwidth] respectively) was measured. Histograms of fluorescence intensity were plotted, and mean statistics extracted.

Author Contributions
CK and JH designed the study; CK performed experiments; CK, AH and ATM performed plasmid construction; CK and JH prepared the manuscript with input from AH and ATM.

Notes
The authors declare no competing financial interest.