Regulatory tools for controlling gene expression in cyanobacteria

Abstract

Cyanobacteria are attractive hosts for converting carbon dioxide and sunlight into desirable chemical products. To engineer these organisms and manipulate their metabolic pathways, the biotechnology community has had to develop genetic tools to control gene expression. Many native cyanobacterial promoters and related sequence elements have been used to regulate genes of interest and heterologous tools that use non-native small molecules to induce gene expression have been demonstrated. Overall, IPTG-based induction systems seem to be very leaky and have low fold induction ratios in cyanobacteria. A variety of other induction systems have been optimized to enable tighter control of gene expression. There are substantial differences in performance of many of these promoters and induction systems in cyanobacteria compared to E. coli, and we hypothesize that these differences are due to fundamental differences in physiology between organisms. This review is not intended to summarize all known products made in cyanobacteria and nor the performance (titer, rate, yield) of individual strains, but instead will focus on the genetic tools used to control gene expression the inherent aspects of cellular physiology that influence and control gene expression in cyanobacteria.

Introduction

To meet growing societal needs and reduce the use of fossil fuels as chemical feedstocks, researchers are engineering photoautotrophs to create chemical products from low-cost, renewable inputs: carbon dioxide, water, and sunlight. Harnessing the power of photoautotrophs has substantial benefits compared to using heterotrophs because carbon dioxide is cheap, overabundant form of carbon that would not compete with edible sugars derived from crops¹. When selecting a photoautotrophic host, synthetic biologists are faced with two tradeoffs – production cost certainty and ease of genetic engineering. Modern agricultural practices provide solid estimates of plant cultivation and harvesting costs, while the uncertainty of large-scale cyanobacteria cultivation remains a barrier to commercialization. Conversely, cyanobacteria can be genetically modified in the lab with mutation cycles measured in weeks rather than months for most crops. For this reason, metabolic engineering and synthetic biology in plants has progressed more slowly than in microbes and other heterotrophic hosts. That said, the growing number of fully sequenced plant genomes and the emerging ability to make precise insertions, deletions, and targeted mutations with Cas9 and similar nucleases will likely revolutionize the field of metabolic engineering in plants. Similar breakthroughs in cyanobacteria cultivation techniques will be needed to leverage the other physiological and processing advantages of cyanobacteria. These advantages include faster growth rates, greater salt tolerance, ability to grow on marginal lands, and ability to use waste streams as nutrients.

There is great diversity in cyanobacterial morphology, physiology, and environmental preferences. There are marine species, Prochlorococcus and Synechococcus, that are abundant in the ocean and could be responsible for as much as 25% of marine carbon fixation². There are freshwater cyanobacteria, as well as cyanobacteria found in glaciers, soil, deserts, and hot springs³. Different cyanobacteria species are cultivated as single cells or as multicellular aggregates; in salt or fresh water; in the presence or absence of reduced nitrogen (i.e. nitrogen fixing or non-nitrogen fixing). Given this wide diversity, researchers have selected a few model cyanobacteria to study and engineer in detail. Synechocystis sp. strain PCC 6803 is a unicellular, freshwater cyanobacteria that has been studied for decades as a model of photosynthesis⁴. Another freshwater cyanobacterium, Synechococcus sp. strain PCC 7942 has been extensively studied especially as a model of the circadian clock. There have been recent efforts to study and engineer Synechococcus sp. strain PCC 7002, a halotolerant, unicellular cyanobacteria, because of its faster doubling time compared to PCC 6803 and PCC 7942⁵. Anabaena (Nostoc) sp. strain PCC 7120 is a cyanobacteria capable of fixing nitrogen through the formation of heterocysts⁶, a trait that would reduce (if not eliminate) the need for providing reduced nitrogen to cultures. In the remainder of this chapter, we will largely focus on gene regulatory tools used in these model organisms (Table 1) because these species are where regulatory tools have been developed and deployed. Most regulatory tools were built to interrogate natural phenomena found in cyanobacteria or to engineer cells to produce valuable chemicals. The future will tell us whether existing tools will transfer to other novel cyanobacteria or if corresponding tools will need to be created to leverage advantageous traits in non-model species.

Table 1.

Cyanobacteria strains mentioned in this chapter

Cyanobacterial Strain Name	Abbreviation
Anabaena (Nostoc) sp. PCC 7120	PCC 7120
Gleobacter violaceus	G. violaceus
Leptolyngbya sp. strain BL0902	BL0902
Nostoc punctiforme	N. punctiforme
Prochlorococcus marinus subsp. pastoris str. CCMP1986	CCMP1986
Prochlorococcus sp. MED4	MED4
Synechoccus PCC 6301	PCC 6301
Synechococcus WH7803	WH7803
Synechococcus elongates PCC 7942	PCC 7942
Synechococcus elongatus UTEX 2973	UTEX 2973
Synechococcus sp. strain PCC 7002	PCC 7002
Synechocystis sp. strain PCC 6714	PCC 6714
Synechocystis sp. strain PCC 6803	PCC 6803
Synechocystis sp. strain WHSyn	WHSyn
Thermosynechococcus elongatus BP-1	BP-1

There are many successful examples of making native and foreign metabolites in cyanobacteria, and comprehensive lists of performance metrics (titer, rate, yield) can be found elsewhere^7–10. Unfortunately, titers and productivities lag significantly behind model heterotrophs engineered to produce the same compounds. It remains to be seen whether this discrepancy is due to differences in metabolism between photoautotrophs and heterotrophs, or arises from gaps in knowledge and engineering tools. While yeast, E. coli, and other heterotrophs have been the subject of metabolic engineering for several decades, a significantly smaller community of researchers has only recently pursued work with cyanobacteria. This difference can be seen in the standardized media and cultivation equipment used with heterotrophs and the diverse set-ups used to grow cyanobacteria. Furthermore, culturing cyanobacteria requires optimization of many additional environmental variables including light source and intensity, timing of light/dark cycling, CO₂ concentration, and vessel geometries. Despite these obstacles, there have been a number of significant steps taken to increase production of target chemicals including enhancing flux through desired pathways, eliminating competing pathways, and developing tolerance to high product concentrations⁷. All three of these strategies are heavily reliant on the ability to manipulate gene expression.

The well-developed toolsets for gene expression combined with the fast generation times of bacteria has enabled significant process in engineering them to create non-native chemicals and protein products. There has been significant work to build and expand toolsets to control gene expression in cyanobacteria, because often tools developed in other organisms (E. coli) often do not behave similarly. We will cover many examples where directly using genetic tools from other organisms into cyanobacteria has yielded unpredictable and undesirable results. There are a great many factors and steps that facilitate gene expression (Figure 1), some of which may be unique and cyanobacteria-specific. Many of the successful gene expression tools in cyanobacteria have been created by systematically optimizing parameters (e.g. DNA and protein sequences) from underperforming embodiments. In this chapter, we will discuss several unexplored areas of cyanobacterial physiology (the transcriptional landscape, a highly over-represented repeat sequence, presence of multiple copies of the chromosome, as well as prevalence of small proteins), that could be playing significant roles in controlling gene expression and cofounding transfer of existing regulatory tools. With future research in these areas we believe improved gene regulatory tools will be built and greatly accelerate the use of cyanobacteria as a chemical production platform.

Figure 1.

Overview of key factors governing gene expression.

Promoters

Endogenous Promoters

Many native promoters have been used to express heterologous genes in cyanobacteria and comprehensive lists can be found elsewhere⁸. There has been extensive study of many of the photosystem promoters^11–13, but these promoters have limited utility for many metabolic engineering applications because their activity is light-responsive. Although many native promoters have been tested on reporters or genes of interest there has been little systematic study of native promoters to determine relative strengths or sequence elements that could be useful in producing predictable expression of a heterologous gene.

A few native promoters from PCC 7002 have been used to compare expression levels from a common reporter. Potential promoters with a range of expression levels as measured by RNA-seq were chosen and expression of heterologous yellow fluorescence protein (YFP) was tested. There was no correlation between expression level of native sequence and experimental measured YFP levels, and most native promoters produced no more YFP than the background wild type fluorescence¹⁴. This suggests that native promoters cannot be simply fused to heterologous genes without additional design.

Similarly, native promoters from PCC 7942 that produced high levels of protein (measured by proteomics) were used to attempt to increase expression of limonene synthase for the production of limonene. Varying levels of limonene were seen with different promoters, but only one native promoter produced as much limonene as the heterologous P_trc promoter with a synthetic RBS and none outperformed this strain. A heterologous pea psbA promoter produced about 13-fold more protein and produced the most limonene¹⁵. Many native cyanobacterial promoters have been used, but sometimes they contain unknown regulatory elements that have unintended effects on gene expression.

Genetic Instability

Using endogenous promoters and regulators can cause genetic instability. It’s been suggested that genetic instability, where strains suddenly drop in productivity or performance, is much more widespread than is reported and that researchers may not be incentivized to investigate failed results¹⁶. One published example in PCC 7942 was an ethylene producing strain containing the ethylene forming enzyme (efe) that produced ethylene but looked yellow-green. Healthy, green/blue colonies soon appeared that no longer made ethylene, and these colonies were shown to have insertions at sites remarkably similar to the HIP1 sequence (see below) that caused a truncated protein¹⁷. This genomic instability of the ethylene-producing strain was not seen in PCC 6803 although this could be due to the use of a codon-optimized version of the gene¹⁸.

Another example of genetic instability was seen in a strain of PCC 7002 engineered to produce mannitol. Cultures would sometimes suddenly stop producing mannitol and start to grow at a faster growth rate closer to WT. One mutant was isolated that had an inactivating point mutation in one of the two heterologous enzymes¹⁹. Spontaneous inactivating mutations were also seen in a PCC 6803 strain that contained a gene to made lactic acid and also expressed a transhydrogenase²⁰. Mutations causing single amino acid changes that lowered enzyme activity was seen in a PCC 7942 strain engineered to produce isopropanol²¹.

The causes of these mutations and genetic instability is unknown and could be related to DNA repair and mutations caused by expressing a heterologous pathway. Another factor causing genetic instability in cyanobacteria is the maintenance of many copies of the chromosome (see Polyploidy section). There are countless examples of the difficulty and inability to obtain fully segregated strains, and variable productivities and phenotypes are seen when strains are not fully segregated. These observations motivate the development and use of more tightly-controlled gene expression regulation and highlight the advantages of using heterologous sequence elements that don’t contain unknown native cyanobacterial control mechanisms.

Heterologous Promoters

Heterologous promoters have been tested to control gene expression to avoid issues with native regulation and potential strain instability due to repeated sequences. Widely used promoters in E. coli such as λP_L (BBa_R0040), λP_R (BBa_R0051), P_Lac (BBa_R0010) have been shown to have little to no detectable expression in PCC 6803²². Very little fluorescence could be detected from reporters expressed from P_lac, P_R, P_trc10, and P_tet in PCC 6803²². Changing elements such as the −10 element to the consensus TATAAT has improved expression from heterologous promoters that originally worked well in E. coli but not in cyanobacteria²³. However, it’s also been suggested that a fully consensus promoter is undesirable and may lead to low activity because of the large number of contacts between RNA polymerase and the promoter²⁴. Heterologous promoters may also perform differently in different cyanobacteria strains. Heterologous promoters P_trc, P_LlacO1, P_conII, P_J23101, and P_J23119 were all tested in PCC 7942 and P_conII and P_J23119 produced the highest expression levels²⁵.

Heterologous artificial promoters have been tested in cyanobacteria and show variable results. Members of the Anderson promoter library (designed in E. coli, part of the BioBrick collection from the iGEM Registry) has been tested in both PCC 7002²⁶ and in PCC 6803²⁷. In PCC 7002 there was relatively poor correlation when constructs were tested in both cyanobacteria and E. coli (R²= 0.434²⁶). Promoter libraries are a useful tool to get a large range of expression levels, but it is necessary to test them in a new host because they do not seem to perform similarly across organisms.

Differences in cyanobacterial transcription from E. coli

Heterologous promoters may not function as expected due to the fundamental differences in transcription. Cyanobacteria have a strong consensus −10 element but seem to lack consensus or motifs in the −35 region²⁸. It has been suggested that cyanobacteria could be reliant on sigma factors in the place of a consensus −35 element. A study of native promoters in MED4 also observed the conserved −10 element but lacked a consensus −35 element. The authors suggested that the −35 regions may be divided into conserved subclasses²⁹.

Cyanobacterial promoters have been divided into three groups based on the DNA sequence elements they contain and which sigma factors recognize them. Type I promoters (σ⁷⁰) contain both a −10 element (consensus 5’-TATAAT-3’) and a −35 element (consensus 5’-TTGACA-3’). Type II promoters contain the −10 element but lack a −35 element, instead containing target sites for other transcriptional activators. An example of a type II promoter is glnB P2 in PCC 6803 that contains the −10 element and the GTAN₈TAC binding motif recognized by NtcA which mediates enhanced transcription under nitrogen-limiting conditions. Type III promoters do not have consensus −10 or −35 elements and most likely are involved in responses to various stresses and stimuli through binding of type III sigma factors³⁰.

Sigma factors are important proteins that facilitate binding of the core RNA polymerase with promoters to initiate transcription, and they are divided into three groups and recognize the corresponding type promoters. σ⁷⁰ (Group 1) is the primary sigma factor in E. coli responsible for initiating transcription at housekeeping genes and cyanobacteria have a conserved homolog, sigA^31,32. Group 2 sigma factors are nonessential, and respond to environmental stimuli. Examples are the heat-shock responsive SigD, dark-responsive and nitrogen starvation-responsive SigB, and less well-studied SigC and SigE. Group 3 sigma factors have alternative structures and respond to stress and include SigF, SigG, SigH, and SigI. SigF has been implicated in control of pili genes but little else is known about others³³. The lack of knowledge of these important transcription factors as well as the coordinated networks between them make predicting transcriptional levels based off of promoter sequence extremely difficult.

Differences in core transcription machinery may also be responsible for the observed differences in expression of promoters between E. coli and cyanobacteria. The cyanobacterial RNA polymerase (RNAP) is composed of different subunits than that of E. coli; instead of on β’ subunit (RpoC) like in E. coli cyanobacteria have both γ (RpoC1) and β’ (RpoC2)³³. Additionally, RpoC2 has large insertions between domains G and H. The reasons for this difference in architecture has not been explained or described in terms of gene expression control.

Other differences between cyanobacterial and E. coli RNAPs have been observed by testing purified RNAP complexes in vitro. RNAP from PCC 7942 and BP-1 both showed lower levels of abortive transcription and less miscincorpoation than E. coli RNAP³⁴. Additionally, transcription of the same transcript was significantly slower in with cyanobacterial enzymes (taking 164 seconds to complete the transcript compared to 26 seconds). These cyanobacterial RNAPs also paused more frequently and cleaved its transcript faster when NTPs were absent. Mn²⁺ is typically thought to be toxic because RNAP tends to misincorporate noncognate bases, but cyanobacteria maintain approximately 2 orders of magnitude higher intracellular levels of Mn²⁺ than E. coli because of its key role in photosynthesis³⁵. It was proposed that cyanobacterial RNAP maintains high fidelity even in the presence of high Mn²⁺ levels with expense of slower elongation³⁴. If similar trends are also seen in vivo, these fundamental differences in transcription could have profound effects of gene expression and native mechanisms used to control gene expression.

Differences in RNAP-interacting factors may also be responsible for observed differences in transcription between cyanobacteria and other model organisms. The E. coli protein DksA has been shown to bind RNAP and help stimulate transcription from λP_R³⁶_. No homologs for DksA have been found in cyanobacteria. Additionally, ppGpp affects gene expression at many promoters including λP_R³⁶_. ppGpp has been shown to be a regulator in cyanobacteria just as it is in model heterotrophs like E. coli, but alternate signals can affect its synthesis. In E. coli ppGpp levels rise as part of the stringent response during stresses. Contrastingly, ppGpp has been shown to be involved to adaption to darkness in PCC 7942³⁷ and heterocyst formation in PCC 7120³⁸.

Many promoters, including P_lac, require cAMP receptor protein (Crp) for transcription. These promoters contain a Crp binding site and positioning of this binding site relative to the −10 element is important to ensure that Crp and RNAP are on the same side of the DNA and can interact to cause activation. The E. coli P_lac promoter has a Crp-binding site −61.5 bases from the transcription start site³⁹, which is 10.5 α-helical turns from the −10 element. A cAMP receptor protein has been identified in PCC 6803 (SYCRP1) and was shown to bind the consensus E. coli Crp sequence in the presence of cAMP⁴⁰. Several SYCRP1-dependent promoters have been identified as well as their putative Crp-binding sites, but not all have proper spacing to ensure Crp and RNAP localize to the same side of the DNA⁴¹. Study of these important transcription factors in cyanobacteria may reveal global differences in specificity and function and help explain differential gene expression observed between organisms.

Global analyses have revealed intriguing aspects of the transcription landscape in cyanobacteria. Most of the genome seems to be transcribed with 88% having reads from RNA-seq align to either strand and 82% of the non-coding regions being transcribed in PCC 7942²⁸. Transcriptome reads mapping to almost the entire genome (>=94%) was seen for PCC 6803, PCC 7002, and CCMP1986⁴². This shows remarkable similarity to plant chloroplasts where over 99% of the plant chloroplast genomes from rice, maize, Arabidopsis, and green algae Chlamydomonas reinhardtii⁴². The model for transcription in chloroplasts involves haphazard initiation and termination resulting in many variable precursor transcripts that are then processed to their final 5’ and 3’ ends by endo and exoribonuceleases. To our knowledge there has been no in-depth study to see if a similar process could be occurring in cyanobacteria.

Other differences in transcription in cyanobacteria compared to E. coli include a distinct pause site approximately 63 nucleotides downstream of the transcription start site as observed by RNA pol ChIP data in PCC 7942²⁸. This is slightly different than has been reported in E. coli where the majority of RNAP is associated with the promoter⁴³. Additionally, a preference for an adenine was seen at the transcriptional start site in PCC 7942²⁸. These differences in both core RNAP machinery as well as factors that influence transcription initiation may help explain differences in promoter strength between E. coli and cyanobacteria.

Known Transcriptional Regulators

Global regulators play major roles in controlling gene expression, and there has been a small subset of global transcriptional regulators studied in cyanobacteria. One global transcriptional regulator, NtcA, has been shown to regulate nitrogen metabolism. NtcA belongs to the CAP (catabolite gene activator or cyclic AMP receptor protein) family and has been identified in nearly all cyanobacteria strains mentioned in this chapter⁴⁴. DNase footprinting revealed that NtcA bound to GTAN₈TAC which takes the place of the −35 element in NtcA-responsive promoters⁴⁵. Activation of these promoters is facilitated through an increase in the binding constant of RNAP and an increase in the rate constant for isomerization from the close to open complex⁴⁴. The consensus sequence for PCC 7120 NtcA was reported to be slightly different (TGTN_{9 or 10}ACA) and proposed to be both a positive and negative transcriptional regulator, like many other members of the CAP family. This PCC 7120 consensus sequence is also located upstream of the large subunit of RuBisCO, rbcL, suggesting it may be playing other regulatory roles.

Another transcriptional regulator involved in sensing nitrogen compounds, NtcB, is often located near NtcA and nitrate assimilation genes (nir operon)⁴⁶. In PCC 7120 NtcB has been shown to bind upstream of the nir operon and is a booster to enhance expression of the nir operon in combination with NtcA⁴⁷. There have been conflicting reports of conditions where NtcB activation is seen. In PCC 7942 NtcB activation of the nir operon is only seen in the presence of nitrite while in PCC 6803 and PCC 7120, activation is seen independently of the presence of nitrite⁴⁸.

Other transcriptional regulators have been shown to involved in regulating genes involved in the CO₂ concentrating mechanism. CcmR and CmpR are both LysR family transcriptional regulators that sense and respond to CO₂ concentrations. Transcriptional activator, CmpR, has been identified in PCC 7942 and PCC 6803 and has been shown to bind to the upstream region of the cmp operon that encodes a high-affinity bicarbonate transporter⁴⁹. Ribulose-1,5-bisphosphate and 2-phosphoglycolate were able to enhance binding of CmpR to its targeting, suggesting that these metabolites are how this regulator senses CO₂ levels^50,51. The transcriptional repressor, CcmR, represses a suite of targets including sbtA, bicA, porB, as well as itself⁵². Its binding motif as well as novel regulon members have been identified through network analysis of PCC 7002 grown under many different conditions⁵³. α-ketoglutarate and NADP⁺ have been identified as co-repressors of CcmR⁵¹.

Other transcriptional regulators sense both nitrogen and carbon and mediate transcriptional responses. PCC 6803 has two cyanobacterial AbrB-like proteins (cyAbrBs); cyabrB1, which appears to have an overlapping or complementary regulon to NtcA⁵⁴, and cyabrB2, which regulates parts of the high affinity bicarbonate transport system and overlaps with CcmR and CmpR regulons⁵⁵.

Many transcriptional repressors respond to iron and the redox status of the cell. The SufR transcriptional repressor has been identified in PCC 6803, PCC 7002, PCC 7120, N. punctiforme, BP-1 as well as others. SufR represses transcription of the suf operon that encodes proteins involved in the assembly of iron-sulfur clusters. This repression is removed under conditions of oxidative stress or iron limitation⁵⁶. FurA is a global cyanobacterial regulator that binds a 19 to 23 bp consensus sequence in the presence of iron. It can act as either an activator or repress in PCC 7120⁵⁷ and similar results have been found for PCC 7942⁵⁸. PedR is a LuxR type transcriptional regulator that senses the redox state of the photosynthetic electron transport chain and facilitates gene expression change⁵⁹.

This is just a small subset of the major transcriptional regulators in cyanobacteria. There remain many uncharacterized transcriptional regulators, but elucidating both their sensing mechanism, targets, and functions is complicated by their many complementary and overlapping targets. Recently, 32 putative transcriptional regulators were studied in PCC 6803 by screening knockouts grown under a variety of conditions and using liquid chromatography-mass spectrometry based metabolomics⁶⁰. There are clearly many unknown factors that could be influencing expression of a gene of interest.

Induction/Repression Systems

Whether studying native functions or implementing synthetic biology strategies it is desirable to be able to turn on and off gene expression with desired timing and strength. Induction and repression systems are well-established in model heterotrophs and can respond to an environmental shift or presence/absence of a small molecule. Unfortunately, directly moving these systems into cyanobacteria has not been universally successful (discussed below). Therefore, significant efforts have been made to adapt and optimize them for use in cyanobacteria.

It is extremely difficult to compare induction systems both within and across cyanobacteria strains and species. Different reporters have bene used, integrated at different chromosomal locations, with different selectable markers. Comparisons between different strains is difficult due to the different culturing conditions with varying media compositions, light intensities, temperatures, culture vessels and CO₂ concentrations. Most of the engineered genetic tools have been designed and optimized with fluorescent reporters as the output due to their ease of quantification. Most papers report fold-induction ranges between uninduced and induced strains, which gives little indication of overall ranges of expression. Additionally, these fold inductions take different forms: fluorescence of reporter proteins, protein quantification with western blots, transcript levels via quantitative PCR, measurements of product titers, or changes in cellular physiology.

A compilation of the fold induction data from all of the induction systems outlined in the following two sections can be seen in Figure 2. We’ve divide them into both transcription-based and translation-based induction systems. There is a wide range of induction ratios seen with various inducers and in various strains. Unfortunately, this comparison does not show overall expression levels due to the indirect measurements of induction in various publications. Even systems with low fold induction ratios can be extremely useful and may have tighter repression than systems with a larger dynamic range of induction.

Figure 2.

Fold induction ratios from inducible systems in cyanobacteria. We’ve color-coded bars by inducer (IPTG, aTc, metals, and theophylline), and separated them by induction mechanism (transcription vs translation). The organism and citation can be found below the bar.

Induction via transcription

There are many native cyanobacterial transcripts that have been shown to respond to environmental stimuli: high-light responsive psbA promoter, copper-regulated petE promoter, nitrate/nitrite-inducible nirA promoter, and the nickel-responsive nrsA promoter⁶¹. Some of these promoters have been used to induce expression of a gene of interest. A native PCC 6803 cpcG2 promoter that is up-regulated in green-orange light and down-regulated in red light was used to create a green-light inducible lytic system in PCC 6803⁶². This light-responsive promoter was placed in front of a T4-derived lysis gene and was shown to induce cell lysis upon exposure to green light resulting in a reduced growth rate. Additionally, reporters driven from Ni²⁺-inducible P_nrsB and Co²⁺-inducible P_coaT showed 50 and 70-fold induction respectively, but have very limited ranges of inducer concentrations⁶³. A Zn²⁺ responsive system was built in PCC 7002 that utilizes the PCC 7942 SmtB repressor and SmtB-regulated promoter P_smtA and achieved 19-fold induction⁶⁴.

Systems that respond to these large environmental shifts or natural inducers have a limited utility for induction of proteins of interest because native targets would be affected as well. Often, it is ideal to have no other cellular or environmental changes besides the induction of your protein of interest, so cellular effects can be attributed to the protein or interest and not global shifts in regulation and metabolism. Efforts have been made to use non-native induction systems and optimized to achieve activity in cyanobacteria.

Directly moving these induction systems from other organisms has been unsuccessful and there has been significant difficulty with IPTG-inducible promoters. The widely used IPTG-inducible, P_trc did have high expression in PCC 6803, but only led to 1.6-fold induction when LacI was present and IPTG was added²². Leaky expression from the P_trc promoter was also seen without ITPG when a tagged protein was purified in PCC 6803⁶⁵. In PCC 7942 an IPTG induction system using P_trc was sufficiently leaky to complement a mutant strain even in the absence of the inducer⁶⁶. Leaky expression of heterologous enzymes was sufficient for isobutanol production in PCC 6803 without inducer⁶⁷. However, there are other examples in PCC 7942 where 360-fold induction or higher was seen with P_trc⁶⁸.

Many versions of IPTG inducible systems were very leaky in the absence of IPTG, resulting in low fold induction ratios. The P_tac (IPTG-inducible combination of E. coli’s P_trp and P_lacUV5) was inducible in PCC 7120, but induction was only 5-fold⁶⁹. Induction of heterologous genes in the 2,3-butanediol pathway in PCC 7942 using P_LlacO1 was only 1.2 to 1.6-fold and showed very significant leaky expression in the absence of IPTG⁷⁰. Use of IPTG-inducible trp-lac promoter to control a sucrose permease gene in PCC 7942 showed 6-fold induction as measured by quantitative reverse transcriptase PCR⁷¹. A variant of P_lac, P_A1lacO-1 that contains a second lac operator sequence, showed stronger repression than P_trc, but became leakier at higher cell densities¹⁸. However, there are examples where similar fold inductions (~2.5-fold) are seen in both E. coli and cyanobacteria⁷².

A systematic study varying the spacing and number of the lac operator sequences in PCC 6803 showed that there was weak repression when only a single lac operator sequence was used (P_trc1O and similar variants). However, when the second operator was added (P_trc2O and similar variants) there was very significant repression, as high as 400 times, when LacI was not present, but these strains could not be induced. They observed the same reliance on the 11 bp periodic placement of the operator sequences as seen in E. coli, ensuring LacI and DNA looping can occur on the same side of the DNA. However, very significant differences in repression and induction were observed, that remain unexplained⁷³.

There seems to be a trade-off between using heterologous IPTG-inducible systems from E. coli and native cyanobacterial induction systems. Different induction systems were used to drive expression of enzymes in ethylene metabolism in PCC 6803 and it was shown that P_trc produced high expression and ethylene levels, but was not tightly regulated like the same construct expressed in E. coli. Contrastingly native metal-inducible promoters (P_petE, P_coa, and P_smt) were tightly repressed, but had low expression even when induced¹⁸. Similarly, a native cyanobacterial promoter controlled by a metal-sensing repressor, P_smt, had the highest induction ratio in PCC 7942 when compared with P_LtetO1, P_ConII-ribo, P_trc, P_LlacO1, and P_BAD²⁵.

Directly moving induction systems from E. coli has not always led to favorable expression levels or tight control of expression levels, so several induction systems have been optimized and modified for use in cyanobacteria. Two IPTG-inducible systems have been optimized for PCC 7002²⁶. Initial systems were optimized by changing the −10 elements, the spacing between the −10 element and the LacO operator sequence, as well as the spacing between the −35 and −10 element. Both constructs were titratable between 1 μM – 10,000 μM IPTG but had different ranges of expression. cLac143 had higher expression and was somewhat leaky in the absence of IPTG (~5% of the constitutive version), but had a 48-fold range of expression. cLac94 had a 38-fold range expression, but does not reach the same expression level when fully induced. These two constructs used the cyanobacterial phycocyanin truncated promoter (P_cpt) flanked by LacO operator sequences to drive expression of the gene of interest. LacI expression was controlled by P_MB2, a synthetic promoter derived from BBa_J23119. Tighter repression was seen by using lacI W220F that has been shown to be less leaky possibly due to tighter DNA binding⁷⁴. Induction was also seen when P_cpt was replaced with P_trc, but a smaller dynamic range was achieved. These optimized IPTG-inducible systems have some of the largest fold induction ranges of those tested so far.

Anhydrotetracycline (aTc) induction systems have been developed for PCC 7002⁷⁵, PCC 7120⁷⁶, and PCC 6803²³. The aTc induction system in PCC 7002 had a 32-fold induction and was titratable with concentrations of aTc between 10 ng/mL and 1000 ng/mL aTc. Although the 32-fold induction is not as high as some other inducible systems, the tightness of this system has advantages compared to other leakier but higher expression systems. The aTc induction system was critical to achieve inducible repression using CRISPR interference in PCC 7002⁷⁷.

The aTc induction system in PCC 7120 achieve >1,200-fold induction and had a dynamic range with between 0 ng/mL and 2,000 ng/mL⁷⁶. There was a decrease in growth of PCC 7120 with 2000 ng/mL aTc, but the growth rate appeared to be unaffected with 200 ng/mL aTc. This system was adapted to respond to nitrate by placing the nirA nitrate-specific promoter in front of tetR and over 200-fold induction was seen in the absence of the inducer when cells were grown in nitrogen-free conditions in comparison to media supplemented with nitrate.

The aTc induction system for PCC 6803 was extensively optimized and achieved 230-fold induction under light-activated heterotrophic growth²³. Expression was titratable with aTc concentrations between 0 ng/mL - 4 ng/mL aTc. This induction system was created by systematically altering the −10 element sequence and the bases between the −10 element and the transcription start site. A guanine in the 2^nd position downstream of the −10 element was shown to be important for high promoter strength, that might be caused by altered lifetime of the promoter-RNAP closed complex.

These systems had some of the largest fold induction ratios (Figure 2), but their use for long term induction is limited due to aTc’s sensitivity to light. The dynamic range of induction in PCC 7002 dropped significantly after both 24 and 48 hours exposure to light⁷⁵. Induction persisted for at least 72 hours in PCC 7120, possibly because the cyanobacterial cells were absorbing light and protecting from aTc degradation⁷⁶. aTc has also been shown to be sensitive to temperature, with a half-life of only 15 hours at typical cyanobacterial cultivation temperature (37°C) compared to 30 hours 30°C⁷⁸. The lability of the inducer in the culture conditions is often forgotten and could be responsible for some of the different dynamics seen above.

Induction via translation

Induction can also occur at the level of translation, where the translation of a transcript can respond through a molecule that alters the occlusion or accessibility of the ribosome binding site. Theophylline-responsive riboswitches have been tested and used in a number of diverse cyanobacteria. A set of six riboswitches originally designed for and shown to be effective in a diverse group of Gram-negative and Gram-positive bacteria⁷⁹ have been tested in cyanobacteria. In PCC 7942, riboswitch E* (a slight change in the RBS from E) showed 190-fold induction of a luciferase reporter while maintaining minimally leaky expression⁸⁰. Expression had a wide range of induction with concentrations between 0 and 2 mM theophylline. This riboswitch was successfully used to regulate circadian clock kaiC expression.

The same set of theophylline riboswitches was tested in a group of 4 diverse cyanobacteria strains (PCC 7942, BL0902, PCC 7120, and WHSyn). As was seen in the original diverse group of bacteria, various fold inductions, leakiness, and gene expression ranges were seen between the six riboswitches. Fold induction ratios in cyanobacteria seemed significantly lower than those reported for the diverse set of other bacteria with cyanobacteria fold induction ranges reaching maximally 30-fold⁶¹ whereas several of the previously tested bacterial strains reached 100 or 2000-fold⁷⁹.

It is thought that cyanobacterial small RNA (sRNA) play a large role in controlling gene expression. sRNAs are very prevalent with 371 reported in PCC 6803⁸¹, but only a few of these sRNA seem to be widely conserved (NsiR1, PsrR1, Yfr1, HliR1, PsiR1, and PsiR2)⁸¹. The lack of development of sRNA tools for cyanobacteria is surprising, with only two reports of synthetic sRNA’s used to mediate gene expression. A trans-acting sRNA system based on the E. coli IS10 RNA-IN/OUT system was adapted for use in PCC 7002 and achieved 70% attenuation⁷⁵, slightly less than the 90% attenuation in E. coli⁸². sRNA control has also been implemented in cyanobacteria through the use of riboregulators. Riboregulators consist of a cis-repressed mRNA (crRNA) that forms a stem loop that occludes the RBS. A trans-activating RNA (taRNA) is complementary to part of the stem loop and hybridizes with the crRNA and enhances ribosome binding. By driving the crRNA from the trc promoter and the taRNA from the Ni²⁺ inducible promoter PnrsB, they saw 13-fold induction of the GFP reporter upon the addition of Ni²⁺ in PCC 6803⁸³. Induction systems based on the level of translation have proved to be a useful tool in cyanobacteria and provide an avenue for gene expression control independent of the complex and unstudied aspects of cyanobacterial transcription machinery.

Ribosome bindings sites

The ribosome binding site (RBS) is a major factor controlling the amount of translation of a given gene. The complementarity of the RBS to the 16S rRNA in the ribosome is a major determinate of protein expression as well as the accessibility of this site given the secondary structure of the surrounding RNA sequence. Most cyanobacteria mentioned here (PCC 7002, PCC 6803, PCC 7942, PCC 7120, and PCC 6301) have the portion of the 16S rRNA that pairs with messenger RNA with a sequence of ACCTCCTTT which would suggest optimal pairing with mRNA with a central region of the RBS being AGGAGG (the consensus prokaryotic sequence).

Thermodynamic models have been developed to predict translation efficiency and have been quite successful in both E. coli and other organisms⁸⁴. The most widely used software is the RBS Calculator⁸⁵, but other translational rate calculators include the RBS Designer⁸⁶ and UTR Designer⁸⁷. These calculators use statistical thermodynamic models that calculate free energies for interactions between the mRNA and rRNA and utilizes RNA folding software to predict mRNA secondary structure. These calculators are capable of both designing sequences for a specified translation rate (forward engineering) and predicting the location and strength of a given RBS (reverse engineering). Each of these calculators tested predicted translation initiation rates and observed fluorescence or β-galactosidase activity of reporters in E. coli and saw good correlations using the forward engineering mode (R² = 0.84 for the RBS calculator⁸⁵, R² = 0.87 for UTR Designer⁸⁸, and R² = 0.87 for the RBS Designer⁸⁹). These calculators have been very useful and enabling, but there are limitations including the effects of antisense RNA, RNA binding proteins, translational coupling, or differences in translational elongation⁹⁰.

There are many examples showing how manipulation of the RBS has enabled desirable changes in cyanobacteria protein expression. Using an optimized RBS in PCC 7120, L0908, PCC 6803, and PCC 7942 increased expressed of both CFP and YFP by several fold⁹¹. Although there are limited examples, RBS predictions do not seem to be as reliable in cyanobacteria as they are in cyanobacteria. Fluorescence from a 11-member library of RBS’ in front of a YFP reporter was measured and compared with the predicted translation initiation rate using the RBS calculator, UTR Designer, and RBS Designer²⁶. Once an outlier that had a large increase in transcript level compared to all other members was removed, R² values were respectively 0.74, 0.78, and 0.04. Altogether, RBS design software is not perfect, but is an extremely useful tool to both study native regulation and achieve desirable expression levels of heterologous proteins.

CRISPR interference

CRISPR interference (CRISPRi) has proved to be a powerful method to repress native and heterologous gene repression in bacteria. CRISPRi utilizes the native RNA based defense system that bacteria use to target invading nucleic acids⁹². The system most widely used is derived from the Streptococcus pyogenes type II system that utilizes a single nuclease protein (Cas9) and a small guide RNA (sgRNA) complementary to the target sequence. CRISPRi utilizes a deactived Cas9 (dCas9) that has point mutations in its active sites so it can no longer make double-stranded breaks in DNA. Genes targeted with CRISPR interference are thought to be repressed because of tight binding of the dCas9/sgRNA complex to the DNA that causes a steric clash, dislodging RNAP⁹³. This heterologous system has been implemented in both PCC 7002, PCC 6803, and PCC 7942 and has been used to re-direct flux toward desired products. Several of these cyanobacteria have a native CRISPR system or multiple systems, but there has been little exploration in their function or mention of cross-talk with CRISPRi.

In PCC 6803 a number of constitutive and inducible promoters were tested to drive expression of dCas9 and sgRNA and their efficiencies were measured by the output of reporter green fluorescent protein (GFP)⁹⁴. They observed significant repression of GFP whenever both dCas9 and the sgRNA was present, and the best combination was using the aTc-inducible P_L22. They observed 96% of the maximal GFP fluorescence without aTc, and only 6% when aTc was added, but there was still significant repression with intermediate levels of aTc. Other promoters caused maximal repression (>99%) even without the presence of the inducer, highlighting that low levels of expression are necessary to achieve inducible repression. They targeted native phaE for repression and saw no PHB accumulation under nitrogen-starvation conditions which normally accumulates under these conditions and functions as a carbon and redundant storage polymer. They also targeted glgC, an enzyme involved in glycogen synthesis and showed that under nitrogen starvation the glgC repression strain only accumulated 25% of the glycogen compared to WT. They showed efficient repression of four targets at the same time. Repression of one target in PCC 6803 (slr0091) was only 2-fold whereas other targets were about 10-fold. This weak repression could have been because this gene is very weakly transcribed than the others, or because it is part of an operon.

An inducible and titratable system for CRISPRi was made in PCC 7002⁷⁷. Repression was seen immediately whenever dCas9 and the sgRNA was present, and the promoters and ribosome binding sites were changed to attempt to relieve this repression in the absence of the inducer. The system with the best titratable range had both dCas9 and the sgRNA under the control of the aTc-inducible promoter, P_EZtet⁷⁵. While this system only showed ~70% of the maximal protein expression without the inducer, it was titratable with intermediate levels of aTc and only ~10% of the maximal protein expression with maximum aTc concentrations. Significant repression of targets was seen even when the promoter of dCas9 was removed and when the sgRNA was truncated down to only 15 bases. Native phycobilisome (cpcBACDEF) and carboxysome genes (ccmK₁K₂LMN) were also targeted and showed characteristic decreases in absorbance at 635nm and requirement for high CO₂ concentrations for growth respectively. This repression scheme was used to increase flux towards lactate by targeting glutamine synthetase (glnA) and over 2-fold more lactate was made in this strain than the baseline engineered strain. Interestingly, increased lactate production was also seen in a strain that contained the sgRNA targeting glnA but lacked dCas9.

A CRISPRi system was built in PCC 7942 to target native genes and increase succinate production. Even when dCas9 was under control of inducible P_smt leaky expression was sufficient to repress targets in the absence of the inducer. Repression of glgC, sdhA, and sdhB (separately) using CRISPRi increased succinate production by ~12.5-fold²⁵.

These CRISPRi systems have enabled both increased titers of desired products as well as provided platforms to manipulate and study native gene regulation. Combined, these three systems show that CRISPRi is extremely effective at repressing gene expression, stable (effective for maximum time tested in PCC 7942: 21 days²⁵), and reversible^77,94. These studies also show that sgRNA’s can efficiently target both the template and non-template strand, but most effective repression is seen when the target is near the transcription start site or near the beginning of the coding sequence. One remaining challenge is the fact that many bacterial and cyanobacteria genes are located within operons, limiting the application of CRISPRi. Nonetheless, CRISPRi is a very promising tool, enabling regulation of gene expression in trans.

RNA stability and degradation

Of the many steps in the central dogma of biology, RNA turnover is the least considered when designing gene expression modules. In cyanobacteria, RNA turnover is rapid and may play a substantial role in determining gene expression levels. Global half-lives of transcripts have been determined in Prochlorococcus MED4 using microarrays, and the median half-life was 2.4 minutes⁹⁵. The observation that RNA is very labile and that RNA is turned over rapidly in bacteria has been long observed, but the half life reported for MED4 is on the shorter end compared with other recent global RNA studies (B. subtilis ~ 5 minutes⁹⁶, Bacillus cereus 2.4 minutes⁹⁷, E. coli 6.8 minutes⁹⁸; 4.7 minutes⁹⁹; 2.5 minutes¹⁰⁰, Chlamydia trachomatis trachoma 15–17 minutes¹⁰¹, Mycobacterium tuberculosis 9.5 minutes¹⁰²). The reason for this especially rapid turnover remains unexplained although it has been suggested that RNA turnover could be linked with cell doubling time. This seems unlikely when the doubling times and average/median global transcript turnover is compared⁹⁵.

This rapid turnover of RNA in cyanobacteria could be a result of its somewhat unique composition of ribonucleases (RNases). Cyanobacteria have homologs of both RNase E (endoribonuclease that often initiates turnover in E. coli) and RNase J (5’->3’ exoribonuclease responsible for initiation of mRNA degradation in B. subtilis). Both RNase E and RNase J appear to be essential in PCC 7002 because attempts to delete either were unsuccessful¹⁰³. This combination of RNase E and J exists in several other bacteria including Sinorhizobium meliloti, Mycobacterium smegmatis, Rhodobacter sphaeroides, and Streptomyces coelicolor. Studies in these organisms have shown that both RNase E and RNase J play important roles in rRNA maturation^104,105 as well as in the production of antibiotics¹⁰⁶ and SAM metabolism¹⁰⁷. A global model of RNA degradation in organisms containing both RNase E and J has yet to be proposed and universal tools for modulating RNA turnover are not available.

Another difference between the cyanobacteria RNA degradation machinery and other more well-characterized systems in the apparent lack of a 5’ pyrophosphohydrolase, RppH¹⁰³. RppH is part of the Nudix (nucleoside diphosphates linked to x) protein family that removes pyrophosphate from the 5’ end of transcripts leaving a monophosphate¹⁰⁸. This processing event affects transcript degradation because both RNase E and RNase J can sense the 5’ of a transcript and cleave more efficiently in the presence of the 5’ monophosphate^109–111. While no proteins with 5’ pyrophosphohydrolase activity have been identified in cyanobacteria, RppH could still be present and be an important element controlling transcript degradation. All RppH proteins identified so far have been part of the Nudix family, but it is difficult to determine the substrates of these protein family members based on amino acid sequence. RppH homologs have been identified previously by purifying these proteins and testing their activities in vitro¹¹². Cyanobacteria possess many proteins annotated as members of the Nudix family and it is yet to be seen if one may have the activity of RppH. Recent reports show that RppH may not be the only enzyme involved in the processing of the 5’ end, and that a yet-identified protein may be responsible of initial cleavage from a 5’ triphosphate to a 5’ diphosphate and serve as the precursor for RppH¹¹³. Given that ribonucleases play a substantial role in regulating gene expression, it is important to identify the full complement of RNA processing enzymes in cyanobacteria.

Transcriptional Termination

Transcription termination has not been widely studied in cyanobacteria. Observations that the cyanobacterial genome is nearly fully transcribed, points to potential differences in transcriptional termination from other model hosts. In bacteria, termination is either rho-dependent, where the protein Rho recognizes a target sequence and dissociates RNAP, or rho-independent where an RNA structure in the transcript destabilizes RNA-RNAP interactions. In rho-dependent termination Rho recognizes a Rho-utilization site (RUT) in recently transcribed RNA, wraps the RNA inside the Rho hexamer, moves along the transcript, and is thought to pull the RNA out of the RNAP active site. The site of termination is not well defined and can be as far as 120 bases from the RUT site¹¹⁴. It is unlikely that cyanobacteria use rho-independent termination because PCC 7942, PCC 6803, and PCC 7002 do not have an obvious homolog of Rho²⁸. Many assume that cyanobacteria rely on rho-independent termination also known as intrinsic termination. Transcripts that are terminated via intrinsic termination have a 3’ secondary structure that is composed of a GC-rich palindromic sequence followed by stretch of T’s¹¹⁵. The stretch of T’s is thought to induced transcriptional pausing, allowing time for the hairpin to form which in turn dislodges key interactions between three nucleic acid binding sites in RNAP (the RNA:DNA T-A hybrid, RNA-binding site of just added ribonucleotides, and upstream RNA-binding site of RNA exiting RNAP)¹¹⁴. Intrinsic termination structures can be found for almost 50% of protein-coding genes in E. coli¹¹⁶.

It was originally presumed that cyanobacteria must use rho-independent termination because of the lack of a Rho homolog, but genome wide predictions of hairpin formation near stop codons showed that termination in cyanobacteria may not involve hairpin formation. In E. coli there is a significant sharp drop in free energy approximately 30 bases downstream of stop codons indicating the presence of a hairpin. Genome-wide free energy calculations were made for many organisms including PCC 6803, and no sharp dip in free energy was seen¹¹⁷. The lack of dips in free energy values was also seen in several other genomes including Mycoplasma genitalium, Mycoplasma pneumoniae, Helicobacteri pylori, Borrelia burgdorferi, Methanococcus jannaschii, Archaeoglobus fulgidus, Methanobacterium thermoautotrophicum, Aquifex aeolicus, Pyrococcus horkoshii, Mycobacterium tuberculosis, and Treponema pallidum¹¹⁷. Dips in the free energy prediction near the 3’ end of transcripts have, however, were seen in PCC 7942²⁸.

Transcriptome studies suggest that cyanobacteria transcribe a large portion of non-coding regions as well as sequences within coding sequences, but on the opposite strand⁴². Cyanobacteria have intriguing parallels in terms of their transcriptional landscapes when compared with plants chloroplasts. It is a well-established in chloroplasts that transcription is very inefficient, and it has been proposed that transcription initiation and elongation are very haphazard, creating a wide array of varying length transcripts⁴². Many transcripts in chloroplasts have been shown to be variable in length with a mixture of 3’ extensions¹¹⁸.Transcripts are thought to be processed by proteins to have the proper 5’ and 3’ ends. This phenomenon has been analyzed globally using RNA-sequencing and also been studied on an individual transcript basis. Similar results have been shown in plant mitochondria where variable 3’ extensions are trimmed back to stem loop structures by exonucleases¹¹⁹.

It’s been observed that many chloroplast genes have inverted repeats at the 3’ end of the transcript (including psbA, rbcL, petD, and atpE)¹¹⁸. These 3’ inverted repeats have been shown not to cause termination like 3’ secondary structure does on many bacterial transcripts, but instead serve as processing sites. Without these 3’ inverted repeats these transcripts are rapidly degraded¹¹⁸.

Transcriptional termination can play a role in controlling gene expression of genes that are located in operons. Terminators are not necessarily efficient and can allow various amounts of read-through. In PCC 7002 the phycocyanin operon (cpcBCADEF) encodes six genes that encode structural subunits of the phycobilisome, linker proteins and assembly enzymes, and has been shown to make eight to ten different length transcripts¹²⁰ (Figure 3). There are predicted stable stem loop structures at several locations within this transcript that could serve as terminators or processing sites. Support for their role as inefficient terminators includes that only 1–2% of the transcripts include cpcE and cpcF located at the 3’ end of the operon¹²⁰. The structural components of the phycobilisome are encoded in the first two genes of the operon, followed by two linker proteins, and finally enzymatic lyases for attachment and their decrease in expression levels seems to match the stoichiometry needed for phycobilisome function²⁸.

Figure 3.

Organization and transcripts observed from cpc operons in PCC 7002 and PCC 7120. Genes are color-coded for their functions (red: structural components, blue: linker proteins, and green: attachment enzymes). Promoters are marked as well as regions that predicted to form significant secondary structures. Observed transcripts are marked below with the thickness indicating relative abundance.

Similar regulation of the phycobilisome operon is seen in PCC 7120, except with slight differences in operon organization and the number of different transcripts made. The PCC 7120 operon contains ten open reading frames (cpcBACDEFG1G2G3G4), and several different transcripts are seen^121,122. The major transcript includes cpcBA, but full-length transcripts as well as partial transcripts have also been observed, including some that suggest an internal cpcC promoter. Although the presence of processing events could not be ruled out, it was suggested that these transcripts arose through partial termination¹²¹. Most of the locations where partial transcriptional termination is occurring are also predicted to form strong secondary structures. This partial read-through seems to be controlling expression levels of important parts of this light harvesting complex.

Partial termination may be playing a large role in regulating gene expression from the cpc operon, an operon extensively studied because of its important role in light harvesting and photosynthesis. The need to for analysis of transcriptional termination in cyanobacteria has been mentioned before¹²³, and perhaps difference in transcriptional termination could play major roles in controlling gene expression.

Advances in Genome Editing

One significant advantage of working in cyanobacteria is that many of the commonly used strains are naturally competent and readily take up foreign DNA (PCC6803¹²⁴, PCC 7002¹²⁵, PCC 7942¹²⁶, and others). DNA-uptake in cyanobacteria is thought to involve type IV pilus apparatus that may be responsible for DNA translocation as a single strand¹²⁷. Many established protocols make use of this natural competency and involve incubating cyanobacteria cells with a plasmid or linear DNA piece containing regions homologous to the chromosome prior to plating on media with a selective pressure corresponding to a selection marker in the foreign DNA.

Traditionally, genomic integrations have bene made by providing a plasmid or PCR product that contains the desired sequence surrounded by homology arms. Higher efficiency has been seen with longer homology arms, but they can be as short as 250 bp¹⁴. Several studies have examined the length relationship between homology arm length and integration efficiency, and generally 500–1000 bases of homology are used.

Integrating constructs onto the genome may be more stable than maintaining them on a plasmid, but deciding the location of integration can be challenging. Ideally genes could be inserted in neutral sites that have no effect on physiology when grown in any conceivable environmental condition. Historically in PCC 7002 desB was used as an integration site¹⁴, but it’s expression at low temperatures and role in lipid desaturation¹²⁸ make it an unattractive candidate. Many have used glpK, encoding glycerol kinase, as a neutral insertion site because of its original annotation as a pseudogene. Re-sequencing of PCC 7002 showed that glpK is intact and an error was made in the original sequencing¹²⁹. Genomic integrations have also been introduced between A0935 and A0936 (termed NS1)¹⁴. Given the abundance of transcription from the cyanobacteria genome, it may not be possible to identify a truly neutral site. That said, systematic studies of gene expression around the genome are not available for all cyanobacteria and many other hosts, a knowledge gap that could provide immediate help in designing engineered strains.

Several methods have been developed to obtain markerless genome edits, but many of these systems require multiple transformation and homologous recombination steps. Levansucrase, encoded by sacB, confers sensitivity to sucrose, is one of the most popular counterselectable markers and has been adapted for uses in many organisms¹³⁰, including cyanobacteria. sacB was used as early as 1990 in PCC 7120 by integrating a selectable plasmid containing sacB onto the chromosome with a single crossover event and then selecting for a replacement event with the desired sequence¹³¹. Unfortunately this technique produced a significant amount of spontaneous mutants as well as a mixture of five different insertion sequences at the target locus. Recently, a faster sacB system has been developed in PCC 6803 that utilizes a single vector containing kanamycin resistance gene and sacB nested within a split gene of interest with partial homology regions. This method requires two homologous recombination events: (1) insertion of the entire cassette and (2) recombination fusing the gene of interest and excising the kanamycin resistance-sacB region, but the time needed to produce mutant strains is greatly reduced without the need for a second conjugation or transformation step¹³². Another one of the most popular markerless counterselectable markers, the rsp12 (strA) system that relies on the resistance and sensitivity of different alleles to streptomycin, has been used in PCC 7942¹³³. This system unfortunately requires working in a mutant rsp12 background. A novel counter selection system utilizing the sensitivity to acrylic acid in the presence of native acsA gene has been developed in PCC 7002¹³⁴.

The adaptation of the CRISPR system as a tool for genome editing has been an enabling technology in bacteria and higher organisms. CRISPR Cas9 systems are useful in cyanobacteria to allow genomic integrations or changes without the use of selectable markers and can be used to alleviate problems with segregation. One of the first examples of a CRISPR Cas9 genome editing system was in UTEX 2973¹³⁵. The system was based on one developed for use in Streptomyces lividans that uses a plasmid that is not functional at 34°C or higher. This plasmid was needed because Cas9 was shown to be toxic in UTEX 2973 when cells were transformed with a replicating RSF1010-based plasmid. This toxicity was still seen when the promoter and ribosome binding site of Cas9 was removed. Growing cells at a non-permissive temperature (38°C) while forcing them to maintain the plasmid due to the presence of the antibiotic, appeared to allow enough transient expression for genome editing and genome edits appeared to be segregated in the first patch. The system is all contained on a single plasmid encoding Cas9, the gRNA, and repair template and can be introduced into cyanobacteria with triparental mating with a conjugal and helper plasmid. The editing plasmid could be cured from the edited strain by re-streaking without the antibiotic, but unfortunately this took 10 rounds of patching.

Another CRISPR Cas9 system was used in cyanobacteria was in PCC 7942. Cas9 and repair templates were provided on two separate plasmids and they saw successful integration at higher efficiencies than traditional transformations and found that they could use shorter homology arms (400bp vs 700 bp)¹³⁶. They found that Cas9 cleavage seemed to help with segregation, but strains were still not segregated until 3 rounds of passaging on the antibiotic. Cas9 was expressed from a plasmid that appeared to only be transiently present because it was not detected after 9 days. Notably, Cas9, and tracrRNA, crRNA were all expressed from native Streptococcus pyogenes promoters. This CRISPR Cas9 system was effective, but genomic edits still contained an antibiotic resistance marker, limiting its utility to make some precise edits.

Cpf1, a type V CRISPR nuclease, has also been used for genome editing in cyanobacteria. The Cpf1 system differs from type II Cas9 CRISPR system in that Cpf1 processes pre-crRNA¹³⁷ and introduces a staggered cut 17 bases from the PAM, and the PAM (TTN) is located 5’ of the crRNA¹³⁸. Cpf1 appeared less toxic than Cas9 because more colonies were recovered even when compared to Cas9 that lacked a promoter. Successful editing events were shown in UTEX 2973, PCC 6803, and PCC 7120 using a RSF1010 based plasmid containing Cpf1, gRNA, and repair template¹³⁹. Only ~20% of the initial colonies were edited and segregated, but subsequent patching on the antibiotic increased the recovery of segregated mutants. They showed markerless point mutations, knock-outs, and insertions. Continued advancements to improve the efficiency and minimize the time needed to make genome edits will help accelerate the development of gene regulatory tools.

Use of plasmids

In other model organisms, plasmids are used to express heterologous protein. Their use is much less widespread in cyanobacteria both because of the lack of characterized plasmids that can replicate in cyanobacteria as well as the difficulties introducing plasmids into the cell. The natural competence of cyanobacteria is advantageous for introducing foreign DNA with selectable markers, but unfortunately is not useful for introducing plasmids. The mechanism of DNA uptake and natural competence in cyanobacteria is relatively unknown, but likely contains core DNA uptake genes as well as hundreds of regulatory genes as in other model organisms used to study natural competence¹⁴⁰. DNA uptake is thought to involve an extendable (and potentially retractable) type IV pilus that translocates single-stranded DNA that is then decorated with protective proteins and then acted on by recombinases. This process imports only single-stranded DNA, preventing its utility for transformation of plasmids.

Both integrative and replicative plasmids have been used in cyanobacteria, but their use lags significantly behind their use and utility in other organisms. There has been a number of shuttle plasmids built that contain both E. coli and cyanobacteria-based origins⁸. E. coli origins are more standard but only a limited number of cyanobacterial origins have been used: pDU1 in PCC7120, pDC1in N. punctiforme, pUH24 and pANS in PCC 7942, pAQ1 in PCC 7002, pBA1 in PCC 6301, and pFDA in Fremyella diplosiphon UTEX481^8,91. In most of these cases native origins have been used and re-introduced on a new vector in the same host. However, pDU1 was originally isolated from Nostoc PCC 7524 but shuttle vectors containing pDU1 have shown to replicate in many other strains of cyanobacteria including PCC 7120¹⁴¹, Anabaena M-131¹⁴¹, Fischerella muscicola PCC 7414¹⁴², Chlorogloeopsis fritschii PCC 6912¹⁴², Chroococcidiopsis spp. Str. 029, 057, and 123¹⁴³, and Oscillatoria MKU 277¹⁴⁴. There is a significant lack of understanding of the maintenance, copy number control, and replication mechanisms of cyanobacterial plasmids and if these aspects are similar or different from plasmids native to other bacterial species.

The most widely used plasmid in cyanobacteria is broad host range plasmid RSF1010⁸. RSF1010 is a broad host range plasmid in the IncQ compatibility group that can replicate in over 30 Gram-negative species¹⁴⁵. This plasmid on its own is not transferable via conjugation, but it contains the necessary elements (oriT and mob mobilization genes) to be transferred in combination with a ‘helper plasmid.’ Conjugation protocols are standard in many strains of cyanobacteria and typically consist of biparental or triparental matings^22,146,147. Plasmid based expression has significant advantages, but there are not many widely used or characterized plasmids that replicate in many cyanobacterial hosts.

Standardization of genetic tools

One of the reasons gene expression regulatory tools have only been built in tested in a few strains is because of the lack of standardization of cyanobacteria genetic tools. There have been significant efforts to standardize genetic tools in synthetic biology. The BioBrick ‘Registry of Standard Biological Parts’ contains thousands of promoters, ribosome bindings sites, terminators, coding sequences, and origins that are surrounded by standardized restriction enzyme sites for simple assembly of parts. There have been a few cyanobacterial plasmids built using the BioBrick system but they have not gained much popularity. The plasmid, pPMQAK1, was built by adding the BioBrick multiple cloning site and antibiotic resistant cassettes to the backbone of the broad-host range plasmid RSF1010. The plasmid was shown to replicate in and E. coli, PCC 6803, PCC 7120, and N. punctiforme²².

A broad-host-range vector system based off of RSF1010 was built with and used to create cyanobacterial shuttle vectors⁹¹. This system utilizes GC-adaptors flanked by restriction enzyme sites, so that elements can be digested out of donor plasmids, purified, and assembled using Gibson assembly. Four different sites with overlapping regions were used that could be used for an E. coli origin, cyanobacterial origin, antibiotic resistance marker, and expression cassette for the gene of interest. These elements can also be replaced with homology regions to direct homologous recombination onto the chromosome. This design scheme and sequence elements have been integrated into a web service, CYANO-VECTOR (http://golden.ucsd.edu/CyanoVECTOR/) that can be used to design and export GenBank files.

The lack of standardization of parts for gene expression in cyanobacteria likely stems from the widespread practice of integrating cassettes onto the chromosome of cyanobacteria. With the ease of Gibson assembly¹⁴⁸, most labs design and assemble custom plasmids. Genomic integration is a popular method of expressing genes of interest because transformation of plasmids to date has been achieved through conjugation with helper plasmids which takes approximately the same amount of time and reagents as using natural competence and homologous recombination. An effort to make gene regulatory tools accessible across strains and labs would greatly enhance current tool building and engineering efforts.

Unexplored Physiology

The development of predictable gene regulatory tools in cyanobacteria may be limited by unexplored physiology. There are several cyanobacteria-specific elements that remain relatively unexplained that could be influencing gene expression and could explain why directly moving elements from other organisms has been somewhat unsuccessful. We highlight three aspects of cellular physiology that may be playing significant roles in influencing gene expression.

Highly iterative palindrome 1 (HIP1)

The extreme prevalence of one sequence in cyanobacterial genomes remains unexplained. The palindrome 5’-GCGATCGC-3’or the extended version, 5’-GGCGATCGCC-3’, is extremely overrepresented in many cyanobacterial genomes and has been named highly iterative palindrome 1 (HIP1). This sequence was first described in PCC 6301 in 1993 as the two flanking sequences surrounding a deletion event that conferred a tolerance to cadmium¹⁴⁹. Since then the overrepresentation of this sequence has been discovered in many cyanobacterial strains and has been shown to have been gained once during evolution and then lost in marine cyanobacteria¹⁵⁰.

HIP1 has been shown to propagate through nucleotide substitutions. Comparisons of several conserved regions between E. coli and PCC 7942 show that E. coli contains partial HIP1 sequences (on average 5.2 of the 8 nt) that require minimal changes to the HIP1 sequence that result in synonymous or conservative amino acid changes¹⁵¹. Deletions flanked by HIP1 sequences have been observed many times. Direct tests comparing deletion frequencies in E. coli and PCC 7942 show that excision frequencies surrounded by HIP1 sequences are 2-fold higher than those surrounded by control palindromic sequences. Additionally, the frequency of excision was similar in both E. coli and PCC 7942, suggesting that this phenomenon is not due to a cyanobacterial-specific recombinase¹⁵¹. It has been suggested that these deletion events could be due to HIP1 sequences forming hairpins or other structures that result in replication slippage and illegitimate replication. A similar hypothesis has been suggested for repeated sequences found in radiation resistant Deinococcus radiodurans, that is thought to survive double-stranded breaks through extensive recombination of its many chromosomes¹⁵².

It’s possible that HIP1 could be involved in gene expression control, but potential mechanisms and regulatory partners remain elusive. Comparative genomics of many cyanobacterial genomes show that is not preferentially located in coding or noncoding regions and has no preference for reading frame. Additionally, attempts to see trends between HIP1 prevalence and location within a transcript, GC content, codon usage, and transcript abundance have been unsuccessful¹⁵³. So far, no HIP1-specific proteins or complexes have been identified via electrophoretic mobility-shift assays¹⁵¹. We urge the cyanobacterial community to be cautious of DNA motif finders such as MEME¹⁵⁴ because they will most likely find HIP1 enrichment no matter what native cyanobacterial sequences are entered. The overabundance of the HIP1 sequence remains unexplained and could be influencing gene expression or other processes that have large implications for growth and study of cyanobacteria.

Polyploidy

It has long been reported that many strains of cyanobacteria contain multiple copies of its chromosome, but it’s implications on gene expression control is often ignored. PCC 7942 and WH7803 have been shown to be oligoploid containing 3–4 copies of its chromosome per cell irrespective of growth phase¹⁵⁵ or light-dark/continuous light cycles¹⁵⁶. Other strains such as PCC 6803 have been shown to be highly polyploid with as many as 218 copies of its chromosome per cell in exponential phase. In PCC 6803 the number of copies of its chromosome changes and only ~ 60 copies per cell are present in linear and stationary phase. Previous studies report only 12 copies of the chromosome in PCC 6803¹⁵⁷. PCC 6803 has the highest ploidy level of any bacterium besides the very large Epulopiscum sp. that contains tens of thousands copies of its chromosome¹⁵⁵. Discrepancies in chromosomal as well as plasmid copy number may come from lysis efficiency issues as well as the method of detection (PCR vs total DNA quantification). There has been also been mention that these chromosome copy number discrepancies could be due to cultivation of these strains in many labs over the past 20 years that could be resulting in mutations.

The ploidy of cyanobacteria could confer resistance to UV radiation; both PCC 6803 and PCC 7942 were more resistant to UV radiation than E. coli. Additionally, PCC 6803 (highly polyploid) is much more resistant to X-ray irradiation than PCC 7942 (oligoploid)¹⁵⁸.The ploidy of cyanobacteria leaves many unanswered questions regarding gene expression. It is unknown when genes are expressed from all chromosomes or selectively from different chromosome copies. The polyploid nature of cyanobacteria has unknown implications on gene expression.

The presence of multiple copies of the genome is challenging for the introduction of foreign genes and making targeted mutations. It has only been recently appreciated that it is imperative to show segregation of strains, to ensure predictable and reliable expression patterns and production results. In some early papers using cyanobacteria to create products, there is no mention of segregation while in others the authors acknowledge that the strains were indeed not segregated, making it almost impossible to re-create these strains and verify these results. PCR, using primers flanking the inserted gene of interest, is often used to demonstrate segregation of strains, but this is not always sufficient to demonstrate segregation. There have been cases where the PCR product made with flanking primers indicates the proper mutated size, but it is still possible to amplify the native WT sequence using an internal primer¹⁰³. A combination of both internal and external primers is necessary to demonstrate segregation. The presence of multiple copies of the genome in cyanobacteria is challenging, and the cyanobacteria community must remain vigilant to ensure thorough and meticulous strain construction.

Role of small proteins

Small proteins often escape annotation by automated pipeline systems despite their important functions, and many of these small proteins have important roles in photosynthesis. These small proteins are sometime known as sproteins, SPs, or μ-proteins with arbitrary cutoffs ranging from < 100, <80, and < 50 amino acids. These small proteins have been implicated in many processes such as being secreted as signals, serve as toxins, membrane components, chaperones of metals or nucleic acids, regulators, or stabilizing factors of large protein complexes¹⁵⁹.

Issues with small protein annotation likely arise because short DNA sequences are likely to contain both start and stop codons. Random DNA sequences have been modelled in yeast, and the frequencies that different length sequences would contain ORFs showed that a large number of short ORFs could be occurring by chance¹⁶⁰. Due to this, most gene annotation algorithms have set a cutoff for ORFs of greater than 100 amino acids to lower then number of false positive annotations¹⁶¹.

Only 80.8% of the entire PCC 6803 predicted proteome has been detected even when the data was combined from over 50 independent proteomic studies¹⁶². Hydrophobicity as well as low or condition-specific expression (especially proteins encoded on plasmids) may cause the inability to detect expression¹⁶², but a major class of proteins that is not-represented is small proteins. Proteomic approaches fail to identify many small proteins for several reasons. Small proteins contain few or no trypsin cleavage sites, necessary for many mass spectrometry analyses. Additionally, many small proteins be lost in the filtering process where two positively charged peptides are necessary to match to the database. Modifications such as phosphorylation, acetylation, glycosylation, and oxidation may prevent detection¹⁶² (all of which have been observed in PCC 6803).

Many short proteins have been shown to play important roles in photosynthesis including 19 proteins < 50 amino acids that interact with photosystem II (psbM, psbT, psbI, psbL, psb30, psbJ, psbM, psbX, psbY, psbN, psbF, and psbK¹⁶³). Other small proteins are involved in photosystem I (psaM, psaJ, and psaI) as well as electron transport (petL, petN, petM, and petG). The smallest protein currently annotation is petN (29 amino acids)¹⁵⁹. The presence and expression of many of these small genes was completely unknown and only found through purification of photosystem II complexes. Purification of PSII from PCC 6803 revealed a total of 31 distinct proteins, 12 of which were less than 8kDa, and 5 of which were novel¹⁶³. Six of these small proteins remained unidentified, highlighting the challenges of both predicting and studying these small proteins. The important role these small proteins play is highlighted by the fact that many are conserved in cyanobacteria, algae and plants.

There have been attempts to predict the presence and expression of small proteins in cyanobacteria by examining transcriptomic data and comparing the conservation of possible coding sequences across different species and organisms. 146 small proteins were predicted to be shared in both PCC 6803 and PCC 6714 and a smaller subset of 42 proteins shared between both cyanobacteria strains and Arabidopsis thaliana, and 29 proteins shared between the cyanobacteria strains and E. coli¹⁶⁴. Expression of a test set of 5 of these genes were confirmed with immunoblotting. Two of these small proteins were originally thought to be non-coding RNAs¹⁶⁴. Challenges to discovering the expression of these small proteins include the time it takes to make markerless mutations such as a C-terminal tag as well as their regulated expression under specific environmental conditions. There have been a number of short proteins that have been found to play import roles in photosynthesis, and we postulate that other unknown and uncharacterized sequences may play large roles in cellular physiology.

Conclusions

There have been significant advances in gene regulatory tools in cyanobacteria, but lack of knowledge of basic cellular processes as well as cyanobacteria-specific physiology may still play a large role in controlling heterologous gene expression. Endogenous cyanobacteria promoters have been used to drive gene expression as well as heterologous promoters. Adapting native sequences for expression of heterologous transcripts can cause problems with genetic instability. Many groups have tried using promoters widely-used in E. coli, but the majority of them are leaky and have low fold induction values, especially IPTG-based systems. Induction systems have been used both on the level of transcription and translation with varying results. It is difficult to compare induction data and measures of tightness of repression in different publications, and currently there is no method to compare overall expression levels between publications and strains. The CRISPR system has expanded the collection of cyanobacteria gene regulatory tools both in use as a repression system (CRISPRi) and for genome editing.

Gene expression tools often do not operate as expected in cyanobacteria, and we explore several relatively under-studied areas relating to cyanobacteria physiology that could be causing these inconsistencies. It’s been observed that most of the cyanobacterial genome is transcribed, but implications for gene expression and gene expression tools remains unknown. Additionally, there has been extremely little study of cyanobacterial transcription termination. There seem to be intriguing similarities in the transcriptional landscape between cyanobacteria and plant chloroplasts, and continued study in this area might be quite fruitful. We conclude with three topics that are cyanobacteria-specific that may be influencing gene expression: the unknown role of the highly prevalent repeat sequence (HIP1), the maintenance of many copies of chromosome, and the role and prevalence of small proteins.

Future Trends

The ability to do metabolic and synthetic biology in cyanobacteria is heavily reliant on the quality and availability of gene expression regulatory tools. We suspect that the cyanobacterial community will continue to develop novel and more dependable gene expression regulatory systems that will enable rapid acceleration of the use of cyanobacteria for biotechnology purposes.

Due to the transcription landscape in cyanobacteria, it is likely that tools to control gene expression on the level of translation may allow stricter control. Leaky expression of genes of interest is a phenomenon seen across many cyanobacteria species. There is significant need for a tightly regulated induction system that has little to no leaky expression.

There are many intriguing aspects of cyanobacterial physiology that may play substantial roles in controlling gene expression. We hope researchers will draw parallels from studies of plant chloroplasts and help unravel the complex gene regulatory networks in cyanobacteria.