Comparative analysis of cyanobacteria species reveals a novel guanidine-degrading enzyme that controls genomic stability of ethylene-producing strains

Bo Wang (  bo.wang.2@vanderbilt.edu ) Vanderbilt University https://orcid.org/0000-0002-6047-1221 Yao Xu Vanderbilt University Xin Wang Miami University https://orcid.org/0000-0002-7174-3042 Joshua Yuan Texas A&M University Carl Johnson Vanderbilt University Jamey Young Vanderbilt Univerisity https://orcid.org/0000-0002-0871-1494 Jianping Yu National Renewable Energy Laboratory https://orcid.org/0000-0003-0466-3197


Introduction
Despite the practical applications of guanidine as a protein denaturant (when applied at high concentrations) 1 and as an ingredient in slow-release fertilizers 2 , little is known about the fate of guanidine in biological systems. Guanidine has been detected in human urine at concentrations of 7-13 mg L − 1 (0.12-0.22 mM) 3 , but its biosynthetic pathway remains elusive 4 . A recent study also revealed that a variety of microorganisms, including E. coli, produce guanidine through unknown mechanisms under nutrient-poor growth conditions, suggesting that guanidine metabolism is biologically signi cant and is prevalent in natural environments 5 .
While nonenzymatic decomposition of guanidine under physiological conditions is extremely slow 6 , soil microbes are able to degrade guanidine using heretofore unknown metabolic pathways 7 . Recently, it was reported that a wide range of microorganisms possess a class of guanidine riboswitches that control the expression of downstream genes, a majority of which encode proteins involved in nitrogen metabolism, nitrate/sulfate/bicarbonate transporters, and small multidrug resistance (SMR) transporters 5,8−11 . The SMR transporters were found to be responsible for exporting guanidine out of cells 5,12 . A previously annotated "urea carboxylase" was reported to carboxylate guanidine to form carboxyguanidine 5 , which is degraded by a carboxyguanidine deiminase followed by further degradation by allophanate hydrolase 13 . Another class of enzymes regulated by guanidine riboswitches are annotated as "agmatinases" in the arginase superfamily 5,9,14,15 , which catalyze the breaking of C-N bonds in the guanidyl moiety of agmatine, releasing urea 16 .
There is no current explanation for why these enzymes evolved regulation in response to free guanidine.
To date, the only known enzyme that produces guanidine is the ethylene-forming enzyme (EFE) that catalyzes formation of ethylene and guanidine simultaneously from α-ketoglutarate (AKG) and arginine 17 .
Due to biotechnological interests in developing an alternative pathway for renewable production of ethylene, which is the most highly produced organic compound in the petro-chemical industry, the efe gene from Pseudomonas syringae (a plant pathogen) has been introduced into a variety of microbial species 17 . Some hosts, e.g., Pseudomonas putida KT2440 and the cyanobacterium Synechocystis sp. PCC 6803 (hereafter Synechocystis 6803), have been able to accommodate stable, high-level expression of EFE and thereby sustain enhanced production of ethylene [17][18][19][20][21][22] . Other species, such as cyanobacterium Synechococcus elongatus PCC 7942 (hereafter Synechococcus 7942) and Synechococcus elongatus PCC 11801 (hereafter Synechococcus 11801), however, have not been able to tolerate high-level expression of EFE, and the recombinant strains suffered severe growth inhibition [23][24][25] that was rescued by spontaneous chromosomal mutations that abolished the expression of functional EFE 23,24 .
In this study, we report a novel guanidine-degrading enzyme discovered through comparative analysis of multiple cyanobacterial species. We show that guanidine possesses signi cant toxicity to cyanobacterial cells and destabilizes their genome in response to recombinant EFE expression. Synechocystis 6803 is able to degrade and utilize guanidine as a nitrogen source through the activity of an enzyme encoded by the gene sll1077, which was previously annotated as an agmatinase in the arginase superfamily. We posit that Sll1077 is more likely a "guanidinase", because it degrades guanidine rather than agmatine to urea. This result is consistent with the nding that there is a conserved sequence motif of the guanidine riboswitch upstream of the sll1077 ORF in the genome of the wild type Synechocystis 6803 strain. Synechococcus 7942 lacks a homologous enzyme in its genome and is unable to mitigate guanidine toxicity. We nd that heterologous expression of Sll1077 in a recombinant Synechococcus 7942 strain confers the ability to degrade guanidine into non-toxic urea. Co-expression of Sll1077 and EFE in Synechococcus 7942 stabilizes the genome of the resultant strain and leads to sustained production of ethylene from light and CO 2 .

Results
Varied guanidine degradation capabilities are present in different cyanobacterial species. Given that the impacts of guanidine on microorganisms are unclear, we studied guanidine degradability and toxicity in two model cyanobacterial species: Synechocystis 6803 and Synechococcus 7942. In our preliminary experiments with Synechocystis 6803, when nitrate was gradually replaced with guanidine in the culture medium, the guanidine concentrations declined over a period of four days in all cases under photoautotrophic cultivation conditions (Fig. S1). In order to rule out the possibility of photochemical degradation, Synechocystis 6803 cells were resuspended in the nitrate-deprived mBG11 medium with or without 5 mM guanidine (detailed in Materials and Methods). In parallel, Synechococcus 7942 and heatkilled Synechocystis 6803 cells were also resuspended in the nitrate-deprived culture medium supplemented with 5 mM guanidine. We found that while Synechocystis 6803 cells grown in nitrate-deprived medium exhibited an expected chlorosis phenotype and were still able to double the amount of biomass, cells grown in the guanidine-supplemented medium were able to maintain their green pigmentation and reached a higher cell density after 6 days of photoautotrophic cultivation (Fig. 1a,b). Noticeably, the Synechocystis 6803 cells exposed to exogenous guanidine had a slower growth rate than those not exposed to guanidine during the rst day, probably due to the toxicity of guanidine (Fig. 1a,b). By contrast, the cultures inoculated with heatkilled Synechocystis 6803 or live Synechococcus 7942 cells did not show a typical chlorotic phenotype, and showed continuous decline of biomass over the period of 6 days (Fig. 1a,b). While the guanidine content in the culture with Synechococcus 7942 or heat-killed Synechocystis cells did not decline, the continuous increase of biomass in the culture of live Synechocystis 6803 cells coincided with a steady decrease of the guanidine concentration in the culture medium (Fig. 1c). To this end, we hypothesized that a guanidinedegrading metabolic pathway may exist in Synechocystis 6803 but not in Synechococcus 7942.
Sll1077 is responsible for guanidine degradation in Synechocystis 6803. A comparative proteomic study of the wild-type Synechocystis 6803 and the guanidine-producing (efe-expressing) strain, JU547 26 , showed that the expression of Sll1077, a putative agmatinase, increased by 10-fold in strain JU547 compared to that in the wild-type Synechocystis 6803 (Table S1). Agmatinase cleaves the C-N bond within the guanidyl moiety of agmatine, which releases putrescine and urea 27 . Interestingly, expression of sll1077 is predicted to be under the control of a guanidine riboswitch based on analysis of the RNA sequence upstream of its ORF ( Fig. 2a, b) 5 . We hypothesized that Sll1077 might be involved in the metabolism of guanidine in Synechocystis 6803 (Fig. 2c). Knockout of sll1077 in Synechocystis 6803, leading to strain PB805W (Δsll1077), did not have any apparent physiological effects on the cells under normal growth conditions (data not shown), or under nitrate-deprived conditions ( Fig. 2d-e). Nevertheless, under nitrogen-deprived guanidine-supplemented culture conditions, the cell growth of Δsll1077 was severely inhibited compared to the wild-type Synechocystis 6803 and the degradation of the light harvesting components, i.e., phycobilisomes (absorbance at 630 nm) and chlorophyll a (absorbance at 680 nm), in Δsll1077 was remarkably retarded compared to the wild-type Synechocystis 6803 or Δsll1077 cultivated in nitrogendeprived medium (Fig. 2d-f). Further analysis revealed that the guanidine degradation capability was abolished in the Synechocystis Δsll1077 strain (Fig. 2g), a phenotype similar to that of wild-type Synechococcus 7942 (Fig. 1a). In addition, during the rst day, the biomass of strain Δsll1077 incubated with guanidine increased to a much less extent relative to other parallel cases; in the next few days, the biomass of strain Δsll1077 incubated with guanidine underwent an autolysis process and the light harvesting complex gradually deteriorated ( Fig. 2d-f).
Overexpression of sll1077 in Synechocystis 6803 was achieved through optimizing the ribosome binding site (RBS) at the 5' region as well as tailoring the 3' region of the expression cassette (Fig. 3a). Among the six tested RBSs, RBSv309 in strain PB809W rendered the strongest expression level (Fig. 3b). While removal of the XhoI restriction site between the sll1077 and the 6 x His tag sequence at the 3' region in strain PB812W did not have any apparent effect on the sll1077 expression level, adding the rrnB T1T2 terminator (from E. coli) to the 3' region signi cantly improved the expression of sll1077 in PB816W (Fig. 3a,b). Strain PB816W was able to degrade guanidine at a rate approximately 80% faster than the wild-type Synechocystis 6803, which led to a faster cell growth rate in nitrate-deprived medium (Fig. 3c,d). Interestingly, although removal of the 6 x His tag sequence from the 3' end of sll1077 did not affect the protein expression level (Fig. 3b), it increased the guanidine degradation rate by about two times and substantially increased the cell growth rate of PB817W (Fig. 3c,d), suggesting that the C-terminus 6 x His tag negatively impairs the guanidine-degrading enzyme activity of Sll1077.
In order to verify that guanidine is degraded by Sll1077 to form urea, according to the enzymatic mechanism of the agmatinase/arginase superfamily 27 , Sll1077-His was puri ed from the crude cell lysate of Synechocystis strain PB816W (Fig. 3b). Puri ed Sll1077-His showed an apparent molecular weight of ~ 45 kDa which is consistent with the predicted molecular weight of 43.8 kDa (Fig. 4a). Incubation of puri ed Sll1077-His with guanidine at 30 o C resulted in hydrolysis of guanidine and release of urea ( Fig. 4b-d). It is noteworthy that no reducing factors, such as ATP or NAD(P)H, are required to drive the guanidine hydrolysis enzymatic activity of Sll1077, which seems to be more energy-e cient compared to the previously reported guanidine carboxylation pathway 5,13 (Fig. 4e).
Expression of sll1077 improves tolerance of Synechococcus 7942 to guanidine. In order to examine if expressing a recombinant enzyme, Sll1077 from Synechocystis 6803, could endow the guanidine degradation capability in a host strain that does not naturally degrade guanidine, we expressed sll1077 in Synechococcus 7942, resulting in strain GD7942 (+ sll1077). While the cell growth of Synechococcus 7942 was already inhibited by guanidine at concentrations as low as 0.3 mM and was severely inhibited by 1 mM guanidine under photoautotrophic conditions (Fig. 5a), the sll1077-expressing strain GD7942 gained signi cant tolerance to exogenous guanidine. Particularly, the cell growth of strain GD7942 was not apparently repressed by as much as 1 mM guanidine present in the culture medium, and was only slightly inhibited by 2 mM guanidine (Fig. 5a). We further examined the fate of the exogenous guanidine in the culture medium containing 1 mM guanidine. As expected, while no guanidine degradation occurred in the culture of wild-type Synechococcus 7942, the guanidine added into the culture medium of the GD7942 strain was completely degraded over 4 days of photoautotrophic cultivation (Fig. 5b). Since the wild-type Synechococcus 7942 does not have any urea biosynthesis or degradation pathways 28 , it was expected that urea would be accumulated in the GD7942 culture. Indeed, along with the degradation of guanidine, urea gradually accumulated in the culture supernatants to concentrations of about 1 mM by end of day 4 ( Fig. 5b), which is consistent with the pathway annotation 28 and enzymatic reaction stoichiometry (Fig. 2a). We further found that supplementing 5 mM urea into the culture medium of Synechococcus 7942 did not show any apparent impact on the cell growth under either nitrate-deprived or nitrate-replete culture conditions (Fig. S2), suggesting that Synechococcus 7942 is highly tolerant to urea.
Sll1077 prefers guanidine rather than agmatine as the substrate. To examine the substrate preference of Sll1077 towards guanidine and agmatine, the crude cell extract of Synechococcus 7942 and GD7942 was incubated with 5 mM of either guanidine or agmatine at 30 o C. Surprisingly, we found that the crude cell extract of GD7942 was able to degrade guanidine but not agmatine. The concentration of guanidine incubated with the GD7942 cell lysate decreased by about 2 mM, and concomitantly about 2 mM urea was produced in the reaction mix over the examined 12 h time period (Fig. 5c,d). We therefore propose denominating Sll1077 as a "guanidinase" instead of an agmatinase.
Co-expression of Sll1077 and EFE enhances genomic stability and sustains high-level ethylene formation in Synechococcus 7942. Given that the EFE reaction produces not only ethylene but also toxic guanidine, which might be responsible for the genomic instability observed upon expression of EFE alone in Synechococcus 7942 23,24 , we examined if co-expressing Sll1077 and EFE in the Synechococcus 7942 host strain would render a stable genome and thereby sustained production of ethylene. We found that following the genetic transformation of Synechococcus 7942 and colony-restreaking on BG11 agar plates, the recombinant efe-expressing strain, EFE7942, grew considerably slower than wild-type and the initially formed colonies appeared yellow-greenish; subsequently, large and dark-green colonies grew up on the background of the smaller colonies (Fig. 6a). Cultivation of these "large" and "small" colonies in the liquid culture revealed that cells from the small colonies, but not from the large ones, retained photosynthetic ethylene productivity. Subsequent colony PCR and DNA sequencing results con rmed that cells from the small colonies retained the correct EFE expression cassette on their genomes, whereas the large colonies consisted of cells with mutations around the EFE expression cassette which abolished expression of EFE (Fig. S3). It is noteworthy that restreaking single small colonies onto fresh mBG11-agar plates supplemented with spectinomycin repeatedly resulted in a mixture of large and small colonies after 1-2 weeks of incubation at 30 o C, indicating a constant selective pressure caused by the expression of EFE. In contrast, co-expression of Sll1077 with EFE in Synechococcus strain GD-EFE7942 resulted in uniform colony sizes on agar plates at 30 o C (Fig. 6a), and colony PCR and DNA sequencing con rmed that these cells were able to maintain the intact EFE expression cassette on their genome (Fig. S5), indicating relief of the selective pressure caused by the expression of EFE. Because EFE exhibits highest enzyme activity in the temperature range of 20-25 o C and becomes unstable at temperature above 30 o C 29,30 , we decided to routinely maintain strain EFE7942 at 35 o C to suppress the EFE activity and thereby prevent spontaneous mutations from occurring.
The wild-type Synechococcus 7942 strain and the efe-expressing strains EFE7942 and GD-EFE7942 were then compared in regard to their cell growth rates and ethylene productivities in liquid cultures at 30 o C under photoautotrophic culture conditions. Initially, strain EFE7942 grew considerably slower than the wild-type Synechococcus 7942 strain, but gradually grew faster after subsequent re-inoculations, reaching a growth rate similar to that of the wild-type by day 13. In contrast, the GD-EFE7942 strain exhibited a slightly slower growth rate compared to the wild-type strain throughout the entire 13-day cultivation period (Fig. 6b). In terms of the ethylene production, during the rst 9 days strain GD-EFE7942 showed 3-6 times higher volumetric ethylene productivities compared to strain EFE7942, with more substantial differences occurring at relatively high cell densities when guanidine accumulated to the highest levels in the culture medium ( Fig. 6c, S4). The higher volumetric ethylene productivity of GD-EFE7942 relative to EFE7942 was largely due to the improved cell growth rate and thereby higher cell density (Fig. 6b), yet was also attributed to the improved speci c ethylene productivity (Fig. 6d). During the rst 7 days, the speci c ethylene productivity of GD-EFE7942 was 1.2-1.8 times higher than EFE7942. The difference increased to 2.6 times by day 8, and to 3.3 times by day 9 (Fig. 6d). Starting from day 10, both the volumetric and speci c ethylene productivities of strain EFE7942 dropped substantially and declined to almost zero by day 13 (Fig. 6c,d). The guanidine production in the EFE7942 culture also started to drop signi cantly by day 10 (Fig. S4). Absorbance spectra of the three examined cultures revealed that the abundance of phycobilisome and chlorophyll a in EFE7942 declined signi cantly compared to those in the wild-type strain. Although the phycobilisome level remained low in GD-EFE7942 relative to that of the wild-type strain, expression of Sll1077 restored the amount of chlorophyll a in GD-EFE7942 to a level similar to that in the wild-type strain (Fig. 6e). Further cell growth phenotyping and DNA sequencing analyses revealed that after 13 days of cultivation, approximately half the cells in the EFE7942 culture lost the entire EFE expression cassette, and the other half had DNA mutations on the genome that caused early termination of translation of EFE (Fig. S5, 6f-g). By contrast, the GD-EFE7942 strain exhibited consistent cell growth pro les and ethylene productivities during the ve consecutive batch cultures (Fig. 6b-d), owing to its engineered capability to mitigate guanidine via Sll1077 (Fig. 5b,S4). In addition, the ethylene productivity of GD-EFE7942 is comparable to that of the previously engineered high-level-efe-expressing Synechocystis strains, e.g., strain PB752 in our previous work 31 , under the examined photoautotrophic culture conditions (Fig. S6).

Discussion
Through comparative analysis of cyanobacterial strains, we were able to identify a novel guanidinedegrading enzyme, Sll1077, which breaks down guanidine to form urea and ammonium (Fig. 2a, 4, 5b,c). Sll1077 constitutes a guanidine degradation pathway that does not require ATP, and is completely different from the recently identi ed guanidine carboxylation pathway 5,13 (Fig. 4e). Guanidine carboxylase catalyzes the carboxylation of guanidine using ATP as the driving force. However, the product compound carboxyguanidine is unstable and is readily hydrolyzed to form guanidine and CO 2 in water, forming an ATPconsuming futile cycle 13 . The e ciency of the guanidine carboxylation pathway largely depends on the rate of removal of carboxyguanidine by the carboxyguanidine deiminase which converts carboxyguanidine to ammonium and allophanate 13 . In contrast, the guanidine-degrading enzyme Sll1077 investigated in the current study acts as a deiminase and is able to, without consuming ATP, directly convert guanidine to urea which could be further degraded into CO 2 and ammonium by the urease in most cyanobacterial species, including Synechocystis 6803 28 (Fig. 4). Therefore, the Sll1077-associated guanidine degradation pathway seems more energy-e cient compared to the guanidine carboxylation pathway.
Sll1077 represents a class of novel guanidine-degrading (i.e., "guanidinase") enzymes. Both Sll1077 and Sll0228 in Synechocystis 6803 have been annotated as putative agmatinases since they both have the conserved regions of the agmatinase/arginase superfamily proteins 15,32 . However, their protein sequences show less than 25% identities (Fig. S7). A previous study reported that neither sll1077 nor sll0228 contributes to the arginase activity, while the agmatinase activity in Synechocystis 6803 is mostly attributed to sll0228 33 . A recent study also showed that deletion of sll0228 rather than sll1077 signi cantly impairs the utilization of arginine in Synechocystis 6803 34 . Additionally, from a bioinformatics approach it was found that the expression of sll1077 and its analogs (previously annotated to encode "agmatinase"/"arginase" enzymes) in a wide range of microorganisms is under the control of guanidine riboswitches (Fig. 2a,b; Supplementary Data 1) 5,8,9 . These genes often form operons with other genes, such as hypA, hypB, SsuA_fam (sll1080 in Synechocystis 6803), TM_PBP2 (sll1081 in Synechocystis 6803) and ABC_NrtD_SsuB (sll1082 in Synechocystis 6803) (Supplementary Data 1) 5 . The expression levels of these genes were all enhanced in guanidine-producing Synechocystis strains compared to wild-type controls according to our proteomic data (Table S1) and results from a previous transcriptomic study 35 , which is consistent with the modulation mechanism of guanidine riboswitches 5 . Taken together, these results suggest that Sll1077 may be evolved for a function that is completely different from the degradation of arginine or agmatine. Our ndings that Sll1077 is able to degrade guanidine and that it prefers guanidine rather than agmatine as substrate is consistent with the prediction that its expression is under the control of a guanidine riboswitch (Fig. 2a,b) 5 , which suggests that Sll1077 and possibly its analogs are evolved for the degradation of guanidine. It is likely that guanidine, formed either biologically or abiotically, is present in the natural environment where Synechocystis 6803 lives, and possessing sll1077 has rendered survival advantage. Running a protein BLAST for the Sll1077 peptide sequence (https://blast.ncbi.nlm.nih.gov/) returned over a thousand hits with > 50% sequence identities, all of which have been annotated as arginase/agmatinase family proteins (Supplementary Data 2). Whether these proteins possess the capability to degrade guanidine needs to be studied in the future.
Guanidine causes a disorder of pigment metabolism in cyanobacterial cells. Guanidine is known to interact with the peptide backbone and side-chains of amino acids, and serves as a protein denaturant when applied at high concentrations (2-6 M) 1,36,37 . At concentrations insu cient to completely unravel the protein structure, guanidine could also be detrimental to biomacromolecules. For example, relatively small amounts of guanidine could trigger unfolding of the active site of ribonuclease A and thereby inactivate the enzyme activity and facilitate the proteolysis process 38 . Another example is that millimolar guanidine could signi cantly inhibit ammonium nitri cation in the nitrifying bacteria in soil 39 . In our study, the presence of guanidine in the culture medium, either from exogenous or endogenous sources, severely inhibited cell growth of wild-type Synechococcus 7942 and the Synechocystis Δsll1077 strain (Fig. 1a,b, 2d,e, 5a, 6a,b). These guanidine-sensitive strains exhibited remarkably slow degradation of their light harvesting components under nitrate-deprived and guanidine-supplemented culture conditions (Fig. 1a, 2e,f). Under nitrogen-poor culture conditions, cyanobacterial cells typically undergo a chlorosis process that involves degrading their phycobiliproteins and chlorophyll as a nitrogen source to support cell growth while simultaneously downregulating photosynthesis in order to reduce the generation of damaging oxygen radicals 40 . Impaired cell growth and retarded pigment degradation in both cultures of Synechocystis 6803 + and Δsll1077 + on day 1 (Fig. 2d-f) suggested that induction of nitrogen chlorosis was disrupted by guanidine under the examined culture conditions. Furthermore, the biosynthesis of phycobiliproteins and chlorophyll was severely inhibited in strain Synechococcus EFE7942, whereas the biosynthesis of chlorophyll was restored through heterologous expression of Sll1077 in strain GD-EFE7942 (Fig. 6e), which provided additional evidence that guanidine hampers the biosynthesis and remodeling of photosynthesisrelated pigments in cyanobacteria.
While the wild-type Synechococcus 7942 is sensitive to guanidine and fails to accommodate high-level expression of EFE (Fig. 1a,b, 5a, 6), our discovery of the guanidine-degrading activity of Sll1077 was leveraged to generate a derivative strain of Synechococcus 7942 that exhibits enhanced genomic stability and stable high-level production of ethylene in prolonged culture, which has not been achieved in prior studies (Fig. 6, S5) [23][24][25]41 . It is noteworthy that co-expression of Sll1077 with EFE substantially attenuate, but does not completely eliminate, the accumulation of guanidine in cultures of the engineered Synechococcus GD-EFE7942 strain (Fig. S4). Although this seems already su cient for rendering genomic stability and sustained stable ethylene production in GD-EFE7942 (Fig. 6, S5), as well as the Synechocystis strain PB752 26 , it could be possible to obtain a more e cient guanidine-degrading enzyme, perhaps through directed evolution of Sll1077, in order to achieve faster degradation of guanidine and further reduce its toxicity in the future. In summary, this study has advanced our understanding of the biological routes of guanidine metabolism in nature and has demonstrated a new approach for enhancing biosynthesis of target molecule(s) by reducing toxic byproduct(s), focusing upon the speci c example of stabilizing ethylene production in engineered microorganisms.

Materials And Methods
Bacterial strains and growth conditions. E. coli NEB5α (New England BioLabs, MA, USA) served as the microbial host for cloning and maintaining all recombinant plasmids, and was routinely grown in LB medium. Synechocystis and Synechococcus strains were typically grown in a modi ed BG11 medium (mBG11) as described before 26 Table   S2.
Construction of recombinant plasmids. All enzymes and cloning kits were purchased from New England Biolabs, MA, USA, unless otherwise speci ed. Kits for DNA puri cation were purchased from Qiagen, MD, USA. Plasmid pPB305 was constructed by PCR ampli cation of the DNA fragments of sll1077U, sll1077D, and cat, and Gibson Assembly into plasmid pBlueScript II SK (+) which was digested with KpnI and SacI. The DNA fragment containing gene sll1077 was PCR ampli ed from the genomic DNA of Synechocystis 6803 and inserted between the NdeI and XhoI restriction sites on pET30a(+), so that Sll1077 will be tagged with 6xHis, resulting in plasmid pPB300. pPB306 was constructed by PCR amplifying sll1077-His from pPB300 and inserting it between the NdeI and SalI restriction sites on pSCPTH (Wang, 2013) using Gibson Assembly Kit. pPB306d was constructed by deleting the lac promoter region on the pBluescript vector backbone via digesting pPB306 with SacI and SapI restriction enzymes and then blunt-ended using T4 DNA polymerase and self-ligated using Quick DNA ligase. pPB307, pPB308, pPB309, pPB310, pPB311 were constructed by replacing the RBS in pPB306d using the Site Directed Mutagenesis Kit. pPB312 was constructed by deleting the "CTCGAG" (XhoI) nucleotides between the sll1077 coding sequence and the 6xHis tag on plasmid pPB309. pPB316 was constructed by inserting the rrnBT1T2 terminator (from E. coli NEB5α) downstream of sll1077 on pPB312. pPB312 was digested with SalI, dephosphorylated and then assembled with the terminator rrnBT1T2 using Gibson Assembly Kit. pPB313 was constructed by deleting the 6xHis tag and "CTCGAG" (XhoI) between the sll1077 coding sequence and stop codon TAA of pPB309. pPB317 was constructed by inserting rrnBT1T2 downstream of sll1077 on pPB313, which was digested with SalI and dephosphorylated, using the Gibson Assembly Kit. Plasmid To express the efe gene in Synechococcus 7942, 1.43 kb of the BbvCI/XhoI fragment containing psbAp::efe-FLAG from pJU158 was blunt-ended and ligated to the SmaI site of the neutral site 1 vector pAM1303, resulting in pEFE-FLAG-NS1.
To overexpress the sll1077 gene in Synechococcus 7942, 1.69 kb of the BamHI/SalI fragment harboring the sll1077 expression cassette from pPB317 was cloned into the BamHI/SalI site of a neutral site 4-targeting vector pCX0104-LuxAB-FT 42 to generate pGD7942-NS4. The DNA sequence of genes of interest were all conformed by DNA sequencing. Primers used in constructing all plasmids are detailed in Table S2.
Genome engineering of cyanobacteria. Transformation of Synechocystis was accomplished via natural transformation as described previously 43 . Brie y, the wild-type Synechocystis 6803 strain was grown in mBG11 medium until the OD 730 reached approximately 0.4. Then, 2.5 mL of culture was condensed to about 0.2 mL via centrifugation and resuspension with the same culture medium. Cells were transferred into a 1.5 mL Eppendorf tube and mixed with 1-2 µg DNA of integration plasmid. The sample was incubated under low light for about 5 hours, and mixed once in the middle of the incubation. Cells were then spread onto BG11 plates supplemented with appropriate antibiotics. Strains PB805W -PB812W, PB816W, PB817W were constructed by transforming wild-type Synechocystis 6803 with integration plasmids pPB305 -pPB312, pPB316 and pPB317. Strains PB816H and PB817H were constructed by transforming an efe-expressing strain, Synechocystis PB752, with the integration plasmids pPB316 and pPB317, respectively.
Transformation of Synechococcus 7942 was completed following a previously established protocol 44 . Transformation of Synechococcus 7942 with integration plasmids pEFE-FLAG-NS1 or pGD7942-NS4 resulted in strain EFE7942 and GD7942, respectively. The efe expression cassette was PCR ampli ed from the genomic DNA of EFE7942 strain using primers NS15 and NS16, and inserted into the neutral site 1 of the genome of Synechococcus GD7942, resulting in strain GD-EFE7942. The complete segregation of genomes was veri ed via colony PCR, followed by DNA sequencing of the PCR products ampli ed using primers (listed in Table S3)  HisProbe™-HRP Conjugate (Thermo Fisher Scienti c, MA, USA) was used as the antibody (at 1:500 dilution) to detect the Sll1077-His. The chemiluminescent blots were imaged using FluorChem Q imager (ProteinSimple, CA, USA).
In vitro enzyme activity assay. His-tagged Sll1077 i.e., Sll1077-His, was rst puri ed from Synechocystis PB816W. PB816W was grown in 250 mL mBG11 medium under 50 µE m − 2 s − 1 until an OD 730 of about 3, and then cells were harvested via centrifugation at 3220 × g, 24 o C for 10 min followed by removal of supernatants. The cell pellets were stored at -80 o C. Cells were subsequently resuspended with 10 mL of cold 0.1 M potassium phosphate buffer (pH7.0) supplemented with DTT (0.2 mM) and Halt Protein Inhibitor Cocktail (Thermo Fisher Scienti c, MA, USA), and lysed by sonication in an ice-water bath using a Q500 Sonicator (Qsonica L.L.C, CT, USA) programed for 100 cycles of 3-sec-on-3-sec-off at an amplitude of 20%.
The cell lysate was centrifuged at 4 o C, 8000 × g for 10 min, and then the supernatant containing soluble proteins was run through His GraviTrap (GE Healthcare) to purify Sll1077-His following the user manual.
Brie y, the puri cation column containing 1-mL Ni sepharose was rst equilibrated with 10 mL binding buffer (20 mM sodium phosphate, 500 mM NaCl, 45 mM imidazole, pH 7.4), and then was loaded with the approximately 10 mL cleared cell lysate. After all of the lysate went through the Ni sepharose, the sepharose was washed twice, with 10 mL and 5 mL of the binding buffer, respectively. Ultimately, 3 mL elution buffer (20 mM sodium phosphate, 500 mM NaCl, 500 mM imidazole, pH 7.4) was applied to the puri cation column to elute Sll1077-His. 0.4 mL puri ed Sll1077-His (3.5 mg mL − 1 ) was mixed with 30 µL guanidine (1 M) dissolved in 5.6 mL reaction buffer (the same as the above binding buffer). As a control, 0.7 mL BSA (2 mg mL − 1 ) was mixed min. Then the cell extract (supernatants) were used for the following in vitro assay: 1.54 mL 100 mM Tris•HCl (pH8.0), 20 µL MnCl2, 20 µL NH4Cl, 400 µL cell extract, and 20 µL 500 mM guanidine•HCl or agmatine•HCl (with a total reaction volume of 2 mL). All the components but the cell extract in each reaction mix were mixed together and incubated in a 30 o C water bath for about 15 min before the cell extract was added into the reaction mix to start the assay. In the control experiments, cell extract were replaced by the sample was taken from the reaction mixes at 0, 2 and 12 h time points, and were immediately mixed with 50 µL 2N HCl to quench any enzymatic activity. 50 µL 2N NaOH was then added the samples to neutralize the pH followed by storage at -20 o C. After all samples were collected, 150 µL of each sample was used for quanti cation of urea using GC-MS and another aliquot of 150 µL was used for quanti cation of guanidine and agmatine using HPLC. For GC-MS quanti cation of urea, each sample was mixed with 600 µL methanol, vortexed, added 150 µL chloroform, vortexed, 450 µL water, vortexed, and then centrifuged at 17000 × g for 2 min. The aqueous layer (~ 1.2 mL) were transferred into a clean Eppendorf tube, air-dried over night and then lyophilized before being derivatized with MTBSTFA + 1% TBDMCS (Regis Technologies, Inc.) at 70 o C for 30 min, and subsequently run on GC-MS for analysis of urea. A series of concentrations of urea standards were dissolved in the in vitro enzyme assay buffer, lyophilized and derivatized side by side with the enzyme assay samples in order to establish a calibration curve to quantify the urea. For HPLC quanti cation of guanidine and agmatine, samples were subjected to methanol/chloroform extraction and air-dried, and then resuspended with 750 µL of water before being loaded on to HPLC using a method described below.
Quanti cation of guanidine and agmatine using HPLC. Guanidine was quanti ed using a protocol modi ed from a previous method 26  and centrifuged again, followed by washing with 20 mL and 1 mL washing buffer. The supernatants were discarded and cells were frozen at -80 o C. Three biological replicates were included for each strain. Comparative proteomic analyses of Synechocystis 6803 and JU547 was conducted following our previously published method 46 . The sample preparation and amount of peptide loaded to the capillary column varied from that in the previous method. Brie y, cell pellets taken out of -80 o C were lysed by sonication with a program of 12 cycles of 10 seconds-on-2-minutes-off on ice. The supernatants were collected via centrifugation and the protein concentrations were analyzed using Bradford assay (Thermo Scienti c, Rockford, IL). Then, 75 µg of total protein for each sample was used for downstream proteomic sample preparation following the same procedure as described before.   Gene sll1077 is responsible for guanidine degradation in Synechocystis 6803. a, The secondary RNA structure of the guanidine riboswitch upstream of the guanidine exporter encoded by the ykkC gene in Bacillus subtilis. Nucleotides in red font are >97% conserved in type I guanidine riboswitch5. b, The secondary RNA structure of predicted guanidine riboswitch upstream of the sll1077 gene in Synechocystis Overexpression of Sll1077 accelerates guanidine degradation and promotes biomass growth in Synechocystis 6803. a, Strategies for enhancing the overexpression of gene sll1077 in Synechocystis 6803. Gene sll1077 was overexpressed driven by the tac promoter, with its RBS at the 5' region and the His tag and terminator at the 3' region optimized. b, SDS-PAGE and western blotting (His tag) showing the improved expression of Sll1077 in Synechocystis. c, Guanidine degradation pro les of Synechocystis 6803 and sll1077-overexpressing strains. d, Cell growth curves for Synechocystis 6803 and sll1077-overexpressing strains, indicated by readings of OD730 of cell cultures. Data represent means and standard derivations from three biological replicates. Figure 4 Con rmation of the guanidine-degrading enzyme activity of Sll1077 through an in vitro enzyme activity assay. a, SDS-PAGE showing the cell extract from Synechocystis 6803, PB816W and puri ed Sll1077-His. b, TBDMS derivative of urea. Red font indicates the urea backbone. The boxed portion indicates the main ion detected by GC-MS. c, Ion counts of ion 231 for TBDMS derivative of urea standard or the product of guanidine incubated with either Sll1077-His (Reaction) or bovine serum albumin (Control). d, Mass spectra of the peak at 16.095 min in c. e, Guanidine degradation pathways identi ed to date. Pathway I was reported previously, and pathway II is demonstrated in the current study.  Expression of sll1077 in Synechococcus 7942 supports sustained high-level ethylene production. a, Colonies of strains EFE7942 and GD-EFE7942 formed on agar plates at 30 oC. DNA sequencing results revealed that for strain EFE7942, the smaller colonies indicated by cyan triangles harbored the correct EFE expression cassette, while the bigger colonies denoted by red triangles harbored mutated EFE expression cassettes; for strain GD-EFE7942, colony sizes were uniform and DNA sequencing identi ed no mutations around the EFE expression cassette. b, Cell growth curves in liquid cultures at 30 oC. Cultures were re-inoculated into fresh media every three days. c, Volumetric ethylene productivities of strains EFE7942 and GD-EFE7942. d, Speci c ethylene productivities of strains EFE7942 and GD-EFE7942. Cultures were re-inoculated every three days. Data represent means and standard deviations from two biological replicates. e, Absorbance spectra of cultures shown in b-d at day 1 and day 2. Absorbance was normalized to the absorbance at 730 nm. f, Two colonies of the wild-type strain 7942, ten colonies of strain EFE7942 and ten colonies of strain GD-EFE7942 were randomly picked from plates spread with diluted day-13 cultures shown in b-d and were subjected to colony PCR using primers anking the efe-insertion site on the genome. Red arrows indicate the expected PCR product size for strains EFE7942 and GD-EFE7942; black arrows indicate the expected PCR product size for the wild-type 7942 strain. g, DNA sequencing of the PCR products obtained in f revealed mutations of the EFE expression cassettes on the genomes of all ten randomly picked EFE7942 colonies, whereas no mutations arose within the genomic region of the EFE expression cassette in any of the ten GD-EFE7942 colonies.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.