A Synthetic Biology Toolbox for Nitrogen-Fixing Soil Microbes

Abstract

The soil environment adjacent to plant roots, termed the rhizosphere, is home to a wide variety of microorganisms that can significantly impact the physiology of nearby plants. Microbes in the rhizosphere can provide nutrients, secrete signaling compounds, and inhibit pathogens. These processes could be manipulated with synthetic biology to enhance the agricultural performance of crops grown for food, energy, or environmental remediation, if methods can be implemented in these non-model microbes. A common first step for domesticating non-model organisms is the development of a set of genetic engineering tools, termed a synthetic biology toolbox. A toolbox comprises transformation protocols, replicating vectors, genome engineering (e.g., CRISPR/Cas9), constitutive and inducible promoter systems, and other gene expression control elements. This work validated synthetic biology toolboxes in three nitrogen-fixing soil bacteria: Azotobacter vinelandii, Stutzerimonas stutzeri (Pseudomonas stutzeri), and a new isolate of Klebsiella variicola. All three organisms were amenable to transformation and reporter protein expression, with several functional inducible systems available for each organism. S. stutzeri and K. variicola showed more reliable plasmid-based expression, resulting in successful Cas9 recombineering to create scarless deletions and insertions. Using these tools, we generated mutants with inducible nitrogenase activity and introduced heterologous genes to produce resorcinol products with relevant biological activity in the rhizosphere.

Graphical Abstract

Introduction

Plant growth-promoting bacteria can be used for developing beneficial microbial communities and improving agricultural yields. In particular, associative nitrogen-fixing bacteria (diazotrophs) have the potential to offset fertilizer usage and provide fixed nitrogen to fertilizer-intensive cereal crops such as wheat, corn, or sorghum, making them appealing targets for developing improved microbial inoculants.^1,2 However, diazotrophs remain poorly characterized, with few tools for robust genetic engineering essential to meet applied goals. Well-characterized genetic engineering toolboxes can accelerate the pace of trait enhancement in non-model microbes.^3,4 Therefore, in this study, we investigated three diazotrophs for tool development and production of relevant products for the rhizosphere: Stutzerimonas stutzeri A1501 (formerly known as Pseudomonas stutzeri), Azotobacter vinelandii DJ and a new isolate of Klebsiella variicola.

S. stutzeri A1501 has previously been studied extensively for its ability to fix nitrogen, and the regulation of this fixation has been studied in great detail. ^5–7 Despite these studies, its synthetic biology tools remain limited. Previous work achieved genetic modifications via a triparental conjugation strategy, with deletions achieved via homologous recombination of antibiotic selection markers.^8,9 Similarly, A. vinelandii DJ is a very well-studied obligate aerobe with an unusual metabolism that allows nitrogen fixation at atmospheric conditions.^10. It is an exceptional model due to its natural competency under metal-starved conditions ¹¹ as well as expression of three different nitrogenase enzymatic complexes (MoFe, VFe, and FeFe cores). ^12–14 It was used previously to elucidate the structure^15,16 of the nitrogenase enzyme complex. Using homologous recombination with antibiotic selection, as well as sacB counterselection, A. vinelandii mutants that secrete high amounts of ammonia when grown in batch conditions have been produced.^17–19 Finally, while Klebsiella pneumoniae has been a model for studying biological nitrogen fixation for decades,^20,21 safety concerns over the use of these strains prevented agricultural use. However, more recently, many Klebsiella variicola isolated from maize and sorghum have been sequenced and found to lack human/animal virulence genes, opening the path for using such K. variicola for genetic engineering and agricultural applications.^22.23 For example, KV37-1036 was used as nitrogen-fixing bacterial supplement in field studies.²⁴ Genetic edits of this strain relied on screening for insertions using the integration of an antibiotic resistance marker using a single crossover, followed by a sacB or other counterselection strategy to remove the plasmid backbone. ²⁵ With these preliminary studies, K. variicola shows promise as a malleable chassis organism and an effective nitrogen-fixing symbiont for fertilizer-intensive crops.

While these three bacterial species have all been studied in some capacity, the ability to rapidly modulate expression in these diazotrophs more quickly and reliably would allow for further investigation of plant-microbe interactions and improving nutrient exchange. This study aimed to validate tools for advanced genetic engineering in diazotrophs. First, we characterized synthetic biology tools in these organisms using fluorescent reporters, focusing on broad-host plasmids, constitutive promoters, and inducible systems. Once we validated robust protein expression using these tools, we developed a genome-editing workflow for these diazotrophs. Using a Cas9/λRed-mediated recombineering strategy, we demonstrated two modes of action for plant growth promotion. In the first mode, we created strains with changes in nitrogenase expression to be used as tools to investigate the nitrogenase regulation mechanisms. In the second mode, we demonstrated the production of alkylresorcinol compounds using the sbARS2 type III polyketide synthase (PKS) gene from sorghum.²⁶ Production of alkylresorcinols is an important precursor step to the production of sorgoleone, which has antifungal, antimicrobial, and herbicidal activities that could lend benefit to the rhizosphere of a variety of crop species.^27–31

Results and Discussion

Isolation of Nitrogen-Fixing Bacteria from Sorghum Roots

Mechan et al. (2023) recently reported the isolation of diazotrophs from the epicuticular wax and the aerial root mucilage of sorghum.²² Similarly, we isolated diazotrophs from the aerial root mucilage of sorghum (Supplementary Table 1), including two Klebsiella pneumoniae strains, two strains in the genus Paenibacillus, one Erwinia endophytica strain, and an isolate of Klebsiella variicola (isolate D4). K. variicola D4 was isolated on nitrogen-free, solid artificial mucilage media.³² Molecular confirmation of nifH was performed as described for all organisms³³ and initial acetylene reduction assays were performed in semi-solid nitrogen-free medium (0.2% agar) to confirm nitrogenase presence and activity. All strains were assayed for the following plant growth promoting traits: auxin production using Salkowski’s reagent,³⁴ organic and inorganic phosphate solubilization,³⁵ siderophore production,³⁶ and 1-aminocyclopropane-1-carboxylate (ACC) deaminase activity.³⁷ The K. variicola strain produced ~43 μg/mL IAA (Indole-3-Acetic Acid) in vitro and solubilized both organic and inorganic phosphate. The strain did not produce siderophores or display ACC deaminase activity on plates. The plant growth-promoting activity of all isolated strains is summarized in Supplemental Table 1. Because the strain was an excellent nitrogen fixer and also displayed strong plant growth-promoting traits, the genome of K. variicola D4 was sequenced (NCBI Accession PRJNA793885). Since this isolate has no virulence gene for plants or animals, it was used as one of the chassis for this study along with the well-characterized S. stutzeri A1501 and A. vinelandii DJ. A summary of this isolate is shown in Table 1.

Table 1. Characteristics of the Klebsiella variicola strain used in this study.

Data obtained from this isolation work as well as the Pathogenwatch platform for virulence loci (https://pathogen.watch).

Classification:	Bacteria; Pseudomonadota; Gammaproteobacteria; Enterobacterales; Enterobacteriaceae; Klebsiella/Raoultella group; Klebsiella
Species:	Klebsiella variicola subs. variicola
Taxonomy ID:	244366
Virulence score:	No virulence loci
Genome length:	5,365,114 bp
Protein-coding genes	5,019
Plasmids:	1
GC content:	57.6%
tRNAs	67
Antibiotic resistance:	Ampicillin 100 μg/mL, Chloramphenicol 12.5 μg/mL, Streptomycin 50 μg/mL.
Host:	Aerial root mucilage of Sorghum bicolor
Place of collection:	Live Oak, FL, USA.
Year of collection	2020
Auxin Production	43.45 μg/mL
Siderophore Production	−
Organic Phosphate Solubilization	+
Inorganic Phosphate Solubilization	+
ACC Deaminase Activity	−

Plasmid and Promoter Characterization in Nitrogen-Fixing Bacteria

The most critical element of a synthetic biology toolbox is the ability to introduce and stably express protein from heterologous DNA. Therefore, we first aimed to characterize the compatibility and relative expression of broad-host plasmids in these non-model organisms. We used four broad host plasmids previously identified in the literature: RK2,³⁸ BBR1,³⁹ RSF1010,⁴⁰ and pRO1600,^41,42 as well as two plasmids generally more specific to Enterobacteriaceae, p15A⁴³ and ColE1.⁴⁴ The transformation efficiencies of different plasmid backbones are summarized in Table 2. All six plasmid backbones were transformed consistently in K. variicola D4, indicating its compatibility with these Enterobacteriaceae plasmids. While all plasmids generated antibiotic-resistant colonies of S. stutzeri, only BBR1 and RSF1010 backbones were transformed reliably. Vectors containing RK2, pRO1600, p15A, and ColE1 origins were transformed into S. stutzeri with very low efficiencies with no plasmid backbone detectable by PCR, suggesting that surviving colonies may be the result of spontaneous mutation or uncommon ectopic integration events of the antibiotic resistance marker. The original cosmid from which the RK2 origin is derived can replicate in S. stutzeri A1501,^6,45 so additional replication machinery from the cosmid may be necessary to restore functionality. Finally, A. vinelandii was able to maintain RK2, BBR1, and pRO1600 backbone plasmids, albeit with slower growth in minimal media (Figure S1).

Table 2: Transformation efficiencies of different plasmid backbones in three diazotrophs.

Bolded values indicate plasmids that transform and express reliably. Standard deviations are from biological duplicates.

Transformation Efficiencies (CFU/μg DNA)	K. variicola	S. stutzeri	A. vinelandii
RK2	(2.0 ± 0.4)* 103	(2 ± 2) * 101	(4 ± 1) * 104
BBR1	(2 ± 1)* 103	(1.3 ± 0.1) * 104	(5.0 ± 0.5)* 104
RSF1010	(1.3 ± 0.7)* 103	(4 ± 1) * 104	(3 ± 2)*101
pRO1600	(1.3 ± 0.1) * 104	(3 ± 2)* 100	(7 ± 0.5) * 103
p15A	(1.20 ± 0.01) * 104	(2 ± 1) * 101	0
ColE1	(7 ± 2) * 103	(3 ± 0.4) * 101	0

After validating stable vector backbones, we also verified the function of 5 constitutive promoters using fluorescent reporters. Using a subset of the Anderson promoter library,⁴⁶ we confirmed that we could get a range of constitutive expression levels. Reporter expression in K. variicola on broad host plasmid origins was consistent with values reported in E. coli. In contrast, expression in A. vinelandii and S. stutzeri was more inconsistent with the values reported in E. coli (Figure 1). These results indicated we could successfully achieve constitutive expression in all three organisms with reliably maintained plasmids. We sought additional promoters from a degenerate promoter library for further granularity in transcriptional control. The BBR1 backbone was functional in all three organisms, so all subsequent promoter testing used this backbone. Since the J23101 promoter from the Anderson library showed reliable constitutive expression in all three organisms, we mutagenized this promoter by PCR using degenerate oligos (5’tttacaNNNNNNNNNNNNNNNNNtattatNNNNNN’3) to generate a library for cloning. The oligo design kept the −35 and −10 regions of the J23101 promoter intact to maximize the number of clones with functional constitutive expression based on our initial results. While we could not capture the full diversity of this library through screening using 96-well plates, we could demonstrate ranges of expression between the original strong promoter and empty vector. A subset of isolates that capture the range of the fluorescence of sequenced promoters is shown in Figure 2. From these promoter sequences, we demonstrate that the interspatial region between the −35 and −10 regions can impact expression significantly. We also analyzed a set of 11 different ribosome binding sites (RBSs) after the J23101 promoter to find additional ways to modulate protein expression. The strength of these RBSs was predicted by using the Salis lab calculator⁴⁷ and resulted in a dynamic range of over 1.5 orders of magnitude in fluorescence, thereby demonstrating the ability to modulate both transcription and translation in the three species (Figure S2).

Figure 1: Constitutive promoter fluorescence in different broad-host backbones in K. variicola, S. stutzeri, and A. vinelandii.

Predicted promoter strengths are from original work in E. coli.⁴⁶ Error bars show standard deviation from biological triplicates.

Figure 2. Promoter modulation in K. variicola, S. stutzeri, and A. vinelandii using degenerate oligos results in dynamic ranges in expression.

A subset of promoters from each organism was taken to capture the full range of expression from the generated library. The red bar in each plot indicates the original fluorescence values from the starting promoter (Anderson Promoter J23101). Error bars show standard deviation from biological triplicates.

Inducible Promoter Systems Are Variable Across Different Bacteria

Inducible promoters allow for conditional expression of heterologous proteins – a critical capability for designing synthetic circuits and tailored metabolism. Therefore, we validated the function of the 12 inducible systems previously developed in the “Marionette” strain of E. coli.⁴⁸ We also validated two additional plasmid-based systems from the same paper (acrylic acid and erythromycin biosensors) and a rhamnose-inducible cassette from previous work in soil bacteria, bringing the total number of tested induction systems to 15.⁴⁹ As the regulatory proteins responsive to the Marionette inducers were previously evolved to limit crosstalk, they offered a good starting set of inducible systems to onboard into new organisms using a broad host backbone. The results for the tests of these 15 induction systems are shown in Figure 3. Almost all these inducible systems worked comparably in K. variicola as E. coli, with the exception of the naringenin and choline induction systems (Figure S3). Again, results were more variable in S. stutzeri and A. vinelandii, with fewer induction systems transformable and functional in these organisms. Many induction systems in A. vinelandii showed nonfunctional induction on plasmids, with only 3 successful cassettes. However, some of the induction systems for A. vinelandii were functional when the regulators and responsive expression cassette were integrated into the genome of the organism (Figure S4). This suggests that plasmid systems are nonideal for A. vinelandii and may not yield an accurate representation of possible expression from the chromosome. Additionally, in both S. stutzeri and A. vinelandii, the fold-change with induction tended to be smaller than in K. variicola D4, indicating more significant differences in regulation from E. coli, which is to be expected given their larger evolutionary distances from E. coli. By starting from a large pool of initial inducers, we could find functional inducer systems for all three diazotrophs for further design work, even when many cassettes were nonresponsive to induction.

Figure 3. Induction plots of inducer systems in K. variicola, S. stutzeri, and A. vinelandii.

The same set of 15 inducers was tried for all three organisms on a BBR1 backbone: DAPG, Cuminic Acid, 3OC6-HSL, Vanillic Acid, IPTG, Anhydrotetracycline (aTc), Arabinose, Choline, Naringenin, DHBA, Salicylic Acid, 3OC14-HSL, Acrylic Acid, Erythromycin, and Rhamnose. The shaded region represents the standard deviation from biological triplicates. Only induction systems that were transformable and functional on plasmids in each organism are shown (nonfunctional inducer data in Figure S3).

Developing Gene-Editing Systems Using Synthetic Biology Parts

With these characterized expression systems for three diazotrophs, we next aimed to develop robust genome editing tools, with an outline of our gene-editing approach shown in Figure 4. Using the design from previous work in E. coli and P. putida,⁵⁰ we developed a λRed/Cas9 recombineering protocol for scarless genome editing with positive counterselection against the original wild-type DNA for these organisms. Based on our broad host plasmid compatibility, we used a two-plasmid system using RK2/BBR1 backbones for K. variicola and A. vinelandii and BBR1/RSF1010 backbones for S. stutzeri. One plasmid contained constitutively expressed Cas9 and arabinose-inducible λRed genes. The other plasmid contained the gRNA for the given target site, allowing for many different guides to be introduced sequentially to make subsequent edits by cycling of the gRNA plasmid. Using these methods and recombineering conditions, we generated scarless mutants in S. stutzeri and K. variicola. However, we could not successfully create genome edits in A. vinelandii using a Cas9-mediated system, likely due to a lack of protein production of our plasmid-based systems (Figure S5). Therefore, we continued with our genome-editing pipeline with only S. stutzeri and K. variicola.

Figure 4. Schematic for Cas9/λ-Red Recombineering in S. stutzeri (a) and K. variicola (b and c).

For S. stutzeri, the repair template was first integrated into the genome using antibiotic selection for a single crossover event. In a second step, the gRNA plasmid was introduced via electroporation to induce positive selection for clean knockouts in cells. In contrast, genome editing in K. variicola utilized a one-step protocol for simultaneous introduction of the gRNA plasmid as well as the repair template, with the deletion workflow in (b), and the insertion workflow in (c). In all cases, electroporation was used to introduce DNA into each organism.

First, we aimed to create mutants of nitrogen-fixing organisms that abolished nitrogen-fixation activity. We deleted nifH, a critical subunit in the nitrogenase enzyme complex, in K. variicola and S. stutzeri to create a negative control, non-fixing strain. This strategy has successfully been used previously to abolish fixation in many other diazotrophs.^51,52 Using our Cas9 editing strategy, we were able to generate strains of both organisms that were unable to fix nitrogen (Figure 5). This initial mutation demonstrated the success of our genome editing tools, so we next moved to test other systems for deregulated nitrogen fixation.

Figure 5. Nitrogenase activity can be induced in K. variicola with arabinose and abolished in both K. variicola and S. stutzeri.

(a) A schematic of nitrogenase regulation through the nifLA operon in K. variicola and S. stutzeri. Acetylene reduction is used to approximate nitrogen fixation activity. (b) The ΔnifL::araC-araBAD-nifA mutant changes the nitrogenase system to be arabinose-inducible in Burk’s media. In both K. variicola and P. stutzeri, ΔnifH strains show dysregulation of nitrogen fixation. P-values indicate the results of a two-way ANOVA on log-transformed data with use of the Bonferroni correction (P<0.05 threshold). Error bars show standard error from biological triplicates.

Previous work in A. vinelandii, K. variicola strain KV37-1036, and other diazotrophs shows that replacement of nifL with a constitutive promoter results in increased expression of nifA, which in turn increases expression of the nitrogenase complex by activating transcription of the nif genes required for nitrogenase synthesis.^24,53 These modifications have resulted in an overproduction of ammonia, which can increase nitrogen availability for symbiont crops. Using our Cas9-editing strategy, we created a ΔnifL::J23101 promoter mutant in both organisms to recapitulate these findings. While this change resulted in a significant difference in nitrogen-fixing activity, we did not observe the deregulation from extracellular ammonia concentrations as expected. Rather, complete removal of the nifL CDS resulted in abolishment of nitrogen fixation activity, with fixation levels comparable to a ΔnifH strain (Figure S6). Interestingly, fixation did not occur even in conditions where extracellular ammonia was absent, indicating that these strains did not fix nitrogen even when it was essential. Through the construction of these initial mutants, we observed that K. variicola was more responsive to our genetic engineering tools, with a higher rate of transformation success. Since K. variicola was most amenable to Cas9-mediated editing, we decided to further explore other gene expression mutations for deregulating ammonia production. In particular, we cleanly knocked out nifL with no promoter, as well as made a knockout of the ammonia importer (ΔamtB) in a nifL knockout strain. However, none of the genomic knockouts fixed nitrogen more robustly than the wild-type strain, even in the absence of a fixed nitrogen source. Other previous work with nitrogenase deregulation in K. variicola has left parts of the N and C terminus of the protein CDS intact to minimize operon disruption,²⁴ which may explain the differences in fixation between our observations and prior literature.

With our K. variicola isolate, we instead aimed to have nitrogenase respond to plant-relevant signals. Arabinose is a carbohydrate source found in both root exudates and mucilage of aerial roots,^54,55 making it a useful metabolite to use as a proof-of-concept for plant-inducible controls. We knocked out nifL in K. variicola and replaced it with an arabinose promoter cassette. This strain did not fix nitrogen when arabinose was absent, even in nitrogen-free conditions, but fixed nitrogen at levels comparable to wild type upon arabinose induction (Figure 5). When ammonia is present in the media, this strain still sees some inducible activity, albeit with weaker expression upon the same level of arabinose induction. This fixation occurs even at a low arabinose concentration relative to the maximum induction seen with sfGFP (1.33 mM vs. 40 mM), indicating that using these signals to induce gene expression could be promising even in sub-optimal conditions.

In both K. variicola and S. stutzeri, our Cas9 knockout strategies had high rates of success. The one-step method for K. variicola demonstrated 93% efficiency for the creation of ΔnifL::araC-P_araBAD-nifA and 100% efficiency for creating the ΔnifH strains. Similarly, the two-step method showed 67% deletion efficiency for creating a ΔnifL::J23101 strain in S. stutzeri and a 100% efficiency for creating a ΔnifH strain, indicating that our λRed/Cas9 two-plasmid system is usable in several diazotrophs for reliable editing. The editing efficiencies in S. stutzeri is comparable, though slightly lower on average, to the editing efficiencies seen in the two-step protocol previously in P. putida, where editing efficiencies were approximately 93% across 4 deletions. K. variicola seems to show more robust editing efficiencies using the one-step protocol compared to P. putida, as the one-step protocol only had an editing efficiency of approximately 5% when used prior.⁵⁰

Heterologous Production of Alkylresorcinols in K. variicola

With K. variicola as our most promising chassis organism for rapid genome editing and reliable expression, we demonstrated gain-of-function in this organism using genomic insertion of the sbARS2 gene from Sorghum bicolor as a first step towards generating relevant bioactive rhizosphere compounds. The sbARS2 gene encodes a type III PKS that generates alkylresorcinol products from the condensation of long-chain CoA species with malonyl CoA. In S. bicolor, generating this alkylresorcinol is a key step in the pathway to generate sorgoleone, a desirable bioactive end product.⁵⁶ In the rhizosphere, sorgoleone acts as both an herbicide as well as a nitrification inhibitor, making it a desirable product to improve fertilization efficiency for crops.^28,29 By integrating sbARS2 into the genome of K. variicola under arabinose induction (genotype yciA-araC-P_araBAD-sbARS2), we demonstrated GC-MS peaks consistent with the production of resorcinol products that are absent in the wild-type and uninduced strains (Figure 6). These alkylresorcinols are also secreted into the supernatant, indicating that they would be released into the rhizosphere environment by our engineered K. variicola strain. Due to the promiscuity of sbARS2 for different CoA species, we saw a variety of alkylresorcinol chain lengths at different retention times (Figure S7). The heterologous production of alkylresorcinol products is the first major committed step to generate sorgoleone as well as other metabolites in K. variicola, which could be beneficial in the rhizosphere environment for engineering novel plant-microbe interactions. The lack of an antibiotic marker also facilitates repeated integrations, allowing for several engineering strategies to be employed at once for maximizing the benefits of engineered bacteria. Through this proof-of-concept integration, we observed 29% efficiency for the integration of sbARS2.

Figure 6. Alkylresorcinol generated in K. variicola eluted between 14 and 21 minutes with sbARS2 integrated into the genome.

(a) The native alkylresorcinol synthesis reaction catalyzed by sbARS2 in S. bicolor. (b) In the supernatant extract of K. variicola, the MS fragmentation pattern shown was not seen in uninduced samples and indicates production of C14 alkylresorcinols as the dominant product, with a small amount of C12 and C16 products also produced (smaller peaks highlighted in yellow). The chemical structure from a similarity search of the NIST database is shown.

Discussion

Working with non-model microbes often requires several rounds of trial and error to find systems that are robust for different organisms, and diazotrophs are no exception. The variability of expression from broad-host plasmids, even when taken from close relatives, makes it difficult to predict which plasmids will yield the best expression in different organisms. This inconsistency slows progress when evaluating multiple organisms simultaneously, as many parts need to be redesigned at each step; for example, BBR1 was the only reliable plasmid origin for all three organisms in our hands. Even with transformable plasmids, expression strength remains varied, making transient plasmid systems difficult to tune for many different cases or organisms. Despite starting with a highly expressed promoter in all three organisms, as was the case with the promoter library, there was a wide range in responses to promoter modulation even with a strong initial promoter. For example, the original sequence was the strongest library member in A. vinelandii, but middling in K. variicola. Evaluating modular synthetic biology parts in several organisms eases the onboarding of heterologous expression cassettes for future work; the efficacy of these initial tools in an organism is often a harbinger of future expression, allowing for the selection of an ideal chassis.

Genome integration remains a reliable strategy for ensuring stability, and the Cas9/λ-Red strategy presented here is a reliable counterselection to avoid background wild-type cells without the use of antibiotic markers. Cas9 positive selection can also be propagated over several passages for complete removal of wild-type cells or heterozygous colonies, which allows for more robust genomic engineering and iterative knockouts that can recycle antibiotic cassettes. However, since Cas9 is constitutively expressed in our plasmids, these systems may not be reliable for organisms with higher basal expression, as Cas9 expression can be toxic at high levels even without a gRNA.^61,62 Having a dual-inducible system may allow for this system to be portable to other organisms, streamlining onboarding of new strains. While we attempted to create one-plasmid, inducible systems, we were ultimately unable to provide a strong enough selective pressure to create homozygous knockouts, and heterozygous colonies reverted to wild-type when patch plated over several generations. Successful transformation of S. stutzeri with the Cas9 and pgRNA editing plasmids was also less reliable, and transformations often needed to be repeated to yield edited colonies. The one-step protocol for Cas9/λ-Red recombineering in S. stutzeri yielded no colonies when attempted.

Nonetheless, using previously studied broad host plasmids and promoters, we were able to successfully transform and evaluate the performance of several constitutive and inducible expression systems in three different soil microbes. From these initial tests, we were further able to construct systems for scarless genome integration in two of these nitrogen-fixing organisms. Additionally, we demonstrated the validity of these knockouts by modulating the nitrogen fixation activity of these strains. Finally, as a proof-of-concept for functional gene insertions, we produced alkylresorcinol metabolites as potential sorgoleone precursors in K. variicola through scarless genome integration, which may be useful for slowing denitrification or deterring growth of fungal pathogens in the soil environment. Future work could involve developing chain-length specificity and quantifying release of these alkylresorcinols into the media, as well as developing bacteria that respond to relevant root exudates for robust fixation and association. Synthetic biology tools for diazotrophs will be essential for modulating their nitrogenase activity and secondary metabolite production in engineered rhizospheres.

Methods

Plasmids, bacterial strains, and growth conditions

The bacterial strains and plasmids used in this study are shown in Table S1 and Table S2, respectively. E. coli DH5α, K. variicola D4, and S. stutzeri A1501 were grown in LB medium at 37°C. For antibiotic selection, LB medium was supplemented with kanamycin (50 μg/mL, Kan50), gentamicin (30 μg/mL, Gent30), carbenicillin (34 μg/mL, Cm34) or tetracycline (10 μg/mL, Tet10) for plasmid maintenance. A. vinelandii was grown in Burk’s Medium (0.89 g/L K₂HPO₄, 0.22 g/L KH₂PO₄, 20 g/L sucrose, 0.2 g/L MgSO₄•7H₂O, 0.119 g/L CaCl₂, 7.35 mg/L CuSO₄, 69.4 mg/L Na₂B₄O₇•10H₂O, 48.4 mg/L MnSO₄•2H₂O, 2 mg/L Na₂MoO₄•2H₂O, 29.9 mg/L ZnSO₄•7H₂O, 14.4 mg/L FeSO₄•7H₂O, 19.3 mg/L Na₂EDTA•2H₂O) at 30°C. Burk’s medium was supplemented with kanamycin (0.5 μg/mL, Kan0.5) and carbenicillin (50 μg/mL, Carb50) for plasmid maintenance in A. vinelandii, as well as with NH₄OAc (15 mM) for testing constitutive and inducible expression from plasmids in A. vinelandii. Burk’s semisolid media (Burk’s + 1% agar) was used for acetylene reduction assays for K. variicola and S. stutzeri.

Isolation of sorghum diazotrophs and plant growth-promoting trait characterization

Wild-type diazotrophs were isolated on nitrogen-free, solid artificial mucilage media (Fahraeus medium supplemented with 113.7 μg/mL arabinose, 354.4 μg/mL glucose, and 377 μg/mL fucose) by streaking mucilage onto plates, incubating at 28°C for 2-4 days, and using 16S primers to identify individual colonies (forward primer 5′-ACTCCTACGGGAGGCAGCAGT, reverse primer 5′-TACGGTTACCTTGTTACGACTT). Molecular confirmation of nifH (IGK3- GCIWTHTAYGGIAARGGIGGIATHGGIAA, DVV- ATIGCRAAICCICCRCAIACIACRTC) was performed as described for all organisms.³³

All strains were assayed for the following plant growth promoting traits: auxin production using Salkowski’s reagent,³⁴ organic and inorganic phosphate solubilization,³⁵ siderophore production,³⁶ and 1-aminocyclopropane-1-carboxylate (ACC) deaminase activity.³⁷ For auxin production, strains were incubated overnight at 28°C in TY broth supplemented with additional tryptophan before being incubated with Salkowski reagent and quantifying indole on a Biotek plate reader. For phosphate solubilization, strains were incubated on plates at 28°C of 10 g/L glucose, 0.5 g/L (NH₄)₂SO₄, 0.5 g/L yeast extract, 0.3 g/L NaCl, 0.3 g/L KCl, 0.03 g/L FeSO₄ * 7 H₂O, 0.3 g/L MgSO₄ * 7 H₂O, and 0.03 g/L MnSO₄ * 4H₂O for 2-4 days supplemented with organic or inorganic sources of phosphate (0.2 g/L lecithin with 1 g/L CaCO₃ or 5 g/L Ca₃(PO₄)₂, respectively). Production of a cleared plate halo (solubilization positive) was recorded. For siderophore production, strains were incubated on TY-CAS/HDTMA plates for 48 h at 28°C and production of orange halos (siderophore positive) was recorded. For ACC deaminase activity, strains were incubated at 28°C for 2-4 days on plates with ACC as the sole nitrogen source (0.1% sucrose, 0.1% 1-aminocyclopropane-1-carboxylic acid (MP Biomedicals), 0.1% KH₂PO₄, 0.1% K₂HPO₄, 0.05 % MgSO₄, 0.013% CaCl₂, 0.013% FeSO₄), and growth was recorded.

Molecular Techniques

For cloning, plasmids were isolated using a QIAprep Spin Miniprep Kit, and oligonucleotides were purchased from IDT. Q5 High-Fidelity 2X Master Mix (NEB) and GoTaq Green (Promega) were used for PCR amplification and colony PCRs, respectively. Gibson Assembly⁵⁷ was used to create knockout and insertion plasmids for genomic modifications, and these plasmids were subsequently used as PCR templates to create linear repair templates for genome editing. Blunt end ligation (T4 PNK/T4 DNA Ligase, NEB) was performed to swap out guide sequences in gRNA plasmids. The sbARS2 coding sequence was codon-optimized from the Sorghum bicolor sequence (accession number XM_002441794) using an online codon optimization tool (Integrated DNA Technologies). The optimized sequence was synthesized with an N-terminal histidine-tag (Twist Biosciences). Plasmids and colony PCRs were sequenced either through Sanger Sequencing (Functional Biosciences) or through next-generation sequencing (Plasmidsaurus).

Transformation

All transformation methods were adapted from Choi and Schweizer.⁵⁸ For the transformation of K. variicola, A. vinelandii, and S. stutzeri, 3 mL of stationary phase cultures were spun down at 7,000 x g for 2 minutes. Cells were washed twice with either ice-cold 300 mM sucrose (A. vinelandii and S. stutzeri) or 10% glycerol (K. variicola), then resuspended in 300 μL of their respective wash solutions. 50 μL of washed cells were mixed with 1-5 μL DNA (100 ng/μL - 1 μg/mL). The mixture was transferred to 1 mm electroporation cuvettes and electroporated using a Bio-Rad Micropulser set to 1.8 kV (Ec1 preset). Cells were recovered in 1 mL LB media (K. variicola or S. stutzeri) or 1 mL Burk’s medium (A. vinelandii) for 1 hour following electroporation, then 100-200 μL of cells were spread on Burk’s media plates with the appropriate antibiotics. Transformation efficiencies for different plasmid backbones were calculated by dividing the average number of CFUs in the outgrowth by the amount of DNA added to the cuvette.

Additionally, triparental mating was used to transform S. stutzeri A1501 with an integrated suicide vector for Cas9 editing. S. stutzeri harboring pMVV205 (pBBR1-λRed/Cas9) were grown overnight in 5 mL LB-Gent30, along with E. coli cells harboring the suicide vector (a ColE1 backbone grown in LB-Kan50) and E. coli cells harboring conjugation helper plasmid pRK600 (LB-Cm34).⁴³ The following morning, 100 μL was taken from each stationary phase culture and mixed with 700 μL LB. Cells were spun down at 7000 x g for 2 minutes and resuspended in 1 mL fresh LB twice. Finally, the pellet was resuspended in 50 μL LB and placed on a 13 mm mixed cellulose ester membrane on an LB plate. Cells were grown at 37°C for 8 hours, resuspended in 200 μL sterile PBS, and plated on LB Gent30/Kan50 to select for S. stutzeri cells with a successful single-crossover event of the antibiotic marker.

Quantifying Promoter Expression

Strains containing sfGFP expression plasmids were grown overnight in LB Kan50 for S. stutzeri and K. variicola or for 72 hours in Burk’s medium supplemented with Kan0.5 for A. vinelandii. After outgrowth, stationary phase cells were inoculated at an OD600 of ~0.1 in 200 μL fresh LB or Burk’s+NH₄OAc with appropriate antibiotics in a Corning 96-well plate with black walls and flat clear bottom. Samples were incubated on a shaking platform for 24 hours (K. variicola and S. stutzeri), or for 60 hours (A. vinelandii), then OD600 and fluorescence measurements were taken using an M1000 Tecan plate reader, using excitation/emission values of 485/510 nm for samples expressing sfGFP at a manual instrument gain of 90. Fluorescence/OD600 values are reported at 24 hours for K. variicola and S. stutzeri and 60 hours for A. vinelandii.

Genome editing in K. variicola and S. stutzeri

For genome editing of K. variicola, cells were transformed using a one-step λRed/Cas9 recombineering protocol. K. variicola cells containing pJMR52 (pRK2-λRed/Cas9) were grown overnight in 5 mL LB-Gent30. The following day, the expression of the λRed genes was induced with 0.2% L-arabinose, and cells were incubated for 45 min at 37°C. These cultures were used to prepare electrocompetent cells, as described above. Cells were transformed with ~500 ng pgRNA plasmid and ~500 ng linear repair template by electroporation and allowed to recover in 1 mL LB for 3 h at 37°C. The recovered cells were selected on LB Gent30/Tet10. Transformants were screened for the desired knockout or insertion using colony PCR with primers flanking the gene of interest. After verifying the presence of the mutation, cells were passaged with antibiotics for the gRNA and Cas9 plasmids for 48 hours to remove the remaining wild-type DNA. Following passaging, all positive hits were screened for the desired knockout or insertion. Colonies were also screened with a secondary colony PCR to check for the complete removal of wild-type DNA. Plasmids were cured from K. variicola by growing the cells overnight for 48 hours in LB media without antibiotics and plating on LB agar. Single colonies were screened for loss of plasmid-based antibiotic resistance by patch plating.

For genome editing in S. stutzeri, cells were edited with a two-step recombineering protocol. Cells containing pMVV205 (pBBR1-λRed/Cas9) were grown overnight in LB-Gent30. In the first step, S. stutzeri cells were then transformed with a suicide plasmid containing the repair template of interest using conjugation, as described above. These colonies were subsequently grown overnight in 5 mL LB-Gent30, Kan50, then induced with 0.2% L-arabinose and incubated for 1 hr at 37°C. Cells were transformed with 1 μg pgRNA plasmid, then recovered for 4 hours in 1 mL LB at 37°C before plating on LB Gent30, Tet10. Transformants were subsequently screened and cured of plasmids identically to K. variicola. Guide sequences for Cas9 gRNAs in both strains were chosen with CasOT to minimize off-target activity.⁵⁹

Acetylene Reduction Assay

Nitrogenase activity was estimated by acetylene reduction assay (ARA) gas chromatography using a GC-2010 (Shimadzu) equipped with a 60 m Rt-Alumina BOND/KCL column (Restek). K. variicola and S. stutzeri were grown in LB medium overnight. Bacteria were collected by centrifugation (5 minutes, 3000 × g at 4°C) and suspended in 1 X PBS buffer, then resuspended to 0.1 at 600 nm (OD₆₀₀=0.1). An aliquot of 3 mL of Burk’s semi-solid medium (Burk’s + 1% agar) was added to 10 mL vials (Supelco-27303U). The medium was inoculated with 30 μL of the cells in PBS. Inoculated and control vials were incubated at 30°C for 72h. From each vial, 1 mL of air in the headspace was removed with a syringe, 1 mL of acetylene (99% v v⁻¹) was injected, and the vials were incubated at 30°C for 16 hours. The ethylene produced by nitrogenase activity was quantified using a standard curve with known ethylene concentrations. The protein content in the Burk’s semi-solid medium was determined using the Bradford’s reagent and quantified using a standard curve of albumin.⁶⁰

Extraction of Alkylresorcinols

To produce alkylresorcinols, K. variicola strain yciA::sbARS2 was grown overnight. The following morning, 5 mL of LB media were inoculated to an OD₆₀₀ of ~0.05 and induced at an OD₆₀₀ of 0.7 with 0.6% w/v l-arabinose. 24 hours after inoculation, cells from the 5 mL culture were collected via centrifugation and separated from the supernatant. The metabolites from the cell-free supernatant were extracted into 4 mL ethyl acetate. After vortexing the ethyl-acetate/supernatant mixture vigorously for 3 minutes, the mixture was placed on a shaking platform for 2 hours, then centrifuged at 1000g for 10 minutes. The top layer of ethyl acetate was removed and evaporated under reduced pressure. The product was resuspended in 200 μL of fresh ethyl acetate for GC-MS analysis.

Gas chromatography/mass spectrometry (GC/MS) analysis of alkylresorcinols

Gas chromatography/mass spectrometry (GC-MS) was performed on a coupled Shimadzu GCMS-QP2010S. 1 μL of alkylresorcinol extracts were injected into an HP-5ms Ultra Inert column (5% phenyl-methylpolysiloxane; 30 m length; 0.25 mm inner diameter; 0.25 μm film thickness; Agilent) with a split ratio of 1:10. The extracts were separated using the following temperature program: 120 °C (2 min), 16 °C min-1 to 200 °C, 6 °C min-1 to 320 °C, 320 °C (15 min). The column flow rate was 1.50 mL min-1 with an injection temperature of 320 °C. Helium was used as the carrier gas with an inlet pressure of 1.2 bar. The mass spectra were recorded at an ion source and interface temperature of 230 °C and 275 °C, respectively. The spectra were collected from the m/z range of 50-500.