Quantitative Dissection of Agrobacterium Virulence to Generate a Synthetic Ti Plasmid

Abstract

Agrobacterium is not only a costly plant pathogen but is also an essential tool for plant transformation. Though Agrobacterium-mediated transformation (AMT) has been heavily studied, its polygenic nature and complex transcriptional regulation make identification of the genetic basis of transformational efficiency difficult through traditional genetic and bioinformatic approaches. Here, we use a bottom-up synthetic approach to systematically engineer the tumor-inducing plasmid (pTi), wherein the majority of virulence machinery is encoded. Using a validated toolkit to control Agrobacterium gene expression in planta, we perform a quantitative dissection of AMT to investigate the contributions of critical vir-genes at different expression levels. We construct a synthetic pTi capable of transient plant and stable fungal transformation and characterize bottlenecks and solutions for complex polygenic synthetic pTi designs. Our reductionist approach demonstrates how bottom-up engineering can be used to dissect and elucidate the genetic underpinnings of complex biological traits, laying the foundation for future engineering to establish full synthetic control over the critical process of AMT.

Introduction

The genetic basis of pathogenesis is challenging to study due to its highly polygenic nature as well as the influence of both host and environmental factors. While advances in comparative and functional genomics have generated myriad hypotheses on how virulence and adaptations to specific hosts evolve, , it is still challenging to isolate and validate specific genetic features that determine these traits. In an ideal system, one would be able to systematically evaluate and build a holistic understanding of how each gene contributes to and influences virulence. However, epistatic effects often complicate the conclusions drawn from traditional top-down approaches that rely solely on knockouts and complementation.

As an alternative, bottom-up approach, synthetic biology enables the introduction of synthetic regulatory control on a defined set of genetic elements. This is crucial for two major reasons. First, the development of minimal, controllable systems allows for specific hypotheses to be tested to better understand how evolution has solved a myriad of problems. Second, this knowledge allows for subsequent data-guided engineering to optimize and leverage the system for biotechnological purposes. Such a strategy has been widely implemented in reconstituting relatively linear metabolic pathways, , but apart from a few notable exceptions, it has rarely been applied to more complex biological phenomena because of the numerous and tremendous intrinsic challenges associated with reconstituting complex biological traits in a reductionist manner. , To reverse engineer complex traits, one must identify the genes necessary and sufficient for a given biological process, as well as have the appropriate genetic tools applicable to the organism of study. Given these significant hurdles, many initial designs from genetic engineering often perform poorly compared to their native systems, but nonetheless offer unique insights into the underlying facets of complex biological traits. ,

A challenge unique to studying the genetic bases of pathogenesis is that any synthetic regulatory elements utilized must be robust in situ, i.e., in the context of the various environments that the pathogen faces during infection, where few genetic toolkits have been rigorously validated. Despite these challenges, work with both plant- and mammalian-associated bacteria has demonstrated that synthetic genetic constructs can be introduced to promote non-native interactions, , indicating the feasibility of a complete synthetic engineering of pathogenesis. Nonetheless, genetically recapitulating complex biological phenomena within a host-associated environment has largely remained out of reach.

Plant pathogenic members of the Rhizobiaceae family (hereafter referred to as Agrobacterium or agrobacteria) are capable of causing crown gall or hairy root diseases and have been extensively studied due to their unique mechanisms of virulence. Virulence involves genetic transformation of eukaryotic hosts, which has been leveraged for many critical biotechnological uses, e.g., plant transgenesis. Central to virulence is an oncogeneic tumor-inducing plasmid (pTi) that carries a ″transfer DNA” (T-DNA) and vir genes. The hallmark of A. tumefaciens virulence is the transfer of a protein-conjugated, single-stranded DNA molecule into host cells and the integration of this DNA into the genome. When genes from this T-DNA are expressed by plant cells, the gene products synthesize phytohormones that result in the formation of a tumor. The infecting bacterial population is hypothesized to gain a fitness advantage in the tumor because of access to novel nutrients, whose synthesis in the host is catalyzed by enzymes encoded on the T-DNA. When scientists domesticated virulence by swapping the tumorigenic genes within the T-DNA region with genes of interest, a new era of plant genetics was ushered in. Today, the T-DNA borders and genetic payloads to be delivered are most often housed on a smaller plasmid referred to as a binary vector, enabling easy genetic manipulation through Agrobacterium-mediated transformation (AMT) of multitudes of plant and fungal species. − However, many agriculturally important crops still remain difficult to transform. Thus, there remains an imperative to develop novel strains of Agrobacterium that will enable scientists to expand the genetic potential of plants.

Our basic understanding of AMT and nearly all of the engineered agrobacterial strains used for AMT are derived from a limited number of A. tumefaciens strains and pTi variants. Yet, it has long been recognized that interactions among strains, Ti plasmids, and host species influence the efficiency of AMT. By mining this natural diversity, strains with improved plant transformation properties for different plant species have previously been developed. , More recently, groups have developed strains that contain additional vir alleles, harbored either on the binary vector (superbinary vectors) or on an additional stand-alone plasmid (ternary vectors). − These strains demonstrate that altering the regulation of vir genes can enhance transformation of otherwise recalcitrant plants. − Precisely how these tripartite interactions influence transformation efficiency remains largely unknown. The high number of possible genetic interactions required for AMT complicates research efforts to improve transformation by A. tumefaciens. Complicating studies further, oncogenic plasmids vary in the composition and sequence of vir genes, the regulation of these genes, , and chromosomal genes implicated in virulence vary in sequence across agrobacterial strains. − Furthermore, the impact of changes in expression level between different vir genes is somewhat masked by a master regulator, VirA/G, which controls the expression of all known vir genes. − This epistatic regulatory schema makes it difficult to evaluate whether differences in virulence are a consequence of the presence of a specific vir gene or its strength of expression. , Thus, to fully capture the impact of these many individual genetic variables on AMT, a bottom-up synthetic genetic approach is required to precisely control genetic interactions and systematically evaluate the contribution of each gene to the transformation. However, due to the sheer size of combinatorial genetic space within pTi and the technical challenges associated with synthetic engineering of complex biological phenomena in planta, no such effort has been reported.

Despite the many technical challenges associated with engineering synthetically encoded AMT, a deeper understanding of this complex process may elucidate molecular constraints to the transformation of plants. Here, we overcome these challenges by (1) developing a set of genetic tools that allow for the reliable control of agrobacterial gene expression within the plant environment, in order to (2) quantitatively characterize the genetic determinants underlying AMT efficiency, to ultimately (3) design synthetic vectors, divorced from native regulation, capable of AMT. This work represents a critical first step in better understanding AMT as we lay the framework for understanding highly specific genotype-to-phenotype connections underlying a complex host-microbe interaction.

Results and Discussion

Developing a Genetic Toolkit to Control Bacterial Gene Expression in Planta

A recurring challenge in synthetic biology has been translating genetic circuits developed in vitro into more heterogeneous environments in situ. Environmental changes can have dramatic impacts on genetically engineered organisms, as demonstrated in scale-ups to large fermentative tanks or living medicines in patients. , Many in vitro synthetic biology designs take advantage of small-molecule-inducible promoters, which offer a range of expression options from a single design, compared to static expression levels from a constitutive promoter. However, dynamic environments such as plant tissue may interfere with inducible promoter systems by making signaling molecules biologically unavailable through degradation or sequestration, limiting their utility. Recent work by multiple groups have characterized inducible promoters in Agrobacterium, though not all were evaluated while the bacteria were in planta; moreover, there has been a dearth of well-characterized constitutive promoters in Agrobacterium. −

To better understand how to control bacterial gene expression within plants, we evaluated the activity of 16 synthetic constitutive and 4 inducible promoters in bacterial cells grown in rich media, as well as infiltrated into the leaf tissue of Nicotiana benthamiana and Arabidopsis thaliana. In both plants, bacterial constitutive promoter activity correlated highly between leaf tissues as well as between observed in vitro and in planta activity (Figures A and S1). We then chose five constitutive promoters with a range of expression strengths (P_J23114, P_J23117, P_J23101, P_J23100, and P_J23111) to complement a virE1+virE2 (hereafter referred to as virE12) deletion mutation in the common Agrobacterium fabrum (formerly A. tumefaciens) laboratory strain GV3101:pMP90 (hereafter simply referred to as GV3101), which is derived from A. fabrum C58. Previous work using inducible promoters showed that transformation of N. benthamiana was highly sensitive to virE12 expression. Using transiently expressed GFP from a medium-strength plant promoter in N. benthamiana as a measure of transformation, we observed that all promoters stronger than the weakest, P_J23114, were able to complement leaf transformation back to wild-type levels (Figure B). Proteomics analysis of the ΔvirE12 complementation strains confirmed that the expression of VirE12 correlated with RFP expression from the same promoters (Figure C). These data reveal that relatively weak constitutive promoters may be sufficient to reconstitute vir gene expression from the pBBR1 origin plasmids.

1.

Characterizing a synthetic biology toolkit in planta: (A) Activity of constitutive promoters driving RFP from pGingerBK plasmid backbone. The x- and y-axes show RFP production from A. fabrum C58C1 3 days after infiltration in N. benthamiana or Arabidopsis, respectively (n = 12). The color palette displays the activity of the same promoter in vitro (n = 8). (B) Transient expression of GFP from agro-infiltrated N. benthamiana leaves (AU) log2 transformed is shown on the y-axis. Different constitutive promoters used to complement a virE12 deletion mutant are shown as box and whisker plots with individual data points overlaid (n = 64). Transformation by wild-type GV3101 and N. benthamiana leaf without infiltration controls is shown. (C) Proteomic spectral counts of VirE12 are shown when virE12 is expressed from different constitutive synthetic promoters in vitro (n = 3). Rows (D–G) show characterization of P_LacO, P_TetR, P_NahR, P_{JungleExpress}, respectively. The left panel shows activity of inducible promoters driving RFP from pGingerBK plasmid backbone in N. benthamiana (n = 12), Arabidopsis (n = 12), and in vitro (n = 8). Inducer either not added (“None”0 mM IPTG), added at the half maximal induction concentration determined in vitro (“Mid”0.2 mM IPTG), or at the maximal induction concentration (“High”2.0 mM IPTG). The middle panel shows the complementation of a virE12 deletion by different inducible promoters as measured by transient GFP expression in N. benthamiana is shown on the y-axis after log2 transformation (n = 64). Inducer either not added (“None”), added at the half maximal induction concentration determined in vitro (“Mid”), or at the maximal induction concentration (“High”). Dashed lines in the middle panels represent wild-type GFP output. The right panel shows proteomic spectral counts of VirE12 when expressed from different inducible promoters (n = 3). Inducer either not added (“None”), added at the half maximal induction concentration determined in vitro (“Mid”), or at the maximal induction concentration (“High”).

Inducible promoters enable the dynamic control of gene expression strength and thus could reduce the number of genetic designs needed to evaluate the impact of gene expression on AMT. We evaluated the expression of RFP from four inducible promoter systems (P_LacO, P_TetR, P_{Jungle Express}, and P_NahR) in culture media as well as in N. benthamiana and A. thaliana leaves, where the inducing compound was mixed with a bacterial suspension before infiltration into leaf tissue. While each of these systems displayed inducible expression in culture media (Figure S2), only P_LacO and P_TetR showed consistent inducibility in both host plant environments­(Figures D–G and S3). Conversely, the P_{Jungle Express} promoter showed poor induction in both plant species, and P_NahR was expressed even in the absence of an added inducer within A. thaliana leaf tissue. These results demonstrate the importance of validating each promoter in its intended environment. For example, although functional in vitro, P_{Jungle Express} performed extremely poorly in planta. One explanation of this observation is that the crystal violet inducer of P_{Jungle Express} may be rapidly bound to plant tissue and thus is not biologically available. Conversely, the salicylic acid inducer of P_NahR can be endogenously produced by plants as an immune response to pathogens such as Agrobacterium, and thus may not be ideal for exerting orthogonal control of gene expression within different plants.

After testing all four inducible promoter systems to complement a ΔvirE12 mutation, only the LacI inducible promoter, with the highest amount of added ligand tested, was able to recover transformation back to wild-type levels (Figure D). As P_LacO showed the best plant orthogonality and ability to complement a virE12 mutation, further designs requiring inducibility utilized this IPTG-inducible promoter. While proteomics from cultures indicated that the levels of VirE2 expressed from inducible promoters were similar to the constitutive promoters (Figure C–G), the plant-specific utility of individual promoters suggests that they may be less useful for designing reliably functioning genetic circuits across plant environments.

A Quantitative Understanding of the Genetic Contributions to AMT across Plants and Fungi

To systematically assess the contributions of individual vir genes to plant transformation, we developed a quantitative virulence assay to measure the efficiency of T-DNA transfer into plant cells. To accomplish this, we first generated internal, in-frame deletion mutants of known functional nonregulatory vir gene clusters in A. fabrum GV3101: virB1–11, virC1+virC2 (hereafter referred to as virC12), virD1+virD2 (hereafter referred to as virD12), virD3, virD4, virD5, virE12, virE3, virF, virH1, virH2, and virK (Figure A). Using a transient GFP expression assay in N. benthamiana leaves, we observed that the deletion of virB1–11, virC12, virD12, virD4, or virE12 resulted in over 90% reduction in transformation efficiency (Figure B). Furthermore, loss of virD5, virE3, virH1, virH2, or virK significantly reduced transformation efficiency compared to wild-type, while deletion of virD3 or virF showed no significant reduction in transformation efficiency (Figure B). Plasmid complementation of these deletions using the relatively weak promoter P_J23117 on a pGingerBS backbone plasmid restored wild-type transformation efficiencies in all deletion strains except for virB1–11, virD4, virD5, and virK (Figure B). These results serve as a benchmark that, for the first time, allow for a relative comparison of vir gene importance in AMT.

2.

Quantitative assessment of vir gene impact on transformation: (A) The virulence gene cluster from pTiC58. Known nonregulatory vir genes are shown in red, while regulatory vir genes are shown in blue. All other genes are gray. (B) Effect of individual vir gene cluster deletion on N. benthamiana transformation is measured by transient expression of GFP shown in log2 transformed AU in white, and complementation of the phenotype driven by P_J23117 is shown in black (n = 64). The dashed gray line shows transformation by wild-type GV3101, while the dashed black line shows transformation by GV3101 expressing RFP from P_J23117 as a control. (C) Picture of a N. benthamiana leaf expressing GFP delivered by a virD12 deletion mutant complemented from an IPTG-inducible promoter with varying levels of induction indicated. Below shows transient GFP measurement with different IPTG levels, shown in log2 transformed AU on the y-axis, with the concentration of IPTG used to induce the promoter on the x-axis (n = 64). The dashed line shows the GFP level of N. benthamiana leaves infiltrated with wild-type GV3101. (D) Complementation of vir gene deletion mutants that showed trends from Figure S4 using different strength constitutive promoters. Transient N. benthamiana-expressed GFP shown in log2 transformed AU is shown on the y-axis (n = 64). The dashed line shows wild-type level transformation, colors of the box plots represent the strength of constitutive promoters from Figure A. (E) Effect of individual vir gene cluster deletions on R. toruloides transformation as measured by total number of transformants (n = 4).

To explore the effect of different expression levels on the transformation efficiency, we then complemented each mutation with the inducible P_LacO promoter. Three phenotypes were observed: (1) the virB1–11, virC12, virD12, and virE12 complementation strains had increasing transformation efficiency with increased induction; (2) the virD3, virD4, and virD5 complementation strains had decreasing transformation efficiency as induction increased, and (3) the virE3, virF, virH1, virH2, and virK complementation strains showed no response to increasing induction (Figures C and S4). Some of the relative decreases in transformation observed as virD5 expression increases may be due to toxicity to the bacterium; however, similar toxicity was not observed with increased expression of virD3 or virD4 (Figure S5). Previous work has shown that overexpression of virD5 resulted in acute toxicity in eukaryotes, where it is localized to the nucleus, and may cause DNA damage. , Based on these results, we used constitutive promoters stronger or weaker than P_J23117 to tune and optimize the expression of each vir gene cassette (Figure D). Strong expression of virD12 improved transformation compared to the wild type by 135%. These results are in line with previous reports that overexpression of virD12 improves transformation. Conversely, lower expression of virD4 improved transformation 72% over the wild type. There was no significant improvement of transformation by increasing the expression of virC12, though expression from the stronger P_J23100 and P_J23101 promoters decreased the transformation. Expression of virD5 and virK from the weak P_J23114 promoter was able to restore the wild-type level transformation efficiency. Overall, these results demonstrate that the transformation efficiency is highly sensitive to the expression strength of nearly all vir genes we evaluated, necessitating precise tuning for optimal DNA transfer.

Unlike other gene clusters, which were all complemented back to at least wild-type levels of transformation, we were able to achieve only ∼2% of wild-type transformation in a ΔvirB1–11 genetic background. Expressing virB1–11 from the strong P_J23101 promoter improved transformation over complementation using P_J23117. However, complementation from the strongest promoter tested, P_J23111, resulted in a significant reduction in the level of transformation, suggesting that high-level expression may be toxic to the bacterium. The virB operon encodes the type 4 secretion system (T4SS), and previous studies genetically reconstructing secretion systems have demonstrated the difficulty associated with engineering efficient transport, suggesting that expressing the T4SS may represent the bottleneck in engineering efforts.

In an attempt to improve virB complementation, we explored whether breaking the cluster into segments would improve our ability to complement the virB cluster. We knocked out virB1–5 and virB6–11 individually and attempted to complement these smaller mutations. Both of the smaller mutations predictably abolished the transformation (Figure S6A). Using the P_LacO inducible promoter, virB1–5 showed a linear improvement in transformation with increased IPTG concentrations. However, virB6–11 complementation plateaued at the median inducer concentration tested, with the highest level of induction causing a sharp decrease in transformation (Figure S6B). The decrease in transformation is likely due to the extreme toxicity associated with virB6–11 being expressed without the other T4SS genes, which greatly compromised growth (Figure S6C). Constitutive complementation assays revealed that the optimal promoters for complementing these deletions were the middle strength P_J23101 for virB1–5, which yielded ∼60% of wild-type transformation, and the relatively weak P_J23117 for virB6–11, which yielded ∼25% complementation (Figure S6D). Based on these data, we designed synthetic virB1–11 complementation vectors that express virB1–5 using three different promoters (weak-P_J23117, medium- P_J23101, and strong- P_J23100), and virB6–11 from the weak constitutive promoter, P_J23117, with this cassette both downstream and upstream of virB1–5 (Figure S7A).

None of these vectors could, however, complement as well as when virB1–11 was expressed in its entirety from a single promoter. The vectors driving virB1–5 from the strong P_J23100 performed particularly poorly (Figure S7A). To assess the performance of our synthetic complementation of virB1–11 against native P_virB, we cloned the entire virB1–11 operon in addition to its intergenic upstream and downstream DNA into a promoterless vector backbone. While this vector was able to complement transformation above the ΔvirB1–11 parent, it was still significantly less than both virB1–11 expressed from P_LacO and virB1–11 driven from P_J23101 (Figure S7B). These results may suggest that the virB cluster and other vir genes must be expressed from the same vector in order to maintain the proper stoichiometry for efficient transformation. Splitting the virB cluster into two separate parts (virB1–5 and virB6–11) performed worse than expressing the entire cluster together from a single promoter, suggesting keeping the native operon intact improves overall performance. By separating the genes, we could have disrupted the precise timing and protein ratios required to successfully assemble the Type IV secretion system. Thus, we kept the full virB1–11 cluster together in our final synthetic pTi designs. Future optimization efforts may need to avoid splitting the operon. Instead, better approaches might include fine-tuning the Ribosome Binding Sites for each gene on a single transcript, or using synthetic scaffolds designed to maintain strict protein ratios.

To explore how conserved the genetic requirements for transformation are across kingdoms and to gain insight into the evolutionary conservation of vir gene function, we evaluated the impact of individual vir gene deletions on the transformation of the basidiomycete yeast Rhodosporidium toruloides (Figure E). Similar experiments have been conducted previously in Saccharomyces cerevisiae and Aspergillus awamori, but these studies were limited to a subset of vir genes and did not comprehensively test the full complement of nonregulatory loci. Consistent with our results in in planta transient transformation, deletion of the virB1–11 cluster, virC12, virD12, and virD4 completely abolished fungal transformation. In contrast, virE12, which is nearly indispensable for plant transformation, was markedly less important in the fungal context, aligning with earlier reports showing that virE2 mutants can still mediate T-DNA transfer into fungal cells. , In contrast to previous reports, deletions of virD5, virF, and virH1 each caused significant decreases in the R. toruloides transformation. The impact of virD5 deletion was notable as it led to a ∼95% decrease in transformation (Figure E), more severe even than virE12 deletion, suggesting that VirD5 may play a critical role in promoting transformation within the fungal nucleus by possibly stabilizing or trafficking the T-complex through its kinetochore-binding activity. , In contrast, deletion of virD3 doubled the R. toruloides transformation efficiency by the GV3101 strain, representing the first reported phenotype for this gene (Figure E). While VirD3 has no proposed molecular function and has been reported to be dispensable for plant transformation, these results may explain why virD3 is maintained in only a subset of Agrobacterium lineages and is absent from many Rhizobium plasmids.

Screening Natural Diversity to Assess the Impact of Disparate vir Gene Homologues on AMT

In many synthetically engineered metabolic pathways, multiple homologues of an enzyme are evaluated to identify the optimal design to enhance flux toward the final product. To take a similar approach, we sampled the natural diversity of agrobacteria and systematically tested homologues of nonregulatory vir genes for their ability to improve transformation efficiency. It has recently been shown that at least 9 distinct lineages of pTi/pRi plasmids exist across the diversity of agrobacteria. , (Figure A). To measure the effect allelic variation plays in plant transformation, we synthesized phylogenetically diverse vir gene alleles from each of the 9 pTi/pRi (Table S1) families and evaluated their ability to complement GV3101 deletion mutants of virB1–11, virC12, virD12, virD4, virD5, virE12, virE3, virH1, virH2, and virF in a N. benthamiana transient expression system (Figure B–K).

3.

Impact of vir gene allele on N. benthamiana transformation: (A) Cartoon showing the localization of different vir gene products within the bacterial and plant cell during the AMT process. Gene products colored blue represent vir gene alleles that outperformed the wild-type C58 allele in complementation assays. (B–K) Plots showing the effects of alleles of the indicated gene clusters in A. fabrum GV3101 deletion mutants complemented with constitutive promoters from a BBR1 origin plasmid. Box plots in orange show alleles that are statistically worse than the C58 allele, box plots in blue show alleles that are statistically superior than the wild-type allele, and box plots in white show alleles that are statistically indistinguishable from the wild-type allele. Statistical significance was determined using a Bonferroni-corrected t test (p-value < 0.05, n = 64).

Of these clusters, we identified replacement alleles of virC12 (91% improvement), virD4 (13% improvement), virD5 (35% improvement), and virE3 (76% improvement) that resulted in improved complementation compared to that of the wild-type allele (C58). For virD5, 4 out of 9 alleles exceeded the wild type (Figure F). For virE3, 4 out of 9 tested also improved transformation compared to the native C58 allele (Figure H). However, for the critical vir genesi.e., virC12 (Figure C), virD12 (Figure D), and virE12 (Figure G)the majority of homologues significantly reduced transformation. These results suggest that while homologues exist that can potentially improve transformation rates, AMT relies on multiple interactions between vir genes. Co-evolution between vir genes may limit the compatibility of distantly related homologues.

To more specifically test whether phylogenetic distance from the wild-type allele impacts the ability of a vir gene to function in a non-native system, we correlated phylogenetic distance to the ability of a homologue to complement the C58 deletion mutant. With the exception of virE12, there were no significant correlations between the phylogenetic distance and ability to complement (Figure S8). Surprisingly, distantly related alleles were able to complement many of the vir gene mutants to the level of the wild-type allele. Many of the vir genes appear to be under purifying selection (d _N/d _S < 1) across much of their coding sequences (Figure S9). This selective pressure may conserve critical residues needed for protein–protein interactions across evolutionary times, but further analysis will be required to identify whether such residues exist.

Given that we identified multiple homologues across 4 vir gene clusters that could improve transformation, we then asked if these homologues could be combined to further improve transformation. To this end, we generated a suite of plasmids, called pLoki, that contained either the critical genes virC12, virD12, virD4, and virE12 (pLoki1) or these critical genes with the addition of virD5 and virE3 (pLoki2) (Figure S10A–C). We constructed a total of 20 variants of both pLoki1 and pLoki2 that explored all possible combinations of both wild-type alleles (C58) and the alleles of virC12, virD4, virD5, and virE3 that performed best from our initial screen under the control of promoters that optimally complemented deletions. These vectors were used to complement a deletion that spanned from virC2 to virE3 in A. fabrum GV3101 (Figure S10D).

The pLoki1 variant containing all C58 alleles restored ∼25% of the wild-type transformation in a transient N. benthamiana expression assay, whereas the pLoki2 variant containing all C58 alleles restored ∼65% (Figure S10E). Across all pLoki variants, none that contained a non-native allele outperformed the pLoki plasmids that contained only C58 genes (Figure S10D). However, looking across pLoki variants, we observed that vectors containing virD5 derived from pTiBo542 were significantly superior to those harboring the wild-type allele (Figure S10F), although the improvement was minimal and not superior to the all wild-type vector. Strains that contained virD4 from pTiBo542 or virE3 from pTiT60/94, however, were both worse than strains with the corresponding wild-type allele (Figure S10G,H). Further understanding the molecular basis and evolutionary constraints in swapping vir genes may help to direct future studies in harnessing the natural diversity of vir genes to improve AMT.

Engineering a Synthetic pTi

To engineer predictable phenotypic control over AMT, the genotypic and regulatory makeup of a synthetic pTi must be composed of a defined set of genes expressed from promoters that are robust to the regulatory influence induced by the plant environment. Based on our quantitative assessment of vir gene importance for N. benthamiana transformation (Figure ) and using previously optimized promoters (Figure ), we first sought to identify the minimal set of genetic elements capable of plant transformation. Our initial design (pDimples0) contained a minimal set of essential vir genes (virB1–11, virD12, and virD4) based on both our findings and previous work, with the virB genes expressed as a single operon controlled by P_LacO, and the other genes controlled by optimally determined constitutive promoters (Figure A). This vector was then introduced into A. fabrum C58C1, a strain of A. fabrum that has been cured of its pTi, also harboring a binary vector containing a plant GFP reporter on the T-DNA.

4.

Synthetic engineering of pTi: (A) Genetic design of pDimples vectors. Variants of pDimples1 that only have either virC12 or virE12 (pDimples0.5), as well as variants of pDimples2 that have only virE3 or virD5 (pDimples1.5) were also constructed. (B) GFP produced by transient transformation of N. benthamiana leaves via synthetic pTi plasmids harbored in A. fabrum C58C1 (n = 64). The Y-axis shows log2 transformed GFP (AU). (C) Fluorescence microscopy of 6 mM N. benthamiana leaf punches infiltrated with A. fabrum C58C1 harboring minimal synthetic pTi plasmids as well as a binary vector for plant expression of a nuclear-localized mScarlet. (D) Transformation of R. toruloides by synthetic pTi plasmids harbored in A. fabrum C58C1. The average number of transformants obtained in three transformations is shown by synthetic pTi strains, as well as a wild-type strain of A. fabrum GV3101 and A. fabrum C58C1 harboring a binary vector but no synthetic pTi. Error bars show standard error. (E) Fluorescence microscopy of 6 mm N. benthamiana leaf punches infiltrated with R. rhizogenes D108/85 harboring a binary vector for the expression of a nuclear-localized mScarlet without (top) or with (bottom) pDimples1.0.

When this strain was infiltrated into N. benthamiana leaves, there was no measurable increase in GFP signal compared to leaves infiltrated with avirulent A. fabrum C58C1 carrying only the binary vector (Figure B). To further explore the minimal genetic requirements, we then generated two additional variants that added critical genes virC12 (pDimples0.5-virC12) or virE12 (pDimples0.5-virE12) upstream of the virB cluster. While pDimples0.5-virC12 was unable to achieve any measurable N. benthamiana transformation, pDimples0.5-virE12 generated a GFP signal above the control (Figure B). This finding was corroborated by experiments which expressed a nuclear-localized mScarlet from the T-DNA. C58C1 containing the minimal pDimples0.5-virE12 were able to induce bright nuclear fluorescence in N. benthamiana leaves, indicating successful T-DNA transfer into plant nuclei (Figure C). These findings experimentally define a minimal set of genes required for transient AMT of N. benthamiana leaves and a starting point for the rational engineering of synthetic pTi vectors.

To iterate upon and further optimize this design, we added both virC12 and virE12 upstream of the virB cluster (pDimples1.0), which dramatically improved transformation efficiency to 6.3% of that of the wild type (Figure B,C). We then sought to evaluate whether the addition of either effector vir genes virE3 or virD5, both of which significantly decreased transformation when deleted in A. fabrum GV3101, could improve transformation compared to pDimples1.0. Either gene was cloned between virC12 and virB clusters to generate pDimples1.5. While pDimples1.5-virE3 improved transformation over pDimples1.0 to 8.3% of the wild type, pDimples1.5-virD5 did not improve over pDimples1.0 (Figure B). When both virE3 and virD5 were added to create pDimples 2.0, transformation efficiency reached 9.1% of that of wild type. While this was significantly improved from pDimples1.0, it was not significantly better than the addition of virE3 alone (Figure B). These vectors were introduced into GV3101 with a deletion from virA-virE3 constituting the majority of the vir genes and their essential positive regulators. When these complementation strains were compared to pDimples vectors harbored in C58C1, there was no significant difference in the transformation ability of strains with each vector (Figure S11A). This suggests that, at least in the context of transient expression within N. benthamiana, other genes on pTi may not play a significant role in the transformation process.

Attempts to optimize the expression of virB via complementation assays showed that P_LacO was an optimal choice to control the expression of the T4SS. The choice of the inducible P_LacO also allowed us to control the magnitude of transformation with the amount of IPTG that was coinfiltrated (Figure S11B). As the ability of pDimples vectors to restore transformation was significantly less than that of the pLoki vectors that contained the minimal vir machinery without the T4SS (∼10% versus 75% restoration of wild-type A. farbrum GV3101 transformation), we concluded that suboptimal expression of the T4SS was likely a limiting factor. A possible bottleneck could be the availability of virD4 that acts as a bridge between VirD2-conjugated T-DNA and the rest of the T4SS.

As virD4 acts in concert with the T4SS to extrude the T-DNA, we sought to see if virD4 expression limited transformation by replacing the very weak P_J23114 promoter with the slightly stronger P_J23117 promoter. However, this resulted in a significant decrease in transformation, indicating that the bottleneck exists elsewhere (Figure S11C). While no pDimples vector was able to restore wild-type level transformation to A. fabrum C58C1, pDimples1 and pDimples2 outperformed any attempt to complement a virB1-11 deletion. These results are consistent with our hypothesis that a specific ratio between the T4SS genes and other vir genes is required for the optimal transformation. Exploring this relationship further will likely be key in debottlenecking future pTi engineering efforts.

Because Agrobacterium is a critical tool for the transformation of many fungi, we evaluated the ability of the pDimples vectors to transform the R. toruloides. Unlike those in N. benthamiana, a small number of transformants were observed with pDimples0.5-virC12 strains, while no transformants were observed with pDimples0.5-virE12 (Figure D). This is consistent with our individual knockout data (Figure E) and previous reports that virE12 is not as important for fungal transformation as it is for plant transformation. While only 5% of wild-type transformation efficiency was achieved with pDimples1.0, the addition of virD5 dramatically increased the transformation efficiency to 40% of wild type (Figure D). These results complement our earlier finding that VirD5 is critical for R. toruloides transformation and highlight that, contrary to previous thought, it is likely that virD5 has a far more fundamental role in AMT than simply as a determinant of host range (Figure E). However, the addition of virE3 by itself did not improve Rhodosporidium transformation, nor did it improve the transformation efficiency when added in combination with virD5. As with N. benthamiana experiments, fungal transformation was dependent on the presence of IPTG to induce virB1–11 expression (Figure S11D), with R. toruloides transformants confirmed by colony PCR (Figure S11E). Our synthetic pTi with differing minimal sets of vir genes demonstrates how a bottom-up engineering approach can define how AMT of fungi and plants differ, offering new opportunities to further dissect the contributory role of each vir gene in fungal AMT.

After demonstrating that pDimples are sufficient to enable T-DNA transfer in a C58C1 A. fabrum background, we were interested in determining if the synthetic pTi contained all necessary components to reconstitute transformation in a divergent strain of Rhizobium rhizogenes. To this end, we introduced pDimples1.0 into R. rhizogenes D108/85, a nonpathogenic strain without a native pRi or pTi plasmid and diverged from A. fabrum ∼200 million years ago. When wild-type R. rhizogenes carrying a binary vector with a T-DNA encoding nuclear-localized mScarlet was infiltrated into N. benthamiana leaves, no red nuclei were observed (Figure E). Yet with the addition of pDimples1.0, red nuclei were observed that produced significantly more fluorescent signal than the parent strain (Figures E and S11F). Together, our results demonstrate that the design and construction of synthetic pTi can be used to (1) identify the core set of genes that are necessary and sufficient for AMT, (2) describe the contributory role of accessory vir genes, (3) respond to user-defined modulation in vir gene expression, and (4) confer transformation capacity to a deficient strain of R. rhizogenes.

Metabolic Burden of Virulence Machinery Results in Unstable Synthetic pTi

Recent work showed that the copy number of BBR1 origin plasmids, used in pDimples vectors, was substantially higher at ∼50 copies/cell than we initially anticipated based on copy numbers from other bacteria. , This likely creates a substantial metabolic burden on cells and strong selection pressure for mutations that compromise the function of the synthetic pTi. Observation of A. fabrum C58C1 transformed with pDimples1.5-virD5 revealed two distinct colony morphologies (big and small), with the small colony variant showing significantly greater transformation efficiency compared to the larger variant, suggesting mutational breakage of pDimples in big colonies (Figure A). In an attempt to lower the metabolic burden of the synthetic pTi, we generated eight copy number variants of pDimples1.5-virD5 using either a BBR1 or pSa origin and, in doing so, also exchanged the original spectinomycin resistance gene for kanamycin resistance. Decreasing the copy number of wild-type BBR1 from 52 to 34 with a T148A RepA mutation significantly improved N. benthamiana transient transformation from ∼4% to ∼8% of wild-type GV3101, and lowering the copy number further (L138F:20 or E182 V:8) resulted in transformation efficiencies of 1% or less (Figure B). While pDimples1.5-virD5 using a wild-type pSa origin (copy number 5) resulted in transformation efficiencies ∼2% of GV3101, mutants that increased copy number (K155N:18, D181Y:24, or V152E:31) significantly improved transformation, with the K155N variant achieving ∼15% of GV3101 transformation and significantly outperforming any BBR1 origin pDimples1.5-virD5 variant (Figure B).

5.

Engineered pTi are unstable and prone to rearrangement: (A) Morphological phenotype of A. fabrum C58C1 transformed with pDimples1.5-virD5 after 3 days of growth on selective solid media. Infiltration of big colony (red arrows) variants resulted in significantly less transient expression of GFP in N. benthamiana than small colonies (right panel, n = 64). (B) Transient expression of GFP in N. benthamiana (n = 64) by pDimples origin variants. BBR1 origin variants are shown in blue, while pSa origin variants are in red and ordered by ascending origin copy number which is specified in the x-axis. (C) Synteny analysis of plasmid sequences recovered from big colony variants of pSa origin pDimples compared to original plasmid. Colored bars indicate conserved gene presence, while gray genes show genes that were lost due to recombination, and red “T” shows the presence of terminator.

Though modifying the origin and copy number of the synthetic pTi can improve transformation efficiency, our designs still fall well short of the wild-type efficiency. Evaluating the colony morphology of C58C1 harboring our pDimples origin variants revealed the same pattern of small and big colony variants as was observed in the original pDimples1.5-virD5 vector (Figure S12). In pSa origin variants of the synthetic pTi, the “small” colony variants apparently decreased in size as the copy number increased, suggesting increased metabolic burden (Figure S12). Counterintuitively, as the copy number of BBR1 variants decreased, the “small” colony variant also decreased in size, suggesting that at lower copies, the BBR1 origin may be less stable in Agrobacterium (Figure S12). To establish whether the “big” colony variants indeed were the result of mutations in the synthetic pTi plasmids, pDimples was sequenced from C58C1 big colony variants of pSa wild-type, pSa K155N, BBR1 wild-type, and BBR1 T148A pDimples1.5-virD5 strains, revealing that all plasmids isolated from “big” colonies had undergone large deletions eliminating the majority of vir genes from the plasmid, likely due to a recombination event between homologous terminator sequences (Figures C and S13). To mitigate the apparent toxicity and resultant selection for recombination events in synthetic pTi, future designs will likely need to take additional inspiration from natural pTi/pRi plasmids by featuring inducible control of both vir gene expression and plasmid copy number, , as well as reducing repetitive sequences to prevent recombination.

Beyond assessing the stability of our synthetic pTi designs, we also aimed to determine whether Agrobacterium strains harboring these plasmids exhibited substantial differences in global protein expression compared with GV3101, as these differences could inform future strain engineering. To this end, we performed shotgun proteomics on GV3101 and C58C1:pDimples1.5-virD5 strains cultured either in Luria–Bertani (LB) medium or under virulence-inducing conditions for 24 h, supplemented with 2 mM IPTG (Figure A). Under inducing conditions, 153 proteins were differentially expressed between the two strains (Figure B,C and Supplementary Dataset 1). Among the upregulated proteins were numerous Vir proteins, along with those involved in protein secretion, folding, and degradation, as well as genes associated with succinoglycan biosynthesis and the pAT plasmid (Figure C). Expectedly, many of the downregulated proteins corresponded to pTi-encoded genes absent from the C58C1 background as well as those involved in arginine metabolism, NAD biosynthesis, and palatinose utilization (Figure D). Notably, several vir gene products were expressed at levels an order of magnitude higher than those in GV3101 under the same induction conditions (Figure E). This dramatic increase may explain the upregulation of protein processing pathways and suggests either that vir gene expression in this system is excessively high, leading to metabolic burden, or that in vitro induction conditions poorly recapitulate the in planta environment (Figure S14A).

6.

Proteomics reveals targets for synthetic pTi improvement: (A) Experimental design of proteomics experiment where either GV3101 or C58C1:pDimples1.5-virD5 grown either in LB or induced for 24 h under vir-inducing conditions. (B) Volcano plot showing global differential abundance of proteins of C58C1:pDimples1.5-virD5 and GV3101. Filled circles show significantly differentially abundant proteins (p-value < 0.05 after HB correction), while empty circles show nonsignificant proteins. Blue diamonds show vir gene products found on pDimples1.5-virD5 that are significantly differentially abundant compared to GV3101, while orange diamonds show proteins expressed from the native pTi that are not found in pDimples that are significantly differentially abundant. Venn diagrams in panel (C) show proteins more abundant in pDimples1.5-virD5 compared to GV3101, while panel (D) shows proteins less abundant under inducing conditions. The large green circle represents all proteins found on the disarmed pTiC58 found in GV3101, while the blue circle represents the vir gene products encoded on pDimples1.5-virD5. (E) Detectable vir gene product expression in GV3101 (white bars) and C58C1:pDimples1.5-virD5 (gray bars) under inducing conditions. Error bars show standard error. (F) Top panel: Volcano plot showing global differential abundance of proteins of C58C1:pDimples1.5-virD5 grown in LB or under inducing conditions. Filled circles show significantly differentially abundant proteins (p-value <0.05 after HB correction), while empty circles show nonsignificant proteins. Orange diamonds show vir gene products found on pDimples1.5-virD5 that are significantly differentially abundant compared, while blue diamonds show vir gene products that are not significantly differentially abundant. Bottom panel: Abundance of significantly differentially expressed Vir proteins expressed from C58C1:pDimples1.5-virD5 when grown in inducing conditions (black bars) or in LB (white bars). Error bars show standard error. Sample sizes for proteomics experiments were n = 4.

To truly break free of the necessity of inducing conditions (low pH, with phenolic compounds and reducing sugars present), we may need additional genes involved in virulence that are encoded on Agrobacteria chromosomes, sometimes termed chromosomal virulence genes. Of the 376 proteins differentially regulated in the synthetic pTi strain when grown either in LB or inducing conditions, 5 vir gene products were statistically differentially expressed, with VirC1, VirD1, VirD2, and VirE2 all upregulated in LB, while VirB10 was substantially downregulated in LB conditions (Figure F). While not statistically significant with BH correction on account of not being detectable in multiple LB samples, VirB5 was also seemingly heavily downregulated in noninduction conditions (Figure S14B). These results may reflect a lack of protein stability at the elevated temperatures in the LB condition, or the lower expression of osmoprotectant-producing ChvA and ChvB (Figure S14C). As chvA mutants have been shown to be less virulent, our proteomics results suggest that additional expression or transcriptional rewiring of chromosomal genes may be required in order to achieve highly efficient condition-independent AMT. , Expression of additional chromosomal genes, combined with balancing vir gene expression and its associated metabolic burden, will likely represent the most significant hurdles to engineering highly efficient synthetic AMT in the future.

Conclusion

Here, we leveraged a comprehensive and quantitative understanding of each vir gene cluster to build synthetic pTi plasmids that define the minimal transferable gene set required for AMT of both plants and fungi. Optimization of this set will allow us to better understand host-specificity between natural strains of Agrobacterium and engineer laboratory strains with superior transformation properties. Furthermore, our analysis of how allelic variation of vir genes impacts transformation suggests that there are likely untapped genetic resources to improve AMT. Overall, this work will also serve to guide related research studying host-microbe interactions, specifically those of plant-associated bacteria. For example, recent research that developed minimized versions of the nitrogen-fixing pSymA in the root nodule-associated legume symbiont Sinorhizobium meliloti could be furthered by evaluating the impact of individual gene expression.

We compared bacterial synthetic biology parts both in vitro and in planta, revealing that while constitutive synthetic promoters will likely perform similarly in different environments, the performance of inducible systems may be highly variable. Further characterization of synthetic regulatory elements in situ will enable precise engineering. However, by using these tools to replace the master regulatory VirA/G system with synthetic regulation, we may not only gain precise control of individual gene expression, but could also insulate the bacteria from host mechanisms that interfere with gene expression, which has been previously observed. , In fungi, current AMT methods require long vir gene induction times in conditions that may not be optimal for the growth of certain fungi, which could be bypassed using synthetic pTi. , Thus, separating AMT induction from its native inducing conditions (i.e., low pH, sugar, and phenolic compounds) may also provide unique opportunities for improving fungal transformation. Achieving complete divorce, however, may require additional global transcriptional rewiring of chromosomal genes that contribute to AMT.

While our synthetic pTi establishes a foundational proof-of-concept for the bottom-up engineering of AMT, it currently exhibits reduced stability and lower transformation efficiencies compared with native systems. This is likely driven by the severe metabolic burden of constitutive virulence expression and high plasmid copy numbers, which exert a strong selective pressure for recombinational breakage. To mitigate this toxicity and resultant selection for recombination events, future engineering efforts must focus on mitigating this burden by incorporating inducible control over both the plasmid copy number and global vir gene expression, alongside the removal of repetitive terminator sequences to enhance genetic stability. Because the current work focuses on establishing the minimal foundational architecture of a synthetic pTi, our assays utilized rapid, high-throughput transient expression in model systems such as N. benthamiana and stable transformation in R. toruloides. Translating these synthetic minimal systems into stable plant transformation pipelines, particularly for recalcitrant crop species, represents a critical next step. Future studies will need to couple these synthetic plasmids with long-term tissue culture and regeneration protocols to fully validate their utility in crop engineering.

By mobilizing the transformation phenotype via pDimples into R. rhizogenes, we open the door to another promising avenue of AMT engineering: transferring the complex vir machinery to other bacteria. As A. fabrum can elicit plant immunity that impedes transformation, multiple efforts have been made to circumvent this either through mutation of known immunogenic loci or the addition of immune-suppressing systems. As strains of R. rhizogenes have recently been shown to improve the transformation of solanaceous plants, future development of synthetic pTi may further improve the performance of these strains. This work lays the foundation for developing synthetic pTi that functions in bacteria that elicit minimal immune responses across plant species, potentially enabling the transformation of those that have traditionally been recalcitrant to genetic modification.

The engineering of synthetic virulence machinery necessitates careful consideration of biosafety. Because our synthetic pTi system operates in a ’disarmed’ context - where the native tumorigenic phytohormone genes of the T-DNA have been completely replaced by standard reporter genes on a separate binary vector, it poses no tumorigenicity risk to host plants. However, the potential for unintended horizontal gene transfer (HGT) of synthetic pTi to environmental microbes remains a concern. To ensure safe deployment and regulatory compliance, future iterations of these strains should incorporate robust biological containment strategies, such as strict genomic auxotrophies or tying the activation of synthetic virulence machinery strictly to exogenously applied orthogonal chemical ligands not found in natural plant environments.

In synthetic biology, our inability to efficiently transform diverse organisms represents the biggest bottleneck in expanding the scope and range of species that can be engineered. Given the wide diversity of eukaryotes that can be transformed by Agrobacterium, future synthetic pTi may be optimized to target currently untransformable organisms and enable entirely new areas of biotechnology.

Materials and Methods

Media, Chemicals, and Culture Conditions

Routine bacterial cultures were grown in Luria–Bertani (LB) Miller medium (BD Biosciences). E. coli was grown at 37 °C, while A. fabrum was grown at 30 °C unless otherwise noted. Cultures were supplemented with kanamycin (50 mg/L, Sigma-Aldrich), gentamicin (30 mg/L, Fisher Scientific), or spectinomycin (100 mg/L, Sigma-Aldrich), when indicated. All other compounds, unless otherwise specified, were purchased through Sigma-Aldrich. Bacterial kinetic growth curves were performed as described previously. Induction conditions for vir genes were performed as described previously.

Strains and Plasmids

All bacterial strains and plasmids used in this work are listed in Tables S1 and S2. All strains and plasmids created in this work are viewable through the public instance of the JBEI registry (https://public-registry.jbei.org/folders/814). To access the sequence data and plasmid maps, researchers can create a free user account by selecting ’Register’ on the registry homepage. Once logged in, navigating to the provided folder link will grant full access to download sequence files and view annotated genetic parts. Physical strains can be requested via email of the JBEI strain archivist. All plasmids generated in this paper were designed using Device Editor and Vector Editor software, while all primers used for the construction of plasmids were designed using j5 software. − Synthetic DNA was synthesized from Twist Biosciences. Plasmids were assembled via Gibson Assembly using standard protocols, Golden Gate Assembly using standard protocols, or restriction digest followed by ligation with T4 ligase as previously described. Plasmids were routinely isolated using the Qiaprep Spin Miniprep kit (Qiagen), and all primers were purchased from Integrated DNA Technologies (IDT, Coralville, IA). Plasmid sequences were verified using whole plasmid sequencing (Primordium Laboratories, Monrovia, CA). Agrobacterium was routinely transformed via electroporation as described previously.

Construction of Deletion Mutants

Deletion mutants in A. fabrum GV3101 were constructed by homologous recombination and sacB counterselection using the allelic exchange as described previously. Briefly, homology fragments of 1 kbp up- and downstream of the target gene, including the start and stop codons respectively, were cloned into pMQ30K - a kanamycin resistance-bearing derivative of pMQ30. Plasmids were then transformed via electroporation into E. coli S17 and then mated into A. fabrum via conjugation. Transconjugants were selected for on LB Agar plates supplemented with kanamycin 50 mg/mL and rifampicin 100 mg/mL. Transconjugants were then grown overnight on LB media also supplemented with 50 mg/mL kanamycin and 100 mg/mL rifampicin and then plated on LB Agar with no NaCl supplemented with 10% w/v sucrose. Putative deletions were restreaked on LB Agar with no NaCl supplemented with 10% w/v sucrose and then were screened via PCR with primers flanking the target gene to confirm gene deletion.

Synthetic Part Characterization

Characterization of pGinger vectors harbored by A. fabrum in vitro was performed as previously described for other bacteria. Briefly, A. fabrum C58C1 with different pGinger vectors were grown overnight in 10 mL of LB supplemented with kanamycin at 30 °C with 250 rpm shaking and then diluted 1:100 into 500 μL of fresh LB media with kanamycin in a deep-well 96-well plate (Corning). For inducible promoters, chemical inducers were added in 2-fold dilutions before incubation. Cells were then grown at 30 °C for 24 h while shaking at 250 rpm, and then 100 μL was measured for absorbance at OD₆₀₀ as well as for RFP fluorescence using an excitation wavelength of 590 nm and an emission wavelength of 635 nm with a gain setting of 75 on a BioTek Synergy H1 microplate reader (Agilent).

To evaluate the performance of synthetic promoters in planta, strains were grown in 5 mL of LB media with kanamycin at 30 °C with 250 rpm shaking overnight, and diluted 1:5 with fresh media, then grown for an additional 3 h at 30 °C with 250 rpm shaking. Cultures were then adjusted to an absorbance at OD₆₀₀ of 1.0 in agroinfiltration buffer (10 mM MgCl₂, 10 mM MES, 200 μM acetosyringone, pH 5.6), and infiltrated into either N. benthamiana or A. thaliana leaf tissue. When appropriate, chemical inducers were added to the agroinfiltration media immediately before leaf infiltration. Either one or 3 days post-infiltration, 6 mm leaf disks were excised from each Agro-infiltrated leaf using a hole puncher and placed atop 300 μL of water in a black, clear-bottom, 96-well microtiter plate (Corning). GFP fluorescence of each leaf disk was then measured by using a BioTek Synergy H1 microplate reader (Agilent) with an excitation wavelength of 488 nm and measurement wavelength of 520 nm.

Plant Growth Conditions

A. thaliana were germinated and grown in Sunshine Mix #1 soil (Sungro) in a Percival growth chamber at 22 °C and 60% humidity using a 8/16 h light/dark cycle with a daytime PPFD of ∼200 μmol/m²s. N. benthamiana plants were grown according to a previously described standardized lab protocol. All N. benthamiana growth was conducted in an indoor growth room at 25 °C and 60% humidity using a 16/8 h light/dark cycle with a daytime PPFD of ∼120 μmol/m²s. Plants were maintained in Sunshine Mix #4 soil (Sungro) supplemented with Osmocote 14–14–14 fertilizer (ICL) at 5 mL/L and agro-infiltrated 29 days after seed sowing.

N. benthamiana Infiltration and Leaf Punch Assay

A. fabrum strains were grown in LB liquid media containing necessary antibiotics (50 μg/mL rifampicin, 30 μg/mL gentamicin, 50 μg/mL kanamycin, and 100 μg/mL spectinomycin for most strains) to an OD600 between 0.6 and 1.0 before pelleting. Cells were then prepared for infiltration by resuspension in agroinfiltration buffer (10 mM MgCl₂, 10 mM MES, 200 μM acetosyringone, pH 5.6) to a final OD600 of 1.0 and were allowed to induce for 2 h in infiltration buffer at room temperature. When appropriate, chemical inducers (i.e., IPTG) were added during the 2 h induction period. Each strain was then infiltrated into the fourth and fifth leaves (counting down from the top) of eight biological replicate N. benthamiana plants. GFP transgene expression in agro-infiltrated leaves was then assessed by a leaf disk fluorescence assay 3 days post-infiltration. Four 6 mm leaf disks were excised from each agro-infiltrated leaf using a hole puncher and placed atop 300 μL of water in a black, clear-bottom, 96-well microtiter plate (Corning). GFP fluorescence of each leaf disk was then measured using a BioTek Synergy H1 microplate reader (Agilent) with an excitation wavelength of 488 nm and measurement wavelength of 520 nm.

Rhodospordium Toruloides Transformation

Agrobacterium tumefaciens-mediated transformation was performed on Rhodosporidium toruloides IFO0880 with a codon-optimized epi-isozizaene synthase from Streptomyces coelicolor A3(2) (JPUB_013523) as previously described. When appropriate, 2 mM IPTG was added to agrobacterium induction media. Transformants were confirmed via colony PCR specific to integrated T-DNA.

Proteomic Analysis

Proteins from A. fabrum samples were extracted using a previously described chloroform/methanol precipitation method. Extracted proteins were resuspended in 100 mM ammonium bicarbonate buffer supplemented with 20% methanol, and protein concentration was determined by the DC assay (BioRad). Protein reduction was accomplished using 5 mM tris 2-(carboxyethyl)­phosphine (TCEP) for 30 min at room temperature, and alkylation was performed with 10 mM iodoacetamide (IAM; final concentration) for 30 min at room temperature in the dark. Overnight digestion with trypsin was accomplished with a 1:50 trypsin:total protein ratio. The resulting peptide samples were analyzed on an Agilent 1290 UHPLC system coupled to a Thermo Scientific Obitrap Exploris 480 mass spectrometer for discovery proteomics. Briefly, 20 μg of tryptic peptides were loaded onto an Ascentis (Sigma–Aldrich) ES-C18 column (2.1 mm × 100 mm, 2.7 μm particle size, operated at 60 °C) and were eluted from the column by using a 10 min gradient from 98% buffer A (0.1% FA in H₂O) and 2% buffer B (0.1% FA in acetonitrile) to 65% buffer A and 35% buffer B. The eluting peptides were introduced to the mass spectrometer operating in positive-ion mode. Full MS survey scans were acquired in the range of 300–1200 m/z at 60,000 resolution. The automatic gain control (AGC) target was set at 3e6, and the maximum injection time was set to 60 ms. Top 10 multiply charged precursor ions (2–5) were isolated for higher-energy collisional dissociation (HCD) MS/MS using a 1.6 m/z isolation window and were accumulated until they reached either an AGC target value of 1e5 or a maximum injection time of 50 ms. MS/MS data were generated with a normalized collision energy (NCE) of 30, at a resolution of 15,000. Upon fragmentation, precursor ions were dynamically excluded 10 s after the first fragmentation event. The acquired LCMS raw data were converted to mgf files and searched against the latest UniProt A. tumefaciens protein database with Mascot search engine version 2.3.02 (Matrix Science). The resulting search results were filtered and analyzed by Scaffold v 5.0 (Proteome Software Inc.). The normalized spectra counts of identified proteins were exported for relative quantitative analysis.

Bioinformatic Analyses

Sequences of individual vir genes from genomes of all sequenced Agrobacterium were identified and extracted as previously described. MACSE v. 2.07 with the parameter “-prog alignSequences” was used to generate codon alignments for each vir gene data set. The HYPHY v2.2 program “cln” was used to remove identical sequences and stop codons from each alignment. IQ-TREE v. 1.6.12 with the default parameters was used to generate a phylogeny for each data set. The HYPHY program FUBAR with codon alignment, phylogeny, and a probability threshold of 0.9 was used to calculate per-site d_N/d_S and detect signals of positive or purifying selection. Gene cluster alignments and visualizations were done via clinker.

Statistical Analyses and Data Presentation

All numerical data were analyzed by using custom Python scripts. All graphs were visualized using either Seaborn or Matplotlib. , Calculation of 95% confidence intervals, standard deviations, and t test statistics was conducted via the Scipy library. Bonferroni corrections were calculated using the MNE python library. Differential protein abundances were calculated as described previously.

Alleles of homologous vir genes were aligned using MAFFT v. 7.508 and converted into phylogenetic trees using FastTree v. 2.1.11. Phylogenetic distance was calculated using dendropy v. 4.6.1.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acssynbio.6c00064.

• Characterization of constitutive promoters (Figure S1); dose response curves (Figure S2); expression of A. tumefaciens C58C1 RFP (Figure S3); impact of increasing vir gene expression on tobacco transient transformation (Figure S4); growth rate of A. tumefaciens GV3101 vir gene cluster deletion strains (Figure S5); complementing virB1-5 and virB6-11 deletion mutants (Figure S6); optimization of virB1-11 complementation (Figure S7); impact of vir gene allele on tobacco transformation (Figure S8); selective pressure on vir genes (Figure S9); evaluation of combinatorial vir allele complementation (Figure S10); evaluation of refactored pTi plasmids (Figure S11); colony morphology of A. tumefaciens C58C1 transformed with pDimples1.5 origin 180 variants (Figure S12); synteny analysis of BBR1 pDimiples Large Colony Variants (Figure S13); and comparative proteomics of Refactored pTi and GV3101 (Figure S14) (PDF)

• Strains (Table S1) and plasmids (Table S2) (XLSX)

• Supplementary Data set 1 includes proteomics data found in Figure (XLSX)

• Supplemental Table Statistics shows results of statistical tests (XLSX)

Conceptualization, M.G.T., P.M.S.; Methodology, M.G.T., L.D.K, M.J.S., A.N.P., L.W., A.W., G.M.G.; Investigation, M.G.T., L.D.K., G.M.G., L.M.W., A.N.P., M.J.S., K.V., S.S., K.M., S.A., N.F.C.H., D.S., C.T., R.C., S.L., J.C., H.P., N.L., J.W.G., Y.C., A.J.W.; Writingoriginal draft, M.G.T.; Writingreview and editing, all authors: Resources and supervision, D.L., C.J.P., J.M.G., H.V.S A.W., J.H.C., J.D.K., P.M.S.

The authors declare the following competing financial interest(s): A patent on the minimized synthetic pTi to enable AMT has been filed by Lawrence Berkeley National Laboratory with M.G.T., A.N.P., and P.M.S. as inventors. M.G.T., M.J.S., and P.M.S. have financial interest in BasidioBio. J.D.K. has financial interests in Amyris, Ansa Biotechnologies, Apertor Pharma, Berkeley Yeast, Demetrix, Lygos, Napigen, ResVita Bio, and Zero Acre Farms.