Abstract
Agrobacterium tumefaciens is a plant pathogen commonly repurposed for genetic modification of crops. Despite its versatility, it remains inefficient at transferring DNA to many hosts, including to animal cells. Like many pathogens, physical contact between A. tumefaciens and host cells promotes infection efficacy. Thus, improving the strength and specificity of A. tumefaciens to target cells has the potential for enhancing DNA transfer for biotechnological and therapeutic purposes. Here we demonstrate a methodology for engineering genetically-encoded exogeneous adhesins at the surface of A. tumefaciens. We identified an autotransporter gene we named Aat, that is predicted to show canonical β-barrel and passenger domains. We engineered the β-barrel scaffold and linker (Aatβ) to display synthetic adhesins susceptible to rewire A. tumefaciens to alternative host targets. As a proof of concept, we leveraged the versatility of a VHH domain to rewire A. tumefaciens adhesion to yeast and mammalian hosts displaying a GFP target receptor. Finally, to demonstrate how synthetic A. tumefaciens adhesion can improve transfer to host cells, we showed improved protein translocation into HeLa cells using a sensitive split luciferase reporter system. Engineering A. tumefaciens adhesion has therefore a strong potential in generating complex heterogeneous cellular assemblies and in improving DNA transfer efficiency against non-natural hosts.
Introduction
A. tumefaciens is a soil bacterium able to transfer DNA fragments tens of kilobases-long to plant host cells. In its natural environment, A. tumefaciens transfers a pathogenic DNA fragment celled transfer-DNA (T-DNA). A. tumefaciens has been repurposed into a potent gene delivery tool for a broad range of plants, yeasts and fungi1–3. Using disarmed strains lacking pathogenic T-DNA, plant engineers introduce new genes in target organisms such as maize, rice and wheat4–6. Genetic modifications include random integration or targeted editing using zinc-finger nucleases or CRISPR/Cas9 4.
A. tumefaciens uses the type IV secretion (T4SS) system to inject DNA into host cells. The T4SS functions upon host surface contact, so that attachment to target cells promotes T-DNA transfer7,8. For synthetic application, forcing cell-cell interaction, for instance by wounding the plant or by using syringe- or vacuum-driven agroinfiltration, is usually recommended9. The addition of extracellular cellulose increases bacterial adhesion to improve T-DNA transfer efficiency to recalcitrant plant cells10. Altogether, these studies suggest that increased bacterial adhesion favors T-DNA transfer. However, our understanding of how A. tumefaciens adheres to host plant surfaces remains incomplete8,11. Despite this knowledge gap, engineering adhesion has the potential to improve gene delivery to plants or alternative targets, for example by broadening the host range of A. tumefaciens.
In Gram-negative bacteria, outer-membrane proteins called autotransporters help display functional proteins moieties at the cell surface, including adhesins12. Autotransporters are promising candidate scaffolds for surface display of exogenous proteins13. They belong to the type V secretion system (T5SS) family, and are composed of a β-barrel domain fused to a passenger domain with an α-helical linker14. The β-barrel anchors the autotransporter to the outer membrane. The passenger domain translocates by traversing the β-barrel, thereby exposing itself at the cell surface and conferring function. In contrast to other secretion systems, autotransporters are expressed as a single protein. This allows for substitution of the passenger domain to other peptides as a way to switch autotransporter function. For example, the Pseudomonas aeruginosa EstA autotransporter has been used to display a variety of lipases, with applications in whole-cell biocatalysis and screening of enzyme libraries14.
Adhesion plays a crucial function in host microbe interactions, which can be repurposed for synthetic applications. Fusing an alternative receptor to an autotransporter is a convenient way to engineer bacterial adhesion to a specific host target. For example, fusions of non-natural receptors to the E. coli intimin autotransporter efficiently rewires adhesion to a wide range of host cells. Specifically, antigen-binding domain from camelid heavy-chain antibodies VHH fused to intimin provide adhesion properties to target ligands15. Utilizing VHH has many advantages as they are single-chained, short (less than 130 amino acids-long) and have a robust structure16. Moreover, they are more efficiently displayed at the surface of bacteria than human single-chain fragment variables (scFv)17. Finally, screening for VHH forms with affinity to new targets is streamlined, for example via phage display starting from naïve DNA libraries18. Using the intimin-VHH display system, Fraile et al. targeted P. putida to abiotic surface coated with the target antigen19. Salema et al. screened libraries against cancer biomarkers by selecting bacteria binding to live mammalian cells20. Finally, Glass et al. created adhesin toolbox that enables the assembly of complex E. coli communities21. Thus, synthetic adhesion can confer modularity and specificity in assembling bacterial consortia or attaching to alternative hosts22,23.
Here, we aimed at engineering A. tumefaciens adhesion to ultimately improve DNA delivery to alternative targets. While using an intimin scaffold is intuitive, it is not a practical solution. Intimin has a topology belonging to the reverse autotransporter family, also called type Ve secretion system24–26. A. tumefaciens C58 does not possess any annotated reverse autotransporter, and consequently likely does not express the right variants of the associated chaperones that enable proper insertion in the outer membrane and/or translocation of the passenger domain24. Therefore, we explored alternative autotransporters in order to engineer synthetic adhesins at the surface of A. tumefaciens.
Here, we engineered a modular synthetic adhesin display system for A. tumefaciens. To achieve this, we investigated the putative autotransporter Atu5364 which we renamed Aat (as Agrobacterium autotransporter) as candidate scaffold for adhesin display. We successfully repurposed Aat to display multiple functional receptors targeting a variety of ligands, including VHH, lectins and arginine-glycine-aspartic acid (RGD) peptide. We then demonstrated that displaying a nanobody-based VHH anti-GFP by Aat fusion strongly improved A. tumefaciens attachment to alternative hosts including yeast and mammalian cells displaying GFP. Finally, and as a proof of concept, we provide preliminary evidence that synthetic display of VHH improves transfer of type IV secretion system proteins to mammalian host cells.
Results
Identification of an A. tumefaciens autotransporter scaffold
The genome of A. tumefaciens is annotated with two uncharacterized autotransporters of the type Va secretion system (T5aSS), namely atu5354 and atu5364, both located on the cryptic megaplasmid pAtC58. We only found a signal peptide encoded within the sequence of atu5364, but not in atu5354, which we therefore disregarded as a potential candidate scaffold (Figure S1). We then modeled Agrobacterium autotransporter Atu5364 (Aat) using RoseTTaFold and obtained an estimate of its three dimensional conformation (Figure 1A)27,28. The predicted structure shows a C-terminal scaffold (β-barrel and α-helix) holding a passenger formed of repeated parallel β-strand repeats organized in a β-solenoid, a prevalent structure in T5aSS autotransporters29. Aat harbors a proline-rich linker between the scaffold and the folded region of the passenger domain. A potential function for this linker is to remain unfolded and extend to help the passenger domain reach distant targets, for example in attachment30. We did not identify a clear homolog for the passenger domain, but our analysis points toward a function in adhesion (Table S1)31–36. Altogether, the predicted scaffold, signal peptide and long linker make Aat a promising candidate for the autotransporter-based display of non-natural passenger domains.
Protein display in A. tumefaciens
We then moved on to use Aat as a scaffold for protein display in A. tumefaciens. We first fused an HA-tag between the endogenous passenger domain and the linker domain. We expressed the fusion under a cumic acid-inducible promoter to decouple expression from virulence37. After staining with FITC-conjugated anti-HA antibodies, these cells showed low expression with patchy localization at the cell surface (Figure S2B). We noticed that the native passenger of Aat contains a disulfide bond located at the N-terminal end of the native passenger domain, which may prevent translocation (Figure 1A)38. We generated a cysteine-free mutant version of the passenger domain with the HA-tag fusion. Staining and microscopy showed that expression levels of the cysteine-free mutant were much stronger than wild-type (WT) and that the localization was uniform along the bacterium surface (Figure S2C).
To demonstrate the capacity of Aat in displaying alternative functional passenger domains, we fused a VHH domain targeting GFP to the Aat β-barrel scaffold with its proline-rich linker (which we call Aatβ). We initially screened a variety of fusion strategies, which consisted in either replacing parts of the passenger domain (amino acids 35-160, 35-512 or 161-512) or in introducing VHH in front, within or after the folded domain of the passenger (at C35, D161 or A513) (Supplementary figure S3A). The constructs were expressed from binary vectors under the control of the pVirE promoter (Supplementary table 2)39. Incubating bacteria expressing these constructs with GFP shows that the scaffold is properly folded and inserted in the outer membrane. However, these constructs caused frequent cell death (Figure S3B, C). We hypothesize that VHH is folded but accumulates in the periplasm, causing toxicity.
We noticed that VHH contains a disulfide bond that again could prevent passenger domain translocation38. Mutating cysteines from VHH anti-GFP has little effect on affinity40. We therefore cloned a disulfide-free VHH by site-directed mutagenesis of C24A and C98V. We then displayed the disulfide-free VHH anti-GFP (VHHGFP) by fusion to Aatβ (VHHGFP_Aatβ) in A. tumefaciens and compared it to the native VHH. We expressed this construct in a strain constitutively expressing cytoplasmic mScarlet. After expressing and staining VHHGFP_Aatβ with GFP, we observed enhanced fluorescent signal at the bacterium surface compared to the native, disulfide-containing version, while not affecting bacterial viability (Figure 1B, Figure S3C, D). Thus, Aatβ efficiently displays functional VHHGFP.
To illustrate the modularity of the display scaffold, we sought to display alternative receptors. We turned to the arginine-glycine-aspartate (RGD) motif which targets integrin ligands, and two disulfide-free lectins from Pseudomonas aeruginosa, namely LecA and LecB which respectively bind to galactose and fucose41–44. We fused each of the receptors to Aatβ. In each, we also inserted an HA-tag between the adhesin and Aatβ for characterization of display. Upon induction and staining with a green fluorescent anti-HA antibody, all constructs showed fluorescence at their periphery. All maintained WT levels of a constitutively-expressed cytosolic mScarlet demonstrating cell viability (Figures 1C, S4A-E). To further demonstrate the process of engineering adhesins for Aatβ-based display, we displayed the FimH fimbrial tip of uropathogenic E. coli. FimH harbors a disulfide bond between C3 and C44. Mutations of one of FimH cysteines to a serine does not affect affinity to mannose at no- and low shear rates45. Hence, we generated a FimH(C3S) mutant that is deficient in forming disulfide bonds and compared the display efficiency with the WT. Like for VHH display, the removal of the disulfide bond greatly improved display (Figures 1C, right panel and S4F and G). Altogether, this shows that Aatβ is highly modular for passenger display and likely can display a wide variety of short peptides, tags, and other disulfide-free β-stranded proteins. For clarity and demonstration purposes, we subsequently focused on the display of VHHGFP adhesin as it enables quantitative control at the adhesin level, but also at the host ligand level.
To improve the stability of the constructs compared to plasmid-based expression, we generated markerless genomic insertion of display. More specifically, we integrated VHHGFP_Aatβ at the virE2 locus on the Tumor-inducing (Ti) megaplasmid. We next titrated acetosyringone induced in both of the plasmid-borne and chromosomally-integrated version. We stained the bacteria with recombinant GFP to quantify the display efficiency of VHHGFP_Aatβ (Figure 2). Both Ti plasmid-integrated and binary vector-based displayed VHHGFP at high and very high levels, a small difference probably likely due to plasmid copy number. In conclusion, both binary vector-based and Ti plasmid-integrated versions are compatible with VHHGFP display.
Attachment of engineered A. tumefaciens to yeast
Synthetic adhesin display opens the possibility of engineering bacteria with emerging functions, such as the assembly of complex multicellular structures, the destruction of cancer cells, or can be applied in investigations of adhesion to host cells21,46. With these applications in mind, we explored the potential of A. tumefaciens with synthetic Aatβ display in providing attachment to non-plant hosts.
We first focused on testing whether the synthetic VHHGFP_Aatβ construct promotes attachment to yeast cells1,47. By fusing GFP to cell-wall-anchoring proteins, we constructed a GFP-displaying S. cerevisiae in the strain eby100, commonly used in yeast display libraries (Figure 3A)48. Then, we separately induced GFP-display in yeast and expression of VHHGFP_Aatβ in A. tumefaciens. We mixed the two strains and imaged the consortium by confocal microscopy. Confocal sections showed that bacteria displaying VHHGFP strongly bound to the yeast cell wall, to the extent of imprinting their shape into the cell walls (Figure 3B). We then compared the binding to GFP-displaying yeast in the absence of VHHGFP_Aatβ and observed a decrease in the number of bacteria bound per yeast cell (Figure 3C). In order to unambiguously attribute binding to the specificity of VHHGFP to GFP, we compared this to GFP-negative yeast and A. tumefaciens displaying a cysteine-free VHH anti-mCherry (VHHmCherry). The number of bacteria attached per yeast cell only increased when the VHHGFP_Aatβ construct was expressed and when yeast displayed GFP (Figure 3D). This shows that our scaffold can rewire A. tumefaciens to the yeast cells, opening the possibility of generating complex interkingdom assemblies. The low binding of A. tumefaciens displaying VHHmCherry additionally demonstrates the specificity of synthetic adhesion.
Attachment of engineered A. tumefaciens to mammalian cells
To demonstrate the capability of the synthetic Aatβ-based display for therapeutic purposes, we tested how engineered A. tumefaciens could specifically bind human cells. To achieve this, we employed HeLa cells displaying GFP at their surface by fusion to a CD80 transmembrane domain46. We incubated HeLa GFP-display with bare or displaying VHHGFP A. tumefaciens. After washing of unbound bacteria, we acquired confocal images to quantify the average number of bacteria per HeLa cell. We observed an increase in the number of bound bacteria when HeLa displayed GFP and A. tumefaciens displayed VHHGFP simultaneously. However, there was no increase in binding when only the receptor or its ligand were expressed (Figure 4Ai-iii and B). By preincubating A. tumefaciens with GFP, we were able to prevent binding to host cells, demonstrating specificity and the potential for external inhibition of adhesion (Figure 4iv and B). In addition, A. tumefaciens displaying VHHmCherry could not attach to HeLa displaying GFP, further demonstrating specificity (Figure 4Av and B). These results held across cells lines, as A. tumefaciens attached in a similar manner to GFP-displaying HEK293T cells (Figure S5).
Monitoring VirE2 transfer using split NanoLuc
We next aimed at targeted delivery into mammalian cells using synthetic adhesion. Following binding to the target cell and T4SS assembly, VirD4 couples helpers proteins such as VirD2, covalently bound to the T-DNA, and VirE2 to the base of the T4SS machinery for injection into the target cell’s cytosol. Consequently, the cytosol of the target cell is the first compartment through which the T-DNA and helper proteins navigate.
To sensitively monitor protein transfer, we employed a split luciferase approach where the translocated protein is fused to one fragment and the host cell expresses a complementary fragment, Successful injection into the host leads to an increase in luminescence signal from the host49–51. LgBit is expressed in the mammalian target cells and HiBit is fused to a protein that gets translocated by a secretion system. NanoLuc complementation restores luciferase activity in target cells. VirE2 is the most abundant Vir protein and is often fused to the short fragment of split fluorescent proteins by insertion into an internal permissive site52,53. We therefore generated an A. tumefaciens strain expressing a HiBit::VirE2 internal fusion under the VirE promoter. As reporter cell line, we further engineered the GFP-displaying HeLa cell line to constitutively express the LgBit fragment (Figure 5A). We controlled LgBit complementation by transiently transfecting cells with a plasmid driving HiBit::VirE2 expression. We measured a strong increase in luminescence signal, suggesting that HiBit internally fused to VirE2 efficiently complements LgBit (Figure S6).
We next challenged HeLa GFP display expressing LgBit with A. tumefaciens expressing VHHGFP_Aatβ and HiBit::VirE2. We first modulated either inducer concentration or multiplicity of infection. Luciferase signal intensity increased with induction levels at constant multiplicity of infection (Figure 5B), and with multiplicity of infection at constant induction (Figure 5C). This suggests that synthetic adhesion increases VirE2 transfer efficiency to HeLa cells.
We then rigorously investigated the specificity of VirE2 transfer to the synthetic binding. We first verified that the signal was specific to HiBit::VirE2 translocation by measuring luminescence of a native VirE2 in the same system. We measured only background luminescence levels, order of magnitude smaller than for the HiBit::VirE2 fusion. In addition, infecting HeLa without or without GFP display with an A. tumefaciens strain lacking VHH adhesin display but expressing HiBit::VirE2 abolished luminescence. These experiments demonstrate that VHH display is required for the increase in VirE2 transfer. We however found one discrepancy in the fact that we measured a marked increase in VirE2 transfer to HeLa without expression of GFP receptors. To explain this discrepancy, we checked for the possibility of leaky GFP expression. We thus sought to further inhibiting binding by pre-incubating A. tumefaciens VHH with recombinant GFP. Luminescence decreased 2-fold compared to non-induced GFP-display, only partly explaining the GFP-unspecific signal increase (Figure S7A). The remaining signal is likely due to VirE2 transfer independent of adhesion through an unknown mechanism. To further highlight the contributions of this interfering mechanism, we expressed VHHmCherry_Aatβ at the surface of A. tumefaciens. This led to similar luminescence level as VHHGFP_Aatβ (Figure S7A). We controlled that the signal increase was due to VirE2 translocation but not VirE2 secreted in the supernatant. However, lysate of A. tumefaciens expressing VHHGFP_Aatβ and HiBit::VirE2 showed a strong increase in luminescence when incubated with HeLa GFP. In conclusion, we demonstrated that synthetic adhesion promotes VirE2 transfer to host cells, but that VHH display also stimulates VirE2 transfer in an unspecific, ligand-independent manner.
Discussion
Here we investigated A. tumefaciens Aat autotransporter as a candidate scaffold for the display of synthetic passenger domains, aiming at designing a custom, target-specific adhesion system. We repurposed it by fusing the scaffold to lectins, RGD and a VHH. VHH are extremely versatile and can be engineered towards almost any biomarker of interest54. As a proof of concept, we rewired A. tumefaciens using a cysteine-free VHH anti-GFP to target soluble and cell-displayed GFP.
In animal cells, only low Agrobacterium-mediated T-DNA transfection efficiencies have been reported55,56. In recalcitrant plants, several reports demonstrate a positive correlation between adhesion and T-DNA delivery efficiency10,57. Hence, one possible explanation of the limited progress in Agrobacterium-mediated delivery in mammalian cells might be the low affinity of bacteria to such non-natural hosts. We showed that our novel Aatβ scaffold has a strong potential in engineering A. tumefaciens adhesion to non-natural hosts, including mammalian cells. Thus, our design opens the possibility to further examine the possibility of A. tumefaciens-mediated DNA transfer to non-plant hosts. In addition, it provides synthetic biologists with a novel tool to program adhesion of complex multispecies microbial consortia, or the display of enzymes at the surface of other alphaproteobacteria that could be used as additives features in biotechnology or bioremediation14.
We next developed a highly sensitive split luciferase assay that enabled us to monitor VirE2 delivery into mammalian cells. The VirE2 transfer in a cell-cell contact-dependent but mostly GFP-independent manner has yet to be resolved. Bacteria displaying VHH might release VirE2 upon contact, potentially by VHH display-facilitated lysis. Like in plants, VirE2 could then trigger clathrin-mediated endocytosis58. Our results however provide a strong basis for further engineering adhesion to improve A. tumefaciens DNA transfer to non-natural host.
Improved A. tumefaciens-mediated protein delivery is of importance, as scientists repurposed the T4SS of the bacterium for protein delivery and gene editing, as an alternative to using T-DNA. As an example, Vergunst et al. fused VirE2 to the Cre recombinase and leveraged minute amount of proteins transferred to the target cells. They monitored Cre-mediated recombination in plant cells by conferring resistance to cells undergoing recombination59. More recently, Cas9 fusions to the VirF peptide responsible for translocation allowed Schmitz et al. to target both yeast and plant reporter cells expressing gRNA60. Consequently, future experiments could consist in trying such fusions in combination with the adhesin display system to optimize helper proteins delivery at high throughput.
A. tumefaciens is an attractive candidate for gene delivery to human cells. The availability of safe, efficient and precisely targeting gene delivery vectors is currently one of the main limiting factors for gene editing therapies in vivo. Adeno-associated viruses (AAVs) are the current gold standard for gene delivery. Due to their limited packaging capacity, AAVs however remain a bottleneck for the development of therapies based on CRISPR/Cas961,62. On the other hand, wild-type A. tumefaciens injects T-DNA of 25 kb, a size that can easily accommodate several times the CRISPR machinery and repair templates. We thus anticipate that, granted important engineering efforts, A. tumefaciens will constitute a powerful tool as an alternative therapeutic DNA delivery vector into human cells.
Material and methods
Chemicals are purchased from Sigma, unless otherwise stated.
Cloning
In silico cloning was performed using Benchling software and the cloning strategy of individual plasmids is described in Supplementary table 2.
Phusion polymerase (Thermo) was used for PCR using primers from Microsynth (Switzerland) and restriction enzymes (NEB) for digestion. Plasmid, genomic and gel-purified DNA was extracted using commercially available kits. We performed Gibson assembly using NEB HiFi DNA assembly kit, or classical ligation using the T4 Ligase (Thermo). We noted an improvement of the Gibson assembly efficiency when gel-purifying DNA using the Monarch kit (NEB) compared to other kits. Constructs were transformed by heat shock in XL10Gold (Agilent) prepared in 100 mM calcium chloride and 15% glycerol. Bacteria were plated on Luria broth (LB) plates containing the respective antibiotics and plasmids were screened by Sanger sequencing (Microsynth).
Composition of home-made media and agar plates
20x AB salts (per 200mL): 4 g NH4Cl, 1.2 g MgSO4.7H2O, 0.6 g KCl, 0.04 g CaCl2, 10 mg FeSO4.7H2O. Sterile-filtered.
20x AB buffer (per 200mL): 12 g K2HPO4, 4 g NaH2PO4, pH to 7.0 using either KOH or H3PO4, as required, before autoclaving.
Induction medium (IM): 1x AB salts, 0.5% glucose, 2 mM phosphate buffer pH 5.6, 50 mM 2-(4-morpholino)-ethane sulfonic acid (MES)
Agrobacterium minimal medium: 1x AB salts, 1x AB buffer, 0.5% sucrose, antibiotics.
ATGN plates: 1x AB salts, 1x AB buffer, 1% glucose, 1.5% noble agar, antibiotics
ATSN plates: 1x AB salts, 1x AB buffer, 5% sucrose, 1.5% noble agar.
SDCAA: 18.2% Sorbitol, 2% Glucose, 0.67% Yeast Nitrogen Base, 0.5% Casamino Acids, 0.54% Disodium Phosphate, 0.86% Monosodium Phosphate (Add 1.5% Agar for plates.
SGCAA: 18.2% Sorbitol, 0.8% Glucose, 8% galactose, 0.67% Yeast Nitrogen Base, 0.5% Casamino Acids, 0.54% Disodium Phosphate, 0.86% Monosodium Phosphate
Bacterial culture and induction
E. coli were cultured at 37°C in LB containing either 100 μg/mL ampicillin (Huberlab), 50 μg/mL kanamycin, 25 μg/mL chloramphenicol, 50 μg/mL spectinomycin (Chemie Brunschwig).
In this study, we used the disarmed strain A. tumefaciens C58C1 pMP90 (GV3101, see Supplementary table 3, genome accession number: GCA_000092025.163). A. tumefaciens were cultured at 28-30°C in LB containing 60 μg/mL gentamycin (Biochemica) or, when required, either 50 μg/mL kanamycin, 50 μg/mL spectinomycin or 100 μg/mL carbenicillin.
Unless otherwise stated, A. tumefaciens was inoculated at an optical density of 0.1 for 8h in LB containing antibiotics and early stationary cells were induced overnight by addition of one volume of induction medium IM and 100 μM acetosyringone (AS) for virulence induction using the VirE promoter. For cumic acid-inducible constructs, early stationary cells were induced overnight in LB by addition of cumic acid at 10 μM.
Bacterial strain engineering
For replicative plasmids, 1 mL of exponential culture of bacteria was washed 3 times in bidistilled water, concentrated 20 times and electroporated with 100 ng of plasmid in 1 mm cuvette, rescued for 60 min in Super Optimal broth with Catabolite repression (SOC) medium and plated on the corresponding antibiotics plates. Electro-competent bacteria were snap-frozen in 15% glycerol solution.
For markerless genetic engineering of A. tumefaciens, we followed Morton and Fuqua, 201264, using E. coli S17-1 for conjugation and pNPTS138 suicide vector (see supplementary Tables 2 and 3), with the following modifications: we added rifampicin (Axon Lab) at 25 μg/mL during selection and counterselection steps (ATGN and ATSN plates) to better kill donor E. coli. As kanamycin is inhibited by phosphate-buffered media, we increased the concentration to 1200 μg/mL during selection. LB plates containing rifampicin at 25 μg/mL and kanamycin at 300 μg/mL were sometimes more efficient than the aforementioned ATGN plates. Mutants were screened by colony-PCR using primers flanking the knockin or knockout sites and validated by Sanger sequencing.
Bacterial staining, titration and quantification
Bacteria displaying VHH were washed with PBS and stained with recombinant GFP at 100 μg/mL for 10 minutes prior to two PBS washes. Bacteria harboring an HA tag were washed with PBS and stained with anti-HA antibody conjugated with FITC (Abcam ab1208) at 10 μg/mL for 75 minutes in the dark on ice, washed once with PBS. For mScarlet-negative cells, viability was checked by concomitant addition of 10 μg/mL propidium iodide (PI) during staining. Wide field fluorescent pictures of bacteria on a coverslip under 1% agarose PBS pads were taken at 100x and 1.5x lens magnification.
Bioinformatics and modeling
Protein sequences were submitted to the deep learning structure prediction online server RoseTTaFold, provided by the Baker lab: robetta.bakerlab.org 27. Protein sequences were submitted to the online deep neural network software SignalP-5.0 for signal peptide prediction (services.healthtech.dtu.dk/service.php?SignalP-5.0)28.
Average amino acid usage was extracted from kazusa.or.jp/codon/cgi-bin/showcodon.cgi?species=260551.
Engineering of S. cerevisiae
For GFP display, we fused eGFP to Aga2p under the control of a galactose-inducible promoter pGal1. Aga2p form disulfide bonds with Aga1p anchored in the cell wall, resulting in eGFP being anchored in the cell wall.
S. cerevisiae eby100 was retransformed using 1 μg of pGal1 – Aga2p_eGFP following the protocol of the EZ yeast transformation kit II (Zymo). 100 μL of cells were selected on SDCAA plates, which do not contain tryptophan. Colonies were directly selected by induction and visualization with fluorescent microscopy.
A. tumefaciens binding to S. cerevisiae
Yeast GFP display induced overnight in SGCAA were washed twice in PBS to remove GFP in suspension and concentrated 10 times. Induced A. tumefaciens were washed once in PBS and added to concentrated yeast at a 1 to 1 volume ratio for 60-90 min. Five μL of the cell mixture were transferred to and sandwiched between two coverslips, the yeast cells were left to settle down and imaging was performed at 100x with a confocal microscope and 0.3 μm step. We used NIS Elements (Nikon) for three-dimensional rendering of z-stack pictures and cell counting.
Mammalian cell culture and engineering
HEK293T and HeLa cells were cultured in high-glucose Dulbecco’s Modified Eagle Medium (DMEM, Thermofisher) supplemented with 10% FBS (Life Technologies) at 37°C and 5% CO2. A GFP-displaying monoclonal HEK cell line was generated by transfecting PvuI-linearized pXP145 and FACS sorting after 7 days of culture.
For HeLa cells, we produced lentiviruses as described previously46. HeLa cells stably expressing the doxycycline-inducible GFP display46 were further engineered to express the LgBit by transduction (pXP499). After removal of the lentivirus, a polyclonal cell line was obtained by selection with puromycin (Labforce) at 2 μg/mL for one week.
Mammalian cell transient transfection
Mammalian cells were transfected with Lipofectamine 3000 (Life Technologies) overnight with 100 ng of purified plasmid per well of 96-well plates, following the manufacturer’s instructions.
A. tumefaciens binding to mammalian cells
Mammalian cells were washed with medium twice prior to the addition of 10 μL of induced bacteria per well of 96-well plates. Bacteria were homogenized by pipetting and left to adhere for 5 hours at 30°C and 5% CO2 with 100 μM AS. For the prevention of binding using soluble GFP, bacteria were incubated with 100 μg/mL recombinant eGFP for 5 minutes prior to the addition to mammalian cells. Consequently, recombinant eGFP was also present during coculture at a concentration of 10 μg/mL. After coculture, wells were washed 5 times with mammalian culture medium and imaging was performed in the region of the well the closest to the dispensing of the medium. Confocal microscope Z-stacks were acquired over 12 μm and 2 μm steps in three representative fields of view (one biological replicate). On each field of view, we estimated the number of HeLa cells and counted bacteria on maximum intensity projection using ImageJ and Trackmate65.
Split luciferase assays
A polyclonal cell line of HeLa cells stably expressing LgBit and doxycycline-inducible GFP display was used. Cells were seeded in 96-well plates (Costar 3603). 6 h later, they were induced when required with doxycycline at 1 μg/mL for overnight expression of GFP-display. Cells were washed twice with medium to remove shed GFP from the supernatant and 22 μL of induced A. tumefaciens (LB-IM AS 100 μM) were added to 100 μL of culture medium for 5 h.
Wells were washed with OptiMEM (Thermo) twice and 20 μL of a 1:19 mixture of substrate:buffer from the Nano-Glo Live Cell Assay (Promega) were added to 80 μL of OptiMEM per well. After 3 min incubation with the reagent, luminescence activity was acquired using a multiwell plate reader (Tecan Spark) with 5 s integration time per well. Background from wells with only reagent and OptiMEM was subtracted from the values.
Lysis of A. tumefaciens by sonication
Induced cultures were sonicated in 1.5 mL Eppendorf on ice using a Branson 550 sonicator equipped with a microprobe at 30% power. 3 seconds pulse and 10 seconds rest cycles were applied for a total time of 45 seconds of sonication.
Microscopy
For widefield visualizations of bacteria, we used a Nikon TiE epifluorescence microscope equipped with a Hamamatsu ORCA Flash 4 camera and an oil immersion 100x Plan APO N.A. 1.45 objective.
For bacterial adhesion to yeast and mammalian cells, we used a Nikon Eclipse Ti2-E inverted microscope coupled with a Yokogawa CSU W2 confocal spinning disk unit and equipped with a Prime 95B sCMOS camera (Photometrics). We used either a 40x air objective with N.A. of 1.15 to acquire z-stacks for mammalian cells or a 100x oil immersion objective with N.A. of 1.45 for yeast cells.
Production of recombinant proteins
6x-His tagged eGFP on a pET28a vector (see Supplementary table 2) was retransformed into BL21 strain. We induced production with 1 mM IPTG (Fisher bioreagents) at room temperature overnight. We pelleted and lysed bacteria by sonication in lysis buffer (Tris 100mM, NaCl 0.5M, glycerol 5%) and eGFP was purified using fast flow His-affinity columns (GE Healthcare) and eluted with 0.5 M imidazole. We exchanged buffer to PBS using 30kDa ultracentrigation spin columns (Merck) and adjusted the concentration to 1 mg/mL. Aliquots were snap frozen for further use.
Acknowledgments
The authors are grateful to Dr. Ingmar Riedel-Kruse (The University of Arizona) and Dr. Luis Angel Fernández (Centro Nacional de Biotecnologia) for the nanobody anti-GFP construct, to Dr. Bruno Correia, Stéphane Rosset and Dr. Leo Scheller (Ecole Polytechnique Fédérale de Lausanne) for the production and purification of recombinant proteins, S. cerevisiae eby100 strain, NanoLuc and split NanoLuc constructs, to Dr. Csaba Koncz (Max Planck Institute for plant breeding research) for A. tumefaciens GV3101. We are grateful for the funding provided by the Gebert Rüf Foundation, project number GRS-057/16 and the Novartis FreeNovation 2020 program.