From macro to micro: a combined bioluminescence‐fluorescence approach to monitor bacterial localization

Summary Bacterial bioluminescence is widely used to study the spatiotemporal dynamics of bacterial populations and gene expression in vivo at a population level but cannot easily be used to study bacterial activity at the level of individual cells. In this study, we describe the development of a new library of mini‐Tn7‐lux and lux::eyfp reporter constructs that provide a wide range of lux expression levels, and which combine the advantages of both bacterial bioluminescence and fluorescent proteins to bridge the gap between macro‐ and micro‐scale imaging techniques. We demonstrate that a dual bioluminescence‐fluorescence approach using the lux operon and eYFP can be used to monitor bacterial movement in plants both macro‐ and microscopically and demonstrate that Pseudomonas syringae pv phaseolicola can colonize the leaf vascular system and systemically infect leaves of common bean (Phaseolus vulgaris). We also show that bacterial bioluminescence can be used to study the impact of plant immune responses on bacterial multiplication, viability and spread within plant tissues. The constructs and approach described in this study can be used to study the spatiotemporal dynamics of bacterial colonization and to link population dynamics and cellular interactions in a wide range of biological contexts.


Introduction
In recent years, advances in DNA sequencing technology and molecular biology techniques have provided important new insights into the composition and functions of host-associated bacteria for both plants (Bulgarelli et al., 2012;Rybakova et al., 2017) and animals (Turnbaugh et al., 2008;Le Chatelier et al., 2013). However, although these methods are of pivotal importance in furthering our understanding of the biological processes governing microbial communities and their functions, they often provide limited insight into the spatio-temporal aspects of host-microbe and microbemicrobe interactions. Such methods typically aggregate information across spatially distributed populations, with information being collected for a limited number of samples and timepoints (Prosser, 2010;Fierer and Ladau, 2012). However, the output of an interaction between a bacterial pathogen and its host (animal or plant) greatly depends on the type of tissue that is colonized, the local microenvironment (spatial relationship) and the movement of the pathogen over time within the host (spatiotemporal relationship) (Lamichhane et al., 2015). The homogenisation of a limited number of biological samples to isolate macromolecules or to enumerate microbial cells can obscure variation in microbial physiology at the macro-and micro-scale. For this reason, visualizing and studying bacteria in their environment continue to be important (Propheter and Hooper, 2015) and considerable effort has been invested into developing tools to allow visualization of bacteria in the host environment (Seleem et al., 2008;Monteiro et al., 2012).
For example, the spatial relationships of microbemicrobe interactions can be studied using fluorescent in situ hybridization (FISH) coupled with confocal laser scanning microscopy (CLSM) (Earle et al., 2015;Soldan et al., 2019). The main advantage of FISH is the observation of spatial relationships between multiple members of a microbial community without necessarily knowing a priori the composition of the community itself. However, FISH requires fixed samples, thus preventing the study of real-time interactions. For this reason, bacteria tagged with fluorescent proteins (e.g. YFP, GFP) have been extensively used to visualize the spatial and temporal relationships of host-microbe interactions for both detrimental (Godfrey et al., 2010;Cerutti et al., 2017;Donati et al., 2018;Planas-Marquès et al., 2020) and beneficial (Compant et al., 2005;Fan et al., 2011;Mitter et al., 2017) relationships.
However, while fluorescent proteins allow researchers to investigate bacterial colonization at micro-scale resolution, they are not suitable for macro-scale resolution studies, such as analyses of systemic infection. This is partially due to the relatively weak fluorescence of labelled bacteria and the autofluorescence of the host (Wang et al., 2007). This can result in a labour-intensive process of collecting a series of host samples that can be visualized with fluorescence microscopy in an attempt to localize labelled bacteria without knowing a priori their location. Light-sheet microscopy, such as the ultramicroscope from La Vision Bio Tec, allows imaging of tissue samples up to centimetres in size (Reynaud et al., 2015), but even with this technology, imaging a whole plant organ, such as a leaf, would not be feasible for many plant species. Moreover, this specialized equipment is not commonly accessible to many laboratories. Wang et al. (2007) proposed that a brighter green-fluorescent protein variant, called GFPuv, encoded in a broad-host-range high-copy number plasmid could allow monitoring of bacterial disease at whole-plant level using UV light. However, with a few notable exceptions (Rodríguez-Moreno et al., 2009;Fujie et al., 2010), this approach has not been widely adopted for macro-scale imaging. In practise, this is commonly due to limitations in sensitivity and difficulties in imaging acquisition and analysis.
An alternative approach is to use bacteria labelled with the lux operon. Bacterial bioluminescence (Huang et al., 2006;Rico et al., 2010;Brodl et al., 2018;Fleiss and Sarkisyan, 2019) offers several advantages over fluorescent proteins for macro-scale imaging. Bioluminescence has generally little or no background, excitation light is not needed and emission of light depends on bacterial metabolism, so only active cells are visible (Gregor et al., 2018;Planas-Marquès et al., 2020). Consequently, bacterial bioluminescence has become a valuable tool to monitor bacterial localization in plants (Fukui et al., 1996;Bogs et al., 1998;Seleem et al., 2008;Xu et al., 2010) and animals (Yu et al., 2004;Burkatovskaya et al., 2006;Mortin et al., 2007;Seleem et al., 2008) at a macroscopic scale. Advanced imaging technologies, such as electron multiplying CCD (EMCDD) cameras, can be used to detect bioluminescence at a single-cell level, as primarily demonstrated in mammalian cell systems to date (Asai et al., 2008;Iwano et al., 2018). However, this equipment is not widely available within the scientific community. Therefore at present, bioluminescence is most often used for temporal analyses of cell growth, viability and gene expression in bacterial populations using luminometers and macroscopic imaging using standard CCD cameras.
In an attempt to solve the low signal intensity limitation, Gregor et al. (2018) recently developed an enhanced lux operon, called ilux, which could allow single-cell imaging of Escherichia coli. The ilux operon consists of a series of random mutations within the lux operon and an frp gene from Vibrio campbellii inserted at the end of luxE to maximize brightness (Gregor et al., 2018). The frp gene codes for an FMN reductase responsible for generating FMNH 2 which is then oxidized in the process of light production (Gregor et al., 2018). In E. coli, the additional FMN reductase caused a 2.3-fold increase in brightness, suggesting that FMNH 2 generated by the endogenous FMN reductase in E. coli was limiting for lux activity under the conditions tested (Gregor et al., 2018). However, this has not yet been tested in a wide range of bacteria and under a range of environmental conditions. When using either fluorescent proteins or the lux operon as a bioreporter, the system used to tag bacteria has to be chosen carefully. Ideally, the marker genes should integrate into the bacterial chromosome in a single copy and in a neutral position to avoid disrupting endogenous gene functions (Lambertsen et al., 2004). These criteria are satisfied by the bacterial transposon Tn7 (Craig, 1991) which can be used to insert cloned DNA at high efficiency into a specific intergenic site attTn7, present in the chromosome of many Proteobacteria (Koch et al., 2001).
This study aimed to develop a novel library of mini-Tn7 delivery plasmids combining the benefits of both bacterial luminescence and fluorescence for macro-and microscale resolution imaging, which can be used to chromosomally tag Proteobacteria. We assessed both ilux (Gregor et al., 2018) and modified lux operons (luxCDABE and luxCDABE-frp, this study) in different bacterial strains to select the brightest constructs before developing the mini-Tn7 vector library. We demonstrate that these constructs can be used to visualize both local and systemic host-microbe interactions between Phaseolus vulgaris and the bacterial plant pathogen Pseudomonas syringae pv. phaseolicola (Pph) and show that Pph can move systemically through bean leaves by colonizing the vascular system. While we have demonstrated the applicability of these constructs to study spatio-temporal relationships in plant-microbe interactions, this toolbox will be instrumental to study a wide range of host-microbe, microbe-microbe and microbeenvironment interactions in situ.

Results and discussion
Comparison between lux, ilux and lux frp operons To select the brightest bioluminescence operon and achieve a low detection threshold when visualizing bacteria in the environment, we tested whether ilux (Gregor et al., 2018) was brighter than lux in bacteria other than E. coli. We developed a low copy-number plasmid, called pRSJ-p nptII ::ilux (Supporting Information  Table S1) derived from the reporter plasmid pIJ11282 (Frederix et al., 2014), in which lux was replaced with ilux. Additionally, to test the effect of frp on lux-mediated bioluminescence, we included frp from pGEX(−) (Gregor et al., 2018) after luxE in pIJ11282, generating pRSJp nptII ::lux-frp (Supporting Information Table S1). The three plasmids pIJ11282, pRSJ-p nptII ::ilux and pRSJ-p nptII :luxfrp have the same backbone and promoter. pRSJ-p nptII ::ilux did not yield brighter bioluminescence than pIJ11282 in Acinetobacter baylyi strain ADP1, Pseudomonas fluorescens NZ011 and Pseudomonas syringae pv. phaseolicola (Pph) strain 1302A (Supporting Information Fig. S1A). On the contrary, in P. fluorescens NZ011, pRSJ-p nptII ::ilux was approximately 20-fold less bright than pIJ11282 in both M9 minimal medium and King's B (KB) medium (Supporting Information Fig. 1A). We speculate that the mutations introduced into the ilux operon were optimized for light production under the conditions tested in E. coli and similar results are not guaranteed for other bacterial strains. However, pRSJ-p nptII :lux-frp was two-fold brighter than pIJ11282 in P. fluorescens NZ011 without affecting bacterial growth (Supporting Information Fig. S1B), indicating that, depending on the bacterial strain, an additional FMN reductase can increase the amount of light produced. FMN reductases oxidize NADPH to NADP+ in order to reduce FMN to FMNH 2 . To test for the activity of the FMN reductase encoded by the frp gene, we measured the NADP+ concentration of cell extracts of P. fluorescens NZ011 harbouring pIJ11282 or pRSJ-p nptII :lux-frp. We observed a higher concentration of NADP+ in bacterial cells expressing the frp gene, which supports the conclusion that the increased bioluminescence of P. fluorescens NZ011 (pRSJ-p nptII :lux-frp) relative to P. fluorescens NZ011 (pIJ11282) is due to increased FMN reductase activity (Supporting InformationFig. S2).
Generation of an improved mini-Tn7 vector library for bioluminescence and combined bioluminescencefluorescence expression To combine the benefit of using bacterial bioluminescence for macro-scale localization with the advantages of fluorescent proteins for micro-scale localization, we developed a library of mini-Tn7 plasmids containing both a bioluminescent operon and eYFP (Supporting Information Fig. S3). eYFP was chosen over other fluorescent proteins for its higher brightness over eGFP (Shaner et al., 2005) and for its peak emission spectra (527 nm), which would easily allow visualization in an environment with red autofluorescence (e.g. leaves). To provide a wide range of expression levels of the bioluminescent operon, which would allow researchers to modularly choose the best expression level according to their experimental needs, we designed the library to have five different constitutive promoters driving lux transcription. Promoters were selected to be highly conserved, constitutive and to provide a range of expression levels. For these reasons, we chose the recA promoter (Giliberti et al., 2006), named p OXB20 and its derivatives p OXB16 , p OXB13 and p OXB11 (Oxford Genetics Limited, https:// www.oxgene.com/Products; Lynn et al., 2015;Presnell et al., 2019), and the promoter p nptII (Wright and Beattie, 2004). We also included two different constitutive promoters driving the expression of eYFP, namely p lac and p A1/04/03 (Koch et al., 2001). p lac is a common and conserved constitutive promoter, while p A1/04/03 has been widely adopted for gene expression in Pseudomonas (Godfrey et al., 2010)..
To generate the mini-Tn7 vector library, we used the two cassettes with the highest light production in the tested bacterial strains, which were the lux and the modified lux and frp operons (Supporting Information Fig. S1). Initially, we used pUC18-mini-Tn7T-Gm-lux  as a backbone. Plasmid pUC18-mini-Tn7T-Gm-lux was developed to easily insert a promoter of interest into a multiple cloning site (MCS) upstream of luxC. However, this backbone has a different ribosome binding site (RBS) upstream of luxC than the RBS found upstream of luxC in pIJ11282 (Frederix et al., 2014). The latter RBS has been predicted to increase the translation initiation rate more than fourfold (14243 arbitrary units compared to 3371 calculated with the RBS calculator; Farasat et al., 2014;Ng et al., 2015). Additionally, the upstream sequence of luxC, encompassing the MCS in pUC18-mini-Tn7T-Gm-lux was found to negatively affect transcription (Glassing and Lewis, 2015) (Supporting Information Fig. S4A). We therefore constructed a second plasmid (pRS-p OXB20 ::lux-p A1/04/03 ::eYFP) containing the RBS from pIJ11282 without the long MCS of pUC18-mini-Tn7T-Gmlux (Supporting Information Fig. S4B).
We tested both constructs in Pph 1302A and found that the promoter with the alternate RBS and without the MCS of pUC18-mini-Tn7T-Gm-lux (p OXB20 ) resulted in four to seven times more luminescence compared to the original promoter (p OXB20(1) ) in both KB and M9 media, with no consequences for bacterial growth (Supporting Information Fig. S5A and B).
As we found these backbones to be an improvement over constructs based on pUC18-mini-Tn7T-Gm-lux, we also included constructs with the lux operon by itself in our library, without eYFP (Table 1; Supporting Information Table S1). The five different constitutive promoters used encompass a 15-fold range between the weakest promoter p OXB11 and the strongest promoter p nptII in Pph 1302A ( Fig. 1; Supporting Information Fig. S6A). We observed a small but significant decrease in bacterial growth both in vitro and in planta when comparing Pph 1302A with Pph 1302A p nptII ::lux-p A1/04/03 ::eYFP inoculated into the susceptible host plant P. vulgaris cultivar Canadian Wonder (Supporting InformationFig. 6B and Fig. 7A and B). However, we did not detect any difference between the growth of Pph 1302A and Pph 1302A p OXB16 ::lux-p A1/04/03 ::eYFP (intermediate expression) or Pph 1302A p OXB11 ::lux-p A1/04/03 ::eYFP (low expression) in planta (Supporting Information Fig. S7B).
Access to a range of promoters will enable researchers to choose an expression level of the lux cassette that is suitable for their experimental conditions, balancing both signal strength and any fitness cost observed (e.g. purpose of the experiment, equipment used to detect luminescence). While we would expect the constructs to give different expression levels in different bacterial strains, the constructs should maintain a wide expression range as the OXB promoters (Oxford Genetics Limited, UK) were all derived from the constitutive and conserved recA promoter (Weisemann and Weinstock, 1991).

Pseudomonas syringae pv. phaseolicola colonizes the vascular system of Phaseolus vulgaris leaves
We applied a combined luminescence-fluorescence approach to investigate whether Pph systemically infects P. vulgaris leaves. Pph, the causal agent of halo blight of bean, is known to be a seed-borne pathogen and to be able to cause systemic symptoms and spread within infected plants (Taylor et al., 1979), but to date, there is limited knowledge of the routes and mechanisms used by Pph to spread systemically within host plants.
We tagged Pph RJ3 (Jackson et al., 2000) using pRS p nptII ::lux-p A1/04/03 ::eYFP (Table 1, Supporting Information  Table S1) as p nptII was the strongest promoter ( Fig. 1) and p A1/04/03 has previously been successfully used in Pph (Godfrey et al., 2010). Pph RJ3 differs from Pph 1302A as it lacks a > 40 Kb genomic island that contains the gene avrPphB, which encodes a secreted effector protein that elicits effector triggered immunity (ETI) in plants carrying the resistance gene R3 (Jackson et al., 2000). Due to the loss of avrPphB, Pph RJ3 can successfully colonize P. vulgaris cultivar Tendergreen (TG) without triggering a plant immune response known as the hypersensitive response (HR), which acts to restrict bacterial growth. In experiments with bacteria incubated in agar plates, we calculated that the detection limit for tagged cells, in terms of number of cells detectable per unit area, was approximately 100 CFU mm −2 (Supporting Information Fig. S8) and observed a high correlation (0.99) between luminescence and CFU, as also reported by Fan et al. (2008).
We used bacterial bioluminescence to macroscopically observe Pph RJ3 in P. vulgaris cultivar TG leaves. We detected a clear co-localization of Pph RJ3 with the leaf vasculature within 3 days after syringe infiltration ( Fig. 2A). We then used this information to select and collect leaf samples (vasculature cross sections) from leaves infected with Pph RJ3 to be visualized with confocal microscopy. This confirmed that Pph RJ3 was localized inside the leaf vasculature ( Fig. 2B and C). Both data types (bioluminescence and fluorescence) showed the ability of Pph RJ3 to move from the infection site.
We were able to detect Pph RJ3 cells within the xylem vessels ( Fig. 2B and C), and in the collenchyma Table 1. Mini-Tn7 vector library constructed in this study. a
Pph strains have been reported to show substantial variation in their ability to cause systemic symptoms in host plants, while host cultivars have been reported to vary in their susceptibility to systemic symptoms. This has been attributed in some studies to systemic movement of the toxin, phaseolotoxin, within host plants (Mitchell and Bieleski, 1977). Our study confirms that the pathogen itself can move systemically within host tissues, although under our experimental conditions Pph RJ3 was not an aggressive systemic pathogen, consistent with the absence of detectable bacterial biofilms in the xylem and wilting symptoms, which are often the manifestation of systemic bacterial infections (Bae et al., 2015). The constructs and methods established in this study will provide valuable tools to investigate the underlying causes of variation in systemic infection and will enable researchers to better understand the biology and epidemiology of this disease. A. Pph RJ3 p nptII ::lux-p A1/04/03 ::eYFP was syringe-infiltrated in localized areas on the abaxial surface of Phaseolus vulgaris cv. Tendergreen leaves. Images were taken after 0, 1, 3, 5 DPI with the nightOWL LB 983 at 10-min exposure. The black line indicates cross sections used for confocal microscopy imaging (B). The white arrows indicate co-localization of Pph RJ3 p nptII ::lux-p A1/04/03 ::eYFP with the leaf vasculature. The black arrows indicate regions of the leaves that have cts values higher than the maximum detection limit. B. Maximum projections of the leaf vasculature cross section. Pph RJ3 p nptII ::lux-p A1/04/03 ::eYFP colonies (yellow). Xylem autofluorescence (bright grey). Collenchyma autofluorescence (dark grey). co; collenchyma. xy; xylem. Red squares indicate areas shown at higher magnification in panels C-E. Scale bar: 40 μm. C. Pph RJ3 p nptII ::lux-p A1/04/03 ::eYFP colonies (yellow; white arrow). Xylem autofluorescence (grey). D. Pph RJ3 p nptII ::lux-p A1/04/03 ::eYFP colonies (yellow; white arrow). Collenchyma autofluorescence (dark grey). E. Zoom of panel B. Pph RJ3 p nptII ::lux-p A1/04/03 ::eYFP colonies (yellow; white arrow). Collenchyma autofluorescence (dark grey). Scale bar: 10 μm. [Color figure can be viewed at wileyonlinelibrary.com] Bacterial bioluminescence can be used to monitor the effect of plant immune responses on bacterial growth and viability in planta Bacterial bioluminescence has been proposed as a highthroughput method to detect in planta growth of bacterial pathogens (Fan et al., 2008). Fan et al. (2008) showed that Pseudomonas syringae tagged with luxCDABE can be used as a quantitative assay to monitor bacterial growth and that CFU is correlated with bacterial bioluminescence. Moreover, luminescence allows the detection of bacterial growth without performing tissue extraction, serial dilutions, plating, and scoring, thus reducing the time needed for the experiment. We therefore examined whether the luminescent constructs developed in this study could also be used to rapidly and sensitively monitor the effect of different plant immune responses on bacterial growth and viability in planta. As expression of high levels of lux activity could have a detrimental effect on bacterial fitness in the plant environment, we examined whether a relatively weak promoter driving lux expression could be used for this purpose.
We tagged Pph 1302A ΔhrpA (Supporting Information  Table S2), Pph 1302A ΔxerC (Lovell et al., 2009) and Pph 1302A ΔavrPphB (Supporting Information Table S2) using pRS-p OXB13 ::lux (Table 1; Supporting Information  Table S1). Pph 1302A ΔxerC (Lovell et al., 2009) lacks a recombinase (XerC) that is able to excise the genomic island containing avrPphB, thus it stably triggers ETI in P. vulgaris cultivar TG. On the contrary, Pph 1302A is able to excise the island and overcome resistance. Pph 1302A ΔhrpA (this study; Supporting Information Table S2) cannot assemble a functioning type III secretion system (T3SS), thus it cannot secrete effectors to counteract plant defences. Therefore, its growth in planta is suppressed by PAMP-Triggered Immunity (PTI) (Jones and Dangl, 2006). Pph 1302A ΔPphB (this study; Supporting Information Table S2) has a functional T3SS, but lacks the avr gene avrPphB, thus it can suppress PTI and colonize TG leaves. This compatible interaction is named effector-triggered susceptibility (ETS) (Jones and Dangl, 2006).
Colony count analyses performed at 2 DPI agreed with luminescence data [ANOVA, F(2,14) = 56.73, P < 0.0001] identifying a statistically significant effect of strain. However, pairwise comparisons failed to distinguish between plants infected with Pph 1302A ΔxerC and Pph 1302A ΔPphB [t(14) = 1.6, P = 0.06] (Supporting Information Fig. 11A and B). The higher variability of colony count data compared to bioluminescence data ( Fig. 3A-C; Supporting Information Fig. S11B) may be associated with variability introduced during tissue extraction, serial dilution and plating that are avoided in bioluminescence detection. Therefore, the bioluminescent reporter constructs developed in this study could be used to effectively monitor the effect of plant immune responses on bacterial growth, spread and viability in interactions between Pph and P. vulgaris.

Conclusion
Dual bioluminescence and fluorescence reporter constructs have previously been described for both gramnegative and gram-positive bacteria and used to study transcription, metabolic activity and host colonization (Unge et al., 1999;Unge and Jansson, 2001;Perehinec et al., 2007;Benedetti et al., 2012;Kim et al., 2018). Here, we describe the development of a flexible library of mini-Tn7 constructs that can be used to tag Gramnegative bacteria with a combination of lux, lux-frp and lux-eYFP operons with a wide selection of promoter strengths to drive lux expression. The strongest promoter used (p nptII ), together with a modified RBS and removal of the MCS, gives a fourfold to sevenfold increase in luminescence over a previously described mini-Tn7-lux construct. We also show that introducing an additional FMN reductase can positively increase bioluminescence in some bacterial strains and that the relative performance of lux and ilux depends on the bacterial strain used.
We have demonstrated the application of these reporter constructs by showing that Pph is able to systemically colonize the leaves of P. vulgaris. By combining bioluminescence and fluorescence, the process of collecting samples to be imaged with confocal microscopy could be greatly accelerated. Interestingly, Pph did not form biofilms within the xylem vessels, where they were observed to be highly mobile. Instead Pph preferentially colonized collenchymatic cells. While previous studies have shown that Pph causes systemic symptoms in P. vulgaris (Zaiter and Coyne, 1984), the question of whether, and how Pph is able to move systemically through host plants has not been fully addressed. Having shown that Pph can colonize the leaf vascular system it will be of interest to elucidate the molecular mechanisms underpinning vascular colonization by Pph and to investigate the contribution of vascular colonization to the development of systemic symptoms and to seed-borne transmission of Pph.
Our results also show that bacterial bioluminescence can be used to rapidly detect the effect of different plant immune responses on bacterial viability and growth and to monitor the effectiveness of plant immune responses in restricting systemic colonization of plant tissues. Importantly, bioluminescence is a sensitive indicator of bacterial viability and metabolic activity as well as localisation, while the relatively high stability of fluorescent proteins means that they can be used to detect both living and metabolically inactive or dead cells. Thus, it is of interest to note that while ETI and PTI could be observed to restrict the movement and population levels of Pph in infected tissue, we were still able to detect a substantial number of bioluminescent cells 2 days, and even 20 days after infection ( Fig. 3A; data not shown), showing that neither PTI nor ETI eliminated invading bacteria. Bioluminescent imaging may be of particular value in studying diseases such as bacterial leaf streak of wheat caused by Xanthomonas translucens pv. undulosa, or bacterial blight of rice caused by Xanthomonas oryzae pv. oryzae, in which resistance is typically assessed by monitoring lesion length or area, allowing rapid comparative imaging of bacterial spread within host plants even before lesions develop (Kandel et al., 2012;Oliva et al., 2019;Sapkota et al., 2020). We have already confirmed that the constructs described here can be used to transform and produce bioluminescence in Xanthomonas (data not shown).
Mini-Tn7 constructs have previously been used to transform a wide range of Proteobacteria by exploiting the presence of a conserved attTn7 site downstream of the glmS gene, in which the effects of transposon insertion have generally been found to be relatively neutral Choi and Schweizer, 2006). More recently, researchers have shown that the range of bacteria that can be tagged using this system can be further expanded through the introduction of an artificial attTn7 site, which can also be used to extend the use of this system to bacteria in which insertion into the pre-existing attTn7 site has been found to not be neutral (Figueroa-Cuilan et al., 2016). We therefore believe that the constructs and approaches described in this study will be a valuable resource for researchers who are interested in tracking both the spatio-temporal and physiological dynamics of bacterial populations in a wide range of biological contexts.

Plant growth conditions
Phaseolus vulgaris cultivars Tendergreen (TG) and Canadian Wonder were grown at 20-22 C, 70% humidity and with artificial light maintained for 8 h periods within the 24-h cycle. Two-week old plants (first true leaves fully developed) were used for all experiments.

Cloning, restriction enzyme digestion and polymerase chain reaction
Cloning was performed using Gibson assembly (Gibson et al., 2009) and constructs heat-shock transformed into E. coli strain DH5α. Transformants harbouring the lux operon were screened for luminescence using a nightOWL LB 983 in vivo imaging system (Berthold Technologies, Germany). For constructs combining luminescence and fluorescence, transformants were screened for both luminescence and fluorescence using a nightOWL LB 983 in vivo imaging system (Berthold Technologies) and a FastGene blue light LED illuminator (Geneflow Ltd, UK) respectively. PCR was conducted according to the manufacturer's instructions using Q5 high-fidelity DNA polymerase (Thermo Fisher Scientific) unless otherwise specified. Colony-PCR was performed according to the manufacturer's instructions using GoTaq Green Master Mix (Promega Corporation, USA). DNA Sanger sequencing was performed by Source Bioscience (UK). Restriction enzyme digestions were performed according to the manufacturer's instructions (Thermo Fisher Scientific).
Generation of a low copy-number plasmid harbouring ilux and lux constructs pRSJ-p nptII ::ilux was derived from pIJ11282 (Frederix et al., 2014) and pGEX(−) (Gregor et al., 2018). Briefly, pIJ11282 was digested with PstI and SnaBI to remove the lux operon. ilux was amplified with primers OXRS_F1 and OXRS_R1 (Supporting Information Table S3) from pGEX(−). The PCR product was ligated into PstI/SnaBIdigested pIJ11282 and transformed into E. coli. Potential clones were selected in LB agar supplemented with 10 μg ml −1 tetracycline. DNA sequencing was performed with primers listed in the Supporting Information Table S3.
Potential clones were selected on LB agar supplemented with 10 μg ml −1 gentamicin. DNA sequencing was performed using primer OX26 (Supporting Informationy Table S4).
pRS-p nptII ::lux was generated by ligating lux amplified with primers OXRS_F3 and OXRS_R3 (Supporting Information Table S3) from pIJ11282 into the pUC18T backbone amplified with primers OXRS_F4 and OXRS_R4 (Supporting Information Table S3) from pUC18T-mini-Tn7T-Gm-dsRedExpress . Transformants were selected on LB agar supplemented with 10 μg ml −1 gentamicin. DNA sequencing was performed using primers listed in the Supporting Information Table S6.

Construction of Pph 1302A PphB and hrpA knock-out mutants
To generate the Pph 1302A avrPphB knock-out mutant (Pph 1302A PphB-), 1 kb genomic regions flanking PphB were PCR amplified with primers OXRS_F9 and OXRS_R9 (left end) and primers OXRS_F10 and OXRS_R10 (right end) (Supporting Information  Table S3). Primer OXRS_R9 was designed to include the starting codon of avrPphB while primer OXRS_R10 included the stop codon of avrPphB. PCR products were ligated into pK18mobsacB (Schäfer et al., 1994) amplified with primers OXRS_F11 and OXRS_R11 (Supporting Information Table S3). Transformants were selected in LB agar supplemented with 20 μg ml −1 kanamycin. Colony-PCR was performed with primers OXRS_F12 and OXRS_R12 (Supporting Information  Table S3) to validate the assembly. The plasmid was conjugated into Pph 1302A, as described below, and transformants were plated onto LB plates supplemented with kanamycin and nitrofurantoin. Transformants were the result of plasmid integration (a single recombination event). To allow allelic exchange, four colonies were cultured overnight in LB and 20 μl plated onto LB (1.5% agar) supplemented with 10% sucrose. Colony PCR with primers OXRS_F9 and OXRS_R10 (Supporting Information Table S3) was used to confirm the deletion of avrPphB from the Pph 1302A genome.
A similar procedure was used to generate Pph 1302A ΔhrpA. Briefly, hrpA upstream and downstream regions were amplified from the Pph 1302A genome with primers OXRS_F13, OXRS_R13 and OXRS_F14, OXRS_R14 (Supporting Information Table S3) and ligated into pK18mobsacB amplified with primers OXRS_F11 and OXRS_R11 (Supporting Information Table S3). Conjugation and selections of knock-out mutants were carried out as described for Pph 1302A ΔPphB.

Mini-Tn7 delivery by four-parental mating conjugation
Pph 1302A strains were tagged by four-parental mating conjugation (Choi and Schweizer, 2006) with minor modifications. Details of the protocol used can be found in the Supporting Information File S1. Pph 1302A transformants were screened for luminescence (if tagged with lux constructs) and luminescence and fluorescence (if tagged with lux and eYFP constructs) as reported for E. coli. The correct insertion of the mini-Tn7 delivery construct was verified by colony-PCR using primers OXRS_F15 and OXRS_R15 (Supporting Information Table S3). Primer OXRS_F15 anneals in the glmS gene of Pph and primer OXRS_R15 binds in the Tn7R (right end of the Tn7 transposon).

Three-parental mating conjugation
Pseudomonas and Acinetobacter strains (Supporting Information Table S2) were inoculated from frozen glycerol stock onto Luria Bertani (LB) (Sambrook et al., 1989) agar (1.5% agar) and grown for 24 h at 28 C. Single colonies were used to inoculate 10 ml LB cultures, which were incubated at 28 C with shaking at 200 rpm. Bacterial cultures were then used for three-parental mating conjugation as described for four-parental mating conjugation (Supporting Information File S1), using the helper plasmid pRK2013 (Knauf and Nester, 1982).

Plate reader assay
Bacterial cultures were inoculated from frozen glycerol stock onto Luria Bertani (LB) (Sambrook et al., 1989) agar (1.5% agar) and grown for 24 h at 28 C. Single colonies were used to inoculate 10 ml LB cultures, which were incubated at 28 C with shaking at 200 rpm. After 24 h, 500 μl pre-culture was used to set up an overnight 10 ml LB culture. Cultures were centrifuged at 4000g for 6 min and resuspended in 10 ml 10 mM MgCl 2. The 15 μl of resuspended bacterial cultures were added to 135 μl King's B medium (KB) (King et al., 1954) or M9 medium (20% glucose) (CSH protocols) placed into a 96-well assay black plate (Corning incorporated, USA). Luminescence and OD 600 were measured using a Infinite M200 plate reader (Tecan Group Ltd, Switzerland) over 20 h. The temperature was set at 28 C and measurements were taken every 30 min.

NADP+ assay
Bacterial cultures were inoculated from frozen glycerol stock onto Luria Bertani (LB) (Sambrook et al., 1989) agar (1.5% agar) supplemented with 10 μg ml −1 tetracycline and grown for 24 h at 28 C. Single colonies were used to inoculate 10 ml LB cultures supplemented with 10 μg ml −1 tetracycline, which were incubated at 28 C with shaking at 200 rpm. After 24 h, 200 μl pre-culture was used to set up 200 ml LB cultures supplemented with 10 μg ml −1 tetracycline which were incubated at 28 C with shaking at 200 rpm for 6 h. Cultures were processed according to the manufacturer's instructions (NADP/NADPH assay ab176724, Abcam, UK) at a final OD 600 of 3,3.

Determining the effect of bacterial bioluminescence expression on bacterial growth in planta
Pph strains were inoculated from frozen glycerol stock onto Luria Bertani (LB) (Sambrook et al., 1989) agar (1.5% agar) and grown for 24 h at 28 C. Single colonies were used to inoculate 10 ml LB cultures, which were incubated at 28 C with shaking at 200 rpm. After 24 h, 500 μl pre-culture was used to set up an overnight 10 ml LB culture. Cultures were centrifuged at 4000g for 6 min and resuspended in 10 ml 10 mM MgCl 2 (5 × 10 6 CFU ml −1 ). Four plants per bacterial culture were used for the experiment. Briefly, the first two fully developed leaves of 12 two-weeks old plants (P. vulgaris cultivar Canadian Wonder) were syringe-infiltrated with Pph 1302A p nptII :: lux p A1/04/03 , Pph 1302A p OXB16 ::lux p A1/04/03 , Pph 1302A p OXB11 ::lux p A1/04/03 and Pph 1302A (5 × 10 6 CFU ml −1 ) (Supporting Information Table S2). Each leaf was infiltrated four times in four small areas (50-80 mm 2 ). The four spots within the same leaf were considered technical replicates and were averaged for the calculation of the CFU mm −2 . At 2 DPI (Days Post Inoculation) four leaf discs per leaf (four leaves total per bacterial culture) were detached from the four inoculated spots with a 1 cm 2 cork and pooled together. Samples were ground with a TissueLyser (Qiagen) in 1 ml 10 mM MgCl 2 , and serial dilutions (20 μl drops) were spotted onto LB agar supplemented with nitrofurantoin 25 μg ml −1 . Plates were incubated at 28 C for 3 days. At 5 DPI, the same process described above was repeated.
Determining the detection limit of Pph RJ3 p nptII ::lux p A1/04/03 ::eYFP Bacterial cultures were prepared as described above. Serial dilutions (100/1000 μl) were performed in 10 mM MgCl 2 and 100 μl aliquots were spotted onto LB (1.5% agar) square plates. Plates were incubated at 28 C for 1 h before imaging with the nightOWL LB 983 in vivo imaging system (Berthold Technologies) at 20 min exposure. Luminescence values and area (mm 2 ) of each spot were measured for the 10 −1 , 10 −2 , 10 −3 , and 10 −4 dilutions. Plates were then incubated at 28 C for 3 days to allow bacterial growth and colony count to be performed.
Combined luminescence-fluorescence approach to monitor Pph dispersal in planta Syringe-infiltration. Pph RJ3 p nptII ::lux p A1/04/03 ::eYFP (Supporting Information Table S1) was inoculated from frozen glycerol stock onto Luria Bertani (LB) (Sambrook et al., 1989) agar (1.5% agar) and grown for 24 h at 28 C. Single colonies were used to inoculate 10 ml LB cultures, which were incubated at 28 C with shaking at 200 rpm. After 24 h, 500 μl pre-culture was used to set up an overnight 10 ml LB culture. Cultures were centrifuged at 4000g for 6 min and resuspended in 20 ml 10 mM MgCl 2 . The bacterial suspension (5 × 10 6 CFU ml −1 ) was syringe-infiltrated into abaxial areas of the first true leaves of TG plants without causing any damage to the leaf mesophyll. To minimize damage to the leaf while infiltrating, we modified a 10 ml syringe by attaching to it a P20 pipette tip. We removed the last 1 cm of the P20 pipette tip and attached to it a rubber O-ring to ensure a soft contact between the tip and the leaf mesophyll. Plants were incubated for 0,1,3 and 5 days at 22 C and artificial light was maintained for 16 h within the 24-h cycle.
At designated times (0, 1, 3, and 5 DPI), leaves were detached and placed on square Petri dishes. Leaf petioles were wrapped in water-soaked cotton discs to allow plant transpiration. Leaves were incubated in the dark for 10 min before imaging with the nightOWL LB 983 in vivo imaging system (Berthold Technologies) at 10 min exposure. Bacterial bioluminescence was used to visualize Pph localisation and to identify portions of leaves to be imaged with confocal laser scanning microscopy. Leaf vasculatures showing Pph colonization were crosssectioned by hand-sectioning (50-80 μm) without embedding. Samples were imaged with a confocal laser scanning microscope Zeiss LSM 880 (Carl Zeiss, Germany). eYFP was excited at 514 nm (5.50%) and detected in the range of 520-610 nm; Phenolic compounds were excited at 405 nm (1.10%) and detected in the range of 200-450 nm to visualize xylem structure and collenchyma. Confocal stacks were acquired with a Z-step of 50-80 μm using the objectives C-Apochromat 40x/1.2 W Korr FCS M27.
Spray-inoculation. Pph RJ3 p nptII ::lux p A1/04/03 ::eYFP (Supporting Information Table S1) was incubated prior to inoculation as described above. The bacterial suspension (5 × 10 7 CFU ml −1 ) was deposited on restricted abaxial areas of the first true leaves of TG plants by spraying the resuspended bacterial culture through a round aperture (about 1 cm 2 ) cut into a folded aluminium foil. When spraying, close contact between leaves and aluminium foil ensured that there was no contamination of adjacent areas of the leaf. Four inoculated plants were placed inside a closed transparent box (30 × 20 × 30 cm) with two Oasis flower foams (Oasis Floral Products, UK) previously soaked in water. Plants were incubated for 3 days at 22 C and artificial light was maintained for 16 h within the 24-h cycle. Negative controls were represented by leaves inoculated solely with 10 mM MgCl 2 .
After 3 days, leaves were detached and placed on square Petri dishes. Leaf petioles were wrapped in water-soaked cotton discs to allow plant transpiration. Leaves were incubated in the dark for 10 min before imaging with the nightOWL LB 983 in vivo imaging system (Berthold Technologies) at 20 min exposure. Bacterial bioluminescence was used to visualize Pph localisation and to identify portions of leaves to be imaged with confocal laser scanning microscopy. Leaf vasculatures showing Pph colonization were crosssectioned by hand-sectioning (50-80 μm) or imaged as part of leaf discs (50 mm 2 ). Samples were imaged with a confocal laser scanning microscope Zeiss LSM 880 (Carl Zeiss).
eYFP was excited at 514 nm (5.50%) and detected in the range of 520-610 nm; phenolic compounds were excited at 405 nm (1.10%) and detected in the range of 200-450 nm to visualize xylem structure; and chlorophyll was excited at 514 nm (5.50%) and detected in the range of 650-720 nm.
Confocal stacks were acquired with a Z-step of 50-80 μm (for cross sections) or 10-30 μm (for leaf discs) using the objectives LD LCI Plan-Apochromat 25x/0.8 Imm Korr DIC M27. Leaf volume-rendering and three-dimensional models of the confocal stacks were created with the software Imaris 8 (Bitplane AG, Zürich, Switzerland).
Bacterial bioluminescence assays to detect the effect of plant immune responses on bacterial growth and viability Pph 1302A ΔhrpA (this study), Pph 1302A ΔPphB (this study) and Pph 1302A ΔxerC (Lovell et al., 2009) were tagged using pRS-p OXB13 ::lux (Table 1; Supporting Information Table S1) as described above. Bacterial cultures were inoculated from frozen glycerol stock onto Luria Bertani (LB) (Sambrook et al., 1989) agar (1.5% agar) and grown for 24 h at 28 C. Single colonies were used to inoculate 10 ml LB cultures, which were incubated at 28 C with shaking at 200 rpm. After 24 h, 500 μl preculture was used to set up an overnight 10 ml LB culture. Cultures were centrifuged at 4000g for 6 min and resuspended in 10 ml 10 mM MgCl 2 (10 6 CFU ml −1 ). Eight plants per bacterial culture were used for the experiment.
The optimal number of replicates were calculated with a jacknife resampling (Hanna, 1989) derived approach before performing the experiment. Briefly, the first two fully developed leaves of 12 two-week-old plants (P. vulgaris cultivar TG) were syringe-infiltrated with Pph 1302A ΔPphB p OXB13 ::lux (10 6 CFU ml −1 ) (Supporting Information Table S1). Each leaf was infiltrated six times in six small areas (50-80 mm 2 ). The six spots within the same leaf were considered technical replicates and were averaged for the calculation of the optimal sample size. Leaves were detached and incubated in the dark for 10 min before imaging with the nightOWL LB 983 in vivo imaging system (Berthold Technologies) at 2-min exposure. Luminescence values were recorded and used for the calculation of the optimal sample size. A custom R script (Supporting Information File S2) was used to randomly sample 1000 thousand times the generated data (12 values, in this case) with a jackknife approach. Standard deviation (SD) for each 1000 resampling per each sample size (from 12 to 1) was used to assess the difference in SD between sample sizes (12-n), where n is the sample size. The process was repeated 300 times and statistically significance differences of differences in SD between sample sizes were assessed with Benjamin-Hochberg (BH) post hoc test. Eight was assessed to be the optimal sample size as the reduction in SD from sample size 8 to sample size 9 was not statistically significant.
Subsequently, 2-week-old TG plants were syringeinfiltrated with p OXB13 ::lux expressing Pph 1302A ΔhrpA, Pph 1302A ΔPphB and Pph 1302A ΔxerC bacterial cultures (Supporting Information Table S2). The experiment was conducted according to a randomized complete block design (RCBD) (Liu and Berger, 2014) in which the block corresponded to the inoculation time. This was done to account for the passage of time from the first plant inoculations to the last plant inoculations (about 4 h). At time 1, three plants (two leaves per plant) were syringe-infiltrated with the corresponding bacterial cultures. One leaf per plant was detached and incubated in the dark for 10 min before imaging with the nightOWL LB 983 in vivo imaging system (Berthold Technologies) at 2-min exposure. Luminescence and area of each spot within leaves were recorded. Plants were then incubated for 2 days at 22 C and artificial light was maintained for 16 h within the 24-h cycle. The same procedure was repeated for subsequent time points. Within each block (time point and tray) plants were randomized. The same bacterial cultures were used for all time points. Negative controls were represented by plants infiltrated with 10 mM MgCl 2 and infiltrated with the non-tagged Pph 1302A ΔxerC bacterial culture to detect any bioluminescence associated with induction of plant immune responses (Bennett et al., 2005).
After 2 DPI, luminescence and the area where luminescence is detected were measured as described previously, for the remaining leaves. Exposure time was set at 5 min. Immediately after measuring bioluminescence signal, leaves were used to perform colony counts. Briefly, six leaf discs per leaf were detached from the six inoculated spots with a 1-cm 2 cork and pooled together. Samples were ground with a TissueLyser (Qiagen) in 1 ml 10 mM MgCl 2 , and serial dilutions (20 μl drops) were spotted onto LB agar supplemented with gentamicin 10 μg ml −1 and nitrofurantoin 25 μg ml −1 . Plates were incubated at 28 C for 3 days.

Data analysis and statistics
Processing of data was performed using R (RDC Team (R Development Core Team), 2006) version 3.5.3 and ggplot2 (Wickham, 2016). Requirements of linear models, namely normal distribution of the residuals and homogeneity of variance were assessed using diagnostic plots (Harrison et al., 2018). When diagnostic plots were not clearly interpretable, we used Leven's test to assess the homogeneity of variance. Independence of the observations was guaranteed by the experimental design.

Author contributions
R.S., N.S., and G.M.P. designed the study and the experiments. R.S. performed the experiments and analysed the data. M.B.P. performed the NADP+ assay and discussed the experiments with R.S, I.B. and W.H. contributed to the design for pRS p nptII ::ilux and I.B. also contributed to the design of the mini-Tn7 library. R.S. and G.M.P. wrote the manuscript.

Data Availability Statement
The research materials supporting this publication have been deposited in the Oxford Research Archive (ORA) as https://doi.org/10.5287/bodleian:y0r8ErQja and https:// doi.org/10.5287/bodleian:KOeJ8agdn. If you wish to access the biological material that is described in this study please contact gail.preston@plants.ox.ac.uk.