Bacterial Biosensors for in Vivo Spatiotemporal Mapping of Root Secretion1[CC-BY]

Development of a suite of rhizobial lux reporters to map in vivo root exudation, spatially and temporally. Plants engineer the rhizosphere to their advantage by secreting various nutrients and secondary metabolites. Coupling transcriptomic and metabolomic analyses of the pea (Pisum sativum) rhizosphere, a suite of bioreporters has been developed in Rhizobium leguminosarum bv viciae strain 3841, and these detect metabolites secreted by roots in space and time. Fourteen bacterial lux fusion bioreporters, specific for sugars, polyols, amino acids, organic acids, or flavonoids, have been validated in vitro and in vivo. Using different bacterial mutants (nodC and nifH), the process of colonization and symbiosis has been analyzed, revealing compounds important in the different steps of the rhizobium-legume association. Dicarboxylates and sucrose are the main carbon sources within the nodules; in ineffective (nifH) nodules, particularly low levels of sucrose were observed, suggesting that plant sanctions affect carbon supply to nodules. In contrast, high myo-inositol levels were observed prior to nodule formation and also in nifH senescent nodules. Amino acid biosensors showed different patterns: a γ-aminobutyrate biosensor was active only inside nodules, whereas the phenylalanine bioreporter showed a high signal also in the rhizosphere. The bioreporters were further validated in vetch (Vicia hirsuta), producing similar results. In addition, vetch exhibited a local increase of nod gene-inducing flavonoids at sites where nodules developed subsequently. These bioreporters will be particularly helpful in understanding the dynamics of root exudation and the role of different molecules secreted into the rhizosphere.

Due to root secretion, the narrow zone surrounding roots, known as the rhizosphere, is a nutrient-rich region where plants encounter a diversity of microbes, fungi, invertebrates, and the roots of other plants (Turner et al., 2013a). The constant improvement of sequencing techniques has rapidly increased the acquisition of knowledge about rhizosphere communities (Turner et al., 2013b). It is now evident that there is a two-way dialogue, with plants actively shaping their rhizosphere community; this, in turn, profoundly alters plant growth (e.g. by improving plant nutrient uptake; Philippot et al., 2013). Secretion patterns from roots differ between plants (Biedrzycki and Bais, 2009), and, despite much research on the role of chemical signals in mediating belowground interactions (Huang et al., 2014), many factors have not yet been identified (Badri et al., 2009). In addition, spatial and temporal variations in root secretions have never been clearly elucidated. Plant roots secrete large amounts of different compounds into the soil, and about 20% of photosynthate is released through the roots (Kaiser et al., 2015). As a consequence of this, compared with bulk soil, the rhizosphere is a rich source of compounds sustaining bacterial growth. This results in the attraction to the rhizosphere of many different microorganisms, among them both pathogens and plant growth-promoting bacteria (Huang et al., 2014). Different techniques have been used to study this complex environment, such as proteomics, metabolomics, and transcriptomics (for review, see Sørensen et al., 2009;Oburger and Schmidt, 2016), but the results obtained with many of these different methodologies are that they give only a single snapshot, as it has not been possible to follow the same plant during the course of its development. Current two-dimensional and three-dimensional noninvasive imaging techniques to examine the physical architecture of the plant roots using radiation-based techniques of x-ray microfocus-computed tomography or synchrotron tomography have been described (Mooney et al., 2011). These techniques are powerful at revealing root architecture and development but do not yield information about chemical secretion by roots.
In response to plant secretions, bacteria modify the expression of specific genes based on the molecules present in the rhizosphere (Ramachandran et al., 2011). Linking this with a method to monitor gene expression, we have used bacteria as biosensors to detect where and when specific molecules are secreted by plant roots. Bioluminescence is noninvasive and allows the measurement of in situ differences in the secretion of specific compounds in a semiquantitative way. Lux biosensors have already been used successfully (Darwent et al., 2003), but the lack of a simple system for image acquisition and complex experimental settings have so far limited routine use. Improvement of the technologies available and an increased knowledge of the bacterial transcriptomic response to roots now give us the chance to develop a suite of biosensors. These biosensors have been constructed using Rhizobium leguminosarum bv viciae. Rhizobia are a-proteobacteria, ubiquitous within soil and able to establish nitrogenfixing symbioses with specific legumes. The perception of environmental signals plays a pivotal role in the association between plants and bacteria (Pini et al., 2011), and R. leguminosarum bv viciae modifies its transcriptomic profile in different rhizospheres (Ramachandran et al., 2011). Moreover, the association of R. leguminosarum bv viciae with pea (Pisum sativum) has been studied in depth (Oldroyd et al., 2011;Terpolilli et al., 2012;Udvardi and Poole, 2013), making R. leguminosarum bv viciae one of the best candidates for biosensor development in which to investigate the rhizosphere. Pea and hairy vetch (Vicia hirsuta) plants have been used, allowing us to monitor both the rhizosphere and the process of nodulation. Applications of this methodology will be multiple and are not restricted to leguminous plants (e.g. screening plant mutant libraries for those altered in secretion from roots or observing different exudations during seed germination).

Solute Specificity of Biosensors
To spatially and temporally investigate secretion in the rhizosphere, we need biosensors able to detect compounds exuded into this environment. The idea that effective biosensors could be constructed using the control elements for the expression of specific bacterial genes, often encoding components of solute transporters or enzymes with precise substrate recognition, has been described previously (Tecon and van der Meer, 2006;Yagi, 2007;Sørensen et al., 2009). Induction of the expression of genes encoding transporters in the presence of their transported solute led to the identification of the substrates of many ATP-binding cassette (ABC) and tripartite ATP-independent periplasmic transporter systems of Sinorhizobium meliloti (Mauchline et al., 2006). Building on these data, Ramachandran et al. (2011) determined which genes of R. leguminosarum bv viciae strain 3841 (Rlv3841) are differentially expressed in the rhizospheres of pea, alfalfa (Medicago sativa), and sugar beet (Beta vulgaris), leading to the identification of metabolites inducing the expression of genes encoding both transport systems and metabolic enzymes. As many of these genes were induced in the rhizosphere in response to specific solutes, we sought to use their expression profiles to develop a suite of biosensors. These biosensors can be grouped by classification of inducer: (1) sugars and polyols, (2) organic acids, (3) amino acids, and (4) flavonoids (Table I). How each biosensor was selected is described below.
Eleven genes whose expression was induced in the rhizosphere during plant colonization by Rlv3841 of the pea rhizosphere (Ramachandran et al., 2011) were selected for biosensor development (Table I). These include seven solute-specific transport systems (which may transport plant-derived compounds or those from any source [e.g. fungi or other bacteria] within the rhizosphere) and four enzymes from metabolic pathways that are up-regulated during rhizosphere colonization (Table I). To increase the suite of biosensors, an additional two genes were included on the basis of microarray data that show differential expression in response to pea root exudate (Table I; Ramachandran et al., 2011). The 14th biosensor is based on a salicylic acid-inducible export system for salicylic acid (Tett et al., 2014), which is known to be an important signaling molecule in plant defense. While an inducer of gene expression was already known for many of these, the only previous characterization of pRL90085, RL4218, and pRL120556 was that their expression was induced either in the rhizosphere or by pea root exudate (Table II).
To develop the lux-based induction biosensors, the promoter regions upstream of the selected genes were used, often including the whole coding sequence of any upstream regulatory gene. Following PCR amplification, each region was cloned into the luminescence vector (pIJ11268, a plasmid stably inherited in rhizobia) in front of the bacterial luxCDABE operon (Frederix et al., 2014;Supplemental Table S1). As these biosensors are plasmid based, it is possible to transfer them into different bacterial backgrounds (e.g. mutant strains of R. leguminosarum or other species of bacteria), although their expression in heterologous hosts may be limited by regulatory elements (Galardini et al., 2015). Each of the 14 lux reporters was tested on 26 different sugars and polyols and/or on a set of 18 selected compounds (organic acids, amino acids, and plant metabolites) to establish the specificity of the induction of lux expression from each of the bioreporters (Table I;  Supplemental Table S2). Nine of the biosensors were induced by only a single compound tested: the polyols erythritol (pRL90085) and myo-inositol (RL4655; intA); the organic acids formate (RL4393; fdsG), malonate (RL0992; matA), tartrate (RL0996), and salicylic acid (RL1329; salA encoding an export system for salicylic Table I. Characterization of biosensors from R. leguminosarum CUT1 and CUT2, Carbohydrate uptake transporters 1 and 2; MFS, major facilitator superfamily; RLU, relative luminescence units; SBP, solute-binding protein. Inducers are solutes that give the highest fold induction and specific luminescence. Biosensor induction by a solute is described as specific if the specific luminescence is $10-fold that observed from the other solutes tested. The biosensors for erythritol, formate, and g-aminobutyrate (GABA) have relatively high expression with a variety of nonrelated solutes (background), and they are described as being specific for these solutes, respectively, with specific luminescence $4-fold values obtained with other solutes. More than one compound is considered inducing when the fold induction is greater than 40% of the maximum fold induction for a biosensor. Fold induction is the ratio of RLU/OD 600 when grown in the presence of solute (Supplemental Table S2  Genes induced greater than 3-fold, P # 0.05 in the pea rhizosphere (Ramachandran et al., 2011). b Genes induced greater than 3-fold, P # 0.05 by pea exudate (Ramachandran et al., 2011). c Fold induction is less than 40% that of the best inducer and, therefore, not considered to be specific induction. d Genes induced greater than 3-fold, P # 0.05 in the alfalfa rhizosphere (Ramachandran et al., 2011). acid induced by the presence of this molecule; Tett et al., 2014); the amino acids Phe (RL1860; phhA) and GABA (RL0102; gabT); and the flavonoid hesperetin (pRL100185; nodABC; Table I). These solutes are considered to be specific inducers (i.e. the specific luminescence is $10-fold the specific luminescence induced by the other solutes tested, apart from bioreporters for erythritol, formate, and GABA, which have relatively high expression in the presence of a variety of chemically unrelated substrates [background] but show $4-fold the specific luminescence of other solutes). However, this definition of specificity in no way excludes the possibility that these bioreporters also may react with other compounds similar to the molecules that elicit the primary response but that are not tested in this work (e.g. it is known that nod genes are induced by a variety of different flavonoids; Maxwell et al., 1989;Maj et al., 2010). Another two sugar biosensors were induced by two closely related compounds tested (Table I): (1) by two aldopentoses, Xyl or lyxose (RL2720; rbsC); or (2) induction by the disaccharide Suc also was achieved by the trisaccharide raffinose (pRL120556), presumably because either the biosensor recognizes the Suc 1-2 b-linkage in the disaccharide/trisaccharide or raffinose is metabolized to Suc. Another two bioreporters were induced by three closely related compounds: (1) the C4dicarboxylic organic acids succinate, malate, and Asp all cause the induction of a single biosensor (RL3424; dctA) and (2) the polyol mannitol (C6) is the main inducer in the biosensor (RL4218), although a weaker induction is elicited also by its isomer sorbitol and C5 adonitol (Table I). Sorbitol and adonitol are not considered specific inducers, as their specific luminescence is less than 40% that of mannitol (Table I). One bioreporter was induced by six of the tested sugars and polyols, with the main inducer being the monosaccharide Fru (RL0489; frcB). Induction also was seen with the disaccharides lactulose (4-O-b-D-galactopyranosylb-D-fructofuranose) and Suc [a-D-glucopyranosyl-(1→2)b-D-fructofuranoside]. These sugars all share a common structure containing a molecule of b-D-fructofuranose. Furthermore, this biosensor was induced by the polyols mannitol and sorbitol and the monosaccharide Man. In this case, it is likely that induction occurs due to the catabolism of these compounds to Fru. The 14 biosensors were used in the following experiments and henceforth are referred to by the compound that elicits the highest fold change in lux expression ( Table I). Provided that rhizobial survival is not affected, these biosensors should be able to detect the presence of the compound(s) for which they are specific, not only in the rhizosphere, whether they are derived from plants, from other bacteria, or from another source, but also in many other environments.
To discover the limits of detection for each biosensor, the sensitivity (the minimum concentration that causes induction) of the main inducer was determined using a range of different solute concentrations. The sensitivity ranges from the ability to detect levels of 1 mM for mannitol and hesperetin to 10 mM for formate and malonate (Table I).

Data Extraction from Images Using NightCROP
Plants were analyzed using NightOWL, a molecular imaging system linked to a sensitive CCD camera, which allows the measurement of light output from the Lux proteins. Images collected were first analyzed with the IndiGO software, but although it was used to obtain data for nodules and vetch plants (output in counts per second [cps] mm 22 ), a limitation of this software is that it is unable to do an automatic segmentation of the picture, so NightCROP, a custom MATLAB script, was developed. The script uses the SeNeCA algorithm (Tomek et al., 2013) to segment roots from the background in the light intensity channel, discarding those segments smaller than 1,000 pixels. The information on the position of the roots is then used on the luminescence channel to specifically detect the fluorescence signal coming from the roots (Supplemental Fig. S1).

In Vivo Mapping of the Bacterial Colonization of Pea Roots
Before using the suite of bioreporters to map the rhizosphere, we examined the colonization by R. leguminosarum bv viciae of pea roots in planta using Lux mapping. Rlv3841 containing pIJ11282, expressing Lux constitutively under the control of the nptII promoter (Frederix et al., 2014), was used to inoculate pea seedling roots, and the luminescent images revealed the location on the root of metabolically active Rlv3841. Lux expression relies on the bacteria having enough energy in the form of ATP to drive this energetically expensive production of light. Pea roots were imaged every 3 to 4 d until 22 d post inoculation (dpi), with nodules becoming visible to the naked eye at 11 to 15 dpi.
At 4 and 8 dpi, luminescence from Rlv3841 colonization was detected mainly in the elongation zone of the lateral and primary roots ( Fig. 1). At 11 dpi, overall luminescence was reduced, probably due to energy depletion of the bacteria ( Fig. 1), although a signal was still detectable at a lower level (Supplemental Fig. S2). After 15 dpi, there was an increase of the signal concurrent with nodule development, and luminescence was then stable in nodules until 18 dpi, decreasing at 22 dpi, probably due to a general decline of plant health under these growth conditions. This constitutive Lux fusion is an energy sensor that is very effective at showing initial colonization and the total energy available to the bacteria, but it must be used with caution in longer term imaging, due to the depletion of bacterial energy reserves resulting in a loss of signal.

In Vivo Mapping of Metabolites on Pea Roots and in Nodules
The composition of pea root exudate was determined by metabolomic analysis of hydroponically grown plants. Although plants grown in different conditions may differ in the composition of their root exudates, metabolomic analysis of these exudates is able to give important indications on compounds that it may be possible to retrieve from the rhizosphere. With the exception of formate, salicylic acid, and hesperetin, the targets of our biosensors were among the 376 compounds detected in exudate from 23-d-old peas (Table  II; Supplemental Table S3). A microarray experiment comparing Rlv3841 grown with and without the addition of the same 23-d-old pea root exudates used for the metabolomic analysis also was performed. Results from this are compared with a microarray where R. leguminosarum was inoculated into the pea rhizosphere and then harvested 1 dpi from 21-d-old plants (Ramachandran et al., 2011;Table II). Of the 14 genes used to develop biosensors, 11 (all except those detecting malonate, salicylic acid, and GABA) showed increased (greater than 3-fold) expression with added pea root exudate and/or in pea rhizosphere-grown cells (Ramachandran et al., 2011; Table II); however, relative increases in expression in these two experiments do not correlate particularly well. This discrepancy is likely to be due to differences in the concentrations of solutes in the two experiments. Indeed, compounds collected from the root (exudate Figure 1. In vivo spatial and temporal mapping images of pea root colonization and nodulation with wild-type Rlv3841 luminescently labeled with a constitutive neomycin phosphotransferase promoter controlling Lux expression in pIJ11282 (Frederix et al., 2014). A, Images were acquired at 4, 8, 11, 15, 18, and 22 dpi, with nodules visible to the naked eye at between 11 and 15 dpi (scale, 300-12,000 cps). B, Mean luminescence (pixels mm 22 ) with SE shown by error bars. sample) and then diluted into liquid medium are very unlikely to be present at the same concentrations as those in the experimental pea rhizosphere. In addition, the exact chemical composition of root exudates also will reflect the two different plant growth conditions. Despite these caveats, the expression of ;80% of the genes selected to make biosensors is elevated by the conditions of the microarray experiments with added root exudate and/or the pea rhizosphere, which suggests that they will be useful in investigating the pea rhizosphere in situ.
We analyzed lux expression on roots in vivo with each of the 14 biosensors in Rlv3841. To ensure that the expression of each biosensor is due to chemicals released from the plant roots, each bioreporter was spotted onto Fahraeus agar plates (supplemented with pyruvate and ammonium chloride) and incubated for 7 d (no growing plant present). The images (Supplemental Fig. S3) show no luminescence, indicating that the inducing compound is coming from the plant roots. To confirm that the biosensors respond to the same inducers on roots (in vivo) as they do in vitro, flooding experiments were undertaken. Pea roots at 4 dpi with a specific biosensor were flooded with a solution of a compound shown to induce Lux expression in vitro (Table I). Inoculation with the Xyl, Suc, or GABA biosensor and flooding with Xyl, Suc, or GABA, respectively, induced a Lux signal (Supplemental Fig. S4, A-C). Moreover, flooding the Suc biosensor-inoculated roots with GABA, and the GABA biosensor-inoculated roots with Suc, showed no increase in Lux expression (Supplemental Fig. S4, D and E).
Images representative of those obtained on pea roots for the biosensors for Suc (a sugar), myo-inositol (a polyol), malonate (an organic acid), and Phe (an amino acid) are shown in Figure 2, A to D. Luminescence images revealed that each of these four biosensors detected target metabolites during the 22-dpi period, and the results gave not only the location of the detected compound but also, by following the same plant over time, the changes that occur over the course of an experiment. Bearing in mind that the bacterial cells containing the bioreporter need to be metabolically active to generate a Lux signal, it is possible to get a false negative result (i.e. the inducing compound is present at levels above the minimum sensitivity but the cells do not have the energy required to produce the signal). The signal from a constitutive Lux fusion fades over several days because it places a substantial energy drain on cells. However, if Lux is detected, this indicates that the inducing compound is present. Analysis of the localization of luminescence in the images (four or more plants) and of the different temporal patterns of metabolite detection during the colonization and nodulation process (Figs. 2, E-H, and 3) revealed that there were similar detection profiles that could be grouped to aid analysis, although the scale and maximum values observed are different for each bioreporter.
The biosensor for the polyol myo-inositol (able to detect $100 mM myo-inositol Table I) was induced in the rhizosphere, mostly on the primary root and near the tips of lateral roots (Fig. 2, B and F), with a steady decrease in expression over time. Expression of the myoinositol reporter that was seen 15 dpi was mostly in nodules. A similar pattern was seen for the reporters detecting Xyl (able to detect $1 mM Xyl), Fru (able to detect $10 mM Fru), and the flavonoid hesperetin (able to detect $1 mM hesperetin; Fig. 3, A, B, and F; Supplemental Fig. S5, A, B, and F). Biosensors for these compounds were highly expressed in the rhizosphere at 4 to 8 dpi, usually localized at and above lateral root tips, and then, despite a general decrease in luminescence over the whole root, the compounds were detected almost exclusively in nodules at 15 to 22 dpi (Supplemental Fig. S5, A, B, and F). Expression of the organic acid C4-dicarboxylate biosensor indicated that succinate/malate/Asp (able to detect $100 mM/$10 mM/$10 mM succinate/malate/ Asp, respectively) are present in the rhizosphere and then found localized specifically to nodules at 15 to 18 dpi, with levels dropping by 22 dpi (Fig. 4D; A second expression profile, although similar to that described above, is that of biosensors that gave a strong signal on roots but were barely detectable within nodules. For example, the biosensor for malonate (able to detect $10 mM malonate; Fig. 2, C and G) was detected only in the rhizosphere (4-8 dpi), both on primary and lateral roots, with the highest levels appearing just before the root tips (Fig. 2C). The expression of this reporter fell over the time course (Fig. 2G) and was barely detectable in early nodules (11-15 dpi) and undetectable in older nodules (18-22 dpi; Fig. 2, C and G).
A third profile, typified by the Phe biosensor (able to detect $10 mM Phe), showed two peaks of total luminescence ( Fig. 2H) similar to that seen with the constitutive promoter (Fig. 1B), one at 8 dpi, with the signal localized to the root elongation zone of lateral roots, and a second peak both in the rhizosphere and in nodules (15 dpi;Fig. 2,D and H). Although the total signal from the Phe reporter detection fell over time, at 18 to 22 dpi, the luminescence was confined to nodules. By following the pattern of luminescence of a number of individual nodules, we conclude that the Phe concentration peaks in nodules and then falls as the nodule senesces. As new nodules are being initiated constantly over the time course analyzed, there were numerous bright spots, which got brighter as the nodule developed and then faded as nodules got older (Fig. 2D). The tartrate sensor (able to detect $100 mM tartrate) was expressed in a temporal pattern similar to that of Phe, with a similar dip in total levels of detection as nodules form at 11 dpi (Fig. 3D), although, in contrast to the Phe sensor, the tartrate reporter was expressed largely on the primary root (4-15 dpi; Supplemental  cps), myo-inositol (B; scale, 150-5,000 cps), malonate (C; scale, 50-2,000 cps), and Phe (D; scale, 150-15,000 cps). Images were acquired at 4, 8, 11, 15, 18, and 22 dpi, with nodules visible to the naked eye between 11 and 15 dpi. These images are representative of those from biosensors in the wild-type Rlv3841 background, which nodulates peas. E to H, Comparison of mean luminescence intensity per pixel from pea roots inoculated with biosensors in the wild-type Rlv3841 background (dark gray bars) and the Rlv3841 nodC128::Tn5 background (light gray bars). Only wild-type Rlv3841 can form nodules. The biosensors detect Suc (E), myo-inositol (F), malonate (G), and Phe (H). SE values are shown by error bars; asterisks indicate significant differences between biosensors in the wild-type Rlv3841 background and the Rlv3841 nodC128::Tn5 background A fourth profile was high expression of the reporter in nodules once they are formed, with very weak or no luminescence in the rhizosphere on pea roots in general prior to that. The biosensor for Suc typifies this group (Fig. 2, A and E). The total levels of expression of the Suc reporter (able to detect $100 mM Suc) did not peak until the nodules were more mature (15-18 dpi) and then fell as the nodules senesced (22 dpi; Fig. 2E). In the same way, the expression of the GABA reporter (able to detect $500 mM GABA) was hardly detected in the rhizosphere but was induced specifically in nodules, with total levels peaking at 15 to 18 dpi ( Fig. 3E;  Supplemental Fig. S5E).
The fifth profile was seen with a group of biosensors that gave results too low to properly evaluate, because luminescence was routinely detected below a mean intensity per pixel of ;30 (Supplemental Fig. S6). The polyol reporters for erythritol (able to detect $1 mM erythritol) and mannitol (able to detect $ 1 mM mannitol) were barely detected either in the rhizosphere or in nodules of pea plants (Supplemental Fig. S6, A and  B). The reporter for formate (able to detect $10 mM formate) was expressed at very low levels, and the expression of the salicylic acid reporter (able to detect $1 mM salicylate) was not detectable (Supplemental Fig. S6, C and D). This last result is not unexpected, because formate and salicylic acid were not found in the metabolomic analysis of pea root exudate (Supplemental Table S3).
Based on the spatial localization on pea roots of the reporters, we can draw conclusions about metabolites found in the rhizosphere prior to nodule formation (Table III). Xyl ($1 mM), Fru ($10 mM), myo-inositol ($100 mM), Phe ($10 mM), and hesperetin ($1 mM) are largely exuded by the elongation zone of primary and lateral roots (Fig. 2, B and D; Supplemental Fig. S5, A, B, and F). Malonate ($10 mM) was detected on both the uppermost portion of the primary root and the elongation zone of primary and lateral roots (Fig. 2C). Tartrate ($100 mM) exudation was localized exclusively to the uppermost portion of the primary root, with little or no tartrate being detectable on the lateral roots (Supplemental Fig. S5D). Although present in the pea rhizosphere, the localization of C4-dicarboxylates was not clear, because detection in different regions varied over time, with no clear pattern being observed (Supplemental Fig. S5C).
Between 11 and 15 dpi, nodules became visible on pea roots. The reporters indicate that Xyl ($1 mM), Fru ($10 mM), Suc ($100 mM), myo-inositol ($100 mM), C4-dicarboxylates ($10 mM-10 mM; see Table I), Phe ($10 mM), GABA ($500 mM), and hesperetin ($10 mM) were present in nodules at 15 dpi (Table III; Figs. 2, A, B, D-F, and H, and 4, B-E; Supplemental Fig. S7). Malonate ($10 mM) and tartrate ($100 mM) were detectable in nodules at 15 dpi but at very low levels of bioreporter expression (toward the lower limit of detection; Table  III; Fig. 2, C and G; Supplemental Fig. S5D). For both Phe and hesperetin, levels were highest in nodules at 15 dpi (dropping to 18 dpi and dropping further to 22 dpi; Table III ;Supplemental Figs. S5F and S7,C and D). In most of the biosensors, the signal from nodules was reduced strongly by 22 dpi, probably due to the general plant growth conditions, the only exception being the Fru reporter, where the signal increased until 22 dpi (Table III; Supplemental Fig. S7B).

Effect of Nodulation on Rhizosphere Metabolites
Since wild-type Rlv3841 induces nodule formation on pea roots, the effects of nodulation on pea root secretion were investigated by comparing the induction of the biosensors in the wild type with their induction in a mutant unable to induce nodulation (a derivative of Rlv3841 carrying nodC128::Tn5; Downie et al., 1985). Rhizobia produce Nod factors enabling recognition by legumes, and nodC mutants are unable to enter the plant or induce nodule formation (Udvardi and Poole, 2013). By comparing the results obtained with wild-type and mutant backgrounds, it is possible to separate the processes and metabolite changes of root colonization and nodule formation (Figs. 2 and 4; Supplemental Figs. S5, S8, and S9). Notably, in the prenodule formation stage at 4 to 8 dpi, no significant differences occurred between any biosensor in the two different backgrounds, other than with the myo-inositol reporter at 8 dpi (Fig. 2F). This suggests that the presence of Nod factor in itself does not alter either the amount or the localization of root secretions prior to nodule formation for the sugars Xyl, Fru, and Suc, the organic acids malonate, C4-dicarboxylates, and tartrate, or the amino acids Phe and GABA.
With the nodC mutant that is unable to form nodules, there was a significant decrease in the detection of myoinositol at 8 dpi relative to the wild type (Fig. 2F). In a wild-type strain at this time point, nodule initiation has begun, although nodules are not yet visible to the naked eye. The conclusion that myo-inositol (at concentrations $ 100 mM) is present in developing and very young nodules can be drawn from the relative decrease in lux expression of myo-inositol biosensor on roots inoculated with the nodC mutant. There was also a significant decrease in expression of the myo-inositol reporter in roots inoculated with the nodC mutant at each time point from 8 to 18 dpi, suggesting that the myo-inositol detected in the roots inoculated with the wild type is due to nodule formation (Fig. 2F). Indeed, expression of the myo-inositol reporter can be seen clearly localized to nodules (Fig. 2B). Lux output from the Suc reporter (detecting concentrations $ 100 mM Suc) was reduced significantly at 11 to 22 dpi in roots inoculated with the nodC mutant compared with the wild type (Fig. 2E); the expression of this reporter in roots inoculated with Rlv3841 was clearly seen localized to nodules at 15 to 22 dpi, and this is consistent with Suc from the shoot being supplied to nodules to support nitrogen fixation by rhizobial bacteroids (Fig.  2A). Expression of the Phe reporter (detecting concentrations $ 10 mM Phe) was reduced significantly only at 15 dpi in roots inoculated with the nodC mutant compared with the wild type (Fig. 2H), suggesting that Phe is abundant in nodules of this age. For xylose, Fru, C4-dicarboxylates, GABA, and hesperetin bioreporters, levels are significantly lower in the nodC Compound not listed as present in metabolomic analysis of pea exudate.  Supplemental Tables S6 and S7. mutant background at 15, 18, 15, 15-18, and 18 dpi, respectively (Fig. 3, B, C, E, and F), indicating the presence of these metabolites (Fru $ 10 mM, C4dicarboxylates $ 10 mM-10 mM, GABA $ 500 mM, and hesperetin $ 1 mM) in pea nodules.

Effect of Symbiotic Nitrogen Fixation on Nodule Metabolites
Biosensors detecting Suc (sugar), myo-inositol (polyol), C4-dicarboxylates (organic acid), and GABA (amino acid) were used to examine the levels of metabolites within effective and ineffective nodules by transferring each to a nifH mutant background ). The nifH mutant induces normal nodule formation, but it is totally defective for nitrogen fixation. Since nifH encodes one of the components of nitrogenase, the enzyme complex that carries out nitrogen fixation, interruption of this gene by mutation means that no functional nitrogenase is produced by the bacteria.
Malonate catabolism is not required for nitrogen fixation (Karunakaran et al., 2013). Transcriptomic data did not show a significant change in the expression of matABC genes in the rhizosphere of pea plants, but a significant difference was found in the alfalfa rhizosphere (Ramachandran et al., 2011). The malonate biosensor indicates that malonate (at concentrations $ 10 mM) is present in the pea rhizosphere, although over time its level decreases to being barely detectable in nodules at 15 dpi (Fig. 2, C and G). With evidence that $10 mM malonate is present in the rhizosphere of peas, colonization of roots by R. leguminosarum mutants defective for malonate metabolism was explored to see if an inability to metabolize malonate affects this process. Root attachment assays and nodule competition (as a measure of effective root colonization) were assayed using matC and matA mutants that are impaired in malonate transport and catabolism, respectively (Karunakaran et al., 2013). Although both mutants, defective in either malonate transport or catabolism, are less efficient at root attachment than the wild type, there were no significant differences in pea root colonization compared with the wild type (Supplemental Fig. S11). We conclude from this that malonate uptake and its subsequent bacterial catabolism play a part in the initial attachment of R. leguminosarum to pea roots. However, although attachment might be the first step of bacterial colonization, in overall colonization assays, the ability to either take up or metabolize malonate has no effect.

In Vivo Mapping of Metabolites on Roots and within Nodules of Vetch
To investigate a different legume root and its rhizosphere, similar analyses were performed with vetch, on which Rlv3841 also is able to form nodules. Images were acquired at similar time points as for pea (up to 22 dpi), but with vetch plants, the nodules appear to the naked eye earlier, at about 8 dpi ( Fig. 5; Table IV). With the exception of the polyols, erythritol and mannitol, and the organic acids, formate, tartrate, and salicylic acid, all other metabolites were detected in the vetch rhizosphere over the time course (Table IV). For the compounds we failed to detect, we cannot exclude the possibility that they may be present, but at levels too low for their detection (i.e. levels of erythritol at #1 mM, mannitol at #1 mM, formate at #10 mM, tartrate at #100 mM, and salicylic acid at #1 mM). This differs from pea only in the case of the tartrate, which was detected in the pea rhizosphere but not in that of vetch. The metabolites Xyl ($1 mM), Fru ($10 mM), myo-inositol ($100 mM), malonate ($10 mM), C4dicarboxylates ($10 mM-10 mM; Table I), and hesperetin ($1 mM) were detected in the vetch rhizosphere at 4 dpi (Fig. 5, A, B, D-F, and I). These were also all detected on pea roots (Table III), but a difference between these two legume rhizospheres is that Phe was barely detected in the vetch rhizosphere at 4 dpi (Fig .  5G; the limit of detection is $10 mM Phe). In vetch as in pea, Suc (able to detect $100 mM) and GABA (able to detect $500 mM) were hardly detected on the roots prior to nodule formation but were detected within nodules: for Suc, the maximum level was observed at 11 dpi, while that of GABA peaked at 8 dpi (Fig. 5, C and H). This pattern is similar overall to that seen on pea roots for both Suc (Fig. 2E) and GABA (Fig. 3E), taking into account that pea nodules form later (more than 11 dpi). Malonate ($10 mM) was detected in the vetch rhizosphere prior to nodule formation, but once nodules were formed (more than 8 dpi), it was detected only at a very low level (8-22 dpi; Fig. 5E), again, a similar pattern to that of pea (Fig. 2G). Hesperetin ($1 mM) was detected in vetch, with levels peaking at 4 dpi and then falling once nodule development was initiated (8 dpi onward; Fig. 5I). Images from 1 to 22 dpi for the detection of hesperetin on vetch roots (Fig. 6) show that, even at 1 to 4 dpi, prior to any sign of nodule formation, there are foci where hesperetin is detected. At later time points, nodules developed at these exact locations, illustrating the power of this technique in pinpointing the spatial and temporal secretion of this flavonoid (Fig. 6).

DISCUSSION
Owing to the hidden nature of the rhizosphere and its complexity, a major problem encountered in its study is the intrinsic difficulty of sampling (Bais et al., 2006). Most of the techniques used require sacrifice of the specimen or at least its manipulation, making it impossible to noninvasively follow the same sample over time. The bacterial lux gene cassette has been used widely in several different applications, including the visualization of gene expression, as a tool for cellular population monitoring and as a bioreporter target through activation under specific, predetermined conditions (Close et al., 2012). In this work, we have constructed a suite of Lux biosensors able to detect a variety of key sugars, polyols, organic acids, amino acids, and flavonoids that are commonly found in root exudates. The presence of these compounds was confirmed using a metabolomics approach, allowing us to identify 376 compounds present in pea root exudate. Lux-based reporter plasmids have been transferred into Rlv3841, an a-proteobacterium, which is generally associated with leguminous plants and is ubiquitous in soil (Udvardi and Poole, 2013). Pea and vetch plants, both hosts of R. leguminosarum, on which it forms nitrogen-fixing nodules, have been used to test the efficiency of this system as a proof of concept. However, the ability of rhizobia to colonize nonlegume plants (Chabot et al., 1996) should allow the use of these bioreporters in other systems.
A constitutive promoter was used to examine the rhizobial colonization of plant roots. Bacteria colonize the whole of the root system, but the strongest Lux signals are visible from the elongation zone of the primary and lateral roots ( Fig. 1; Supplemental Fig. S2). Heavy colonization at the root elongation zone is to be expected, as this is an area of actively growing root where many metabolites are secreted and exuded. Low Lux signal from the root cap is probably due to the reduced colonization of this area, which generally secretes antimicrobial phytochemicals (Baetz and Martinoia, 2014). The level of Lux signal detected in the rhizosphere was constant until 11 dpi and then was reduced, possibly due to a general decrease in root  Table S8. Nodules are visible to the naked eye from 8 dpi.
exudation caused by plant growth conditions and/or to the physiological status of the bacterial population. Constitutive lux expression drains energy reserves in bacterial cells and reduces light output over time. In a wild-type bacterial background, after 11 dpi, the Lux signal was localized mainly in the nitrogen-fixing nodules formed on the legume roots, which are densely populated with metabolically active bacteria. It is important to consider where bacteria are located. Although the whole root is colonized by rhizobia (Supplemental Fig. S2), these are unevenly distributed (i.e. more bacteria or more metabolically active bacteria are in the root elongation zone). The overall levels of metabolites detected in the rhizosphere before 11 dpi and in nodules (after 11 dpi) for pea roots are summarized in Figure 7.
On pea and vetch roots, Xyl ($1 mM), Fru ($10 mM), myo-inositol ($100 mM), Phe ($10 mM; not detected in the vetch rhizosphere), and hesperetin ($1 mM) were detected largely at the elongation zone (just behind the root tip) of lateral roots, while malonate ($10 mM; at least initially) and tartrate ($100 mM; not detected on vetch roots) were localized mainly on the primary root ( Fig. 2C; Supplemental Fig. S5D). In Rlv3841, malonate transport and metabolism has been shown to have no role in nitrogen fixation in peas (Karunakaran et al., Figure 6. Time course of hesperetin detection on a vetch seedling root from 1 to 22 dpi. Arrowheads indicate spots where luminescence is concentrated and a nodule forms later. 2013), but they do seem to be involved in the attachment of Rlv3841 to pea roots, as mutants in malonate uptake and metabolism show a reduced attachment phenotype, although no change in overall colonization of pea roots is observed. Malonate at $10 mM is localized to the primary root at 4 dpi during initial colonization (although levels #10 mM malonate may well be present on other parts of the root but are below the levels of detection by this method). Given the role of malonate in attachment, it is possible that bacteria attach to the primary root before colonizing the lateral roots, where most of the nodules appear subsequently. C4-dicarboxylates (succinate/malate/Asp with detection levels $100 mM/$10 mM/$10 mM, respectively) were detected in the rhizosphere of both pea and vetch, but they showed no clear spatial pattern on pea roots (Supplemental Fig. S3C).
In pea and vetch nodules, Xyl ($1 mM), Fru ($10 mM), Suc ($100 mM), myo-inositol ($100 mM), C4dicarboxylates (succinate/malate/Asp $ 10 mM-10 mM and Phe $ 10 mM), and GABA ($500 mM) were present (Tables III and IV ; Fig. 7). Although nodules formed on pea and vetch have not been analyzed by metabolomic studies, matrix-assisted laser-desorption ionization mass spectrometric analysis of Medicago truncatula nodules formed by S. meliloti (Ye et al., 2013) revealed many of these same compounds present: Suc, C4dicarboxylates (succinate/malate/Asp), and GABA; however, salicylic acid also was detected, which, if present in pea or vetch nodules, is below the limit of detection (#1 mM) of Lux-based salicylic acid detection. In pea, hesperetin ($1 mM) was detected in mature nodules (Fig. 3F), while in vetch, it was detected only before nodules could be seen with the naked eye or in very young nodules, with none detected in mature nodules (i.e. #1 mM; Fig. 6). Suc is supplied from the shoot to nodules, where it is converted to C4dicarboxylates and supplied to bacteroids as their primary energy source for nitrogen fixation (Poole and Allaway, 2000). In the ineffective nodules of a nifH mutant background, the levels of Suc and C4dicarboxylates detected were lower, suggesting that the plant sanctions the supply of carbon to nodules unable to provide them with nitrogen. While the nifH mutant showed lower levels of the constitutive promoter, presumably because of reduced bacteroid numbers and metabolic activity (Kiers et al., 2003;Berrabah et al., 2015), the levels of Suc, C4dicarboxylates, and GABA were still reduced substantially when the decrease in activity of the constitutive promoter was accounted for. Furthermore, the levels of some metabolites, such as myo-inositol increased dramatically in the nifH mutant, possibly due to environmental/osmotic stress.
The use of these tools has allowed us to draw a spatial and temporal map of key compounds present in the legume rhizosphere and to monitor the relative supply of specific metabolites inside nodules (e.g. Suc, Figure 7. Summary of metabolite detection on pea roots: in the rhizosphere (#11 dpi) and within nitrogen-fixing nodules ($15 dpi). Lines on the left side group similar chemicals into sugars and polyols, organic acids, amino acids, and flavonoids. Colors show levels detected in rhizosphere and nodules: high, red; medium, dark pink; low, pale pink; barely detected, gray; not detected, black.
C4-dicarboxylates, and GABA). We have demonstrated that, with this system, it is possible to follow the same plant for days, gathering data noninvasively, and it is relatively easy to set up. Moreover, it will be possible in the future to expand the set of reporters to include many different compounds. We believe that this is an excellent tool that can be adapted to investigate the roles of specific root exudates in many different plant growth conditions (e.g. stress, both abiotic and biotic). As R. leguminosarum colonizes root systems of nonleguminous plants (Schloter et al., 1997), it will be possible to monitor root exudates of other plant species using this series of biosensors. By combining this methodology with plant mutant collections, screening specific exudate-related phenotypes in genome-wide association studies could be performed. Finally, by coinoculating biosensors with other bacteria and/or fungi, it will be possible to observe the effects of other microorganisms on the plant secretome.

Strain Construction and General Techniques
The promoter region (often including the complete upstream regulator) of each of the candidate genes was amplified using the primers listed in Supplemental Table S9 with Phusion High-Fidelity DNA Polymerase (Thermo Fisher) according to the manufacturer's instructions. Fragments were purified and double digested with KpnI or XhoI (at the 59 end) and XhoI or BamHI (at the 39 end; Thermo Fisher). Restriction fragments were cloned into pIJ11268 (Frederix et al., 2014) digested with the same enzymes. Plasmids (Supplemental Table S1) were transferred into wild-type (Rlv3841), Rlv3841 nodC128::Tn5, Rlv3841 nifH::VSp ), Rlv3841 matA::pK19, and Rlv3841 matC::pK19 (Karunakaran et al., 2013) backgrounds by triparental mating according to Figurski and Helinski (1979). All plasmids are available from addgene (https://www.addgene.org).

Determining Solute Specificity
Each biosensor was grown for 3 d on a UMA slope with antibiotics, resuspended in UMS with no added carbon or nitrogen, and washed three times. Each was then diluted to an OD 600 of 0.01 in a final volume of 5 mL of UMS with a sugar (10 mM) as sole carbon source or with pyruvate (30 mM) as carbon source in the presence of a specific compound (Supplemental Table  S2) for 17 h. Luminescence (in RLU) and OD 600 were measured using the GloMax-Multi+ Detection System (Promega). For each compound, the fold induction is defined as the ratio of RLU/OD 600 when grown in the presence of that compound (Supplemental Table S2 gives concentrations of each solute) to that obtained in control conditions (UMS with pyruvate and ammonia). The solutes that give the highest fold induction and specific luminescence (Table I) are described as inducers. Biosensor induction by a solute is described as specific if the specific luminescence is $10-fold that observed for the other solutes tested. The biosensors for erythritol, formate, and GABA have relatively high expression with a variety of nonrelated solutes (background), and they are described as being specific for these respective solutes, with specific luminescence $4-fold values obtained with other solutes. More than one compound is considered inducing when the fold induction is greater than 40% of the maximum fold induction for a biosensor (Table I). For each biosensor grown on each solute, three independent cultures were measured. Supplemental Table S10 shows the expression of each gene used in biosensor construction in microarray experiments performed under 73 different conditions.

Plant Growth Conditions
Seeds of pea (Pisum sativum 'Avola') and vetch (Vicia hirsuta) were surface sterilized and germinated on distilled water agar plates. Plates with pea seedlings were put into black bags and incubated for 6 d at room temperature. Vetch seedlings were incubated overnight at 4°C and for 3 d at room temperature. Seedlings were transferred to 10-cm square petri dishes containing Fahraeus agar (Somasegaran and Hoben, 1994) covered with sterile filter paper (one seedling per plate for peas, six per plate for vetch). Each biosensor was analyzed with at least four plates for pea (corresponding to four plants) and one plate for vetch (corresponding to six plants). Biosensors were grown on UMA slopes for 3 d at 27°C, washed three times in UMS without any additions, and inoculated directly on the seedling root. Each seedling was inoculated with 5 3 10 7 or 2 3 10 7 colony-forming units (cfu) for vetch and pea plants, respectively. Plates were covered with aluminum foil to prevent exposure of roots to light and placed in growth chambers at 23°C with a 16-h/8-h day/night cycle for 22 d. In flooding experiments, peas were grown for 4 dpi with a biosensor before the plate was flooded with 10 mL of 10 mM solution of substrate and poured off. Plates were imaged before and 5 min, 3 h, and 21 h after flooding. Background luminescence in experiments with plants was evaluated by spotting 2 3 10 7 cfu mL 21 bacteria (the same amount used for pea root inoculation) onto Fahraeus agar plates supplemented with pyruvate and ammonium chloride. Plates were imaged after 7 d.

Image Acquisition
Plates were photographed using a NightOWL camera (Berthold Technologies) 4, 8, 11, 15, 18, and 22 dpi. CCD images of light output were exposed for 120 s. Each CCD image consisted of an array 1,024 by 1,024 pixels, and after acquisition, images were postprocessed for cosmic suppression and background correction. Images were analyzed with the imaging software IndiGO (Berthold Technologies) and with the custom software NightCROP. Night-CROP first segments an image using the SeNeCA algorithm (Tomek et al., 2013) into roots and background using a bright-field image, discarding objects smaller than 1,000 pixels. All subsequent analysis uses the respective fluorescence image. The script subtracts background intensity: background in the image is defined as a set of pixels given by logical inverse of the mask containing roots, subsequently morphologically eroded with a disk structural element of radius 9 to filter out signal from the edges of roots. Then, for each pixel belonging to a root, the mean intensity of background pixels in a 200-3 200-pixel square is subtracted. The output from NightCROP for a given image is expressed as the mean intensity of pixels labeled as roots after the background subtraction. Exact values of parameters (minimum root size, erosion radius, and background size) may be freely selected based on the resolution and nature of the data (Supplemental Fig. S1). Data are expressed as the ratio of luminescence to surface (cps mm 22 ; IndiGO) or intensity (pixels L 21 ; NightCROP).

Software Development and Use
NightCROP was developed as a script for the MATLAB environment (version 8.5.0 r2015a; MathWorks). Before using the script, an image-processing toolbox should be installed in MATLAB. The script is available upon request to the corresponding author and works on any operating system that supports MATLAB.

Extraction of Exudate from Roots
Glass jars (0.5 L) were prepared by filling one-third of the jar with glass beads (6 mm diameter; Atlas Ball & Bearing) and adding water (reverse osmosis highest quality) to cover all except the top layer of beads (;150 mL). Jars were covered with a metal lid containing a foam bung and sterilized by autoclaving. Six sets of 20 sterilized and germinated peas with 1-cm roots were transferred into six sterilized glass jars, with each jar (of 20 plants) representing one biological replicate. The jars were wrapped with black plastic up to the level of the beads, and the peas were grown for 21 d at 20°C for 16 h of light/18°C for 8 h of dark. The liquid was then decanted into sterile glass bottles and sampled for sterility by plating 100-mL aliquots on TY plates. Samples were filtered through a 0.2-mm nitrocellulose filter. Samples for metabolite profiling were freeze dried and resuspended in 800 mL of sterile water prior to downstream analysis.

Metabolite Profiling Analysis
The metabolomic profile of six biological replicates of pea root exudates was analyzed using nonbiased, global metabolome profiling technology based on gas chromatography-mass spectrometry (GC-MS) and UHLC/MS/MS 2 platforms (Lawton et al., 2008;Evans et al., 2009;Terpolilli et al., 2016) developed by Metabolon (www.metabolon.com). Samples from the six biological replicates were extracted using the automated MicroLab STAR system (Hamilton; www. hamiltoncompany.com). Recovery standards (Evans et al., 2009) were added prior to the first step in the extraction process for quality control purposes. The protein fraction was removed using methanol extraction, which allows maximum recovery of small molecules. The resulting extract was divided into two fractions: one for analysis by liquid chromatography and one for analysis by gas chromatography. Organic solvent was removed by placing samples on a TurboVap (Zymark). Each sample was frozen and dried under vacuum. Samples were then prepared for the appropriate instrument, either liquid chromatography-mass spectrometry or GC-MS.
The liquid chromatography-mass spectrometry portion of the platform was based on a Waters ACQUITY UHPLC device and a Thermo Finnigan LTQ mass spectrometer, which consisted of an electrospray ionization source and a linear ion-trap mass analyzer. The sample extract was split into two aliquots, dried, and then reconstituted in acidic or basic liquid chromatography-compatible solvents, each of which contained 11 or more injection standards at fixed concentrations. One aliquot was analyzed using acidic positive ion-optimized conditions and the other using basic negative ion-optimized conditions in two independent injections using separate dedicated columns. Extracts reconstituted in acidic conditions were gradient eluted using water and methanol both containing 0.1% (v/v) formic acid, while the basic extracts, which also used watermethanol, contained 6.5 mM NH 4 HCO 3 . The mass spectrometry analysis alternated between mass spectrometry and data-dependent MS 2 scans using dynamic exclusion.
Samples destined for GC-MS analysis were redried under vacuum desiccation for a minimum of 24 h prior to being derivatized under dried nitrogen using bistrimethyl-silyl-triflouroacetamide. The gas chromatography column was 5% (w/v) phenyl, and the temperature ramp was 40°C to 300°C over a 16-min period. Samples were analyzed on a Thermo Finnigan Trace DSQ fast-scanning single-quadrupole gas chromatograph-mass spectrometer using electron-impact ionization. For metabolite profiling, the identification of known chemical entities was based on comparison with metabolomic library entries of purified standards as described previously (Evans et al., 2009;Yobi et al., 2012).

Biosensor Sensitivity Assay
Bacterial biosensors from UMA slopes (with appropriate antibiotics) grown for 3 d at 28°C were washed three times before resuspension in 3 mL of UMS. Bacteria were added to 50 mL of molten (cooled to 42°C) UMA to give a final concentration of 1 3 10 8 cfu mL 21 . The agar containing bacteria was then poured into a 12-cm square petri dish and allowed to set. Droplets of 25 mL (n = 5) of 10-fold dilutions of solute, with concentrations ranging from 10 mM to 1 mM (including a distilled water control), were spotted onto the agar and incubated at 28°C. The sensitivity of each biosensor was defined as the lowest concentration of solute that gave a signal when imaged using a NightOWL camera at 4 h after spotting.

Accession Numbers
Microarray data were deposited in the ArrayExpress database (www.ebi.ac. uk/arrayexpress) under accession number E-MTAB-4790.

Supplemental Data
The following supplemental materials are available Supplemental Figure S1. Workflow of the NightCROP image-processing script; the image is of a pea inoculated with a constitutively expressed bioreporter at 15 dpi.
Supplemental Figure S2. In vivo spatial and temporal mapping images of pea root colonization and nodulation with wild-type Rlv3841.
Supplemental Figure S3. Luminescence is not detected in the absence of plant roots.
Supplemental Figure S4. In vivo spatial and temporal mapping images of pea roots inoculated with the Xyl biosensor.
Supplemental Figure S5. In vivo spatial and temporal mapping images of pea roots with biosensors detecting the sugars Xyl and Fru.
Supplemental Figure S6. Comparison of pea roots inoculated with biosensors in wild-type Rlv3841 and the Rlv3841 nodC128::Tn5 mutant background.
Supplemental Figure S7. Comparison of mean luminescence of pea nodules from 15 to 22 dpi with biosensors in the wild-type Rlv3841 background.
Supplemental Figure S8. In vivo spatial and temporal mapping images of pea roots with biosensors detecting Suc, myo-inositol malonate, and Phe.
Supplemental Figure S10. Comparison of mean luminescence from pea roots inoculated with biosensors in wild-type Rlv3841 and the Rlv3841 nifH::VSp mutant background.
Supplemental Figure S11. Investigation of the role of malonate during root attachment.
Supplemental Table S1. Strains and plasmids used in this work.
Supplemental Table S2. Conditions used to test the specificity of the bioreporter library.
Supplemental Table S3. Metabolomic data from root exudates of pea.
Supplemental Table S9. Primers used in this work.
Supplemental Table S10. Relative expression in microarray experiments under 73 different conditions of the R. leguminosarum genes whose promoters were used for the construction of biosensors.