Gene expression profiling of Escherichia coli in response to interactions with the lettuce rhizosphere


  • Z. Hou,

    1. Department of Food Science and Nutrition, University of Minnesota, St. Paul, MN, USA
    2. BioTechnology Institute, University of Minnesota, St. Paul, MN, USA
    Search for more papers by this author
  • R.C. Fink,

    1. Department of Food Science and Nutrition, University of Minnesota, St. Paul, MN, USA
    Search for more papers by this author
  • E.P. Black,

    1. Department of Food Science and Nutrition, University of Minnesota, St. Paul, MN, USA
    2. BioTechnology Institute, University of Minnesota, St. Paul, MN, USA
    Current affiliation:
    1. Ecolab Inc, St. Paul, MN, USA
    Search for more papers by this author
  • M. Sugawara,

    1. BioTechnology Institute, University of Minnesota, St. Paul, MN, USA
    2. Department of Soil, Water, and Climate, University of Minnesota, St. Paul, MN, USA
    Search for more papers by this author
  • Z. Zhang,

    1. BioTechnology Institute, University of Minnesota, St. Paul, MN, USA
    Current affiliation:
    1. Metabolix, Inc, Cambridge, MA, USA
    Search for more papers by this author
  • F. Diez-Gonzalez,

    1. Department of Food Science and Nutrition, University of Minnesota, St. Paul, MN, USA
    Search for more papers by this author
  • M.J. Sadowsky

    Corresponding author
    1. BioTechnology Institute, University of Minnesota, St. Paul, MN, USA
    2. Department of Soil, Water, and Climate, University of Minnesota, St. Paul, MN, USA
    • Department of Food Science and Nutrition, University of Minnesota, St. Paul, MN, USA
    Search for more papers by this author


Michael J. Sadowsky, 1991 Upper Buford Circle 439 Borlaug Hall, St Paul, MN 55108, USA. E-mail:



The objective of this study was to examine transcriptional changes in Escherichia coli when the bacterium was growing in the lettuce rhizoshpere.

Methods and Results

A combination of microarray analyses, colonization assays and confocal microscopy was used to gain a more complete understanding of bacterial genes involved in the colonization and growth of E. coli K12 in the lettuce root rhizosphere using a novel hydroponic assay system. After 3 days of interaction with lettuce roots, E. coli genes involved in protein synthesis, stress responses and attachment were up-regulated. Mutants in curli production (crl, csgA) and flagella synthesis (fliN) had a reduced capacity to attach to roots as determined by bacterial counts and by confocal laser scanning microscopy.


This study indicates that E. coli K12 has the capability to colonize lettuce roots by using attachment genes and can readily adapt to the rhizosphere of lettuce plants.

Significance and Impact of the Study

Results of this study show curli production and biofilm modulation genes are important for rhizosphere colonization and may provide useful targets to disrupt this process. Further studies using pathogenic strains will provide additional information about lettuce–E. coli interactions.


Fresh produce is a main component of a healthy diet, and its consumption is encouraged by government and health organizations in the United States (Berger et al. 2010). However, fresh produce, such as lettuce and spinach, is consumed raw and has become one of the major sources of food-borne infections in the US Escherichia coli strain O157:H7 (EHEC O157) is one of the pathogens most commonly linked to these outbreaks (Beuchat 1996; Doyle and Erickson 2008). Lettuce appears to be particularly susceptible to EHEC O157 contamination. Franz and van Bruggen (2008) reported that 29% of all lettuce-related outbreaks were caused by EHEC O157 and 38% of all EHEC O157 outbreaks with fresh produce were associated with lettuce. Infection of humans by EHEC O157 can be caused by very low bacterial numbers (approx. 100 cells g−1) in foods (Armstrong et al. 1996) and can lead to haemolytic uremic syndrome (HUS), a severe form of kidney failure and, in some cases, neurologic damage and death (Franz and van Bruggen 2008). Moreover, non-O157 Shiga toxin-producing E. coli on fresh produce is a rising concern in the United States and world-wide (Mathusa et al. 2010). The recent outbreak with sprouts caused by E. coli O104:H4 in Germany stresses the importance of EHEC associated with plants (Mora et al. 2011).

The interaction of plant pathogens with the rhizosphere of susceptible hosts can serve as a model for understanding the colonization of plant roots by food-borne pathogens. These interactions have been extensively characterized (Brandl 2006). For example, Ralstonia solanacearum uses chemotaxis to seek the best growth microenvironment in the rhizosphere (Yao and Allen 2006), and subsequently attaches to specific locations on the root surface. It was reported that this bacterium formed microcolonies using root exudates as a nutrient source (Garg et al. 2000). Ralstonia solanacearum was also able to invade the root and spread to the vegetative parts of the plant after overcoming the natural plant defence mechanisms. This same general mechanism is thought to also occur when E. coli interacts with the rhizosphere of susceptible plants.

Internalization might be a factor in the persistence of food-borne pathogens on lettuce (Solomon et al. 2002). Jablasone et al. (2005) described the interaction of EHEC O157 and Salmonella on seeds and seedling roots of Arabidopsis thaliana. During seed germination, pathogen numbers increased significantly, reaching high levels of contamination in the rhizosphere of the seedlings. Even after prolonged cultivation in a soil-free system, EHEC O157 and Salmonella could be recovered from the entire plant, including seeds, thus indicating that the pathogens were able to survive in the plant (Jablasone et al. 2005).

Unfortunately, modern molecular tools had not been applied the study of the interaction between E. coli and lettuce. E. coli K12 is probably the most widely characterized bacterium and it is a viable surrogate for pathogens (Ohnishi et al. 1999). The use of strain K12 presents a unique research opportunity as several extensive and well characterized mutant collections of E. coli K12 are available and greatly facilitate genetic investigations by providing single deletion mutations in all nonessential genes (Hayashi et al. 2001).

Here we report on the use of a hydroponic system to obtain a better understanding of the interaction of E. coli with lettuce plants. As bacterial invasion into lettuce plants via the root system has been hypothesized and reported under certain conditions, we used microarray analyses to understand gene expression of E. coli in the rhizosphere of lettuce roots. Microarray data were subsequently used to guide mutant studies that were then used to obtain a more precise mechanistic understanding of the complex plant-microbe interactions that occur in the lettuce rhizosphere.

Materials and methods

Bacterial strains and culture conditions

The strains and plasmid used in this study are listed in Table 1. All E. coli strains were cultured on Luria-Bertani (LB) medium at 37°C (Bowtell and Sambrook 2003). Mutants were maintained on LB medium supplemented with 50 μg ml−1 kanamycin or 100 μg ml−1 ampicillin, as needed. Cells were grown overnight in LB medium and collected by centrifugation at 9000 g for 10 min. Cells were washed twice with lettuce nutrient solution (Caspersen et al. 1999) and resuspended in the same solution for inoculation onto plants.

Table 1. Escherichia coli strains used in this study
E. coli K12 strainPlasmid or genotypeReference
MG1655 (wild-type)Wild-typeJensen (1993)
MG1655 puA66(puA66):kanrThis study
BW25113 (Parent strain for Keio mutant)Wild-typeBaba et al. (2006)
JW1297-1pspA:kanrBaba et al. (2006)
JW1298-1pspB:kanrBaba et al. (2006)
JW1299-1pspC:kanrBaba et al. (2006)
JW0230-1crl:kanrBaba et al. (2006)
JW1025-1csgA:kanrBaba et al. (2006)
JW1024-1csgB:kanrBaba et al. (2006)
JW5106-1ybiM:kanrBaba et al. (2006)
JW1930-1fliN:kanrBaba et al. (2006)
JW5316-1fliO:kanrBaba et al. (2006)
BW25113csgA:kanr(pGFPuv):AmprThis study
BW25113crl:kanr(pGFPuv):AmprThis study
BW25113fliN:kanr(pGFPuv):AmprThis study

Hydroponic growth and bacterial isolation

Romaine lettuce seeds were purchased from Jordan Seeds, Inc. (Woodbury, MN, USA) and surface sterilized using 10% H2O2 for 30 min. The sterilized seeds were incubated on LB agar for 24 h to verify sterilization efficiency. Seeds were aseptically germinated for 4 days in the dark at 25°C on 1% water agar, located in the wells of empty, sterile, pipette tip boxes (Tip One, USA Scientific, Ocala, FL, USA). Seedling trays were transferred into a hydroponic system (Fig. 1) and incubated for 7 days at 25°C under 80% r.h. Plants were grown with a photoperiod of 16 h. An aeration tube was inserted into the nutrient reservoir and air was provided by using an aquarium pump and an air stone (Coralife, Franklin, WI, USA).

Figure 1.

Photographs illustrating the hydroponic system used in this study.

The E. coli K12 strain MG1655 (Jensen 1993) was grown in LB broth and directly inoculated into the hydroponic tip box to a final concentration of 3·3 × 107 CFU ml−1 of nutrient solution. Each hydroponic system contained 300 ml of filter sterilized (0·2 μm Cellulose acetate filter, Millipore, Bedford, MA) nutrient solution consisting of 10 mmol l−1 KNO3, 4·5 mmol l−1 Ca(NO3)2, 1·0 mmol l−1 MgSO4, 1·0 mmol l−1 KH2PO4, 1·0 mmol l−1 Na2HPO4, 1·25 mmol l−1 NH4Cl, 40 μmol l−1 Fe-EDTA, 5 μmol l−1 MnSO4, 4 μmol l−1 ZnSO4, 30 μmol l−1 H3BO3, 0·75 μmol l−1 CuCl2 and 0·5 μmol l−1 Na2MoO4, pH 6·0 (Caspersen et al. 1999). Seedlings were incubated for 72 h in a plant growth chamber at 25°C with a 16 h photoperiod. For controls, the same amount of E. coli K12 was inoculated into the boxes containing plant nutrient solution without exposure to the plant roots, and incubated under the same conditions as described above. After 3 days of incubation, roots were gently rinsed to remove loosely attached bacteria, and roots were cut directly into 100 ml of ice cold stop solution (5% H2O–saturated phenol, pH 4·3, in 95% ethanol) to stop transcription. Roots were shaken in the stop solution for 30 min, at approximately 200 strokes per min by using a horizontal shaker, and bacterial cells were separated from plant cells and debris by filtration through 5·0 μm filters (Sterlitech, Kent, WA, USA). For the controls, 10 ml hydroponic solution only containing E. coli cells was also harvested after 3 days and treated using ice cold stop solution as described above. RNA protect solution (Qiagen, Valencia, CA, USA) was used for samples and control cells to stabilize RNA during extraction, and bacteria were stored in −80°C until used for RNA isolation.

RNA extraction, cDNA synthesis and microarray hybridization

Total RNA was extracted using a modified hot phenol-lysozyme method (Bowtell and Sambrook 2003). Total nucleic acid was precipitated with ethanol and treated with DNase I for 30 min. The concentration of RNA was measured spectrophotometrically at 260 nm, and RNA samples were electrophoresed in 0·8% agarose gels prior to use to ensure RNA integrity. Total RNA (10 μg) was converted to cDNA using the Superscript II Reverse transcriptase and directly labelled with Cy3 or Cy5 dyes (Invitrogen, Carlsbad, CA, USA) prior to hybridization. Microarray slides were hybridized with labelled cDNAs for 48 h at 42°C in a water bath, under dark conditions, as previously described (Cytryn et al. 2007).

Microarray expression analysis

Triplicate E. coli O157 microarrays (six slides) (OciChip,Ocimum Biosolutions Inc., Indianapolis, IN, USA) were used in this study. These microarrays consisted of 6176 50-mer oligonucleotides specific for E. coli strain K12, substrain MG1655. Following hybridization, microarray slides were washed twice with 2 × saline-sodium citrate (SSC) and 0·1% sodium dodecyl sulphate at 25°C for 5 min, twice with 1 ×  SSC at 25°C for 6 min, and three times with 0·1 ×  SSC at 25°C for 3 min. Microarrays slides were scanned with a GenePix 4000B scanner (Axon Instruments, Molecular Devices Corporation, Sunnyvale, CA, USA) and image analysis was carried out with the company's software program using a file that links the corresponding gene to each spot (Chang et al. 2007). Signal intensities for each spot on scanned image files were determined using GenePix® Pro 6.0 software (Molecular Devices Corp.).

Signal intensities were normalized for spot and slide abnormalities using the spatial Lowess algorithm and analysed by mixed-effect anova (maanova) (Kerr et al. 2000). Both Lowess and maanova programs were part of the R/maanova microarray statistical package The resulting variety-by-gene interaction (VG) values of the control and experimental spot intensities were combined with the residual noise from each spot to obtain filtered and adjusted expression values (Cui et al. 2003). Values from maanova were subsequently subjected to significance analysis of microarray (sam) data with the sam package (Tusher et al. 2001). Significantly up- and down-regulated genes were identified based on a 1·5 cutoff threshold, and a global false discovery rate < 5% (P < 0·05). Six arrays representing a total of three biological replicates were analysed between treatments and controls. Controls consisted of replicate microarray slides hybridized to E. coli MG1655 grown in hydroponic plant medium and not exposed to plant roots. The microarray data generated in this study were uploaded to the NCBI database and assigned accession number: GSE33735.

Quantitative reverse transcription PCR

The same RNA which used in microarray analyses was converted to cDNA using Superscript II Reverse Transcriptase as described above. The Quantitative reverse transcription PCR (qRT-PCR) was performed with the primers listed in Table 2, using an ABI 7500 real-time PCR system with sequence detection system software version 1.3 (Applied Biosystems, Foster City, CA, USA) and SYBR green PCR master mix (Applied Biosystems) as previously described (Sugawara et al. 2010). Data were obtained for two technical replicates of each biological replicate and normalized to the gyrB housekeeping gene (Fink et al. 2012). Statistical analyses were carried-out by using the Student's t-test of the xlstat program (2 Tails, Type 2, P ≤ 0·05).

Table 2. Primers used for qRT-PCR
GeneFunctionPrimer pairs (5′→3′)
Crl Transcriptional regulator of cryptic csgA gene for curli surface fibre



fliN FliGMN switch complex on the rotor of the bacterial flagellum




Colonization of lettuce roots by wild type and mutant E. coli MG1655

Wild type E. coli K 12 MG1655, BW25113, and mutants with insertions of a kanamycin gene cassette into individual genes of interest were individually inoculated onto lettuce roots to an initial concentration of 105 CFU ml−1 and incubated for 3 days Mutants with insertions in pspA, pspB, pspC, crl, csgA, csgB, ybiM, fliN, and fliO were from the Keio collection (Baba et al. 2006) and were obtained from the E. coli genetic stock centre For each experimental unit (e.g. plants belonging to one hydroponic system), the roots from ten plantlets were collected, and data were normalized to a CFU/plant root basis. Loosely attached bacteria were removed from roots by gently washing with water, and adhering bacteria were released from roots by shaking with glass beads (0·7–1·2 mm) for 30 min. Bacteria were enumerated on LB agar plates supplemented with 50 μg ml−1 kanamycin.

Observation of root-attached GFP-labeled E. coli strains by fluorescence microscopy

Plasmid pGFPuv (Clontech, Mountain View, CA, USA) was introduced into mutant strains with insertions in the crl, csgA and fliN genes by electroporation (Bloemberg et al. 1997). Strain MG1655 carrying the pUA66-promoter-gfp plasmid (Zaslaver et al. 2006) or the generated mutants carrying the pGFPuv plasmid were inoculated onto lettuce roots using the hydroponic system as described above. Confocal scanning laser microscopy (CSLM) was used to visualize bacterial colonization on root tissue. The CSLM system was a Nikon Eclipse C1si Confocal 60 × water scope (Melville, NY, USA) equipped with a motorized x, y, z stage, a filter for GFP (488 nm) and a 32-channel spectral detector. Digital images of attached bacteria were sequentially collected using Nikon's EZ-C1 acquisition and analysis software supporting standard and spectral confocal acquisition, spectral unmixing and 3D rendering visualization. Roots inoculated with wild-type and mutant strains were taken using CSLM after 1, 3, 6 and 10 days of incubation.


In this study, a novel hydroponic system was used to investigate the interaction of E. coli and lettuce roots in controlled and axenic conditions (Fig. 1). Microbiological tests of seed coats, seed germination and roots indicated that seeds were effectively sterilized without largely affecting the seed germination rate (data not shown).

Microarray expression studies

In our microarray analyses, 324 genes (7·6% of total genes) were differentially expressed between E. coli cells interacting with lettuce roots for 3 days, compared to the free-living controls incubated into the hydroponic system without lettuce roots. The expression of 193 genes (4·5%) increased at least 1·5-fold, and 131 genes (3·1%) were repressed more than 1·5-fold, when compared to the nonroot control systems (Table S1 and S2). The up- or down-regulated genes were categorized according to the Clusters of Orthologous Groups (COG) classification system (Table 3). Forty-five out of the 193 up-regulated genes (23%) were involved in protein synthesis (Table 3), including 19 genes encoding ribosomal protein subunits (rplACDEFMNRSU, rpsBDFJMU, rpmBCF). Similarly, 14 genes involving energy metabolism were induced by growth in the lettuce rhizosphere.

Table 3. Number of Escherichia coli K12 genes up- and down-regulated more than 1·5-fold after exposure to lettuce roots for 3 days
Functional gene groupaNo. of genes
  1. a

    Categories classified using Clusters of Orthologous Groups.

Amino acid biosynthesis21
Biological process30
Biosynthesis of cofactors14
Cell envelope59
Cellular processes1015
Central intermediary metabolism41
DNA metabolism31
Energy metabolism1414
Fatty acid and phospholipid metabolism43
Mobile and extrachromosomal elements34
Protein fate46
Protein synthesis045
Purines, pyrimidines, nucleosides and nucleotides02
Regulatory functions59
Transport and binding proteins710
Hypothetical proteins3429
Unclassified and unknown3137
Viral function10

Stress response-related genes were markedly up-regulated in bacterial cells associated with lettuce roots. In particular, the phage- (pspABCDE) and cold-shock (cspABEG) operons were on average induced 2·5 and 5·0-fold respectively. In contrast, the acid shock response gene asr was induced 19·3-fold in gene expression level (Table 4). The biofilm formation modulator ybiM was also significantly induced during growth on lettuce roots. In addition, crl, a regulator of the csgBA operon for curli production, was also significantly up-regulated (Table 4). Fimbrial genes, such as fimA, were also significantly induced (2·0-fold), as well as genes regulating cell division such as ftsJ, ftsN and ftsW. These genes were up-regulated 1·5, 1·5 and 3·3-fold respectively.

Table 4. Changes in gene expression of selected Escherichia coli K12 genes significantly up-regulated in association with the lettuce root rhizosphere
Protein function and gene IDGene functionFold change
ybiM (B0806)Protein involved in colanic acid production3·8
crl (B0240)Transcriptional regulator of cryptic csgA gene for curli surface fibres1·6
Cell envelope
fimA (B4314)Major type 1 subunit fimbrin2·0
flil (B1941)Flagellum-specific ATP synthase2·2
mrcB (B0149)Bifunctional penicillin-binding protein 1b: glycosyl transferase (N-terminal)2·6
ybiJ (B4045)Putative stress-response protein10·6
Cellular processes
ftsJ (B3179)Cell division1·5
ftsN(B3933)Cell division1·5
ftsW(B0089)Cell division3·3
B1530Repressor of mar operon1·7
B1531Transcriptional activator of defence systems1·8
Stress responses
asr (B1597)Acid shock protein19·3
cspA (B3556)Cold shock protein5·7
cspB(B1557)Cold shock protein4·9
cspF (B1558)Cold shock protein2·5
cspG (B0990)Cold shock protein5·0
cspI (B1552)Cold shock protein2·2
pspA (B1304)Phage shock protein2·1
pspB (B1305)Phage shock protein3·4
pspC (B1306)Phage shock protein2·3
Regulatory function
rseA (B2572)Anti-sigma E (sigma 24) factor, negative regulator1·7
rseB (B2571)Anti-sigma E (sigma 24) factor, negative regulator1·8
rpoE (B2573)RNA polymerase, sigma-E factor1·8
Fur (B0683)Negative regulator1·8
dps (B0812)Global regulator, starvation conditions1·9
Protein synthesis
rpsD (B3296)30S ribosomal subunit protein S42·9
B329730S ribosomal subunit protein S112·5
rpsM (B3298)30S ribosomal subunit protein S132·6
rpmJ (B3299)30S ribosomal subunit protein L361·6
rplR (B3304)50S ribosomal subunit protein L182·4
rplF (B3305)50S ribosomal subunit protein L61·8
B330630S ribosomal subunit protein S82·4
B330730S ribosomal subunit protein S143·0
rplE (B3308)30S ribosomal subunit protein L52·0
rplX (B3309)50S ribosomal subunit protein L243·3
rplN (B3310)50S ribosomal subunit protein L142·4
rpmC (B3312)50S ribosomal subunit protein L291·7
B331550S ribosomal subunit protein L221·9
rplW (B3318)50S ribosomal subunit protein L232·6
rplD (B3319)50S ribosomal subunit protein L41·7
rplC (B3320)50S ribosomal subunit protein L32·0
rpsJ (B3321)30S ribosomal subunit protein S102·3

In contrast, 14 genes involved in energy metabolism, including a gene encoding for glycogen synthesis (glgS), were repressed following growth in the lettuce rhizosphere. The flagella synthesis gene fliO was also down-regulated 1·7-fold in E. coli cells associated with lettuce roots. Similarly, the acrB, emrA and cutA genes, involved in multidrug resistance, were also significantly repressed 3·6, 1·6 and 1·7-fold respectively (Table S2). No genes belonging to the COG category' protein synthesis was down-regulated during the interaction with the lettuce rhizosphere.

Physiological role of selected genes

Knockout mutants of genes of interest, obtained from the Keio collection, and two wild-type K12 strains were evaluated for their ability to survive and attach to the lettuce roots. To ascertain the physiological role of the pspABC and ybiM genes in the pathogen-plant rhizospehere interaction indicated by the microarrays, we selected mutants for different genes belonging to these operons. The crl gene is suspected to be involved in attachment and colonization, and indeed it was up-regulated. However, the crl-controlled genes csgA and csgB were not detected in our microarray experiments. Based on their interrelationship, though, we selected mutants for these genes. Furthermore, to evaluate the importance of motility, we included in our analysis two motility impaired mutants, strains with mutations in the flagella synthesis regulator fliO and the flagellar motor switch fliN. The fliN gene was not included in those genes spotted on the microarray slides used for our experiments.

After 3 days of incubation of bacterial cells with lettuce roots, counts of the wild-type K12 strains (MG1655 and BW25113) were approximately 7·3 log CFU per plant root (Fig. 2). In contrast, final counts of the crl, csgA and fliN mutants were reduced at least 2 log CFU per plant root, compared to wild-type E. coli K12 in the hydroponic system (Fig. 2). The mutants for crl and csgA, regulating curli production, colonized the roots at a significantly lower level compared to the wild-type controls. However, the csgB gene, encoding a curli subunit necessary for curli assembly, colonized the roots to the same levels as did the wild-type strains. Final counts of the fliN mutant were also 2 log CFU per plant root lower than the control strains, but the fliO mutant had similar population numbers as did the wild-type strains. Similarly, strains carrying a deletion in the phage shock protein (pspABC) and biofilm formation (ybiM) gene attained similar counts as did the wild-type strains when inoculated onto lettuce roots.

Figure 2.

Colonization of Escherichia coli K12 and mutants on lettuce roots after 3 days. Gene identification corresponds to specific BW25113 mutants examined. Error bars represent the standard deviations of the means of three independent biological experiments.

Quantitative reverse transcriptase PCR analysis

Because attachment of the crl, csgA and fliN mutants to the root system was clearly impaired relative to the wild-type strains, and to better understand the role of curli and flagella in the plant-microbe interactions, we used a qRT-PCR assay to evaluate the extent of the changes in expression of these genes (Fig. 3). These studies confirmed that crl was significantly up-regulated when wild-type strains interacted with lettuce roots (32·5-fold) (Fig. 3). While the csgA and fliN were up-regulated, relative to controls, gene expression was not significantly increased. (P > 0·05).

Figure 3.

Absolute induction of genes of Escherichia coli K12 identified by microarray analysis associated with lettuce roots for 3 days measured by quantitative RT-PCR. Values were normalized to the constitutively expressed gyrB gene. Error bars represent the standard deviations of the means of three independent biological experiments. Means in each column followed by the same letter are not significantly different according to student t-test (P ≤ 0·05) (image – sample; image – control).

Microscopic cell attachment and colonization visualization

The CLSM image analyses showed that the wild-type E. coli cells increasingly attached and colonized the lettuce roots during the duration of the experiment (Fig. 4, panels a, b, c and d). From day one to six, the bacteria attached and colonized the lettuce roots individually or forming small clusters. After ten days, we observed larger and diffused bacterial colonies on the lettuce roots, some likely in biofilms. In contrast, attachment and growth of the crl, csgA and fliN mutants on lettuce roots after 3 days of incubation were reduced relative to the wild-type control cells (Fig. 4f–h).

Figure 4.

Confocal laser-scanning microscopy images of GFP labelled Escherichia coli attached to lettuce roots. Wild-type cells after: (a) 1, (b) 3, (c) 6 and (d) 10 days of incubation. Wild type (e), crl (f), csgA (g) and fliN (h) mutant cells after 3 days of incubation.


Recent studies indicated that E. coli possess unique characteristics favouring its attachment, survival, growth and internalization into plant tissue (Beuchat 1996; Doyle and Erickson 2008). However, the molecular mechanisms involved in regulating the interaction of this bacterial species with fresh produce are still largely unknown. Genomic approaches provide useful information about such mechanisms and have been successfully used in studying the interaction of E. coli with lettuce leaves (Fink et al. 2012). However, to our knowledge this approach has never been employed to determine the molecular mechanisms involved in the interaction of this micro-organism with lettuce roots. This knowledge is particularly important if we consider that roots are a potential route of plant contamination through the application of contaminated manure, soil or water (Doyle and Erickson 2008). In particular, bacteria attached to the roots can become internalized and further survive postwashing processes (Solomon et al. 2002).

A major problem in dissecting the relationship between a micro-organism, the complex environment represented by the root, and the growing medium is the lack of a system providing perfectly controlled and axenic environmental conditions in a simple laboratory setting. To this purpose, we developed a time and labour-effective hydroponic growth system that allows studying the interaction between lettuce roots and E. coli, both from molecular and physiological standpoints and that can easily be adapted to other leafy green vegetables such as spinach. To our knowledge, this is the very first study investigating the interaction between E. coli and lettuce roots in such a fashion.

In our colonization analyses, a total of 105 CFU of bacteria were applied to each hydroponic system. The data indicated that E. coli K12 was able to colonize lettuce roots reaching cell densities greater than 107 CFU per root system after 3 days. This result supported the idea that in our model system the root environment is a suitable habitat for the survival and growth of E. coli. This finding also corroborates previous observations reported in the literature indicating that root exudates can be used by bacteria as a source of nutrients (Darrah 1991).

Surprisingly, the sustained growth observed was particularly large. One possible explanation is that the hydroponic nutrient solution contributed to supporting bacterial cells growth, despite that there was no carbon source in the hydroponic nutrient solution. Another study conducted in the same conditions by our group on a different E. coli strain resulted in a more modest increase reaching only 105 CFU per root system after 3 days (data not shown). This strain difference may suggest that E. coli K12 might be particularly adept to using some of the plant exudates as carbon sources to support its colonization of the rhizosphere. Unfortunately, this experiment was not specifically designed to differentiate between E. coli cells growing on the roots with direct access to the root nourishment or in proximity of the rhizosphere using released exudates as growth substrates. However, the other observations we made through microarray and microscopy supported the hypothesis that the growth of E. coli in the proximity of the roots may lead to their subsequent attachment.

Our microarray data showed that the shift from a pelagic lifestyle in the hydroponic nutrient solution to the sessile colonization of the lettuce root system resulted in the induction of 45 protein synthesis-related genes out of the 193 positively regulated genes (23·3%). This likely is due to a rapid shift in the environmental parameters (i.e., macro- and micro-nutrients availability), thus inducing a dramatic change in the physiology of the micro-organism. These physiological adaptations to the new environment require newly synthesized proteins involved in attachment, nutrient acquisition, metabolic assimilation and colonization. Interestingly, glgS, a gene involved in glycogen biosynthesis decreased 1·6-folds in attached cells. This decrease in expression might indicate access to more energy efficient carbon sources. We did not observe changes in other carbon source related genes, but it was expected because, as shown by Liu et al., as carbon substrate quality declines, cells systematically increase the number of carbon source related genes expressed (Liu et al. 2005). In our case, the environment which the cells were exposed to (hydroponic solution) was very carbon poor and most likely a large number of these genes were expressed and were kept at similar levels even after the interaction with root surfaces and exudates.

The stress response genes (pspABCDE) were also significantly up-regulated when E. coli interacted with the lettuce roots. These genes are expressed in response to a variety of environmental and intracellular stresses (Model et al. 1997). We hypothesize that the E. coli cells were able to survive and adapt to the microenvironment of the plant roots by using these root exudates as carbon source. The presence of osmolytes in the microenvironment surrounding the roots might has caused osmotic stress thus inducing modifications to the bacterial cell membrane. However, the mutants had the same ability to attach and colonize the roots compared to the wild-type strains. These genes may be important for growth on the roots as they hallmark the adaptation to the root environment. But the data showed that they are not essential for colonization (e.g., their increased number) on the roots.

Microarray analyses indicated that the curli regulator, crl, was significantly up-regulated during growth in the lettuce rhizosphere. The curli produced by E. coli form an extracellular matrix that is important in the initial stages of biofilm formation as it promotes the attachment on abiotic surfaces and mediates the adhesion to host cells (Olsen et al. 1989; Austin et al. 1998; Vidal et al. 1998). Following that Crl is thought to be a thermal sensor that is more stable under low temperature (Bougdour et al. 2004) and highly up-regulates the expression of the curli operon at low temperatures (below 30°C). It is possible that the high level of induction of crl observed under our experimental conditions might be caused by the temperature we employed (25°C). However, it is important to note that in our study, both sample and control bacteria were exposed to the same temperature regardless of the presence of the roots. Therefore, the observed increase of crl expression was likely linked to shift from free-living to root-attached cells.

The crl has also been postulated to activate the transcription of csgA, a gene required for curli production in most E. coli strains (Arnqvist et al. 1992). However, we did not observe any significant change in csgA transcript levels by microarray or qRT-PCR analyses. These data suggest that csgA transcripts were already present at elevated levels in both physiological stages. Interestingly, however, both the crl and csgA mutants were impaired in their ability to colonize roots with < 2 log CFU per root difference in colonization compared to wild-type strains (10CFU per root). These growth experiments were confirmed by our data collected by CSLM (Fig. 4f, g). Recent studies have shown that curli produced by E. coli O157:H7 and S. enterica may play a significant role in the attachment to alfalfa sprouts (Barak et al. 2005; Jeter and Matthysse 2005). Recently, we found that the csgA mutant in E. coli K12 and O157:H7 was impaired in their ability to attach and colonize on lettuce leaves as well (Fink et al. 2012). Our findings in the present study further substantiate the importance of curli synthesis genes for E. coli attachment and colonization of plants.

Interestingly, our microarray data also showed that ybiM, a gene involved in biofilm formation modulation, was also up-regulated in the rhizosphere. This gene regulates the production of colanic acid which causes conversion to a mucoid phenotype and prevents biofilm formation (Zhang et al. 2008). Zhang et al. (2008) found that deletion of ybiM directed the cells towards the formation of biofilm. Not surprisingly, the colonization data reported here showed that there was no difference in root colonization by the ybiM mutant and the wild-type, and a similar result was noted for the role of ybiM when E. coli K12 interacted with lettuce leaves (Fink et al. 2012). Although our results showed that biofilm modulation is an important component for the interaction between E. coli and roots, the molecular mechanisms by which this occurs are still unclear.

Several studies have reported that motility and chemotaxis are also linked to bacterial colonization of plants (Lugtenberg and Dekkers 1999; Lindow and Brandl 2003). Indeed, FliO, involved in flagella synthesis, was down-regulated in this study possibly as a consequence of the shift to more easily metabolized carbon sources (Liu et al. 2005) and growth in the rhizosphere and attachment to plant roots. Moreover, Li and Sourjik (2011) recently found that the fliO mutants were only slightly impaired in their motility and chemotaxis. Therefore it is possible that fliO mutants have no disadvantage in colonizing lettuce roots compared to the wild-type cells in our experimental conditions. It has been suggested that fliN is also essential for flagella assembly and the flagella may facilitate bacterial attachment or colonization to lettuce roots; the direction of rotation of the E. coli flagellum is determined by a ring-like structure switch complex which is formed by FliG, FliM and FliN protein units (Park et al. 2006). Although the qRT-PCR did not show any significant up-regulation in fliN, the mutant lacked the ability to attach to the plant root surface compared to the wild-type. One possible explanation could be that fliN is essential for flagella assembly (Paul and Blair 2006) and these structures were already present on the surface of the bacteria when they were beginning to attach to root surfaces.

In the studies reported here, microarray analyses were used for the first time as a tool to identify putative genes involved in the colonization of lettuce rhizosphere by E. coli. We have determined that E. coli K12 is able to attach and colonize the lettuce rhizosphere. Some genes are essential for this process and in particular genes involved in curli formation and biofilm modulation. The physiological role of these genes was also confirmed and our results are in accordance to results from a previous study we performed on lettuce leaves (Fink et al. 2012). This project is the first to study the interaction between E. coli and lettuce roots on the gene expression level which provided a better understanding of the fundamental mechanisms whereby E. coli, and likely some pathogens, interact with the lettuce rhizosphere. Ultimately, this data may be useful to help develop strategies to reduce the incidence of food-borne disease outbreaks.


We would like to thank Antony Dean for providing E. coli strain MG 1655 and thank Arkady Khodursky for providing the microarray facility. This project was funded, in part, by a grant from the University of Minnesota Healthy Food and Healthy Lives Institute, awarded to F.D.G. and M.J.S.