Subtyping Listeria monocytogenes through the combined analyses of genotype and expression of the hlyA virulence determinant


Knut Rudi, MATFORSK, Norwegian Food Research Institute, Osloveien 1, N-1430 Ås, Norway (e-mail:


Aims: A major challenge for Listeria monocytogenes diagnostics is that this bacterium is ubiquitous in the environment, and that only a small fraction of the lineages are potential human pathogens. The aim of this work was to obtain a better subtyping of L. monocytogenes through utilization of combined analyses of genotype and the expression of the virulence determinant hlyA.

Methods and Results: We investigated the effect of growth temperature and medium on the hlyA expression. The gene expression levels were determined by real-time quantitative reverse transcription PCR. The expression pattern of hlyA was highly diverse among the different strains tested. The expression ranged from repression to a 1000-fold induction for growth at 42°C, as compared with 0°C. The expression patterns were compared with the corresponding genotypes. There were surprisingly low correlations between the expression patterns and the genotype clusterings.This is exemplified for the virulent type strain NTNC 7973 and non-virulent type strain DSMZ 20600. These strains are genetically nearly identical, while the hlyA gene expression patterns are very different.

Conclusions: The hlyA gene expression was highly diverse even within genetically clustered subgroups of L. monocytogenes. Consequently, the gene expression patterns can be used to further differentiate the strains within these genetic subgroups.

Significance and Impact of the Study: A major limitation in the control of L. monocytogenes is that the current tools for subtyping are not accurate enough in determining the potential virulent strains. The impact of this study is that we have developed a subtyping approach that actually targets a virulence property.


Listeria monocytogenes is a Gram-positive, facultative intracellular bacterium that causes invasive, often fatal, disease in susceptible hosts (Cossart 1998). The organism has caused several epidemics over the past decade. Epidemiological studies indicate that contaminated ready-to-eat foods with long shelf-life are important in the transmission of L. monocytogenes (Pearson and Marth 1990). Listeria monocytogenes is widely distributed in nature and is prevalent in decaying vegetation, soils, animal faeces, sewage, silage and water. The bacterium can grow at temperatures between 0 and 45°C, in the pH range of 4·1–9·6 and tolerates a salt concentration up to 10% (Jay 1996).

The three L. monocytogenes serotypes 4b, 1/2a and 1/2b (Vines and Swaminathan 1998) are the main epidemic strains. However, only a minor fraction of these serotypes actually cause epidemics (Vines and Swaminathan 1998). Thus, there is a clear need for new approaches for better discrimination of the virulent L. monocytogenes strains.

The virulence of L. monocytogenes is relatively ill defined, although several virulence genes have been described in detail (Leimeister-Wachter et al. 1990; Gedde et al. 2000; Portnoy et al. 1992; Kuhn and Goebel 1998; Vines and Swaminathan 1998; Kayal et al. 1999; Sibelius et al. 1999). The best characterized virulence gene is the hlyA gene. Initially, this gene was identified because the gene product, listeriolysin O, is responsible both for haemolysis of blood cells and disruption of eukaryote membranes in general (Kingdon and Sword 1970). A possible function of listeriolysin O during infection is to facilitate the escape of L. monocytogenes from phagosomal vesicles within macrophages (Decatur and Portnoy 2000). The hlyA gene may also function in cell to cell transmission and in the activation of endothelial cells (Waseem et al. 1995; Kayal et al. 1999; Gedde et al. 2000).

Initial studies on enzymatic activity, the amount of protein produced and gene expression, have been conducted for the hlyA gene under different environmental conditions (Kathariou et al. 1988; Sokolovic and Goebel 1989; Ripio et al., 1996; Leimeister-Wachter et al. 1992; Park et al. 1992). These results indicate that there is a possible effect of both the growth medium (Park et al. 1992; Ripio et al. 1996) and temperature (Sokolovic and Goebel 1989; Leimeister-Wachter et al. 1992) on the expression level of the hlyA gene. The fact that the haemolytic activity is not directly correlated with the virulence of L. monocytogenes (Kathariou et al. 1988) has been demonstrated. However, recent data indicate that the level of listeriolysin O is strictly regulated during an infection (Decatur and Portnoy 2000). This may suggest that the gene regulation pattern, in addition to the protein stability, is important for the virulence of L. monocytogenes.

The aim of this work was to investigate if the combined analyses of genotype and expression of the virulence determinant hlyA could be used for better discrimination of L. monocytogenes. Fourteen selected strains of L. monocytogenes were investigated with respect to the effect of temperature on the gene expression. In addition, the effect of growth medium was tested for two type strains NTNC 7973 (ATCC 35152) and DSMZ 20600 (ATCC 15313). The gene expression levels were determined by quantitative reverse transcription PCR (RT-PCR) through the application of 5′-exonuclease chemistry (Yajima et al. 1998). The patterns were subsequently correlated with the genotypes, as determined by multilocus sequencing (MLST) (Rasmussen et al. 1995) and amplification fragment length polymorphism (AFLP) (Vos et al. 1995).

The results presented in this work show that the expression and regulation patterns of the hlyA gene are diverse. We found low correlations between the expression patterns and the genotypes for the strains investigated. This may indicate an adaptive expression pattern of the hlyA gene. There is a clear potential, from a diagnostic perspective, for defining a genotype and gene expression combination that gives a better diagnostics of the virulent L. monocytogenes strains.

Materials and methods

Bacterial strains, media and cultures

A total of 14 strains of L. monocytogenes, including two type strains, were used in this study (Table 1). The two type strains were DSMZ 20600 (ATCC 15313) and NTNC 7973 (ATCC35152). These strains were selected according to the criteria that NTNC 7973 is virulent, while the strain DSMZ 20600 is non-virulent (Kathariou and Pine 1991). Strains MF 2A/24 and MF 24A/2/48 isolated from a poultry and a pig/bovine-processing plant, strain MF 2B/17 isolated from a meat-processing plant and strain MF 2C/12 from the fish-processing industry, represent isolates from the food-processing industry. The remaining strains SS 12067, SS 9618, SS 7751, SS 8819, SS 7785, SS 5001, SS 65500 and SS 5223 were kindly provided by Peter Gemer-Smidt at the Statens Serum Institute, Copenhagen, Denmark. These strains were chosen because they represent a selection of different serotypes and virulence properties, and because a thorough DNA sequence investigation has been carried out for the target genes selected in our study (Rasmussen et al. 1991, 1995).

Table 1. Listeria monocytogenes strains used in this study
    Analysed genes‡
  1. *DSMZ, Deutsche Sammlung von Microorganismen und Zellkulturen GmbH, Braunschweig, Germany; NTNC, National Collection of Food Bacteria, Reading, England; SS, Statens Seruminstitut, Copenhagen, Denmark; MF, MATFORSK Norwegian Food Research Institute, Ås, Norway.

  2. †Flagellar antigens were not determined for all strains.

  3. ‡EMBL/GenBank accession numbers. All sequences assigned with AJ were determined in this work. Rasmussen et al. (1991, 1995) determined the remaining sequences.

20600DSMZGuinea-pig (non-virulent type strain)1/2aAJ311991AJ311996AJ311985
7973NTNCGuinea-pig (virulent type strain)1/2aAJ311990AJ311995AJ311984
2A/24MFPoultry processingUnknownAJ311988AJ311993AJ311982
24A/2/48MFPig/bovine processingUnknownAJ311989AJ311997AJ311986
2B/17MFMeat processingUnknownAJ311987AJ311994AJ311983
2C/12MFFish processingUnknownAJ311992AJ311998AJ311981
12067SSHuman epidemic4bX85830X86997X85882
9618SSUnknown4X85825 X85858
7751SSUnknown4X85821 X85861
8819SSHuman epidemic4bX85815X85864X85890
7785SSUnknown1X85909 X85877
5001SSHuman sporadic1/2aX85924X86989X85879
5223SSUnknown4X85819 X85860

The cells were grown to stationary phase at 37°C in tryptone soya broth (TSB) and then adapted to 0°C for 24 h. The cold adaptation was performed because L. monocytogenes is selected at low temperatures in the environment (Hayes et al. 1991). Thus, the cold adaptation mimics the natural conditions and mRNA from high-temperature growth is avoided. Then the cells were pelleted and subsequently resuspended in TSB or brain–heart infusion (BHI) media to a concentration of approx. 108 cells ml−1 in a 20-ml volume. Relatively high inoculums were used, as very slow growth is expected at 0°C. The strains then were grown at 0, 25, 37 and 42°C and samples were analysed at different time intervals (see Results). All the strains used were also plated onto horse blood agar and analysed for haemolytic activity.

RNA purification

Aliquots of 1·5-ml culture were transferred to 1·5-ml microcentrifuge tubes and immediately centrifuged for 5 min at 5000 ×g at 4°C. The supernatant was discarded and the cells resuspended in 100 μl of TE [10 mmol l−1 Tris–HCl (pH 8·0), 1 mmol l−1 EDTA (pH 8·0)] containing 3 mg ml−1 of lysozyme (L-6876; Sigma Chemicals, St Louis, MO, USA). The contents were gently mixed and incubated at room temperature for 5 min. The subsequent RNA extraction was performed using the RNeasy kit (Qiagen, Hilden, Germany) following the recommendations of the manufacturer. A 350 μl volume of lysis buffer RLT containing 10 μl of β-mercaptoethanol ml−1 was added, and the sample vortexed vigorously for 5 s. A 250 μl volume of ethanol was added to the lysate and mixed thoroughly by pipetting. The sample lysate was then added to the RNeasy spin column and centrifuged at 6000 ×g for 15 s. The flow-through fraction was discharged. The column was then placed in a new collection tube and washed twice with 500 μl of wash buffer RPE. The tube was centrifuged at 6000 ×g for 15 s after the first wash, and finally 2 min at 10 000 ×g to ensure dryness of the of the column membrane. The RNA was then eluted twice in 20-μl diethylpyrocarbonate (DEPC)-treated deionized water through centrifugation at 6000 ×g for 1 min each time.

Twenty microlitres of the RNA solution was treated with 3 units of RQ1 DNase (Promega Inc., Madison, WI, USA), 0·7 μl Ribonuclease inhibitor (40 U μl−1) (Promega) for 10 min at 37°C. The enzymatic reaction was stopped by addition of EDTA to a final concentration of 2·5 mmol l−1.

Reverse transcription and TaqMan assay

Aliquots of 2 μl total RNA (corresponding to approx. 100 ng) were used in the cDNA synthesis reactions. The standard protocol for the Reverse Transcription System from Promega was followed.

Each 50 μl reaction volume of TaqMan RT-PCR contained 0·1–1 μl of the reverse transcription reaction; 1 × TaqMan buffer A; 5 mmol l−1 MgCl2; 200 μmol l−1 dATP, dCTP and dGTP; 400 μmol l−1 dUTP, 0·1 μmol l−1 specific probe (hlyA; 5′-CGA TTT CAT CCG CGT GTT TCT TTT CG-3′ and 23S rRNA; 5′-CGG TCG CCT CCT AAA GAG TAA CGG AGG-3′), 0·3 μmol l−1 specific primers (hlyA forward 5′-TGC AAG TCC TAA GAC GCC A-3′, reverse 5′-CAC TGC ATC TCC GTG GTA TAC TAA-3′ and 23S rRNA forward 5′-GTG TCA GGT GGG CAG TTT G-3′, reverse 5′-CAT TCT GAG GGA ACC TTT GG-3′), 2·5 U of AmpliTaq Gold DNA polymerase, and 0·2 U Uracil N-Glycosylase. The hlyA primers and probes, yielding an amplification product of 113 bp, have previously been described by Nogva et al. (2000), while the 23S primers and probes, yielding an amplification product of 77 bp, have been constructed by H.K. Nogva (manuscript submitted). The reaction tubes were MicroAmp optical tubes, and the tube caps were MicroAmp optical caps. All consumables were supplied by Applied Biosystems (Foster City, CA, USA).

The PCR mixtures were heated to 50°C for 2 min in order to inactivate possible traces of contaminating DNA, then heated to 95°C for 10 min to denature the template DNA and to activate the enzyme. The amplification profiles were as follows; 40 cycles of 94°C for 20 s and 60°C for 1 min. Reactions were performed in the ABI Prism 7700. Reaction conditions were programmed and data analysed on a power Macintosh 4400/20 (Apple Computer, Santa Clara, CA, USA) linked directly to the ABI Prism 7700 Sequence Detection System using the SDS 1.6 application software (Applied Biosystems) as described by the manufacturer. PCR products were directly detected by monitoring the increase in fluorescence from the dye-labelled DNA probes. The TaqMan probes consisted of an oligonucleotide with a 5′-reporter dye and a 3′-quencher dye. The reporter dye, FAM (carboxyfluorescein) was covalently linked to the 5′-end of the oligonucleotide. The fluorescence of the reporter was quenched by TAMRA (6-carboxy-N′, N′, N′, N′-tetramethylrhodamin) located at the 3′-end. The same threshold (0·03) was chosen for all the experiments. The different amplifications could then be compared according to the respective threshold cycles. The CT values were plotted against log input to estimate the amplification efficiencies of the two systems (Perkin-Elmer Applied Biosystems User Bulletin, 1997).

Cluster analyses of gene expression data

The software package used was Cluster and Tree View developed by Eisen et al. (1998) for analysis of the gene expression data. The value for the hlyA signals relative to 23S rRNA at both 0 and 42°C were used as inputs in the cluster analyses. Uncorrected data were then used for both the Genes and Array options in the software. Clustering of the data were performed hierarchially. The output data was visualized as a dendrogram through the application of the Tree View program (Eisen et al. 1998).

Definition of terminology

In this work, gene expression is defined as the balance between promoter activity and transcript degradation; induction is defined as the shift in the balance leading to accumulation of transcripts and repression is defined as the shift that leads to decrease in transcript levels.

Amplified fragment length polymorphism analyses

The primers used are described by Vos et al. (1995) for AFLP analyses. Listeria genomic DNA was prepared using the Qiagen tissue kit according to manufacturer's instructions (Qiagen). Genomic DNA was incubated overnight at 37°C with 12 U EcoRI (Promega), and 4 U MseI (New England Biolabs, Beverly, MA, USA), in 1 × NEB-buffer 2 (New England Biolabs) with 1 ng μl−1 BSA. To the digestion mix were added 2 μmEcoRI adapters and 2 μmol l−1MseI adapters in 1 × T4 DNA Ligase buffer (Promega), 50 mmol l−1 NaCl and 1 ng μl−1 BSA with 12 U EcoRI, 4 U MseI and 0·5 U T4 DNA ligase. Incubation was continued overnight at room temperature. After ligation, the reaction mixture was diluted 50-fold in TE (10 mmol l−1 Tris–HCl, 0·1 mmol l−1 EDTA, pH 8·0) and stored at 4°C.

The ligation mix was prepared for PCR by adding 1 × AmpliTaqGold DNA polymerase buffer, 5 mmol l−1 MgCl2, 200 μm dNTP (Promega), primers AFML4 (5′-GAT GAG TCC TGA GTA AC-3′), AFEC4 (GAC TGC GTA CCA ATT CC-3′), and 2·5 U AmpliTaqGold DNA polymerase. PCR reactions were performed on a Perkin-Elmer 9700 thermal cycler using the following cycle profile: the polymerase first was activated at 94°C for 10 min, cycle 1: 2 min at 94°C, 20 s at 66°C, 2 min at 72°C; cycles 2–10: 2 min at 94°C, 20 s at an annealing temperature 1°C lower than the previous cycle, starting at 65°C, 2 min at 72°C; cycles 11–21: 2 min at 94°C, 20 s at 56°C, 2 min at 72°C; cycle 22: 30 min at 60°C. Two microlitres of the amplified product were added to 25-μl loading buffer (24-μl deionized formamide and Gene Scan 500 Tamra standard; Applied Biosystems). All samples were heated for 5 min at 95°C and then rapidly cooled on ice prior to electrophoresis. Amplified fragments were separated on ABI Prism Genetic Analyzer 310.

Fragment profiles were analysed using GelCompare II (Applied Maths, Gent, Belgium). Band search was carried out using a minimum profiling of 3% and a minimum area of 0·3%. A position tolerance of 1% was used in the band matching. The Pearson correction and the Dice binary coefficient were used for the distance calculations. The provided unweighted pair group method using arithmetic averages (UPGMA) and WARD algorithms were used for the cluster analyses (User manual, GelCompare II; Applied Maths).

Multilocus sequencing analyses

Paramagnetic beads were used for the DNA purification following the protocols developed by Rudi et al. (1997, 1998. The PCR primers used in this work are for amplifying the hlyA gene; hlyA-F (5′-GCC GTA AGT GGG AAA TCT GTC TCA-3′) and hlyA-R (5′-GCA ACG TAT CCT CCA GAG TGA TCG-3′), the iap gene; iap-F (5′-GCT GAA AAA CAA GCA GCT CCA GTA GT-3′) and iap-R (5′-TCA AAT GTA GTT GGT CCG TTA CCA CC-3′), and the flaA gene; fla-F (5′-GTT CAA TCT TGC AAC GTA TGC GTC-3′) and fla-R (5′-CCA CTA CCT AAA GTG ATT GTT CCA GCA-3′). Between 30 and 40 cycles were used in the amplification reactions, with a denaturation at 95°C for 30 s, annealing for 30 s at 55°C and extension at 72°C for 30 s. All reactions were initiated with 4-min denaturation at 94°C and ended with 7-min extension at 72°C. Amplified products were sequenced using the Big Dye chemistry according to the manufacturer's recommendations (Applied Biosystems). The sequences then were determined with the Model 310 DNA Sequencer (Applied Biosystems). The sequence data have been deposited in the EMBL nucleotide sequence database (Cambridge, UK) with the accession numbers shown in Table 1.

For the hlyA dataset 275 aligned positions were analysed, for the flaA dataset 235 positions, and finally for the iap dataset 350 positions. Sequences were aligned using the Clustal X software (Thompson et al. 1997) then edited manually using the program GeneDoc (Nicholas and Nicholas 1997).

The data were analysed by using a maximum parsimony approach under the assumption that the correct phylogeny is the tree that can explain the data with fewest possible genetic events (Fitch 1977). Phylogenetic trees were constructed using the software package PAUP* 4.0 developed by D. L. Swoford (Florida State University, Tallahassee, FL, USA) and distributed by Sinauer Associates Inc. (Sunderland, MA, USA). Search for the maximum parsimony tree was performed heuristically. Consensus trees were constructed from 1000 bootstrap replicates (Felsenstein 1985). The tree length skewness, determined by the third-momentum statistics (g1) for 100 000 randomly generated trees, was used to investigate the phylogenetic structure of the data. The results obtained were compared with critical values given by Hillis and Huelsenbeck (1992).

Comparison of gene expression and genotype data

We tested whether the gene expression data were significantly different from the genotype data. This was carried out by applying user-defined trees, as implemented in DNAPARS program in the Phylogeny Inference Package (PHYLIP; Version 3·5) developed by J. Felsenstein (Department of Genetics, University of Washington). The statistical support was tested by using the Templeton and Felsenstein test implemented in this package (Templeton 1983). For situations where there was lack of resolution in one of the trees, we assumed the same strain distribution as in the other tree.


Determination of the accuracy of the quantitative gene expression assay

The hlyA gene expression was measured relative to the level of the 23S rRNA expression. Factors affecting this ratio are critical for the accuracy of the analysis (Williams et al. 1998).

When using RT-PCR for quantifying gene expression for prokaryotes it is essential to control the level of DNA contamination in the RNA preparation. We used a two-step RT-PCR with random hexamer primers for the cDNA synthesis. The negative control in this setup was a reaction without reverse transcriptase. The DNA contamination level was then determined by the ΔCT for the reverse transcriptase-containing reaction as compared with the reaction without reverse transcriptase. A ΔCT > 4 indicates that the DNA contamination level in the signal for that particular target is <10% (from the calculations shown below). This contamination level is within the range of accuracy of the experiment and does not interfere with the RT-PCR signal obtained.

The quantitative range of the assay was determined through a dilution series of cDNA (Fig. 1). The regression curves for hlyA and 23S rRNA had slopes of 3·44 and 3·61 which correspond to amplification efficiencies of 0·95 and 0·89, respectively. The square regression coefficients for the two curves were 0·991 and 0·995. We determined a log linear range for both target transcripts of >4 log. A standard curve, with a slope of 3·50, was used to calculate the relative hlyA gene expression levels within the log linear range determined for both targets for all subsequent experiments.

Figure 1.

Standard curves for (a) hlyA, and (b) 23S rRNA. The curves were generated by 5′-nuclease PCR on triplicates of a 10-fold dilution series of cDNA generated by random hexamers form cold-adapted L. monocytogenes DSMZ 20600. The curves show CT values plotted against log dilution of the cDNA concentration. The straight lines (y = 3·44x + 19·77; panel a), and (y = 3·61x + 12·33; panel b) have square regression coefficients (R2) of 99·1 and 99·5, respectively

The effect of RNA concentration was tested through cDNA synthesis of RNA isolated from the strain DSMZ 20600 adapted to 0°C on a dilution series of RNA over a 2 log range (Table 2). Approximately the same CT value was obtained up to a 20-fold dilution of the RNA. As expected from the amount of input RNA, the CT value decreased when the RNA was diluted 20–100-fold. However, the ΔCT values were approximately equal for all the dilutions. The average ΔCT was 6·94, with a standard deviation of 0·26 (see Table 2).

Table 2.  Effect of RNA concentration on the quantification accuracy
 RNA dilution*
 1 : 11 : 101 : 201 : 501 : 100
  1. *The RNA used was from the cold-adapted DSMZ 20600 strains, for which the RNA has been stored for 5 months.

  2. †The CT values were obtained as described in Materials and methods. Standard deviations were calculated from triplicate independent cDNA syntheses.

  3. ‡The ΔCT value was calculated by subtracting the CT value for hlyA from 23S rRNA CT value for each corresponding cDNA synthesis. The errors are standard deviations for three independent experiments.

CT 23S rRNA†13·41 ± 0·3213·05 ± 0·3013·66 ± 0·3915·46 ± 1·0516·41 ± 1·61
CThlyA20·18 ± 0·2920·00 ± 0·4821·0 ± 0·3122·43 ± 0·9423·06 ± 0·73
ΔCT6·77 ± 0·076·95 ± 0·477·34 ± 0·416·97 ± 0·276·65 ± 1·00

The RNA used in the above-mentioned experiment had been stored for 5  months at −80°C. We obtained a ΔCT value of 4·77 with a standard deviation of 0·12 for cDNA synthesized 2 days after the extraction. A repeated experiment gave a ΔCT value of 4·57 with a standard deviation of 0·86 for cDNA synthesized within a week after the first start of the experiment. These results indicate that storage at −80°C resulted in a differential degradation of the hlyA transcript as compared with 23S rRNA (Alifano et al. 1994). Thus, the cDNA syntheses were always performed within 1 week after the RNA extraction in order to avoid potential errors as a result of differential RNA degradation.

The 23S rRNA and hlyA gene regions were sequenced for all the strains used. There were no mutations in these regions that could have influenced the quantitative results (data not shown).

Effect of growth conditions on hlyA gene expression for L. monocytogenes type strains DSMZ 20600 and NTNC 7973

The hlyA expression level in the two type strains was investigated for temperatures ranging from 0 to 42°C in both TSB and BHI medium for up to 4 h (Fig. 2). There were no significant differences in the growth kinetics for the two strains in the media tested. Cell numbers remained nearly constant at 0°C for all the time intervals tested. The cells were in logarithmic growth after both 2 and 4 h at 25°C, while they reached a stationary phase at about 109 cells ml−1 after 4 h at 37 and 42°C for both strains and media used (results not shown).

Figure 2.

Expression levels of hlyA in L. monocytogenes strain DSMZ 20600 (♦) and NTNC 7973 (▮). Cold-adapted cells were resuspended in TSB (panel a to d) and BHI (panel e to h), at 0°C (panel a and e), 25°C (panel b and f), 37°C (panel c and g), and 42°C (panel d and h). The signal of hlyA relative to 23S rRNA is shown in the graph for 0, 2 and 4 h. Error bars show standard deviations

The data show that the mean hlyA gene expression after cold adaptation for DSMZ 20600 was 48-fold that of NTNC 7973 (Fig. 2A). The experiment was repeated through an independent cold adaptation (in triplicates) of both strains. The mean difference in expression levels obtained in this series was 38-fold (results not shown).

Cells were resuspended in fresh TSB and BHI media after being cold-adapted and incubated further at 0, 25, 37 and 42°C for up to 4 h. The expression level for hlyA converged when grown in TSB medium, while there was no significant convergence when grown in BHI at 0°C. The expression level for DSMZ 20600 decreased at 25°C, while the level remained relatively constant for NTNC 7973 in both media. Expression levels converged to nearly identical expression values for both media at 37 and 42°C.

The gene expression patterns for the strains DSMZ 20600 and NTNC 7973 were also investigated when the bacteria grew as colonies on horse blood agar plates at 4 and 37°C. The expression levels for the two strains were approximately similar after incubation at 4°C for 4 days (Fig. 3) and there was no visible haemolytic activity for the strains at this temperature (results not shown). After incubation at 37°C for 2 days the expression level was approx. 3 log higher for NTNC 7973 as compared with DSMZ 20600 (Fig. 3). Surprisingly, the haemolytic activity (determined by the time taken to reach observable haemolysis on blood agar) was higher at 37°C for DSMZ 20600 as compared with NTNC 7973 (results not shown).

Figure 3.

hlyA gene expression during colony growth of L. monocytogenes strains DSMZ 20600 and NTNC 7973. Cells were grown on horse blood agar plates at 4°C for 4 days (black bars) and 37°C for 2 days (white bars). The signals of hlyA relative to 23S rRNA are shown. Error bars show standard deviations

Diversity of hlyA gene expression patterns

The diversity of hlyA gene expression patterns was determined for 14 selected strains of L. monocytogenes (Fig. 4). All the strains had a haemolytic phenotype on the blood agar plates (results not shown). The expression levels were determined in TSB medium at 0°C and at 42°C after 4 h, as the expression patterns for the two type strains were relatively stable between 2 and 4 h (see Fig. 2).

Figure 4.

Diversity of hlyA expression level among 14 selected L. monocytogenes strains. Cold-adapted cells were resuspended in TSB medium and incubated at 0°C (black bars) and 42°C (white bars) for 4 h. The bars represent the hlyA signal relative to 23S rRNA. Error bars show standard deviations

The strains MF 2C/12 and MF 24A/2/48 had an intermediate expression level at 0°C. The strains MF 2A/24,MF 2B/17, NTNC 7973, SS 5223, SS 7785, SS 7751, SS 8819, SS 5001 and SS 12067 were apparently somewhat repressed at 0°C, while the two strains SS 9618 and SS 6500 were highly repressed at this temperature. The hlyA expression for the strain DSMZ 20600 was (as described previously) relatively high at 0°C, while the expression was partly repressed at higher temperatures. The strains MF 2C/12 and MF 24A/2/48 were uninduced at 42°C, as compared with 0°C. NTNC 7973, SS 5223 and MF 2A/24 were moderately induced at 42°C, while the strains MF 2B/17, SS 7785, SS 7751, SS 8819, SS 9618, SS 6500, SS 5001 and SS 12067 were highly induced.

A gene expression tree was constructed through clustering of the strains according to their expression pattern (Fig. 5). The downregulated strain DSMZ 20600 was grouped separately. There was one group comprising the strains MF 2C/12, NTNC 7973 and MF 24A/2/48 with no or slight difference in gene expression between cells grown at 0 and 42°C. There were two groups of upregulated strains. One group contained the strains MF 2A/24 and SS 5223 with a moderate difference in gene expression between 0 and 42°C. The largest group comprised the strains SS 8819, SS 7751, SS 65500, SS 9618, SS 7785, MF 2B/17, SS 12067 and SS 5001 with a relatively large difference in expression level (more than 10-fold) between the two temperatures. All three isolates from human listeriosis patients (SS 12067, SS 8819 and SS 5001) are within this group.

Figure 5.

Strain clustering according to hlyA expression patterns. The strains used in this work were clustered according to their temperature shift response. The horizontal branches represent the differences in gene expression pattern. The hlyA gene expression level visualized as an intensity scale for each of the strains and conditions tested

Genotype clustering

Basically, the same three clusters of strains were obtained with the AFLP as with the multilocus DNA sequence analyses (Fig. 6). These clusters were also congruent with the serotypes. There was, however, lack of resolution within the group of strains belonging to serotypes 4 and 4b, which could not be separated in distinct groups. In addition, the strains SS 7785, SS 65500 and MF 2A/24 could not be placed within the group of serotype 1 strains with statistical confidence. Finally, the strain SS 5001 cluster with NTNC 7379 in the multilocus DNA sequence tree, while SS 5001 forms a separate branch within the group of 1/2a strains in the AFLP tree.

Figure 6.

The WARD dendrogram for the AFLP data (A) and phylogenetic reconstruction from the DNA sequence data (B). Each L. monocytogenes isolate is represented by the respective strain number. (A) Similar trees were generated with both the WARD and UPGMA clustering algorithms (the only difference is for the strain SS 65500, which is located in the root of the serotype 1 and 1/2a cluster in the UPGMA tree). (B) A maximum parsiony tree was constructed from the informative sites in the alignments of the hlyA, flaA and iap data. Numbers at the nodes indicate the percentage of 1000 bootstrap replicates for which the strains descending from the node was present. Only values above 50% are shown

The resolution is seemingly higher in the AFLP tree (Fig. 6a), as compared with the multilocus DNA sequence tree (Fig. 6b). This may reflect the fact that more informative characters are analysed in the AFLP, as compared with the DNA array data (results not shown).

Comparison of gene expression and genotype

We tested if the gene expression tree (Fig. 5) was significantly different from the AFLP tree (Fig. 6a), and which tree being closest to the tree obtained with the DNA sequence data from the hlyA, iap and flaA genes. A scewedness (g1) of −0·78 indicate that the DNA data is highly structured (Hillis and Huelsenbeck 1992). There are 29 most parsimonious trees with 17 steps for the DNA sequence data. The DNA sequence data explained the AFLP tree with 20 steps, and gave a variance of step difference determined as the step difference at the individual positions between the most parsimonious and the AFLP tree of 2.30 (user-defined trees in the DNAPARS program in PHYLIP). According to the criteria given by Templeton (1983), this is not a significant difference.

The same branching pattern as in the AFLP tree was assumed for the group of strains with no resolution in the gene expression tree (Fig. 5). The DNA sequence data explained the gene expression tree with 36 steps, giving a step difference of 16 from the AFLP tree, and variance in step differences as determined by the step differences at individual positions for the two trees of 5·99. This shows that the gene expression tree is significantly different from both the AFLP and the DNA sequence trees (Templeton 1983).


Potential errors in the gene expression analyses

For several reasons, quantification of gene expression is difficult. One being that RNA is an unstable molecule, another being that different transcripts may have very different degradation kinetics (Alifano et al. 1994). The effect of storage was tested through comparing the ΔCT values for cDNA synthesized from freshly prepared RNA and for RNA stored for 5 months. These results quantitatively showed that the hlyA transcript was differentially degraded during storage.

The question of errors introduced in the purification step was addressed by performing three independent RNA purifications in parallel for each collected datapoint. Another potential analytical error is the variation in the efficiency of the cDNA synthesis step. The reaction is often not very efficient and secondary structure elements may severely affect the reaction efficiency. Furthermore, there may be an RNA concentration effect for the random hexamer incorporation for different targets. To investigate these potentially analytical artefacts, an RNA dilution series was set up. Although there was a saturation effect for the cDNA synthesis for RNA concentrations less than 1 : 20 dilution this saturation did not affect the ΔCT values (see Table 1). We do not know whether the saturation effect is due to enzymatic or to primer limitations.

The choice of target and amplicon is crucial for quantitative analyses. A requirement for quantitative analyses is that the internal reference gene should not be differentially expressed in different environments. For most genes this criterion may be difficult to meet. We chose the 23S rRNA gene as our internal standard because 90–95% (nearly corresponding to total RNA) of the RNA in bacteria is ribosomal RNA. However, the limitation of using 23S rRNA is the relatively large difference in gene expression between the reference gene and the gene of interest (Lantz et al. 1998; Williams et al. 1998). Another potential limitation is the differential degradation of the target transcript during RNA purification or storage (as observed in this work).

Finally, in order to have a quantitative system, it is necessary that the amplification efficiency of the target is similar to that of the reference gene. Thus, standard curves were made over the expected concentration range. The CT value, plotted against the log cDNA concentration, gave an amplification efficiency of approx. 0·9 for both for the hlyA and the 23S rRNA transcript (Freeman et al. 1999). Obviously, if point mutations were located in the primer regions, the amplification efficiency would have been altered in the first few cycles. This could have been misinterpreted as differential gene expression. No point mutations were identified either in the flanking regions or in the primer or amplification region for the six strains used in our study. Thus, the observed differences in amplification efficiencies are not due to point mutations.

The conclusion from these comparisons is that the quantification method used is robust and that the main source of errors in the experiments probably is the cDNA synthesis step (see Table 2). However, as indicated for the standard deviations, these errors are lower than the differences observed in gene expression for the experimental conditions used.

Diverse gene expression patterns

We have shown that the expression pattern for the hlyA gene is highly diverse among different strains of L. monocytogenes. Our selection of 14 strains responded very differently to the environmental conditions tested in this work. The generalization of the hlyA expression pattern, from a few model organisms and experiments, may thus be incorrect.

Studies where the hlyA transcript has been used as marker for viable or dead bacteria (Klein and Juneja 1997; Norton and Batt 1999), or as a mean for the detection of L. monocytogenes (Blais et al. 1997) have also been carried out. From the knowledge gained in this work, such applications could be difficult as the absence of gene expression under the conditions tested could be misinterpreted as the absence of L. monocytogenes, or the presence of dead bacteria.

The lack of correlation between hlyA gene expression and the identified genotypes

The determination of genetic relatedness is difficult, and often associated with artefacts. That is the reason for comparing the AFLP and MLST data. There were no significant differences in phylogenetic trees constructed from these two datasets (Fig 6), indicating that the correct phylogeny is described.

Low correlations were found between the gene expression pattern, and the genotypes identified (compare Figs 5 and 6). For instance, the two type strains DSMZ 20600 and NTNC 7973 were isolated from the same epidemic of listeriosis among laboratory animals in 1924 (Kathariou and Pine 1991). They cluster genetically and belong to the same serotype 1/2a (Ripio, 1996). The gene expression pattern, however, is very different (see Figs 2 and 3), which also results in different haemolytic phenotypes (Jones and Seeliger 1983; Kathariou and Pine 1991). The group of eight highly upregulated strains (see Fig. 5) contained all the genotypes included in this work. This group is also diverse, ranging from the strain SS 7751 with a 24-fold induction, to the strain SS 65500 with an 836-fold induction at 42°C, as compared with 0°C, respectively. There was, however, no correlation between the genotype and the level of induction within this group either. For instance, the two strains SS 65500 and SS 9618, with the highest level of induction, belong to two different genotype clusters. Interestingly, all three strains isolated from human patients (SS 12067, SS 8819 and SS 5001) were within this group.

The most probable explanation for the low correlation between hlyA gene expression and the genotype is that the genetic alterations leading to differential gene expression (e.g. mutations or methylation pattern; Heitoff et al. 1999) are adaptive, leading to a selection for these characters. It is unlikely that the divergence in expression pattern for the strains investigated is due to neutral or negative events, as the expression pattern is apparently regulated, and all the strains have a haemolytic phenotype. Such events would either result in the loss of the haemolytic phenotype, or unregulated expression. Adaptive characters are rare in nature, and can be used to indicate that the character is important for the survival of the organism in changing environments (Kimura 1987).

Recent data by Decatur et al. (2000), show that the level and timing of the listeriolysin O production during infection is crucial for the pathogenicity of the organism. Only a slight alteration of the protein, making it more stable, resulted in a 3 log reduction in LD50 for mice, where L. monocytogenes had been injected intravenously. Thus, the stability of listeriolysin O, and the regulation of the hlyA gene expression is possibly a highly adaptive character during infections. This may be a reason for the adaptive process for hlyA gene expression, as only slight changes in listeriolysin O level may have dramatic effects on the virulence properties of L. monocytogenes (Chakraborty et al. 2000).

Potential for defining expression patterns for pathogenic L. monocytogenes

A major limitation of the current L. monocytogenes diagnostics is that this bacterium is ubiquitous in the environment, and that only a small fraction of the lineages are potential human pathogens. Several studies have been conducted addressing the correlation between the amount and haemolytic activity of listeriolysin O and virulence (Kathariou et al. 1988; Park et al. 1992; Waseem et al. 1995; Conte et al. 1996; Kayal et al. 1999; Sibelius et al. 1999; Chakraborty et al. 2000; Decatur and Portnoy 2000; Nørrung and Andersen 2000). Although the variations in virulence for different L. monocytogenes strains are significant (Nørrung and Andersen 2000; Norton et al. 2001), there are no clear correlations between these parameters in virulence models. However, investigation of the regulation of gene expression may be the key to better characterize the most virulent strains.

The three serotypes 4b, 1/2a and 1/2b are mainly responsible for the listeriosis epidemics (Vines and Swaminathan 1998). It is possible to define these pathogenic serotypes genetically (Rasmussen et al. 1995; Vines and Swaminathan 1998). Therefore, it should also be possible to define gene expression patterns characteristic for human pathogens within these groups.

We have shown that the hlyA expression pattern is very diverse even among genetically closely related strains. These patterns can be used for further differentiation of the genetic subtypes of L. monocytogenes. For instance, the virulent type strain NTNC 7379 is genetically tightly clustered to the non-virulent type strain DSMZ 20600. The hlyA expression patterns, however, are very different. All three human isolates included in this study showed related expression patterns although they belong to different sero and genotypes. A future development in defining expression patterns for virulent stains would require a large-scale comparative study of a set of strains from human sporadic and epidemic cases, in addition to non-virulent strains. Other virulence determinants should also be used in such screenings. Ultimately, the approach presented here by comparing genotype with gene expression could lead to a diagnostics of the most virulent L. monocytogenes strains.


This work was supported by a research levy on agricultural products in Norway and grant no. 14256/140 from the Norwegian Research Council. We want to thank Ellen S. Tronrud and Helga Næs for carefully reading and commenting on the manuscript.