Is biofilm formation related to the hypermutator phenotype in clinical Enterobacteriaceae isolates?

Authors

  • Bela Kovacs,

    1. Department of Urology, Jahn Ferenc South-Pest Hospital, Budapest, Hungary
    Search for more papers by this author
  • Sandrine Le Gall-David,

    1. Equipe Microbiologie, EA 1254, SFR Biosit, Biogenouest, UEB, Université de Rennes 1, Rennes, France
    Search for more papers by this author
  • Pascal Vincent,

    1. Equipe Microbiologie, EA 1254, SFR Biosit, Biogenouest, UEB, Université de Rennes 1, Rennes, France
    2. Bacteriology 1 Department, CHU Pontchaillou, Teaching Hospital, Rennes, France
    Search for more papers by this author
  • Hervé Le Bars,

    1. Equipe Microbiologie, EA 1254, SFR Biosit, Biogenouest, UEB, Université de Rennes 1, Rennes, France
    2. Bacteriology 1 Department, CHU Pontchaillou, Teaching Hospital, Rennes, France
    Search for more papers by this author
  • Sylvie Buffet-Bataillon,

    1. Equipe Microbiologie, EA 1254, SFR Biosit, Biogenouest, UEB, Université de Rennes 1, Rennes, France
    Search for more papers by this author
  • Martine Bonnaure-Mallet,

    1. Equipe Microbiologie, EA 1254, SFR Biosit, Biogenouest, UEB, Université de Rennes 1, Rennes, France
    2. Bacteriology 1 Department, CHU Pontchaillou, Teaching Hospital, Rennes, France
    Search for more papers by this author
  • Anne Jolivet-Gougeon

    Corresponding author
    1. Equipe Microbiologie, EA 1254, SFR Biosit, Biogenouest, UEB, Université de Rennes 1, Rennes, France
    2. Bacteriology 1 Department, CHU Pontchaillou, Teaching Hospital, Rennes, France
    • Department of Urology, Jahn Ferenc South-Pest Hospital, Budapest, Hungary
    Search for more papers by this author

Correspondence: Anne Jolivet-Gougeon, Equipe Microbiologie, EA 1254 Microbiologie, Faculté des Sciences Pharmaceutiques et Biologiques, Université de Rennes 1, 2 Avenue du Professeur Léon Bernard, 35043 Rennes, France. Tel.: (33) 2 23 23 49 05; fax: (33) 2 23 23 49 13; e-mails: anne.gougeon@univ-rennes1.fr; anne.gougeon@chu-rennes.fr

Abstract

In bacteria, complex adaptive processes are involved during transition from the planktonic to the biofilm mode of growth, and mutator strains are more prone to producing biofilms. Enterobacteriaceae species were isolated from urinary tract infections (UTIs; 222 strains) and from bloodstream infections (BSIs; 213 strains). Relationship between the hypermutable phenotype and biofilm forming capacity was investigated in these clinical strains. Mutation frequencies were estimated by monitoring the capacity of each strain to generate mutations that conferred rifampicin resistance on supplemented medium. Initiation of biofilm formation was assayed by determining the ability of the cells to adhere to a 96-well polystyrene microtitre plate. UTI Enterobacteriaceae strains showed significantly higher biofilm-forming capacity: 63.1% (54.0% for E. coli strains) vs. 42.3% for BSI strains (47.7% for E. coli). Strains isolated from UTIs did not present higher mutation frequencies than those from BSIs: contrary to what has been widely described for P. aeruginosa strains, isolated from pulmonary samples in patients suffering from cystic fibrosis, no relationship was found between the hypermutator phenotype in Enterobacteriaceae and the ability to initiate a biofilm.

Introduction

Usually the rate of mutation of a bacterial population is of the order of 10−6 to 107, but numerous studies have described hypermutable strains in clinical and animal isolates (Leclerc et al., 1996; Denamur et al., 2002; Baquero et al., 2005; Le Gall et al., 2009). Polymorphisms in rifampicin resistance genes have been studied by Baquero et al. (2004), who defined four categories of E. coli strains according to their mutation frequencies (f): hypomutable (f ≤ 8 × 10−9), normomutable (8 × 10−9 ≤ f < 4 × 10−8), weak mutator (4 × 10−8 ≤ f < 4 × 10−7) and strong mutator (f ≥ to 4 × 10−7). These categories are based on the distinction of a first group of strains ‘normomutable’ (hypomutators and normomutators) and a second group of strains ‘hypermutable’ (weak and strong mutators), with clearly distinct modal mutation frequencies (this reference only applies for E. coli strains).

It is still not clear whether high mutation frequencies are particularly important for the global evolution of pathogen populations, including the development of biofilms (Conibear et al., 2009). An elevated mutation frequency (hypermutator phenotype) in bacterial populations in biofilms is more likely to occur. In biofilms, endogenous oxidative stress is common and can lead to the accumulation of DNA damage, particularly in the methyl mismatch repair (MMR) genes, which can increase the mutation rate.

In a clinical collection of Enterobacteriaceae strains isolated from urinary tract (UTI) and blood stream infections (BSI), we investigated (1) the capacity to initiate a biofilm, (2) the mutation frequency and (3) whether there could be a relationship between the mutation frequencies and the ability to form a biofilm.

Material and methods

Bacterial strains

Clinical strains of Enterobacteriaceae (n = 435) were isolated from urine and blood samples at the University Hospital of Rennes (France). Only the first episode of bacteraemia or urinary tract infection was considered for each patient (Table 1). Patients with polymicrobial bacteraemia were excluded. Potential clonality of the isolates was studied using pulse-field gel electrophoresis (PFGE) (Allardet-Servent et al., 1989). PFGE patterns were compared by calculating the Dice correlation coefficients using gelcompar ii software (Applied Maths, St-Martens-Latem, Belgium) and were clustered into a dendrogram using the unweighted pair group method (tolerance, 2.0%). Two isolates were considered genetically related if their Dice coefficients revealed ≥ 85% similarity. Only isolates with unique patterns were considered for statistical analysis.

Table 1. Mutation frequencies and ability to form a biofilm (on polystyrene plate) of a collection of clinical Enterobacteriaceae species (n = 435), isolated from blood and urine samples
Mutation frequency (f) statusaEnterobacteriaceae speciesUrinary strains (No of isolates)Blood strains (No of isolates)Total (No of isolates)
No biofilm producersb+Biofilm producersTotalNo biofilm producersb+Biofilm producersTotal
  1. a

    Categorisation according to Baquero's criteria (2004).

  2. b

    Biofilm production was calculated as the ODm/ODc ratio as recommended by Stepanovic et al. (2007). Accordingly, the studied strains were classified into two categories: no biofilm producers (ratio ≤ 1), +biofilm producers ratio > 1). The raw value of the ratio was used for the correlation calculations.

Hypomutable strains
f ≤ 8 × 10−9 E. coli 61117571229
K. oxytoca    3477
K. pneumoniae    1122
C. koseri    2 22
S. marcescens    1122
C. freundii  11   1
M. morganii     111
P. mirabilis    1 11
Total hypo6121813142745
Normo mutable strains
8 × 10−9 < f < 4 × 10−8 E. coli 4554995945104203
K. pneumoniae  131352720
P. mirabilis  337 710
E. cloacae    8199
K. oxytoca 1562139
S. marcescens 2461239
M. morganii 235 116
C. freundii 145   5
P. rettgeri 134   4
C. koseri  111123
H. alvei    1122
E. aerogenes    1 11
S. enterica Typhi    111
Total normo52901428555140282
Weak mutable strains
4 × 10−8 < f< 4 × 10−7 E. coli 21204113193273
P. mirabilis  77   7
K. pneumoniae 1124 46
C. freundii  44   4
E. cloacae    3 33
K. oxytoca  11 112
S. marcescens  11 112
C. koseri  11   1
E. aerogenes    1 11
M. morganii    1 11
Total weak223557222143100
Strong mutable strains
f ≥ to 4 × 10−7 E. coli 2241 15
E. cloacae    2 22
C. freundii  11   1
Total strong2353 38
Total 8214022212390213435

Mutation frequencies

Mutation frequencies (f) were reported as a proportion of rifampicin-resistant colonies to the total viable count. The results were related to the mean value obtained from three independent cultures of about 108 CFU mL−1. When mutation frequencies were ≥ 4.10−8, mutation frequencies were also tested with lower (about 10−7 CFU mL−1) or higher (about 109 CFU mL−1) inocula and, if necessary, with another antibiotic (fosfomycin 30 μg mL−1), to eliminate the possibility of ‘jack-pot’ emergence that might disturb the calculation of mutation frequencies. According to Denamur et al. (2002), strains displaying a > 50-fold increase of the median value of mutagenesis were considered strong mutators and a 10- to 50-fold increase as weak mutators. According to Baquero et al. (2004), a strain was considered normomutable when the mutation frequency was equal or close to the modal point of the distribution of mutation frequencies. The results were categorised as previously described for E. coli strains.

Biofilm formation

The initiation of biofilm formation was assayed using polystyrene microtitre plates, as described previously (Stepanovic et al., 2007), with some modifications. Specifically, one fresh colony of each strain was inoculated into 10 mL tryptic soy (TS) broth and cultured for 2 h. Samples (150 μL) of exponential-growth-phase bacteria were removed and incubated overnight (18 h) at 37 °C in 96-well microtitre polystyrene plates (Falcon Microtest™ 96; Becton Dickinson, Meylan, France). After removal of the medium, crystal violet (0.4% solution; 150 μL) was added to the emptied wells to stain the biofilm (if present). The biofilm was quantified at least in triplicate for each sample. Streptococcus gordonii (strain Challis/ATCC 35105/CH1/DL1/V288) and bovine Salmonella Heidelberg B182 (Le Gall et al., 2009) were used as positive controls. Escherichia coli HB 101 and sterile TS culture broth were used as negative controls. The mean OD570 nm value (ODm) was calculated for three wells. A cut-off value (ODc) was established as three standard deviations above the mean OD570 nm of the three negative controls in each plate. Biofilm production was calculated as the ODm/ODc ratio as recommended by Stepanovic et al. (2007). Accordingly, the studied strains were classified into four categories: no biofilm producer (ratio ≤ 1), + biofilm producer (1 < ratio ≤ 2), ++ biofilm producer (2 < ratio ≤ 4) and +++ biofilm producer (ratio > 4). The raw value of the ratio was used for the correlation calculations.

Statistical analysis

Percentages were compared using a two-sided Fisher's exact test for count data or the Mantel–Haenszel chi-squared test for stratified data.

Results and discussion

The PFGE results for the UTI strains revealed 222 unique clones isolated from 195 patients and for the blood samples revealed 213 Enterobacteriaceae strains isolated from 213 bacteraemic patients. The Enterobacteriaceae species were distributed as described in Table 1.

Mutation frequencies

The mutation frequency distribution was calculated for the whole collection: according to Baquero et al.'s (2004) criteria, 10.3% were hypomutable, 64.8% were normomutable, 23% were weak mutators, and 1.8% were strong mutators. One hundred weak mutators (57 UTI and 43 BSI strains) and eight strong mutators were isolated [five UTI strains (one C. freundii and four E. coli) and three BSI strains (two E. cloacae and one E. coli)] (Table 1). The mutation frequency distribution for E. coli strains was in close agreement with the distribution reported by Baquero et al. (2004) and was superposed on the distribution for the entire Enterobacteriaceae collection (Fig. 1). In this collection of Enterobacteriaceae, a submajority population whose mutation frequency was 1 × 10−8 (1 × 10−8 ≤ f < 2 × 10−8) and a second smaller subpopulation with a mutation frequency 10-fold higher (1 × 10−7 ≤ f < 2 × 10−8) were already described as normomutators and weak mutators by Baquero et al. (2004). Between these two populations, strains have intermediate rates, mainly between 3 × 10−8 and 1 × 10−7. Incorrect assessment of the nature of these mutator strains could distort the measurement of the association with the ability to produce a biofilm. To confirm the distinct modal mutation frequencies and the threshold proposed by Baquero et al. (2004), the analysis has been repeated and limited to strains with the characteristic rate of 1 × 10−8 and 1 × 10−7, respectively, considered as normomutators and weak mutators. In this specific population, 115 strains were normomutators (including 84 E. coli and 31 other Enterobacteriaceae; 63 from BSIs and 52 from UTIs) with a mutation frequency between 1 × 10−8 and 1.98 × 10−8. We also found 27 weak mutators (including 19 E. coli and eight other Enterobacteriaceae: eight from blood cultures and 19 from urine) with a mutation frequency between 1 × 10−7 and 1.91 × 10−7.

Figure 1.

Distribution of the mutation frequencies to rifampicin resistance for 310 Escherichia coli (black circles) and 125 other Enterobacteriaceae (white triangles) clinical strains. The vertical lines indicate the thresholds proposed for E. coli by Baquero et al. (2004). The mutation frequencies of E. coli and other Enterobacteriaceae species showed the same distribution, which was similar to that presented in the Fig. 1 of the paper by Baquero et al. (2004).

Mutation frequency calculations using the criteria defined by Denamur et al. (2002), helped to highlight 10-fold and 50-fold mutator E. coli strains: for the whole collection, 4.1% of the UTI isolates were 10-fold mutators (1.9% of BSI), while 0.9% of the UTI isolates were 50-fold mutators (0% in BSI).

The difference in the mutagenesis value might also be due to the difference in the median value of mutagenesis: Denamur et al. (2002) reported 5 × 10−9, but in this study, it was 2.2 × 10−8. It might also be due to geographical variation or differences in host characteristics: Baquero et al. (2004) showed that Spanish weak mutators tended to have higher frequencies of mutation than the Swedish ones.

Biofilm formation

A higher biofilm-forming capacity was observed in UTI strains than in BSI strains: 140/222 (63.1%) and 90/213 (42.3%), respectively (Fig. 2). The results were similar for the E. coli subpopulation in UTI strains and BSI strains: 87/161 (54.0%) and 71/149 (47.7%), respectively. The Mantel–Haenszel chi-squared test for stratified data (with E. coli in stratum 1, and others Enterobacteriaceae in stratum 2) gave a chi-square statistic = 18.4, a P-value < 0.001 and an Odds ratio = 2.26; 95% CI [1.55–3.28]. Uropathogenic E. coli (UPEC) are known to form biofilms easily (Stickler et al., 1995; Choong & Whitfield, 2000; Tenke et al., 2006). Previous studies (Soto et al., 2007; Rijavec et al., 2008) are in accordance with our results, showing that biofilm was produced by about half of the strains. However, taking into consideration the site of isolation, this percentage only reached 42.3% in BSI strains compared with 63.1% in UTI strains in our series. Soto et al. (2007) found 43% and 40% of biofilm-forming strains from patients with cystitis and pyelonephritis, respectively, while this percentage reached 63% in case of prostatitis.

Figure 2.

Biofilm production level of Enterobacteriaceae strains: Comparison between strains isolated from urinary tract infections (n = 222) and from bloodstream infections (n = 213). Biofilm production was classified using a 4-group classification system (Stepanovic et al., 2007) (see Methods).

As expected with the urease producer Proteus mirabilis, previously described as the predominant organism in biofilms from encrusted catheters (Jacobsen & Shirtliff, 2011) and urinary stones, 10/10 (100%) and 0/8 were biofilm producers in UTIs and BSIs, respectively (P < 0.001). Of the Klebsiella pneumoniae isolates (which are also urease positive), 14/15 (93.3%) and 3/13 (23.1%) were biofilm producers in UTIs and BSIs, respectively (P < 0.001). No significant difference in biofilm production was demonstrated for E coli UTI strains (87/161; 54.0%) vs. BSI strains (7/149; 47.7%). Some biofilm-producing species (i.e. P. mirabilis) are particularly isolated from urine during pyelonephritis associated with bacteraemia, compared with digestive translocation or other origin. The great majority of K. pneumoniae UTI strains are biofilm producers; however, this is not the case for BSI strains (urinary, pulmonary or digestive origin). In our series, only 35.2% of the BSI strains had a proven urinary tract origin.

Is there a relationship between the mutation frequency of a clinical collection of Enterobacteriaceae strains and their capacity to initiate a biofilm?

The bacterial population studied was relevant, because it reflected the bacterial species already described in the literature, responsible for biofilm formation (i.e. on urinary catheters) and/or hypermutable. Hypermutable bacteria, particularly P. aeruginosa, seem to be particularly frequent in respiratory samples from patients suffering of cystic fibrosis (Oliver et al., 2000; Mena et al., 2008; Conibear et al., 2009; Feliziani et al., 2010). Several authors have attempted to explain the large amount of hypermutators in a bacterial population known to readily form biofilms. One explanation has been suggested: bacterial biofilm formation can be induced by DNA damaging agents (involved in mutations) triggering the SOS response, through a connection between stress-inducible biofilm formation and the RecA-LexA interplay (Gotoh et al., 2010). Several damaging agents have been described, including silver nanoparticles used for their antibacterial properties (Radzig et al., 2013), production of oxidative product by other bacteria present in the biofilm (i.e. S.gordonii producing H2O2) (Itzek et al., 2011) or some antibiotics such as fluoroquinolones (Morero et al., 2011). The appearance of persister cells in the biofilm also promotes survival and may be related to the hypermutator phenotype or to the action of fluoroquinolones (Dörr et al., 2009, 2010). A second explanation has been proposed: by definition, hypermutable strains are expected to have higher capacities for adaptation (Jayaraman, 2011), and some mutations might be linked to an increase in biofilm formation. To explain the large amount of hypermutators in biofilms, some authors have involved the formation of persister cells, in a quiescent state, inside urothelial cells or within bladders (Blango & Mulvey, 2010; Wang & Wood, 2011). The fact that the generation of persister cells depends on growth rate adds to the fact that this phenotype is especially found in late cultures and therefore in chronic infections (which is the case with cystic fibrosis).

In this study, we could therefore expect a relationship between the mutation frequency of a collection of clinical Enterobacteriaceae strains and their capacity to initiate a biofilm.

Contrary to what could be expected, there was no significant correlation (Pearson's R correlation coefficient and P-values < 0.05 were considered statistically significant) between mutation frequency and the capacity to initiate a biofilm (R = −0.030, Fig. 3), irrespective of the source of the strain (urine: R = 0.022; blood: R = −0.057) or the group of species (E. coli: R = −0.038; other Enterobacteriaceae: R = −0.020). The ability of the bacterial population to form a biofilm was then studied between the two main peaks of mutation frequencies: 1 × 10−8 (1 × 10−8 ≤ f < 2 × 10−8) and 10-fold higher (1 × 10−7 ≤ f < 1 × 10−8). In the first group, the biofilm production, assessed by the DOm/DOC ratio, ranged from 0.49 to 20.04, with an average of 1.64 and a standard deviation of 2.16. In the second group, these values ​​were, respectively, 0.14, 4.13, 1.38 and 0.97. The number of strong mutators (i.e. MMR deficient type) was very low and the genetic basis was not characterised. Thus, solid conclusions may be obtained for weak mutators, but not for MMR deficient strong mutators (n = 8). Moreover, among the few highly biofilm-producing strains, only one weak mutator type (E. coli) was detected, while seven were normomutators among other strains (three K. pneumoniae, two E. coli, one K. oxytoca and one S. marcescens). These strains were mostly isolated from urine samples (6/7). However, 8 of 10 isolates of P. mirabilis from UTIs showed a weak mutator phenotype, all of which were biofilm producers (one was +biofilm producer, five were ++biofilm producers, and two were +++biofilm producers).

Figure 3.

Relationship between rifampicin resistance mutation frequencies and biofilm production for 435 Enterobacteriaceae strains: 222 from urinary tract infections (a) and 213 from bloodstream infections (b). Biofilm production was calculated as the ODm/ODc ratio as recommended by Stepanovic et al. (2007). The dotted vertical lines indicate the thresholds of mutation according to Baquero et al. (2004).

Several studies, especially in P. aeruginosa isolated from respiratory samples in patients with cystic fibrosis, showed a higher proportion of hypermutable strains in biofilms. Luján et al. (2011) showed that the adaptation of bacteria under these conditions could be favoured by the hypermutator phenotype. The emergence of biofilm production frequently implies increased adhesion between bacterial cells, which could be responsible for errors in estimating the number of colonies counted to measure the frequency of mutation. The relationship between mutation frequency and the capacity to form a biofilm is very complex and could be dependent on the state of growth: García-Castillo et al. (2011) showed decreased mutation frequencies of hypermutators in biofilms compared with planktonic conditions. This process, involved in the maturation of the biofilm rather than during the initiation steps, therefore appears to be an important point for hypermutator phenotype emergence. Not finding a correlation between increased mutation rates and initiation of biofilms in Enterobacteriaceae does not mean that there is not a relevant effect in other stages or aspects of biofilm growth.

Conclusion

In this study, we found no link between a greater capacity to initiate biofilm formation and a higher mutation frequency. This absence of correlation might be directly linked to the fact that strains were mainly isolated from acute infections. Despite the fact that an increased proportion of hypermutable bacteria have often been reported in clinical biofilm-forming strains (Denamur et al., 2002; Jayaraman, 2011) or chronic infections, it remains unclear whether the presence of hypermutators in clinical biofilms is a cause or a consequence of the mutator phenotype.

Acknowledgements

We thank Marie-France Travert for technical assistance with PFGE and Anne-Marie Gouraud for technical assistance with the biofilm assays. This work was supported by the European Urological Scholarship Program (EUSP) of the European Association of Urology (EAU) and by grants from Conseil Regional de Bretagne and from Fondation des Gueules Cassées. The authors declare that they have no conflicts of interest.

Ancillary