Evaluation and implementation of a membrane filter method for Cronobacter detection in drinking water


Correspondence: Lanzheng Liu, Jinan Municipal Center for Decease Control and Prevention, 2 Weiliu Road, Jinan 250021, China. Tel.: +86 0531 81278911;

fax: +86 0531 81278777;

e-mail: lanzhengliu1965@hotmail.com


A membrane filter (MF) method was evaluated for its suitability for qualitative and quantitative analyses of Cronobacter spp. in drinking water by pure strains of Cronobacter and non-Cronobacter, and samples spiked with chlorinated Cronobacter sakazakii ATCC 29544. The applicability was verified by the tests: for pure strains, the sensitivity and the specificity were both 100%; for spiked samples, the MF method recovered 82.8 ± 10.4% chlorinated ATCC 29544 cells. The MF method was also applied to screen Cronobacter spp. in drinking water samples from municipal water supplies on premises (MWSP) and small community water supplies on premises (SCWSP). The isolation rate of Cronobacter spp. from SCWSP samples was 31/114, which was significantly higher than that from MWSP samples which was 1/131. Besides, the study confirmed the possibility of using total coliform as an indicator of contamination level of Cronobacter spp. in drinking water, and the acquired correct positive rate was 96%.


Cronobacter genus, belonging to the Enterobacteriaceae family, comprises seven species (Joseph et al., 2012a) and ubiquitously exists in environmental sources including water, plants, animals, food ingredients, and a wide variety of processed foods (Iversen & Forsythe, 2003; Friedemann, 2007). Cronobacter spp. will cause opportunistic infection, and most Cronobacter cases were reported in adults (Food and Agriculture Organization and World Health Organization, FAO/WHO, 2008). However, the severe illnesses, such as necrotizing enterocolitis, septicemia, and meningitis that associate with high mortality, mainly occurred in neonates and infants (Himelright et al., 2002). The infections of the newborns were caused by strains from major species of C. sakazakii and other two species of C. malonaticus and C. turicensis (Joseph et al., 2012b). The majority of these illness cases have been epidemiologically linked with powdered infant formula (PIF) (Himelright et al., 2002); however, at least in two independent cases, Cronobacter infections were caused by extrinsic contamination of infant formula so far (Noriega et al., 1990; Ray et al., 2007). Thus, other potential contamination sources are also of concerns, such as unsafe drinking water that was implied by a strain of C. sakazakii isolated from PIF reconstitution water that associated with a Cronobacter case in Illinois in 2011 (Schindler & Metz, 1991; Lee & Kim, 2003; Hariri et al., 2013). Theoretically, drinking water is rarely contaminated in pipes on premises, which makes it the preferred supply way to provide safe drinking water, but in reality, the transferred water may not be as safe as expected, due to inadequately maintained water sources (United Nations International Children's Emergency Fund – UNICEF and WHO, 2012). The piped water supply on premises includes municipal water supply on premises (MWSP) and small community water supply on premises (SCWSP). The definition of SCWSP varies largely within and between countries (WHO, 2012). In the present work, SCWSP was defined as local water supply on premises in township(s) or/and village(s), and MWSP was defined as central water supply on premises in a county town or a city, referring to the Chinese national standard (Ministry of Health and Standardization Administration of the People's Republic of China – MH and SAPRC, 2006). In general, SCWSP systems are no match for MWSP systems in regard to providing stable and safe drinking water (WHO, 2012). In the present study, piped drinking water samples from MWSP and SCWSP were screened for the presence of Cronobacter spp. Because there was no reliable reference method for screening Cronobacter spp. in drinking water samples, a new method should be refined. Membrane filter (MF) method was chosen to be developed, which has been widely used in detection of certain kinds of bacteria, such as coliform, fecal streptococci, enterococci, Staphylococcus aureus, and Pseudomonas aeruginosa etc., and is extremely effective to monitor waters including drinking water and a variety of natural waters (American Public Health Association – APHA et al., 1998; International Organization for Standardization – ISO, 2000). A MF method for detection of Cronobacter spp. in PIF has been designed by Miled et al. (2010) and was compared with a most probable number (MPN) method modified from the ISO/TS 22964 method (ISO and International Dairy Federation – IDF, 2006). Thus, a MF method was adapted from previous MF methods considering its suitability for recovering Cronobacter cells with and without stress results from chlorination in drinking water.

Materials and methods

Membrane filter method procedure

A modified MF method was developed to detect Cronobacter spp. in drinking water, after evaluating the impact of certain factors (volume of filtered sample, medium, number of colonies on FMs, temperature, and time; not published) on Cronobacter recovery by pretests that were based on the standard MF method for detecting coliform in water and the ISO/TS 22964 method (APHA et al., 1998; ISO, 2000; ISO and IDF, 2006). The pretests confirmed that preincubation of filter membranes (FMs) on tryptone soy agar (TSA) for 4 h before incubating FMs on Druggan–Forsythe–Iversen (DFI) agar was able to overcome the inhibition of stressed Cronobacter cells by directing application of DFI agar (Osaili et al., 2010), and other pretested factors that referred to literature were not in need of adjustments. The MF method protocol was described in the following paragraphs.

One hundred millilitre or replicates of smaller sample volumes such as four replicates of 25 mL portions of drinking water samples were tested to obtain 20–80 target colonies, and no more than 200 colonies of all types grew on a FM surface (Barnes et al., 1989; APHA et al., 1998). Samples were filtered through mixed cellulose ester FMs with 0.45 μm pore size and 47 mm diameter (Millipore, MA) by autoclaved vacuum pump (PLUS, Millipore). The filtered FMs were cultured on TSA (Land Bridge, Beijing, China) at 36 °C. After 4 h of incubation, the cultured FMs were transferred onto DFI agar (Oxoid, Hampshire, England), incubating for 20 h at the same temperature.

The typical Cronobacter colony on the FMs has a blue-green color. For qualitative analysis of Cronobacter spp., after 24-h incubation, two or more typical Cronobacter colonies were picked and cultured on TSA (Land Bridge) for another 24 h and then verified by Vitek 2 Compact system and Vitek GNI cards (bioMérieux, NC). Species of the verified strains were determined by fusA sequencing; premiers and protocol were available at http://pubmlst.org/cronobacter/info/protocol.shtml; phylogenetic analyses were performed using the maximum-likelihood algorithm in MEGA5, the fusA sequence from the genome of Citrobacter koseri (GenBank accession number CP000822), which closely related to Cronobacter spp. was used as an outlier (Tamura et al., 2011; Joseph et al., 2012b).

For quantitative analysis of Cronobacter spp., typical colonies were counted, and 10 of them were verified by Vitek 2 Compact system and Vitek GNI cards (bioMérieux). The final count was proportionally adjusted based on the verification results.

Evaluation of membrane filter method

Inclusivity (sensitivity) and exclusivity (specificity) test

In the study, 45 pure strains of Cronobacter spp. and 33 pure strains of potential competitive bacteria were applied (Table 1). Six strains were collected ones; the rest 72 strains were all isolated from food and clinical samples in our laboratory and identified by Vitek 2 Compact system and Vitek GNI cards (bioMérieux), and the Cronobacter strains were speciated by fusA sequence analysis (Joseph et al., 2012b). At first, the strains were cultured on nutrient agar (Land Bridge) for 24 h at 36 °C. Then, each culture was serially diluted in sterile saline water: target bacteria were diluted to 1-log colony-forming unit (CFU) per mL, and nontarget bacteria were diluted to 3 log CFU mL−1. One millilitre of each dilution was added into 100 mL sterile saline water followed by filtration and incubation mentioned in the ‘membrane filter method procedure’ section. After incubation, growth characteristics of each bacterium were recorded. Typical colonies yielded by non-Cronobacter strain were identified by Vitek 2 Compact system and Vitek GNI cards (bioMérieux).

Table 1. Evaluation results of the MF method with the pure cultures
SpeciesNo. of total strainsNo. of positive strainsGrowth characteristics
  1. a

    Including a collection strain of ATCC 25944.

  2. b

    Identified as Klebsiella ozaenae correctly by Vitek 2 Compact system and Vitek GNI cards.

  3. c

    Including a collection strain of ATCC 8099.

  4. d

    Weak growth: colony count was more than two log units less than the initial spike level.

  5. e

    The two strains were CMCC 51577 and CMCC 51579; CMCC: National Center for Medical Culture Collections, Beijing, China.

  6. f

    Including a collection strain of ATCC 14028.

  7. g

    The strain was ATCC 6538.

Cronobacter strains
Cronobacter sakazakii 28a28Blue-green colonies
Cronobacter malonaticus 33Blue-green colonies
Cronobacter dublinensis 1010Blue-green colonies
Cronobacter turicensis 44Blue-green colonies
Non-Cronobacter strains
Citrobacter freundii 30Colorless colonies with black center
Klebsiella pneumoniae 10Colorless colonies
Klebsiella ozaenae 11Blue-green coloniesb
Klebsiella oxytoca 10Colorless colonies
Escherichia coli 2c0Colorless colonies with light green center
Escherichia fergusonii 20Colorless colonies
Enterobacter cloacae 10Colorless colonies
Proteus mirabilis 40Brown colonies, brown surrounding medium
Proteus vulgaris 10Weak growthd, colorless colonies
Shigella flexneri 2e0Colorless colonies
Salmonella spp.4f0Colorless colonies with black center
Morganella morganii 10Colorless colonies
Other organisms
Acinetobacter baumannii 10Colorless colonies
Pseudomonas aeruginosa 40Dark brown colonies, some filter members turned into deep green
Aeromonas sobria 10Colorless colonies
Candida albicans 10No growth
Enterococcus faecalis 10No growth
Staphylococcus aureus 1g0No growth
Stenotrophomonas maltophilia 10Weak growth, colorless colonies

Recovery test

Recovery test was adopted to determine recovery rate and repeatability of the MF method to detect Cronobacter spp. due to the lack of standard method to measure the quantity of Cronobacter spp. that naturally exist in drinking water. In the experiments, naturally contaminated drinking water samples were collected and spiked with chlorinated target bacterium, referring to the membrane filter method procedure to detect coliform, which was interfered by large number of competitive microbial communities and chlorination stress (APHA et al., 1998).

In the experiments, Cronobacter sakazakii ATCC 29544 (American Type Culture Collection, MD), the type strain of C. sakazakii, which is the type species of the genus Cronobacter, was adopted as the target bacterium (Iversen et al., 2007a). Three-log CFU per mL suspension of ATCC 29544 was made in sterile phosphate-buffered water by serial 10-fold dilution, after two times of resuscitation in tryptic soy broth (Land Bridge) for 24 h at 36 °C. To simulate the stress imposed on microorganisms during chlorination, 0.8 mL 1 : 100 000 dilution of 5% sodium hypochlorite (wt/vol; Lierkang, Dezhou, China), determined by preliminary tests, was added to 10 mL prepared ATCC 29544 suspension, reducing the log unit of the bacteria by 1. The suspension was vortexed for 3 min at room temperature and neutralized with 1 mL 0.1% sterile sodium thiosulfate, producing chlorinated suspension. Diluted chlorinated suspension was made by diluting the chlorinated suspension 10-fold.

Two liters of drinking water from a SCWSP plant, which had tested positive for Cronobacter spp. 2 weeks early (not shown), was collected in a 4-L sterile polypropylene bottle and dispensed into 100-mL sterile polypropylene bottles preadded with 1 mL of 1% sterile sodium thiosulfate solution. Two samples were spared as unspiked samples to evaluate background colonies, five samples were, respectively, spiked with 1 mL of the chlorinated suspension, and another five samples were spiked with 1 mL diluted chlorinated suspension individually. After 10-min standing, all the spiked samples were filtered as well as the unspiked ones. The filtered FMs were incubated according to the membrane filter method mentioned above, and blue-green colonies were enumerated after incubation. Meanwhile, cell density of Cronobacter spp. in the chlorinated suspension, the diluted chlorinated suspension and the suspension before chlorination were measured by standard plate count method on TSA, incubating at 36 °C for 24 h, and the colony counts were compared with those of spiked samples.

Detection of Cronobacter spp. in drinking water

From July to August in 2010, 144 drinking water samples from MWSP and SCWSP were collected in Jinan, Shandong Province in China. Fifty-eight MWSP samples were collected from users' taps. Eighty-six SCWSP samples were collected from 43 water supply systems: one sample from the plant and a paired one from the end user's tap.

From July to August, 2011, 104 drinking water samples from users' taps were collected in the same area: 73 were MWSP samples and 31 were SCWSP samples.

The water samples were collected in 500-mL sterile glass containers preadded with sodium thiosulfate to neutralize residual chlorine (MH and SAPRC, 2006). The final concentration of sodium thiosulfate in water samples below 1 g L−1 would not stress Cronobacter cells, indicated by pretests (not shown) in which DFI medium containing 1 g L−1 sodium thiosulfate recovered unstressed Cronobacter cells correctly. The qualitative analysis of Cronobacter spp. using the MF method was conducted within 4 h after sample collection. For each MWSP sample, 100 mL drinking water was filtered, and 25 mL drinking water was filtered for SCWSP samples with four replicates. The unused portion of each sample was stored at 4 °C.

For the FM without blue-green colony, it was recorded as negative when the total number of colonies grew on a piece of FM was below 200, and as invalid when the total number of colonies was above 200. If any portion of a sample was invalid, the stored part of the sample was retested with less volume (10 or 5 mL) and more FMs (10 or 20 pieces) according to the result of its first-round test. The total holding time of the samples was no more than 30 h (APHA et al., 1998).

Quantitative test of total coliform

Total coliform of the drinking water samples except the SCWSP samples collected in 2011 were tested using the fermentation technique of the Chinese standard method (GB/T 5750, MH and SAPRC, 2006). Ten millilitre and 1 mL of each sample and 1 mL 10-fold dilutions of the samples were added to tubes with fermentation tubes that contain lactose broth and incubated for 24–48 h at 36 °C, respectively, with five replicates. The presumptively positive tubes that produced gas and acid were confirmed by streak inoculation on eosin methylene blue agar (Land Bridge) and subsequent refermentation in lactose broth (Land Bridge). The number of coliform per 100 mL was estimated based on a 15-tube MPN table.

Pulsed field gel electrophoresis

Pulsed field gel electrophoresis (PFGE) fingerprint patterns were obtained for a total of 32 Cronobacter isolates, using the restriction enzymes XbaΙ and SpeΙ. PFGE of Cronobacter spp. was performed by following the developed PulseNet USA protocol (Brengi et al., 2012). The gel was run for 17–18 h at 6 volts per cm in a CHEF-MAPPER system (Bio-Rad, CA) at an initial switch time of 1.8 s and a final switch time of 25 s. PFGE types were analyzed using BioNumerics software version 6.6 (Applied Maths, Sint-Martens-Latem, Belgium). The unweighted pair group method using arithmetic averages (UPGMA) was applied to compare and to cluster the patterns with averaged similarity matrices involving Dice coefficient from the two experiments (XbaΙ and SpeΙ), and the position tolerance of 1.0% and the optimization of 1.0% were applied. PFGE patterns were interpreted according to the criteria of Tenover et al. (1995).

Statistical analysis

Chi-squared test was used to compare the positive proportions of MWSP and SCWSP samples. The isolation rates of samples from users and corresponding water plants of SCWSP were analyzed with McNemar's test. To study internal consistency, Kappa test was applied. These statistical analyses were performed with SPSS 13.0 (SPSS Inc., Chicago).

Results and discussion

Evaluation of membrane filter method

Evaluation results of the MF method with the pure cultures were given in Tables 1. First, the cultures were differentiated by colony color, 45 Cronobacter strains all grew blue-green colonies, so the sensitivity (positive number divided by total number of Cronobacter strains) was 100%; 32 of 33 non-Cronobacter strains did not yield blue-green colonies; therefore, the specificity (negative number divided by total number of non-Cronobacter strains) was 97%. It was similar to the results that 99% (176/177, Iversen & Forsythe, 2007b) or 100% (95/95, Iversen et al., 2004) Cronobacter strains grew blue-green colonies on DFI agar, and 13% (19/148, Iversen et al., 2004) non-Cronobacter Enterobacteriaceae gave false-positive results. The false-positive strain found in the present work was Klebsiella ozaenae. Thus, blue-green colonies on cultured FMs need to be verified by biochemical tests. After the verification, the sensitivity and the specificity of the MF method were both 100%.

For the spiked samples, at a spike level of 101.0 ± 9.0 CFU (after chlorination) which the ATCC 29544 cells reduced from 1210.0 ± 110.0 CFU (before chlorination) to, the MF method could recover 82.8% chlorinated ATCC 29544 cells in the naturally contaminated drinking water samples, which carried 109.5 ± 8.5 CFU of total background colonies without typical colony of Cronobacter spp., and the relative standard deviation (RSD) was 12.6%; at a spike level of 9.5 CFU, recovery rate was 109.3%, and the RSD was 23.8%. In brief, the MF method properly recovered the stressed ATCC 29544 cells, and the preconditions that 20–80 target colonies yielded and no more than 200 colonies of all bacteria grew on an FM surface, which determined the filtered drinking water volume, were appropriate. In sum, the MF method is suitable for qualitative and quantitative analyses of Cronobacter cells in drinking water.

Drinking water samples

Although a strain of C. sakazakii was isolated from PIF reconstitution water that associated with a Cronobacter case in Illinois in 2011 (Hariri et al., 2013), so far there is no direct evidence for relationship between polluted water and Cronobacter infection. This may result from low contamination level of Cronobacter spp. in municipal drinking water (Schindler & Metz, 1991), which was supported by the present work in which the positive rate of Cronobacter spp. in the MWSP samples was 0.8% (1/131, 0–4.1%, 95% confidence interval, Table 2), significantly lower than that of either type of the SCWSP samples: the positive rate of samples from water plants and users was 27.9% (12/43, 15.3–43.5%, 95% confidence interval, Table 2) and 25.7% (19/74, 16.1–37.1%, 95% confidence interval, Table 2), respectively (< 0.01). A total of 32 Cronobacter strains were acquired, including 11 strains of C. sakazakii, five strains of C. malonaticus, three strains of C. dublinensis, and 13 strains of C. turicensis (Fig. 1). In short, Cronobacter contamination was frequently detected in the SCWSP samples, and the issue might exist around the world, in view of that small community water supplies in rural areas were more prone to contamination (WHO, 2012). In 2010, 46% of world population was still without access to piped drinking water on premises (UNICEF and WHO, 2012), and their drinking water also might be at risk of Cronobacter contamination. Thus, contamination of Cronobacter spp. in drinking water should be concerned and evaluated all over the world, especially in rural areas. Moreover, the relationship between water-borne Cronobacter spp. and Cronobacter infection in infants should be researched in the future.

Table 2. The numbers of drinking water samples positive for Cronobacter spp.
YearSCWSP samplesMWSP samples
From water plantsFrom users
No. of positive samplesNo. of total samplesNo. of positive samplesNo. of total samplesNo. of positive samplesNo. of total samples
Figure 1.

Maximum-likelihood tree of the fusA alleles of the Cronobacter strains (= 32). The tree was drawn to scale using MEGA5 with 1000 bootstrap replicates.

For the SCWSP, the positive proportions of the samples from users in 2010 were not of statistical difference (> 0.05) from either the samples from water plants or the samples from users in 2011. Individual samples contaminated with Cronobacter spp. were presented in Table 3. However, Cronobacter spp. did not inhabit in drinking water pipes all through (κ < 0.4): six water supply systems were detected positive for Cronobacter spp. in the samples from water plants only, and eight water supply systems were detected positive for Cronobacter spp. in the samples from users only; the Cronobacter strains in different water supply systems all did not belong to genetically ‘indistinguishable’ category by the PFGE fingerprint with XbaΙ enzyme and SpeΙ enzyme (Fig. 2). Although Cronobacter spp. were detected in six pairs of samples from water plants and users, no genetically indistinguishable strains were discerned by PFGE fingerprints with XbaΙ enzyme and SpeΙ enzyme (Fig. 2), which implied that diversity of Cronobacter strains existed in water supply systems from water plants to users.

Table 3. Samples contaminated with Cronobacter spp.
TimeSample numberTotal coliform (MPN/100 mL)
  1. a

    Rx was SCWSP sample number, Rx-1 was a SCWSP sample collected from a water plant, and Rx-2 was a SCWSP sample collected from an user.

  2. b

    Ux was MWSP sample number.

R1009-2< 2
R1017-12.4 × 102
R1017-22.4 × 102
R1023-19.2 × 102
R1025-21.3 × 102
R1029-22.4 × 102
R1032-11.3 × 102
R1032-23.5 × 102
R1039-13.5 × 102
R1039-22.4 × 102
Figure 2.

Dendrogram generated from PFGE profiles of Cronobacter isolates (= 32) with the enzymes XbaΙ and SpeΙ. The isolate numbers were consistent with their original samples showed in Table 3.

For the MF method, the data of drinking water samples showed that for MWSP samples, the water volume filtered through a FM can be enlarged to more than 100 mL for specific demands; for SCWSP samples, generally, it is appropriate to filter 25-mL sample through a FM with four replicates, processing 100 mL in total, and under severe contamination conditions, reducing filtered volume and increasing replicates are needed. The result also supported the above conclusion that blue-green colonies on cultured FMs need to be discerned by biochemical tests. For example, in this work, an isolate that yielded blue-green colonies as typical colonies of Cronobacter spp. on the FM was identified as Plesiomonas shigelloides.

Because Cronobacter spp. (fermenting lactose and producing gas) are members of coliform (Gurtler et al., 2005), theoretically total coliform may be used as an alternative indicator of Cronobacter contamination level in drinking water, which was supported by the results obtained in the present work that the correct positive rate was 96% (26/27, P > 0.05): only one sample that its total coliform were below the detection limit (< 2 MPN/100 mL, Table 3) tested positive for Cronobacter spp. This result does not seem to be consistent with the conclusion of the Scientific Panel on Biological Hazards of the European Food Safety Authority that no universal correlation between Enterobacteriaceae and Cronobacter spp. exists, although Cronobacter spp. is a member of the Enterobacteriaceae (The Commission of the European Communities, 2007). In view of that total coliform did not completely indicate Cronobacter contamination, under certain conditions such as using the drinking water for raising infants, Cronobacter spp. should be detected.


The authors thank Yiheng Liu for her contribution in spelling revision and the technicians who helped to collect samples.