Comparative study of Cronobacter identification according to phenotyping methods

Since previous version(s) of the API20E database identified strains as “Enterobacter sakazakii,” this result was considered to be a match for all Cronobacter species. A significant difference (p??0.05, X
² test) was found in the percentage of strains producing a profile that resulted in a match with the different versions of the database. Only 82.3 % of strains resulted in a match with version 5.0, while 90.0 % of strains resulted in a match with previous version(s). These results are summarized in Additional file 1: Table S1. The majority of strains (68.8 %) produced matches with both the old and new versions of the database; however, 6.3 % of strains produced profiles which resulted in match previously but a mismatch following the update to the API20E database. Only one strain (0.4 %) showed a mismatch in the archived records, and a match following the update to the database.

Strains of Cronobacter were misidentified as Enterobacter aerogenes, E. amnigenus, and E. cloacae with the previous version of the API20E database and as E. aerogenes, E. amnigenus, E. cancerogenus, E. cloacae, and Serratia liquefaciens with the updated version. Strains of E. cloacae and E. hormaechei were misidentified as Cronobacter spp. (or ‘E. sakazakii’) with both versions of the database. Though most profiles returned a percent identification for each species, seven profiles did not do so with the updated database, returning a result of “Unacceptable profile”. These results were considered to be mismatches and such results incorporate further uncertainty into identification of isolates producing these profiles.

A total of 61 API20E profiles were observed for confirmed strains of Cronobacter species. All of these profiles were positive for amygdalin fermentation and negative for hydrogen sulfide production. More than 90 % of these profiles were also positive for ?-galactosidase (96.7 %), arginine dihydrolase (93.4 %), citrate utilization (93.4 %), and D-glucose (95.1 %), D-mannitol (91.8 %), L-rhamnose (91.8 %), D-saccharose (93.4 %), D-melibiose (96.7 %), and L-arabinose (93.4 %) fermentation. Additionally, more than 90 % of these profiles were negative for urease activity (95.1 %). The observed reactions for arginine dihydrolase, citrate utilization, hydrogen sulfide production, urease activity, and acid production from glucose, D-mannitol, L-rhamnose, D-saccharose, D-melibiose, and L-arabinose match the original description of the Cronobacter genus [5, 6].

Of the 61 API20E profiles, the most common were 3305373, 3307173, and 3305173 which represented 60, 41 and 40 strains, respectively. Together, these three profiles were observed for more than half of the strains analysed (141/260; 54.2 %). These profiles differed from one another by only two traits; gelatinase activity and inositol fermentation. Each of these profiles resulted in a species identification of “E. sakazakii” with version 4.0 of the database and were identified as Cronobacter spp. with version 5.0. Profile 3307173 was only observed for confirmed strains of Cronobacter, belonging to four different species; C. sakazakii (n?=?34), C. malonaticus (n?=?4), C. turicensis (n?=?2), and C. dublinensis (n?=?1). In addition to C. sakazakii (n?=?32) and C. malonaticus (n?=?8), strains of E. hormaechei (n?=?3) and E. cloacae (n?=?1) were identified with profile 3305173. Strains of C. sakazakii (n?=?49), C. malonaticus (n?=?7), C. turicensis (n?=?4) and E. hormaechei (n?=?6) produced profile 3305373. Therefore, while these profiles were frequently observed for Cronobacter spp., they are not necessarily specific to members of this genus and may be produced by members of the Enterobacter genus.

Though easily distinguished by colony morphology on Druggan-Iversen-Forsythe agar, Enterobacter isolates have been misidentified as members of the Cronobacter genus in the past. For example, two illnesses in Mexico were attributed to C. sakazakii using cultural and molecular identification methods [24]. Subsequent DNA sequence-based analysis of the patient isolates, however, identified them as members of the Enterobacter genus [25]. Similarly, Townsend et al. (2008), used 16S rDNA sequencing to identify 10 clinical strains of Enterobacter hormaechei that had been previously identified as C. sakazakii by phenotyping [26]. The correct identification and genotyping of these isolates revealed that a previously unrecognised noscomial outbreak of E. hormaechei had occurred [26]. These investigations show that clinical Enterobacter isolates have been mistaken for C. sakazakii and though infectious, these organisms are not addressed in the microbiological criteria applied to PIF [1, 25–29]. Although false positive results were observed for E. cloacae and E. hormaechei, these organisms have been isolated from infant formula and can result in neonatal infections [25, 26]. Even though the the risk from these organisms in PIF has not been closely examined, they may present a hazard and their misidentification as Cronobacter spp. could result in rejection of an unsafe batch of formula.

For the ID32E biochemical kit, Fanjat et al. noted that variation in the L-arabinose and ?-maltosidase tests resulted in misidentification of some Cronobacter isolates (then known as “E. sakazakii”) when using version 2.0 of the ID32E database [30]. The update to version 3.0 reflected this variability, resulting in 100 % correct identification of the isolates [30]. Conversely, updating the database to version 4.0 resulted in higher numbers of misidentifications of Cronobacter species (p??0.05, X
² test). With the previous versions of the ID32E database, 88.9 % of strains produced a match (as E. sakazakii); however, this percentage drops to only 43.2 % when the updated version of the database is used (Additional file 2: Table S2).

Over one third of strains (37.9 %) produced a match both in the culture collection records and with version 4.0 of the ID32E database; however, 36.6 % of strains produced a match in the culture collection records, but a mismatch after the database was updated. As with the API20E results, only one strain (0.3 %) showed a mismatch in the archived culture collection records, but a match with the updated database.

The major limitation with the update to the ID32E database is that it attempts to identify strains to the species level, but members of the Cronobacter genus can be difficult to differentiate without highly specific methods. In particular, C. sakazakii and C. malonaticus are so closely related that they cannot be reliably differentiated using 16S rDNA sequencing methods [16]. The ID32E biochemical test panel is simply not specific enough to differentiate the seven species of Cronobacter. If these results are examined to only the genus level, 82.3 % of confirmed Cronobacter strains were assigned to the Cronobacter genus using version 4.0 of the ID32E database. Though this still represents a significant difference (p??0.05, X
² test) in the accuracy of identification from the archived culture collection records, it is considerably more accurate than identification to the species level.

With the ID32E test kit, Cronobacter strains were misidentified as Enterobacter cancerogenus, E. cloacae, and Stenotrophomonas maltophilia with the old version(s) of the database and as Buttiauxella agrestis, Citrobacter freundii, Cit. koseri, E. aerogenes, E. cancerogenus, E. cloacae, Escherichia vulneris, Leclercia adecarboxylata, Serratia liquefaciens, S. rubidea, and Stenotrophomonas maltophilia with the updated database. Meanwhile, strains of Cit. freundii, Cit. koseri, E. hormaechei, Escherichia hermanii, Franconibacter helveticus, F. pulveris, Leclercia adecarboxylata, Pantoea spp., Siccibacter colletis, and Sic. turicensis were misidentified as Cronobacter species with the previous version of the database while Cit. freundii, Cit. koseri, E. hormaechei, E. pyrinus, Esh. hermanii, F. helveticus, L. adecarboxylata, and Siccibacter turicensis were misidentified as Cronobacter species with the updated version of the database. Misidentification of Enterobacter strains as Cronobacter spp. also occurred with the API20E test kit and the implications of such misidentifications were discussed previously. The misidentifications of Franconibacter and Siccibacter species as members of the Cronobacter genus is unsurprising as these species were briefly considered to be part of the Cronobacter genus [13–15, 20]. Forty-eight ID32E profiles returned a species identification with a description of ‘”Unacceptable profile” instead of a percent identification with the updated version of the database. As with the API20E profiles, this produces more uncertainty in the species identification resulting from those profiles.

A total of 155 ID32E profiles were observed for confirmed strains of Cronobacter species. Nearly all (152/155, 98.1 %) of the profiles give a positive result for ?-glucuronidase activity. Similarly, only one profile (0.7 %) has a positive reaction for rhamnose acidification. More than 90 % of the profiles for known Cronobacter spp. produced positive results for ornithine decarboxylase (90.3 %), arginine dihydrolase (94.2 %), ?-glucosidase (98.1 %), malonate utilization (95.5 %), and acidification of maltose (91.0 %), D-glucose (91.0 %), D-saccharose (97.2 %), L-arabinose (93.6 %) and D-arabitol (90.3 %). Additionally, more than 90 % of these profiles were negative of L-aspartic acid arylamidase (97.4 %), ?-galactosidase (92.3 %), and acidification of both L-arabitol (97.4 %), and 5-ketogluconate (94.3 %). The observed reactions for ornithine decarboxyase, arginine hydrolase, maltose utilization and acidification of D-glucose and L-arabinose match the original genus description of Cronobacter; however, the results for the acidification of L-rhamnose and D-arabitol do not match the original description of the genus [5, 6]. These differences may be due to the use of different phenotyping methods when defining the genus. Additionally, the results for 4-nitrophenyl-?-D-glucopyranoside, and acidification of D-arabitol and D-sorbitol differ significantly from the expected phenotypes specified in the proposed ISO standard [10]. While only 58.7 % of Cronobacter ID32E profiles were positive for 4-nitrophenyl-?-D-glucopyranoside, the ISO standard states that 100 % of strains produced a positive result for this test [10]. Similarly, 74.2 % of ID32E profiles were found to be positive for the acidification of D-sorbitol; however, none of the strains included in the ISO standard produced a positive result [10]. Finally, nearly all of the profiles (90.3 %) were positive for the acidification of D-arabitol, but only a few strains of C. dublinensis, C. turicensis, and C. universalis were reported to produce positive results in the ISO standard [10]. The presence of such discrepancies suggests that the biochemical test panel may not be completely accurate in its characterization of isolates. Importantly, biochemical test panels are suggested as an alternative to more traditional biochemical test methods in the draft ISO standard; however, the differences between the expected results given in the standard and the results reported here suggest that the ID32E test kit is not an acceptable alternative method for biochemical identification of suspect Cronobacter isolates [10].

The most common ID32E profile was 34276767250 (n?=?34). Only C. sakazakii strains produced this profile, indicating a match to the species level. In addition to the tests matching the genus description given above, this profile is also positive for palatinose acidification, a characteristic given in the original description of C. sakazakii [6]. The remaining characteristics used for the description of the species are not included in the ID32E test kit. The next two most common profiles were 34276767050 (n?=?22) and 34276763050 (n?=?19). Each of these profiles was observed for three Cronobacter species. Profile 34276767050 was produced by strains of C. sakazakii (n?=?16), C. malonaticus (n?=?4), and C. turicensis (n?=?2). According to the updated ID32E database (version 4.0), these strains were all identified as C. sakazakii; however this result is a match for only 16/22 strains (72.3 %). Similarly, profile 34276763050 was observed for C. sakazakii (n?=?16), C. malonaticus (n?=?2), and C. turicensis (n?=?1). The species identification for this profile reported by the updated database was C. dublinensis. Though correct to the genus level, none of the strains producing this profile resulted in a match to the species level. It should also be noted that these three most common profiles differ from one another in only two tests, ?-glucosidase activity and trehalose acidification.

Both the API20E and ID32E biochemical kits include a test for inositol fermentation. Of the 61 API20E profiles, 62.3 % were positive for inositol fermentation. A total of 159 strains were identified as being inositol positive with this test kit, including 112 strains of C. sakazakii. With the ID32E test kit 80.7 % of the 155 unique profiles were positive for inositol fermentation. Of the 280 strains identified as being inositol positive with the ID32E test kit, 192 were C. sakazakii. As a major pathogenic species in the genus, the high number of C. sakazakii strains showing positive results for inositol fermentation is interesting due to the association of the trait with virulence [31]; however, this result may be due to over representation of clinical isolates of C. sakazakii in the culture collection and further investigation may be necessary for confirmation.

Two of the major limitations of the API20E and ID32E kits are that they rely on a limited number of tests (20 or 32, respectively) and are subjectively based on visual detection of a color change when reporting results. Thus, the interpretation of the colors by an individual researcher could affect the reported profile and subsequent species identification. The Vitek system attempts to avoid these problems using 64 biochemical tests and automated detection of the color changes for more consistent results. Using a subset of strains, the Vitek system identified all 14 Cronobacter strains as members of the “Cronobacter sakazakii group”; Table 2. Although this term has no taxonomic standing and has not been defined, it was taken here as being equivalent to the Cronobacter genus. In contrast, the API20E misidentified 8/14 (57.1 %) of these strains to the genus level and the ID32E misidentified 13/14 (92.9 %) to the species level. Notably, though they were assigned to the ‘C. sakazakii group’ with the Vitek GN card, the type strains for C. condimenti (LMG26250^T), C. turicensis (LMG23827^T), C. universalis (NCTC9529^T) and C. dublinensis (LMG23823^T) were misidentified with both the API20E and ID32E test kits; Table 2. The ID32E kit did identify these strains as members of the Cronobacter genus, but did not correctly identify these type strains to species level.

  

Vitek GN

API20E

ID32E

Strain

Species

Identification

Profile

Species identification (v5.0)

Profile

Species identification (V4.0)

LMG26250^T

Cronobacter condimenti

Cronobacter sakazakii group

3367373

Cronobacter spp.

34217360051

Cronobacter muytjensii

LMG23823^T

Cronobacter. dublinensis

Cronobacter sakazakii group

7347373

Cronobacter spp.

34256166211

Cronobacter malonaticus

8

Cronobacter malonaticus

Cronobacter sakazakii group

3004153

Escherichia vulneris

34774563051

Cronobacter malonaticus

16

Cronobacter muytjensii

Cronobacter sakazakii group

0004153

Serratia plymuthica

30676563051

Cronobacter malonaticus

4

Cronobacter sakazakii

Cronobacter sakazakii group

3305173

Pantoea spp. 3

34074743011

Enterobacter cloacae

5

Cronobacter sakazakii

Cronobacter sakazakii group

3305173

Pantoea spp. 4

14074743211

Enterobacter amnigenus

12

Cronobacter sakazakii

Cronobacter sakazakii group

0005173

Escherichia vulneris

30674773050

Cronobacter malonaticus

1436

Cronobacter sakazakii

Cronobacter sakazakii group

3305373

Cronobacter spp.

34276763011

Cronobacter malonaticus

1437

Cronobacter sakazakii

Cronobacter sakazakii group

3305173

Cronobacter spp.

34276763010

Cronobacter malonaticus

1438

Cronobacter sakazakii

Cronobacter sakazakii group

1304373

Enterobacter gergoviae

14234767010

Cronobacter malonaticus

9

Cronobacter turicensis

Cronobacter sakazakii group

1006523

Cronobacter spp.

30074773050

Enterobacter cloacae

LMG23827^T

Cronobacter turicensis

Cronobacter sakazakii group

7315373

Enterobacter gergoviae

34276767211

Cronobacter malonaticus

NCTC9529^T

Cronobacter universalis

Cronobacter sakazakii group

3205373

Escherichia vulneris

24276777051

Cronobacter malonaticus

1435

Cronobacter universalis

Cronobacter sakazakii group

3304373

Cronobacter spp.

34356767010

Cronobacter sakazakii

162

Escherichia hermanii

Pantoea spp.

1004153

Cronobacter spp.

34676767050

Cronobacter sakazakii

LMG23732^T

Franconibacter helveticus

Cronobacter sakazakii group

1024153

Cronobacter spp.

30675567010

Cronobacter sakazakii

1393

Franconibacter pulveris

Cronobacter sakazakii group

3004173

Cronobacter spp.

04275763310

Enterobacter cloacae

LMG24057^T

Franconibacter pulveris

Cronobacter sakazakii group

3004173

Escherichia vulneris

04275773310

Enterobacter cloacae

NCTC14934^T

Siccibacter colletis

Pantoea spp.

3204153

Escherichia vulneris

04077563310

Buttauxiella agrestis

In addition, the Vitek GN card misidentified members of the Franconibacter genus as members of the “C. sakazakii group”. The type strain of F. helveticus (LMG23732^T) was also identified as a member of the Cronobacter genus by the other two test kits, while the type strain of F. pulveris (LMG24057^T) was misidentified with the API20E test panel; Table 2. As mentioned previously, this not unexpected as current strains in the Franconibacter genus were briefly considered to be a part of the Cronobacter genus before Franconibacter was defined [13, 15]. The false positive results generated by the Franconibacter species with the Vitek system are important for PIF manufacturers, as these species have not been linked to any cases of human illness but occur in similar ecosystems as Cronobacter species [15]. The international microbiological criteria applied to PIF require the absence of all Cronobacter species and this misidentification may cause manufacturers to incorrectly reject a batch of microbiologically safe infant formula.

Of the three biochemical methods evaluated in this study, the Vitek system performed better than the API20E or ID32E for identification of a subset of strains, including the type strains for all Cronobacter species and strains which had been previously misidentified. Though some false positive results were observed, this method was able to identify all tested Cronobacter species as members of the “Cronobacter sakazakii group;” however, relatively few strains were analysed with this method and analysis of a larger number of strains would be necessary to properly determine the accuracy of the Vitek system for identification of Cronobacter isolates. The FDA BAM method does suggest using either the Vitek 2 GN cards or the ID32E biochemical test panel for identification of suspect isolates, but there is a high cost associated with the initial purchase of the Vitek 2 instrument. Thus, while this method did seem to perform better in this small-scale analysis, the technology may unfortunately not be accessible to all PIF manufacturers and food testing laboratories.

As reported here and by others, commercially available biochemical test panels are not sufficient for accurate identification of Cronobacter species and even show difficulties in identifying strains to the genus level [7, 20, 30, 32]. Thus, the use of an alternative method may be necessary for accurate species identification of suspect Cronobacter isolates from PIF. DNA sequence-based identification methods are more reliable, but also more expensive, labor-intensive and have a long turnaround time. As DNA sequencing becomes cheaper and easier, it may be possible to incorporate it into manufacturers’ testing schemes, but this is not currently feasible. Similarly, while many other alternative methods are available for identification of Cronobacter spp., they may not be accessible to all laboratories. For example, real-time PCR methods and MALDI-TOF MS can be used to identify suspect isolates; however, these methods require expensive, specialized equipment and trained personnel that will prevent some laboratories from utilizing such techniques.

Several PCR probe-based identification methods have been proposed for the Cronobacter genus, which may provide an alternative to biochemical phenotyping or DNA sequence-based identification [22, 33–39]. These methods do not require any specialized equipment beyond a PCR thermocycler and gel electrophoresis capabilities, which would be expected in most, if not all, microbiology testing laboratories. Previous analysis comparing PCR probes with Cronobacter genome sequences showed the methods targeting the genes ompA and rpoB were most likely to result in accurate identification [20, 22, 38, 39]. The rpoB method requires a separate PCR primer pair and different amplification conditions for each of the seven Cronobacter species [38, 39]. In contrast, the ompA method only requires one primer pair and therefore is less laborious and has a shorter turnaround time [22]. Though the PCR method targeting ompA was developed to identify “E. sakazakii,” strains previously known by this name were reclassified as members of the Cronobacter genus [5, 6]. The analysis of Jackson et al. showed that the method targeting ompA was able to identify the type strains of all seven Cronobacter species, while excluding closely related strains, particularly those belonging to the Franconibacter and Siccibacter genera [20, 22].

For a thorough examination of the ompA PCR method, in silico analysis of full genome sequences was used. This method enabled the analysis of a large cohort of internationally derived strains that are not centrally available. To determine if this method could be used for identification of Cronobacter from PIF, the genomes of 187 Cronobacter and 36 non-Cronobacter strains were examined for the presence of the primer sequences targeting ompA. Table 3 shows the number of mismatches with each primer and the expected amplicon size for each species. The ESSR-R primer showed a higher number of mismatches than the ESSR-F primer for non-Cronobacter species and all F. helveticus strains showed 7 mismatches with the ESSR-R primer. Four of these strains were used in the previously reported laboratory evaluation of this method and none produced amplicons [20]. It was, therefore, concluded that strains with 7 or more mismatches to the ESSR-R primer would not produce amplicons with this method.

  

# mismatches

 

Species

n

ESSR-F^a

ESSR-R

Predicted amplicon size (bp)^b

  

(22 bp)

(19 bp)

 

Cronobacter sakazakii

101

0

1–2

468

Cronobacter malonaticus

45

0

2

468

Cronobacter dublinensis

14

0

2

469

Cronobacter turicensis

10

0

2

468

Cronobacter muytjensi

9

0

0

469

Cronobacter universalis

7

0

2

468

Cronobacter condimenti

1

0

2

469

Franconibacter pulveris

7

1

8

NA

Franconibacter helveticus

5

2

7

NA

Siccibacter turicensis

4

4

10

NA

Enterobacter asburiae

2

4–6

8–11

NA

Enterobacter hormaechei

2

6–7

8

NA

Enterobacter spp.

2

5–6

7

NA

Citrobacter amalonaticus

1

7

9

NA

Citrobacter freundii

1

6

12

NA

Citrobacter koseri

1

7

13

NA

Citrobacter rodentium

1

7

13

NA

Citrobacter youngae

1

6

13

NA

Enterobacter aerogenes

1

5

13

NA

Enterobacter cancerogenus

1

5

7

NA

Enterobacter cloacae

1

7

8

NA

Enterobacter ludwiggi

1

5

7

NA

Enterobacter massilensis

1

5

10

NA

Enterobacter mori

1

5

7

NA

Enterobacter sacchari

1

7

15

NA

Siccibacter colletis

1

4

10

NA

Yersinia regensburgii

1

5

7

NA

^aPrimer sequences described previously [22]

^bExpected amplicon size: 469 bp [22]

Based on the in silico analysis of the ompA gene, none of the non-Cronobacter species examined would be expected to produce amplicons with this method. On the other hand, all Cronobacter spp. were predicted to produce amplicons of approximately the expected size. Mohan-Nair and Venkitanarayanan reported an expected size of 469 bp, while the in silico analysis predicted amplicons of 468-469 bp [22]. This in silico analysis indicates that this method should be sufficient to identify all Cronobacter species, while excluding members of closely related genera. Though further identification would be needed to identify the isolates to the species level, this would not be necessary for PIF manufacturers to ensure exclusion of all members of the Cronobacter genus from their products [1, 27–29]. As this method is capable of identifying all Cronobacter spp., it could be used in place of the biochemical test panels for more accurate genus-level identification of suspect Cronobacter isolates from PIF.