Prevalence, characterization, and antibiotic susceptibility of Cronobacter spp. in a milk powder processing environment: The first reported case in Serbia

Abstract Cronobacter spp. are opportunistic foodborne pathogens that most often infect neonates and infants through contaminated powdered infant formula. No reports have been published in Serbia on the prevalence of Cronobacter spp. in powdered milk production environments. Consequently, this study aimed to determine the prevalence, molecular characterization, antimicrobial susceptibility, and biofilm‐forming ability of Cronobacter spp. isolated from a milk powder plant. Hundred samples were collected from the production facility. Fifteen Cronobacter sakazakii strains were isolated and identified, giving a contamination rate of 15%. Using multi‐locus sequence typing, the isolates were divided into five sequence types (STs). Cronobacter sakazakii ST4 (50%), ST1 (16.67%), and ST83 (16.67%) were the dominant STs isolated. A novel sequence type (ST759) was identified and registered in the Cronobacter MLST database. The results of the antibiotic susceptibility testing indicated that C. sakazakii strains were susceptible to piperacillin/tazobactam, ampicillin/sulbactam, and amoxicillin/clavulanate, especially to meropenem and cefotaxime. Most of the ST4 showed moderate‐to‐strong biofilm‐forming ability. The presence of clinically relevant isolates (ST4, ST1, ST83, and ST8) revealed that the production plant is likely a potential concern for public health. Finally, finding new sequence types like the one detected in this study (ST759) underlines evolving genetic changes in C. sakazakii.

Within the powdered milk plant, layers are forming on surfaces that are in constant contact with nutrients. This phenomenon can lead to accelerated biofilm production of the observed pathogen (Huang et al., 2020).
Cronobacter was previously reported to be found in PIF samples and the environment of the powdered milk production plant, with the dominance of Cronobacter sakazakii clonal complex 4 (Sonbol et al., 2013). In the powdered infant formula production process, the addition of heat-sensitive material, spray drying, fluidized beddrying, filling, and packing could be the possible links with C. sakazakii contamination (Fei et al., 2015;Sonbol et al., 2013). Therefore, the presence of Cronobacter sakazakii in powdered milk plants and commercial PIF needs to be monitored.
Additionally, to determine phylogenetic relatedness, sequence data from more than one gene can be used to reduce the number of possible doubts caused by genetic recombination or specific selection. Thanks to this, species identification and determination of intra-and intergeneric relationships are much more reliable than in 16s rDNA gene sequencing (Baldwin et al., 2009;Dauga, 2002).
The work of Baldwin et al. (2009) focused on a higher resolution analysis of Cronobacter using seven loci (atpD, fusA, glnS, gltB, gyrB, infB, and ppsA). Multi-locus sequence typing gives a clearer picture of the pathogen itself and potential mutations in the pathogen's genetic material in the form of a different nucleobase arrangement. By identifying Cronobacter in a production facility and finding possible mutations using the MLST analysis, a considerable contribution can be made to public health in the form of disease prevention itself.
The main objective of this study was to determine the prevalence of Cronobacter spp. in the milk powder production plant located in Serbia, to perform genetic diversity analysis by MLST and to explore the association between biofilm formation and antibiotic resistance among food and environmental Cronobacter spp. strains.
The primary hypothesis was that Cronobacter spp. could be present in milk powder plants in Serbia.

| Milk powder plant
The dairy factory investigated in this study produced a variety of dairy products including milk powder for sale. The product was partially agglomerated skim milk powder in 25-kg bags. The product was stored at ambient temperature, preferably lower than 30°C, with relative humidity not higher than 70%. The shelf life of the end product under these storage conditions would be not less than 9 months.
The factory with its 15,700-m 2 production area accommodating 40,000 L of milk as daily processing capacity has been operating ever since 1980s and has implemented the ISO 22000, HACCP, and IFS standards. HACCP plan included steps after pasteurized milk storage to product storage and considered health hazards. Potential microbiological hazards were identified in the following process steps: preheat treatment, concentration, balance tank and spray dryer operations, fluid bed dryer, pipelines, cyclones, and packaging.
Also, a program of environment control in powder manufacturing, packaging, and surrounding areas has been laid down due to environmental contamination with microorganisms through air and aerosols of powder.
Biological hazards in self-control plans encompassed testing against Salmonella spp., Listeria monocytogenes, Escherichia coli, Staphylococcus aureus, molds, and Enterobacteriaceae. There was no process hygiene criterion set for Cronobacter spp. Instead, production facility monitored the processing areas and equipment for Enterobacteriaceae as part of their sampling plan.
Facility records showed occasional presence of Enterobacteriaceae in dried milk, which was mostly attributed to incidental contamination via the large volume of air in drying and cooling operations.

| Sample collection
Powdered milk and environmental samples were collected during 2020 from a milk powder plant located in Serbia. Surface swabs (at least 1,000 cm 2 ) were taken from various locations with sterile moistened Sponge-Sticks (Romer Labs, Newark, USA). Samples were taken from the food contact surfaces and equipment 2 h after cleaning and sanitization. After sampling, the Sponge-Sticks soaked in neutralizing buffer (10 ml) were placed in sterile bags with the location identification label. Immediately after collection, samples were transported to the laboratory in refrigerated coolers and analyzed within the next 3 h.

| Isolation and identification of Cronobacter spp
The presence of Cronobacter spp. in powdered milk samples and en- method. The method consisted of nonselective pre-enrichment, enrichment in a selective medium, isolation, and identification of Cronobacter spp. on chromogenic agar, and confirmation of potential colonies. Nonselective pre-enrichment was performed by homogenizing a 25-g powdered milk sample with 225 ml of buffered peptone water (BPW; Oxoid). Each bag containing an environmental sample (sponge) and neutralizer was added 90 ml of BPW, and the content was thoroughly homogenized by mechanically massaging the sponge. Samples were further incubated at 34°C to 38°C for 18 ± 2 h. Subsequently, 0.1 ml of the BPW suspension was transferred to 10 ml of selective broth, based on lauryl sulfate tryptose broth with sodium chloride and vancomycin (Oxoid). The mixture of suspension and the selective broth was incubated at 41.5°C for 24 ± 2 h. After incubation of the broth mentioned above, the suspension was inoculated utilizing a 10-µL loop to the chromogenic Cronobacter isolation (CCI) agar surface. Chromogenic plates were incubated at 41.5°C for 24 ± 2 h. After incubation, chromogenic plates were examined for typical or suspectable Cronobacter spp.
In order to gain well-isolated colonies for biochemical characterization, selected colonies were streaked onto the surface of a nonselective tryptone soy agar (Oxoid). Biochemical tests included oxidase test, hydrolysis of 4-nitrophenylαd-glucopyranoside, llysine decarboxylase test, l-ornithine decarboxylase test, methyl red test, and Voges-Proskauer reaction. Simultaneously, acid production from d-arabitol, d-sorbitol, d-sucrose, and α-methyld-glucoside was tested.
Presumptive Cronobacter spp. isolates were stored at −80°C in brain heart infusion broth (Oxoid) with 15% glycerol for further confirmation by PCR assay.

| DNA extraction
Isolation of genomic DNA and preparation of the same for real-time polymerase chain reaction (RT-PCR) analysis and MLST were performed using a QIAquick PCR Purification Kit (Qiagen) and 1.5 ml of overnight culture grown in tryptic soy broth (TSB), according to the manufacturer's instructions.

| Confirmation of Cronobacter sakazakii by realtime PCR
The real-time PCR method was used to confirm the presence of C. sakazakii in powdered milk plants in the present study. The following species-specific primers of the cgcA gene described by Hu et al. (2016) were used: cgcA-forward (Agilent) was used for thermocycling and to measure changes in fluorescence. The PCR was initiated by denaturation at 95°C for 3 min, followed by 40 cycles of denaturation at 95°C for 15 s and annealing at 60°C for 60 s. Fluorescence was measured at 60°C negative controls were included, containing all the elements of the reaction mixture except the template. All samples were processed in triplicate. DNA extracted from the pure culture of C. sakazakii ATCC 29544 was used as a positive control.

| Multi-locus sequence typing (MLST) and sequence analysis
The MLST gene set was amplified using primers and PCR condi- was assigned by the database curator to each isolate with a unique allelic profile. In contrast, subsequent isolates with an identical allelic profile were classified into the same sequence type identifier. They considered isogenic strains because they did not show a noticeable difference at all seven loci. All alleles within the MLST scheme were located within the framework and served as references for analysis.
Linkage analysis was performed using the index of association (I A ). It was examined whether the alleles were randomly linked; that is, the mentioned linkage was in equilibrium, which would indicate a freely recombinant population. When the association is in equilibrium, or in other words, when there is a random association between alleles of different loci present, the association index is equal to zero

| Antimicrobial susceptibility testing
The Epsilometer test (Etest) method was chosen based on the conclusions of Ogata et al. (2014), who showed that the Etest was the best option to routinely perform susceptibility testing, especially

| Biofilm-forming ability
The biofilm-forming ability of Cronobacter sakazakii was tested by an indirect quantitative method, the so-called crystal violet method, using microtiter plates. In the performance of the crystal violet method, 96-well, flat-bottomed, polystyrene microplate In the next step, the microplate was dried inverted at room temperature. The crystal violet, which remained bound to the adherent bacteria, was dissolved by adding a decolorizer (a mixture of ethanol and acetone in a ratio of 80:20) in an amount of 250 µl.
The plate was left for 15 min at room temperature to achieve complete decomposition of the dye.
The results were read using a spectrophotometer at 595 nm (Eppendorf). The absorbance values used to characterize the biofilmforming ability of the tested isolates were obtained by reducing each absorbance by the mean absorbance value of pure broth (negative control). Isolates were classified into four categories according to Stepanović et al. (2007).

| Isolation, identification, and prevalence of Cronobacter spp
Cronobacter spp. were isolated from 15 out of 100 (15%) samples collected from a powdered milk plant. All isolates from the samples formed small convex blue-to-blue-green colonies on CCI agar plates.
These colonies were considered as Cronobacter spp. in this research.
The biochemical characterization tests of 15 Cronobacter isolates demonstrated the following reactions: negative oxidase test, a positive test for hydrolysis of 4-nitrophenylαd-glucopyranoside, a negative test for l-Lysine decarboxylase and l-Ornithine decarboxylase, acid from d-sucrose and a methylαd-glucoside, and no acid from darabitol and d-sorbitol. All 15 isolates were confirmed as C. sakazakii by the real-time PCR method.
The highest percentage of C. sakazakii isolates were found in July

| MLST analysis of Cronobacter sakazakii isolated from the powdered milk plant
Using 7-locus MLST, the 15 C. sakazakii strains were analyzed. The analysis on the three of the tested strains (2 strains from dry powder from baghouse, one strain from environmental swabs from evaporator premises) was found difficult due to their weak signal. Hence, they were dropped from the study. After the MLST analysis, the 12 C. sakazakii strains were clustered into five sequence types, shown in

| Antimicrobial susceptibility
The minimal inhibitory concentration (MIC) of the 15 Cronobacter strains is shown in Table 2. Antibiotic susceptibility profiling indicated that all 15 isolates were sensitive to all antimicrobial agents investigated. No drug or multiple drug resistance was detected.

| Biofilm-forming ability
The microtiter plate assay detected the biofilm-forming ability among the 15 C. sakazakii isolates. The cutoff OD (ODc) was defined as three standard deviations (SD) above the mean OD (optical density) of the negative controls. Based on the ODc, the C. sakazakii isolates were classified into four categories: OD of the test isolate ≤ ODc (0.212)-no biofilm production, ODc (0.212) ≤ OD of the test isolate ≤ (2 × ODc (0.424))-weak biofilm production, (2 × ODc (0.424)) < OD of test isolate ≤ (4 × ODc (0.848))-moderate biofilm production, and OD of test isolate > (4 × ODc (0.848))-strong biofilm production. The results indicated that all 15 of the tested isolates were capable of producing biofilm on polystyrene microplates.
According to the results, one of the two isolates with strong biofilmforming ability belonged to ST4 (OD = 1.243). The other one with the strong biofilm-forming ability (OD = 1.299) was given a low signal during MLST analysis, and it was left out of sequencing. Five out of 15 tested isolates manifested moderate biofilm-forming ability. Four out of 5 isolates with moderate biofilm-forming ability belonged to ST4, and the other was identified as ST1. The remaining eight isolates were classified as weak biofilm producers. The distribution of the sequence types with weak biofilm-forming ability was as follows: ST83 (25%, 2/8), ST4 (12.5%, 1/8), ST1 (12.5, 1/8), ST8 (12.5, 1/8), and ST759 (12.5%, 1/8), including two isolates, which were not classified into sequence types, for the same reason as the previous one.

| D ISCUSS I ON
In Serbia, Jošić et al. (2017) analyzed 360 commercial powdered infant formula samples, but C. sakazakii was not detected in these samples. In this research, Cronobacter sakazakii was isolated from a powdered milk plant, and the prevalence of the pathogen was 15%.
Sequence type 4 was isolated from environmental swabs collected in evaporator premises, swabs collected from the spray dryer chamber, and dry powder from the baghouse. It is interesting in this connection for us to note that  found that in cases of neonatal infection, almost half (48.78%, 20 out of 41) of the isolated strains belonged to ST4, followed by ST8 The dominant sequence type in commercial PIF and infants exposed to PIF was ST8, who had diarrheal symptoms. The appearance of ST4 and ST8 in baghouse can be a serious concern because meningitis associated with diarrhea can lead to fatal outcomes in preterm infants and neonates. The original study by Sonbol et al. (2013) reported that 21 out of 72 C. sakazakii strains were in the clinically significant ST4 clonal complex. They described that ST4 and ST1 are not closely related, and ST1 was primarily isolated from powdered milk plants or powdered infant formula production environments. In this research, ST1 (16.67%, 2 out of 12) was only isolated from the spray drier chamber, but as described by Sonbol et al. (2013), in most cases, ST1 appears together with ST4. Chase et al. (2017)  values for meropenem and cefotaxime. These findings might occur due to the less frequent use of these antibiotics in dairy cow therapy compared to the use of β-lactams and β-lactamase inhibitors (Darko et al., 2017;Zhou et al., 2015). Another good sign could be that MIC values for meropenem and cefotaxime vary in a narrow range, in the case of ST4, ST1, and ST83, which were described as the most pathogenic sequence types. The novel sequence type ST759 showed a high susceptibility rate toward all the tested antibiotics. Li et al. (2017) reported that 13 (32.5%) of the isolated Cronobacter strains from spices and cereals were classified as weak biofilm producers, one (2.5%) was a moderate biofilm producer, and one (2.5%) was strong biofilm producer. These authors stated that there was no correlation between biofilm formation and sequence types. In this study, 2 (13.33%) isolates were marked as strong biofilm producers, 5 (33.33%) strains were recognized as moderate biofilm producers, and 8 (53.33%) isolates were categorized as weak biofilm producers. One of the ST4 isolates (M72) showed a strong biofilm-forming ability and the highest MIC values for all the tested antibiotics. This finding could mean that the higher biofilm-forming ability of the bacteria plausibly correlates with increased antimicrobial resistance (Neopane et al., 2018). Most of the ST4 showed a moderate-to-strong biofilm-forming ability, especially in dry powder collected from the baghouse, indicating that the mentioned point is critical in the specific production plant regarding public health risk.
The results of Du et al. (2018) are in correlation with the results of this research. The novel ST759 showed a weak biofilm-producing ability, which, in light of high sensitivity to all the tested antibiotics, may be a favorable sign from the aspect of public health.

| CON CLUS IONS
The present research demonstrated that the prevalence of C. sakazakii in milk powder facility is 15%. Real-time PCR analysis revealed that C. sakazakii was the only species recovered from environmental swabs and from the dry powder in baghouse. MLST analysis showed a moderate-to-high genetic diversity of isolated C. sakazakii strains.
The presence of isolates of clinical relevance, such as ST4, ST1, ST83, and ST8, indicated that the production plant could likely be the potential source of contamination for PIF ingredients. It is encouraging that all isolates were susceptible to antibiotics usually prescribed for children, especially to meropenem and cefotaxime. One of the ST4 isolates (M72) showed a strong biofilm-forming ability and the high- est MIC values for all tested antibiotics. This finding could indicate association between higher biofilm-forming ability and increased antimicrobial resistance. Most of the ST4 showed a moderate-tostrong biofilm-forming ability, especially in dry powder collected from the baghouse, which indicates this point is critical in evaluation of existing HACCP self-control plans.
In this study, a novel sequence type with a unique gene locus structure was isolated (ST759), which requires further investigation to reveal its potential pathogenic potential. This finding underlines evolving genetic changes in C. sakazakii. Whether this is just a static episode or continuous evolution governed by the paragenetic factors present in the milk powder production environment or not remains to be explored yet.

ACK N OWLED G M ENTS
This study was supported by the Ministry of Education, Science and

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.