Three different production levels were investigated for the presence of bacteria in food ice. The ice samples were collected from domestic freezers (DF) for home-level production, from stock boxes (SB) for self-production performed with ice machines in bars and pubs, and from sales packages (SP) for the productions performed by industrial ice facilities. Five ice productions per typology (DF1-DF5, SB1-SB5, SP1-SP5) were sampled in duplicate at a 2-month interval for a total of 60 ice samples, which were transferred into sterile stomacher bags and transported in a thermal insulated box. All producers were located within the Palermo province (Sicily, Italy) and transformed the water supplied by the municipal source.
Approximately 500 g of each ice sample, corresponding to 25–50 ice cubes (considering that, depending on the shape, the individual ice cube weight ranged from 10 g to 20 g) were thawed in a 1-L sterile Dhuram bottle at room temperature, and a volume of 100 mL per each microbial group was subjected to membrane filtration. The microbiological investigation included the following bacterial groups: total mesophilic bacteria (TMB) on plate count agar, incubated at 30 °C for 72 h; total psychrotrophic bacteria (TPB) on plate count agar, incubated at 7 °C for 7 days; and pseudomonads on Pseudomonas agar base supplemented with 10 mg/mL cetrimide fucidin, incubated at 25 °C for 48 h. When the number of colonies exceeded 186, corresponding to the number of squares present on the membrane grid, aliquots of 1 mL from each sample were inoculated directly into agar media. All media and supplements were purchased from Oxoid (Milan, Italy).
Isolation and phenotypic characterisation of ice bacteria
Colonies of bacteria developed on agar media were randomly picked up for each morphology (color, margin, surface and elevation). Almost five identical colonies (or fewer if five were not available or showed confluent growth) were collected from the agar plates. All bacterial cultures were subjected to the purification phase, verified by an optical microscope, after several consecutive subcultures on the same media and under the same growth conditions as those used for plate counts.
A preliminary phenotypic characterisation was applied on the pure bacterial isolates propagated in nutrient broth (Oxoid). The KOH test (Gregersen 1978) was performed to determine the type of cell wall. The catalase test was carried out by addition of H2O2 (5%, w/v) to the colonies. Cell morphology and motility were evaluated by microscopic inspection. The formation of spores was investigated as follows: cell suspensions of pure cultures were treated at 85 °C for 15 min and, subsequently, inoculated in the same media used for isolation and purification and incubated under optimal growth conditions.
Bacterial strain typing and identification
DNA was extracted from cells by lysis, applying the Instagene Matrix kit (Bio-Rad, Hercules, CA) following the manufacturer’s instructions, after overnight growth in nutrient broth at 30 °C. DNA concentration was adjusted to 25 ng/μL before being used as a template for PCR.
Differentiation of bacteria at strain level was performed by random amplification of polymorphic DNA (RAPD)-PCR using three primers (M13, AB111, and AB106), while identification at species level was performed by 16S rRNA gene sequencing with the primer pair fD1/rD1. The methodology applied for the genetic characterisation was previously described by Settanni et al. (2012). The sequences of the 16S rRNA gene fragment were compared with those available in the EzTaxon-e (http://eztaxon-e.ezbiocloud.net/) database, which compares a given sequence to those of type strains only.
The survival test was performed with the bacteria found at the highest levels for the most numerous groups. These microorganisms were first cultivated under optimal growth conditions and then washed twice in Ringer’s solution (Sigma-Aldrich, Milan, Italy), after centrifugation at 5000 g for 5 min. The cells were suspended in autoclaved (121 °C for 20 min) still mineral water (Terme di Geraci Siculo S.p.a., Geraci Siculo, Italy) at the highest concentrations found in the samples analyzed, and the contaminated water was used to produce ice cubes (at −32 °C) containing a single bacterial species using autoclaved stainless steel ice cube trays. After plate counts, the levels of the artificial contamination of the ice cube was ca. 3.00 × 102 CFU/100 mL t.i. for A. lwoffii ICE100, ca. 9.00 × 103 CFU/100 mL t.i. for B. cereus ICE170, ca. 4.00 × 102 CFU/100 mL t.i. for P. putida ICE224 and ca. 4.00 × 103 CFU/100 mL t.i. for S. haemolyticus ICE182.
The concentrations of microorganisms were followed in beverages and drinks characterised by different pH, alcohol volume, sugar content and presence or absence of CO2. To this purpose, the following systems were used: vodka (Keglevich, Plzen, Czech Republic), as an alcoholic (38% vol) neutral (pH 6.0) beverage; whisky (Jack Daniel’s, Lynchburg, Tennessee), as an alcoholic (40% vol) acidic (pH 4.2) beverage; Martini (Martini and Rossi, SpA, Pessione, Italy) as a moderate alcoholic (14.4%, pH 3.8) beverage; peach tea (Conad, Fruttagel SCpA, Alfonsine, Italy) as a sugary (9.7% w/v, pH 3.3) drink; tonic water (Kinley, Sesto S. Giovanni, Italy) and coke (Coca Cola Italia Srl, Sesto S. Giovanni, Italy) as sugary (9.4 and 10.6 w/v, respectively) acidic (pH 2.8 and 2.7, respectively) carbonated drinks. The simulation of drink contamination occurred through addition of artificially inoculated ice cubes (corresponding to 60 mL) to 100 mL of each drink in 200 mL volume sterile cups (Anicrin, Scorzé, Italy). All contaminated systems were left at room temperature until complete ice thawing and then analyzed, depending on the inoculation level, by membrane filtration or plate count. Ringer’s solution (Oxoid) was used as control. Three independent experiments were carried out on three different days.
Microbiological data were subjected to one-way analysis of variance (ANOVA). Pair comparison of treatment means was achieved by Tukey’s procedure at P < 0.05. Differences among the production levels (domestic, restaurant and industrial) were evaluated using the generalised linear model (GLM) procedure.