Following two O121 Shiga toxin–producing Escherichia coli (STEC) outbreaks linked to wheat flour, this study was conducted to gain baseline information on the occurrence of bacterial pathogens and levels of indicator organisms in wheat flour in Canada. A total of 347 prepackaged wheat flour samples were analyzed for Salmonella species, STEC, Listeria monocytogenes, aerobic colony count (ACC), total coliforms, and Escherichia coli. Salmonella spp. and O157 STEC were not detected in any of the samples. L. monocytogenes was identified in two samples (0.6%) at levels below the limit of detection (<0.7 log CFU/g). Non-O157 STEC were isolated from six samples (1.7%) and were characterized for the presence of STEC virulence genes: stx1, stx2, and their subtypes, eae, hlyA, and aggR. One O103:H25 STEC isolate carried virulence genes (stx1a+eae) that are known to be capable of causing diarrhea and/or bloody diarrhea in humans. Of the five remaining non-O157 STEC isolates, four carried single stx2a or stx2c genes and were considered to have the potential of causing diarrhea. The remaining non-O157 STEC isolate (stx2), while not a priority non-O157 STEC, was not available for sequencing; thus, its potential to cause illness is unknown. ACC, total coliforms, and E. coli were detected (≥0.48 log CFU/g) in 98.8, 72.6, and 0.6% of the flour samples. The mean counts of ACC were greater in whole wheat flour compared with the other flour types tested (P < 0.001). The results of this study suggest that the occurrence of O157 STEC and Salmonella is low but that the occurrence of non-O157 STEC in wheat flour with the potential to cause human illness of diarrhea is relatively common. Therefore, the consumption of raw flour could increase the likelihood of STEC infections. Further research is merited for potential risk mitigation strategies within the food production system and with consumers.
O157 STEC and Salmonella were not found.
Non-O157 STEC was found in six samples (1.7%), of which five carried the stx gene only.
One non-O157 STEC (O103:H25) isolate carried virulence genes stx1a and eae.
L. monocytogenes was identified in two samples (0.6%) below the detectable counts.
ACC, coliforms, and E. coli were detected in 98.8, 72.6, and 0.6% of samples.
In recent years, foodborne illness outbreaks linked to wheat flour have been reported in many countries; in these, bacterial pathogens like Shiga toxin–producing Escherichia coli (STEC) and Salmonella species were identified as the hazard of concern. Wheat flour has been implicated in two outbreaks of O121 STEC infections in Canada between 2016 and 2017 (2, 36). Similarly, outbreaks of O26 STEC infections associated with wheat flour occurred in the United States in 2018 to 2019 (17) and 2015 to 2016 (18). Outbreaks of O157 STEC infections in the United States associated with cookie dough (pasteurized egg) in 2009 (16, 38) and dough mix in early 2016 (22) were reported. Wheat flour was also involved in salmonellosis outbreaks in the United States in 2009 (35). These wheat flour–associated outbreaks suggest that STEC and Salmonella spp. are the primary bacterial pathogens of concern in this commodity. Among the STECs, O157 STEC and some non-O157 STEC carrying virulence genes stx2a+eae or aggR can cause severe human illnesses (20). Although Listeria monocytogenes has not been identified as a source of flour-associated outbreaks, dairy and egg ingredients in flour-based raw cookie doughs and batters have the potential to support the growth of L. monocytogenes. L. monocytogenes is, therefore, a pathogen of concern in flour-based raw cookie dough and batter products.
Wheat flour is an agricultural product that can be contaminated with bacterial pathogens during primary production and the milling process (34, 37). The low moisture content (aw < 0.6) of wheat flour does not support the growth of bacterial pathogens; however, some bacterial pathogens can survive in this low-moisture environment (21, 25) and can multiply in higher-moisture raw flour–based products such as batter or cookie dough (43). Ready-to-eat (RTE) wheat flour–based products are usually subject to thermal treatments such as baking, steaming, and frying, all of which would contribute to the inactivation of bacterial pathogens before consumption. However, food safety and epidemiological investigations of recent wheat flour–associated outbreaks found that consumers often taste or eat raw cookie dough and batter during the RTE product preparation process (16, 43). Therefore, bacterial pathogens associated with raw wheat flour have been identified as a potential risk for foodborne illness outbreaks in recent years due mostly to consumer behaviors associated with the preparation of flour-based foods.
To gain baseline information on the occurrence of bacterial pathogens and the levels of indicator organisms in wheat flour on the Canadian market, the Canadian Food Inspection Agency conducted a targeted survey of bacterial pathogens (STEC, Salmonella spp., L. monocytogenes) and organisms that are indicators of overall quality and potential contamination (aerobic colony count [ACC], total coliforms, E. coli) in wheat flour samples collected from retail locations across Canada. In this study, STEC isolates were characterized by whole genome sequencing (WGS), and the levels of indicator organisms in wheat flour were analyzed.
MATERIALS AND METHODS
A total of 347 prepackaged wheat flour samples were collected between 1 April 2018 and 31 March 2019 from retail stores in 11 major cities across Canada. The percentage of samples collected from each city was approximately proportional to the population of the province in which they were located, in relation to the total population of Canada. Cities from which samples were collected, and the percentage collected from each, were as follows: Halifax (4%), Moncton (2%), Montreal (19%), Quebec City (4%), Toronto (32%), Ottawa-Gatineau (7%), Vancouver (11%), Victoria (2%), Calgary (12%), Saskatoon (3%), and Winnipeg (4%). A sample (n = 1) consisted of one package (≥500 g per package) or multiple consumer-sized packages (<500 g per package) to reach a total weight of at least 500 g from a single lot. Samples were randomly collected at retail from among the available product types so as to be as representative as possible of the market share of the product types collected.
All samples were analyzed in ISO/IEC 17025 accredited Canadian Food Inspection Agency laboratories. Samples were analyzed for bacterial pathogens (STEC, Salmonella spp., and L. monocytogenes) as well as indicator organisms (ACC, total coliforms, and E. coli) using methods published in Health Canada's Compendium of Analytical Methods for the Microbiological Analysis of Foods (28) (appendix I). All media used in microbial analysis are specified in the Health Canada's Compendium of Analytical Methods (28) and were prepared as per suppliers' instructions. Analyses for bacterial pathogens (STEC, Salmonella, and L. monocytogenes) were conducted on 125 g of sample. Enumeration of L. monocytogenes was performed for L. monocytogenes–positive samples using a 10-g analytical unit. Analysis of ACC, total coliforms, and E. coli was performed using Petrifilm enumeration methods (3M Canada, London, Ontario, Canada) that required a 10-g analytical unit.
Enumeration of indicator organisms (ACC, total coliforms, and E. coli)
Total ACCs were determined by preparing succeeding serial dilutions (1 in 10) of the flour sample in 0.1% peptone water, followed by plating onto 3M Petrifilm aerobic count plates as described in method MFHPB-33. Total coliforms and E. coli counts were determined by preparing succeeding serial dilutions (1 in 10) of the flour sample in 0.1% peptone water, followed by plating onto 3M Petrifilm E. coli count plates as described in method MFHPB-34.
Salmonella spp. isolation and identification
Briefly, following enrichment, samples were first screened for Salmonella spp. using a BAX system assay (MFLP-29), a rapid PCR-based screening method that targets Salmonella spp., or using a VIDAS UP Salmonella method, an automated enzyme immunoassay based on the enzyme-linked fluorescent assay (MFLP-40). Any presumptive-positive enrichments were further analyzed by selective plating, isolation, and identification using cultural confirmation method MFHPB-20.
L. monocytogenes isolation, identification, and enumeration
Briefly, following enrichment, samples were first screened for L. monocytogenes using a BAX system assay (MFLP-28), a rapid PCR-based screening method that targets L. monocytogenes. Presumptive-positive enrichments were further analyzed by selective plating, isolation, and identification using cultural confirmation method MFHPB-30. Samples that were confirmed positive for L. monocytogenes by MFHPB-30 were further analyzed by enumeration method MFLP-74. Briefly, a new 10-g portion was taken from the positive flour sample, diluted in four volumes of 0.1% peptone water, and plated across the surface of three plates of two different types of media, as described in the method. If no L. monocytogenes was confirmed by enumeration, counts were reported as <0.7 log CFU/g, which is the limit of detection of the enumeration method.
STEC enrichment, identification, isolation, and characterization
Samples were analyzed for STEC using method MFLP-52 with an expanded multiplex PCR and cloth-based hybridization array system (7, 8) and a modified enrichment step involving a presoak prior to enrichment (6). Specifically, samples were subjected to an additional resuscitation step by first preenriching 125 g of each sample in buffered peptone water in a stomacher bag using a ratio of 1 mL of buffered peptone water to 1 g of sample followed by incubation overnight at 35°C. Following this resuscitation step, a suitable volume of modified tryptic soy broth (mTSB; Oxoid CM0989, Oxoid Ltd., Basingstoke, Hampshire, England) was added to 125 g of the flour and buffered peptone water mixture to bring the ratio of sample to mTSB to 1:9. The sample was enriched in mTSB broth with delayed addition of the antimicrobials of vancomycin and cefsulodin (24) as described in MFLP-52, after which method MFLP-52 was followed as published (8). Briefly, the enrichment broth was screened for the genes stx1 and stx2 by PCR. stx-positive enrichment broth cultures were subsequently plated onto selective agar, and colonies were screened batchwise by the stx PCR to identify stx-positive isolates. The isolates were subsequently confirmed by an expanded multiplex PCR and cloth-based hybridization array system (7, 9), which simultaneously identifies the presence of virulence genes stx1, stx2, eae, hlyA, and gene targets for O157 and top six priority non-O157 STECs (O26, O103, O111, O145, O45, and O121).
Genomic characterization of STEC isolates
All STEC isolates derived from positive samples were subject to characterization by WGS to determine serotype and assess the presence of the selected virulence genes (stx subtypes, eae, hlyA, aggR) (14, 32). Bacterial isolates were cultured in brain heart infusion broth (Oxoid Ltd.) for 3 to 6 h at 37°C. Genomic DNA was extracted using the Maxwell 16 Cell SEV DNA purification kit (Promega, Madison, WI) and quantified using the Quant-it High-Sensitivity DNA assay kit (Life Technologies Inc., Burlington, Ontario, Canada). Sequencing libraries were constructed from 1 ng of gDNA using the Nextera XT DNA sample preparation kit and the Nextera XT index kit (Illumina, Inc., San Diego, CA) and sequenced on the MiSeq system using a v3 reagent kit (Illumina, Inc.). Raw sequencing read quality was assessed with FastQC version 0.11.8 (1). Quality trimming was performed with BBDuk from BBTools version 38.22 (12) with a trim quality of 10 and removal of reads below 50 bp in length. Error correction was performed using tadpole version 8.22 (12) in “correct” mode with default parameters. Sequences were checked for contamination using ConFindr 0.5.0 with default parameters (33). Contigs were assembled from the trimmed and error-corrected reads using SKESA version 2.3.0 with the vector percent argument disabled (40). The presence of genes encoding Shiga toxin 1 (stx1) and 2 (stx2), intimin (eae), enterohemorrhagic E. coli hemolysin (hlyA), and O-serogroup markers within the assembled genomes was determined based on detection of e-probe sequences (7, 9, 32). Shiga toxin subtypes were detected from raw reads using the V-typer tool (14). Serotypes were identified using Serotype Finder version 1.1 (https://cge.cbs.dtu.dk/services/SerotypeFinder/) with default parameters (85% identity, 60% minimum length) (29).
Nucleotide sequencing access number
Raw data have been deposited at DDBJ/EMBL/GenBank under BioProject PRJNA454819 and PRJNA435747. Accession numbers are listed in Supplemental Table S1.
Wilson confidence interval (two-sided 95% CI [lower limit, upper limit]) was used for the calculation of estimated prevalence of bacterial pathogens. The counts of ACC and total coliforms equal to and higher than the limit of detection (0.48 log CFU/g) were transformed to log CFU per gram, and the mean of the log-transformed values was calculated. The 95% CI interval of the mean was calculated by subtraction or addition of 1.96 × standard error of the mean (SEM) (95% CI [(mean − 1.96SEM), (mean + 1.96SEM)]). The normality of the log-transformed data (>0.48 log CFU/g) was assessed using a qnorm plot in Stata 15/SE. The log-transformed mean counts of ACC and coliform were compared between product types using two-sample t tests with unequal variances to make pairwise comparisons (42).
RESULTS AND DISCUSSION
General product information
Of the 347 wheat flour samples collected, 229 (66%) were domestic, 41 (11.8%) were imported from five countries, and 77 (22.2%) were of unknown origin. In terms of production practices, 74 (21.3%) were organically produced and 273 (78.7%) were conventionally produced. The flour samples consisted of all-purpose (52.2%), whole wheat (33.7%), and other pastry (14.1%) flour types. The flour samples consisted of 36 different brands: 210 (60.5%) from five major brands and 137 (39.5%) from the remaining 31 brands.
ACC, total coliforms, and E. coli in wheat flour
In all flour samples analyzed, the ACC ranged from below the limit of detection (<0.48 log CFU/g) to 5.7 log CFU/g. Almost all (343 of 347, 98.8%) of the samples had counts ≥0.48 log CFU/g, of which 7 (2%) had ACC counts ≥5 log CFU/g. The largest proportion of all flour samples (mode) was found to have ACC counts in the range of ≥3 to <4 log CFU/g (Fig. 1A). The total coliform counts ranged from below the limit of detection (<0.48 log CFU/g) to 3.96 log CFU/g. A large portion (252 of 347, 72.6%) of the samples had counts ≥0.48 log CFU/g; of these, 12 (3.5%) had total coliform counts ≥3 log CFU/g. Contrary to what was observed with ACC, total coliforms were reported at levels below the limit of detection (<0.48 log CFU/g) in 95 (27.4%) of the 347 samples. Therefore, samples were distributed throughout the total coliform ranges of below the limit of detection, ≥1 to <2, and ≥2 to <3 log CFU/g (Fig. 1B). For samples with counts ≥0.48 log CFU/g, log transformation improved the fit of the distribution when plotted using <qnorm> in Stata, by visual assessment of the curves for both ACC and total coliforms. Significantly different from the ACC and total coliforms, E. coli was found to be below the limit of detection (<0.48 log CFU/g) in most (345 of 347, 99.4%) of the flour samples (Fig. 1C), and in only two samples (0.6%) was it detected at 0.70 and 0.88 log CFU/g.
Of the 347 flour samples tested (Table 1), 343 had detectable levels of ACC (≥0.48 log CFU/g), with a mean ACC of 3.45 log CFU/g (95% CI [3.37, 3.53]); 252 samples had detectable levels of total coliforms (≥0.48 log CFU/g), with a mean total coliform count of 1.93 (95% CI [1.84, 2.02]) log CFU/g. The flour samples were grouped in two ways: by production practice (organic or conventional) and by product type (all-purpose, whole wheat, or other pastry). Comparisons were made for the mean of ACC and total coliform counts of the following flour product types: (i) organic and conventional, (ii) whole wheat and all-purpose, all-purpose and other pastry, and other pastry and whole wheat (Table 1). Of the different flour product types, the mean ACC was greater in organic products compared with conventional products. The mean ACC was greater in whole wheat flour compared with all-purpose and other pastry flours. The differences observed between the mean ACC in the compared flour types were statistically significant (P < 0.001). The mean total coliform count was greater in organic samples compared with conventional samples, and the difference was statistically significant (P < 0.05). However, the mean total coliform counts did not statistically differ (P > 0.05) between the compared flour product pairs of whole wheat flour and all-purpose flour, and other pastry flour.
Flour milling is a mechanical process that removes the outer layer of the wheat kernels and grinds the interior endosperm into flour. The removal of the outer grain layers (bran and germ) during the milling process results in the reduction, but not elimination, of bacteria (3). This study suggested that, in general, counts of ACC and total coliforms in wheat flour decreased by 1 log CFU/g after the milling process. However, E. coli was introduced at low levels during the conditioning process, in which wheat is treated with water prior to grinding; the equipment used to handle the wheat being the likely source of contamination (3). Therefore, the quality of the incoming wheat and the milling process can influence the quality and safety of final flour products. In this study, we selected three quality indicators (ACC, total coliforms, and E. coli) to explore their potential as indicators of contamination with bacterial pathogens in wheat flour.
ACC, total coliforms, and E. coli were detected (≥0.48 log CFU/g) in 343 (98.8%), 252 (72.6%), and 2 (0.6%) of the 347 wheat flour samples, respectively. These results are similar to those reported in other flour studies conducted in Australia (3) and the United States (41): ACC were commonly detected in wheat flour products, total coliforms were detected (above the limit of detection) in 82% of the wheat flour samples (3), and E. coli was at very low levels (3, 23, 41). Wheat flour is a raw agricultural product; thus, the presence of ACC and total coliforms in this commodity has little public health significance. These microorganisms were discussed as potential quality indicators in the Australia flour studies (3, 41). On the other hand, the U.S. flour study mentioned that there was little value in total coliforms as a quality indicator for processed wheat flour but suggested that the monitoring guideline of ACC >5 log CFU/g be used as a warning signal of increased microbial counts in the milling system (41). E. coli was also observed as having value in monitoring steps of the milling process because the conditioning process resulted in increased counts in the final wheat flour product compared with the premilled wheat, due to the introduction of E. coli from contaminated equipment used to handle wheat during the conditioning process (3).
The mean counts of ACC and total coliform log CFU per gram (Petrifilm method) obtained from this study are similar to or lower than those reported in the United States (41). The U.S. study conducted by the North American Millers' Association found that North American wheat flour produced between 2003 and 2005 had a mean ACC of 3.79 ± 0.70 log CFU/g (n = 6,598) and mean total coliforms of 2.65 ± 0.78 log CFU/g (n = 2,467) (Petrifilm method). Our study also found that the mean ACC was greater in whole wheat flour than in all-purpose flour (P < 0.001), which is expected because whole wheat flour is ground from whole grain with the bran contents (outer layer of wheat). The Australian study found that the counts of ACC and total coliforms were 1 log CFU/g higher in the outer layer (bran and germ) than in the flour (3). The U.S. study also found that the whole wheat flour had higher mean counts of ACC (4.41 ± 1.15 log CFU/g) and total coliforms (3.64 ± 0.62 log CFU/g) (n = 435) compared with all wheat flour (3.79 ± 0.70, 2.65 ± 0.78, n = 6,598) (41). In this study, the mean ACC and total coliform counts were greater in organic versus conventional samples, with a statistically significant difference observed at P < 0.001 and P < 0.05, respectively. Organic wheat is expected to have different microbial loads compared with conventional wheat due to differences in production practices (39). There is little literature on the counts of ACC and total coliforms in organic wheat flour versus conventional wheat flour types. The organic flour samples in our study consisted of whole wheat (70.3%), all-purpose (25.7%), and other pastry (4.1%) flour types, whereas the conventional flour samples consisted of all-purpose (58.7%), whole wheat (24.2%), and other pastry (17.1%) flour types. An attempt was made to analyze the ACC and total coliform results by production practice (organic, conventional) and by product types (whole wheat, all-purpose); however, the number of samples by product type within a production practice category were too small to allow for statistical inference. The results did, however, show that within both the organic and conventional categories, the mean of ACC was higher in whole wheat flours than all-purpose flours (data not shown). Different from ACC, the mean total coliform counts were higher in whole wheat flours compared with all-purpose flours only in the organic category. In addition, the mean total coliform counts were higher in organic whole wheat compared with conventional whole wheat flour (data not shown). These results suggest that organic production practices may have impacts on total coliforms in whole wheat flour compared with conventional production practices. Further studies specifically designed to investigate organic and conventional production practices may be able to address this issue.
Bacterial pathogens in wheat flour samples
Bacterial pathogens Salmonella spp. and O157 STEC were not detected in any of the 347 samples (0%, 95% CI [0, 1.1]). Non-O157 STEC were isolated from 6 (1.7%) of the 347 samples (95% CI [0.8, 3.7]). L. monocytogenes was detected in 2 (0.6%) of the 347 samples (95% CI [0, 1.1]).
L. monocytogenes in wheat flour samples
In this study, L. monocytogenes was isolated from 2 of the 347 wheat flour samples, and the contamination levels were below the limit of detection of the quantitative method (0.7 log CFU/g) in both samples (Table 2). The two L. monocytogenes–positive samples were both conventional all-purpose flours. The ACC levels detected in these two L. monocytogenes–positive samples were ≥3.4 log CFU/g, which is higher than the mean ACC of all-purpose flours (3.32 log CFU/g, 95% CI [3.24, 3.40]). The total coliform count was below the limit of detection in one L. monocytogenes–positive sample and was low (<1 log CFU/g) in the other positive sample. E. coli was not detected (<0.48 log CFU/g) in either of the L. monocytogenes–positive samples. The two samples were found to be from the third (brand C) and seventh (brand J) most frequently sampled brands from among a total of 36 brands sampled. The results suggest that L. monocytogenes can survive in the low-moisture environment of flour. L. monocytogenes is widely distributed in the environment and has been isolated from a wide variety of agricultural food products. Flour is a low-moisture agricultural product in which L. monocytogenes may survive for an extended period of time. Raw flour–based products, such as batter and baking dough, may support the growth of L. monocytogenes due to their higher moisture content. Currently, in Canada there is no policy or standard for L. monocytogenes in flour and the L. monocytogenes policy applies to RTE foods (27). Because wheat flour is not considered a RTE food, the public health significance of finding L. monocytogenes in this commodity remains to be assessed, and any potential health risk would be further mitigated by a cooking step.
Non-O157 STEC in wheat flour samples
The six non-O157 STEC isolates derived from wheat flour samples were serotyped and further characterized by WGS for the presence of Shiga toxin genes stx1, stx2, and their subtypes, the intimin gene eae (30), enterohemorrhagic E. coli hemolysin gene hlyA (4), and aggR, a transcriptional regulator that controls several genes, including aggregative adherence fimbriae adhesins (5) (Table 3). One non-O157 STEC isolate (O103:H25) possessed the stx1a and eae virulence genes. This gene profile (stx1a+eae) is categorized as being a low risk (level 4) for potential human illnesses of diarrhea and/or bloody diarrhea (20). Four of the five remaining non-O157 STEC isolates, including O8:H19 (stx2a), O159:H19 (stx2a), Ount:H19 (stx2a), and Ount:H2 (stx2c), each carried a single subtype of stx2 (not stx2d) gene and were categorized as being the lowest risk (level 5) for potential human illnesses of diarrhea (20). The other STEC isolate (O type not identified, stx2) was not one of the top priority STECs (O157, O26, O103, O111, O145, O45, and O121) (7); because it was unavailable for sequencing, a risk level could not be assigned.
The STEC-positive samples were all found to have detectable levels of ACC and coliforms (≥0.48 log CFU/g), but none were found to have high levels (ACC >5 log CFU/g, coliforms >3 log CFU/g). E. coli was not detected in any of the STEC-positive samples (Table 3). All of the non-O157 STEC–positive samples were found to be conventional all-purpose flours. Four of the STEC-positive samples were found to be brand A, which was the brand most frequently sampled in this study (Table 3); however, all four samples were from different production lots bearing different best-before dates. The other two STEC-positive samples were found to be brand B, which was the second most frequently sampled brand in this study (Table 3); but both samples were from different production lots bearing different best-before dates. Therefore, all six STEC-positive samples were from different production lots from producers of brand A or brand B.
A limited number of STEC-positive samples were identified in this study, which precludes us from identifying a correlation between the presence of bacterial pathogens and the levels of indicator organisms. A study has suggested that the levels of STEC in contaminated flour samples were below 1 MPN/100 g (23). Another study found that isolation of STEC was independent of the presence of E. coli in flour samples (34). Therefore, it is not surprising that non-O157 STEC–positive samples had E. coli below the limit of detection (<0.48 log CFU/g). Of the non-O157 STEC identified from this study, O103 STEC (stx1a+eae) is one of the top six priority non-O157 STEC in Canada (15) and the United States (11, 26). The virulence gene profile of stx1a+eae was present in 168 (95.5%) of 176 O103 STEC clinical isolates in the European Union (19) and in 73 (97%) of 75 O103 STEC clinical isolates in the United States (11). Clinical cases of O103 STEC (stx1a+eae) were reported to be associated with diarrhea and/or bloody diarrhea (11, 19). Therefore, O103 STEC (stx1a+eae) could represent a potential health risk to humans. O8:H19 STEC (stx2) has been reported in association with clinical diarrhea (19), whereas O159:H19 STEC has been reported to be a rare non-O157 STEC serotype associated with clinical diarrhea illnesses (11). The results suggest that wheat flour can be contaminated with non-O157 STEC; therefore, if it is consumed raw there is a potential for human illnesses.
Recently, the characterization of STEC isolates from food sources by WGS has been recommended, with the aim of characterizing the virulence genes. Based on the virulence gene profile and in association with clinical illness severity, the potential health risk of human illness associated with STEC in foods has been categorized into five risk levels, from level 1 (highest) to level 5 (lowest) (Table 4) (20). The prevalence of non-O157 STEC was found to be 1.7% (95% CI, [0.8, 3.7]) in this study, which is lower than those found in two Swiss retail flour studies (10, 31). A Swiss study (10) identified 5 (9.6%) STEC strains from 52 wheat flour samples, and the STEC isolates carried various subtypes of single stx gene that were categorized as being the lowest risk level (level 5) for potential human illness (10). In another Swiss study, 1 STEC isolate (4.8%) was identified from 21 wheat flour samples, and the STEC isolate (O103:H2) carried the virulence genes (stx1a+eae) that belonged to a lower risk level (level 4) for potential human illness (31). Clinical cases of O103 STEC (stx1a+eae) associated with diarrhea and bloody diarrhea have been reported (19). O103 STEC (stx1a+eae) was also found in our study; the prevalence of STEC that belonged to the lower risk level 4 was found to be 0.3% (Table 4). The other four non-O157 STEC isolates (1.2%) identified from this study belonged to the lowest risk levels (20) (Table 4). In contrast, wheat flour samples collected from specific lots of interest and consumer complaints during the O121 STEC outbreaks between 2016 and 2017 in Canada (13) had a higher overall STEC prevalence (15 of 114, 13.2%; 95% CI [8.1, 20.6]) and a higher prevalence (6 of 114, 5.3%; 95% CI [2.4, 11.0]) of non-O157 STEC possessing virulence genes stx2a+eae belonging to the highest risk level (level 1) (Table 4). These observations are not surprising because the samples were collected in support of outbreak investigations and were, consequently, biased toward samples in which STEC pathogens were expected to be detected. For example, several samples were from the same lot, manufactured during the same time periods, originating from consumer complaints, etc. The source of contamination of the wheat flour samples implicated in the foodborne illness outbreaks could not be identified (13).
The occurrence of O157 STEC and Salmonella spp. in wheat flour is considered to be low. The regular occurrence of non-O157 STEC posing the lowest level of risk to human health cannot be ruled out. Given that non-O157 STEC that is classified as low risk (level 4) can cause bloody diarrhea and diarrhea illnesses, this commodity can be considered a food safety concern because the consumption of raw wheat flour or undercooked wheat flour can expose consumers to STEC infections. Consideration should be given to educating farmers, millers, consumers, and those in the food preparation industry about the risk associated with raw flour. Further research and understanding of sources of contamination of wheat flour could prove useful in designing further surveillance and control measures.
We gratefully thank the staff of the Canadian Food Inspection Agency's Food Microbiological Laboratory Network for their technical assistance. We also thank all staff involved in the implementation of this project.
Supplemental material associated with this article can be found online at: https://doi.org/10.4315/JFP-20-297.s1