Postpasteurization contamination (PPC) of high temperature, short time-pasteurized fluid milk by gram-negative (GN) bacteria continues to be an issue for processors. To improve PPC control, a better understanding of PPC patterns in dairy processing facilities over time and across equipment is needed. We thus collected samples from 10 fluid milk processing facilities to (1) detect and characterize PPC patterns over time, (2) determine the efficacy of different media to detect PPC, and (3) characterize sensory defects associated with PPC. Specifically, we collected 280 samples of high temperature, short time-pasteurized milk representing different products (2%, skim, and chocolate) and different fillers over 4 samplings performed over 11 mo at each of the 10 facilities. Standard plate count (SPC) as well as total GN, coliform, and Enterobacteriaceae (EB) counts were performed upon receipt and after 7, 10, 14, 17, and 21 d of storage at 6°C. We used 16S rDNA sequencing to characterize representative bacterial isolates from (1) test days with SPC >20,000 cfu/mL and (2) all samples with presumptive GN, coliforms, or EB. Day-21 samples were also evaluated by a trained defect judging panel. By d 21, 226 samples had SPC >20,000 cfu/mL on at least 1 d of shelf life; GN bacteria were found in 132 of these 226 samples, indicating PPC. Crystal violet tetrazolium agar detected PPC with the greatest sensitivity. Spoilage due to PPC was predominantly associated with Pseudomonas (isolated from 101 of the 132 samples with PPC); coliforms and EB were found in 27 and 37 samples with spoilage due to PPC, respectively. Detection of Pseudomonas and Acinetobacter was associated with lower flavor scores; coagulated, fruity fermented, and unclean defects were more prevalent in d-21 samples with PPC. Repeat isolation of Pseudomonas fluorescens group strains with identical partial 16S rDNA sequence types was observed in 8 facilities. In several facilities, specific lines, products, or processing days were linked to repeat product contamination with Pseudomonas with identical sequence types. Our data show that PPC due to Pseudomonas remains a major challenge for fluid milk processors; the inability of coliform and EB tests to detect Pseudomonas may contribute to this. Our data also provide important initial insights into PPC patterns (e.g., line-specific contamination), supporting the importance of molecular subtyping methods for identification of PPC sources.

Some gram-negative bacteria, including Pseudomonas spp., can grow at refrigeration temperatures and cause flavor, odor, and texture defects in fluid milk. Historical and modern cases exist of gray and blue color defects in fluid milk due to Pseudomonas, and several recent reports have detailed fresh cheese spoilage associated with blue-pigment-forming Pseudomonas. Our goal was to investigate the genomes of pigmented Pseudomonas isolates responsible for historical and modern pigmented spoilage of dairy products in the United States to determine the genetic basis of pigment-forming phenotypes. We performed whole genome sequencing of 9 Pseudomonas isolates: 3 from recent incidents of graypigmented fluid milk (Pseudomonas fluorescens group), 1 from blue-pigmented cheese (P. fluorescens group), 2 from a historical blue milk spoilage incident (Pseudomonas putida group), and 3 with no evidence for blue or gray pigment formation (2 from P. fluorescens group and 1 from Pseudomonas chlororaphis group). All 6 isolates collected from products with a gray or blue pigment defect were confirmed to produce pigment using potato dextrose agar or pasteurized milk. A subset of 2 isolates was selected for inoculation into milk and onto the surface of a model cheese for subsequent color measurement. These isolates produced different colors on potato dextrose agar, but produced nearly identical color defects in milk and on model cheese. For the same subset of 2 isolates, the gray color defect in milk was produced only in containers with ample headspace and not in full containers, suggesting that oxygen is vital for pigment formation. This work also demonstrated that a Pseudomonas isolate from cheese can produce a pigment defect in milk, and vice versa. Comparative genomics identified an accessory locus encoding tryptophan biosynthesis genes that was present in all isolates that produced gray or blue pigment under laboratory conditions and was only previously reported in 2 P. fluorescens isolates responsible for blue mozzarella in Italy. Because this locus was found in genetically distant isolates belonging to different Pseudomonas species groups, it may have been acquired via horizontal gene transfer. These data suggest that several past and present gray-or blue-pigmented dairy spoilage events share a common genetic etiology that transcends species-level identification and merits further investigation to determine mechanistic details and modes of prevention.

Pseudomonas species are well recognized as dairy product spoilage organisms, particularly due to their ability to grow at refrigeration temperatures. Although Pseudomonas-related spoilage usually manifests itself in flavor, odor, and texture defects, which are typically due to production of bacterial enzymes, Pseudomonas is also reported to cause color defects. Because of consumer complaints, a commercial dairy company shipped 4 samples of high temperature, short time (HTST)-pasteurized milk with distinctly gray colors to our laboratory. Bacterial isolates from all 4 samples were identified as Pseudomonas azotoformans. All isolates shared the same partial 16S rDNA sequence and showed black pigmentation on Dichloran Rose Bengal Chloramphenicol agar. Inoculation of one pigment-producing P. azotoformans isolate into HTST-pasteurized fluid milk led to development of gray milk after 14 d of storage at 6°C, but only in containers that had half of the total volume filled with milk (∼500 mL of milk in ∼1,000-mL bottles). We conclusively demonstrate that Pseudomonas can cause a color defect in fluid milk that manifests in gray discoloration, adding to the palette of color defects known to be caused by Pseudomonas. This information is of considerable interest to the dairy industry, because dairy processors and others may not typically associate black or gray colors in fluid milk with the presence of microbial contaminants but rather with product tampering (e.g., addition of ink) or other inadvertent chemical contamination.

after). Overall, our findings suggest that commonly used "broad stroke interventions" may have a limited effect on reducing PPC. Our case study also demonstrates the inherent complexities of identifying and successfully addressing sanitation problems in large and complex fluid milk processing facilities. For example, broad changes to sanitation practices without improvements in PM and sanitary equipment design may not always lead to reduced PPC. Our data also indicate that although short-term evaluations, such as pre-and post-tests for employee training, may suggest improvements after corrective and preventive actions, extensive microbial testing, ideally in combination with isolate characterization, may be necessary to evaluate return on investment of different interventions.

Spoilage of HTST- (high-temperature, short-time) and vat- pasteurized fluid milk due to introduction of Gram-negative bacteria post-pasteurization remains a challenge for the dairy industry. While processing facility level practices (e.g., sanitation practices) are known to impact the frequency of post-pasteurization contamination (PPC), the relative importance of different practices is not well defined, affecting the ability of facilities to select intervention targets that reduce PPC and provide the greatest return on investment. Thus, the goal of this study was to use an existing longitudinal dataset of bacterial spoilage indicators obtained for pasteurized fluid milk samples collected from 23 processing facilities between July 2015 and November 2017 (with 3 to 5 samplings per facility) and data from a survey on fluid milk quality management practices, to identify factors associated with PPC and rank their relative importance, using two separate approaches: (i) multimodel inference and (ii) conditional random forest. Data pre-processing for multimodel inference analysis showed (i) nearly all factors were significantly associated with PPC when assessed individually using univariable logistic regression and (ii) numerous pairs of factors were strongly associated with each other (Cramer’s V ³0.80). Multimodel inference and conditional random forest analyses identified similar drivers associated with PPC; factors identified as most important based on these analyses included cleaning and sanitation practices, activities related to good manufacturing practices, container type (which is a proxy for different filling equipment), in-house finished product testing, and designation of a quality department, indicating potential targets for reducing PPC. In addition, this study illustrates how machine learning approaches can be used with highly correlated and unbalanced data, as typical for food safety and quality, to facilitate improved data analyses and decision-making.