Annual Change in the Composition of Bulk Tank Milk Microbiota in Northern Kanagawa Prefecture, Japan

Simple Summary

A cooling system called a bulk tank is used by farmers to temporarily store milk prior to shipment, and the milk stored in the tank is called bulk milk. Producing high-quality bulk milk leads to producing high-quality milk and dairy products. However, bulk milk is difficult to control because it is affected by the environment in which the cows are raised, the sanitary conditions of the milking equipment, and many other factors, and is composed of many types of bacteria. Additionally, few studies have examined the bacterial composition of bulk milk in Japan. Therefore, the objective of this study was to analyze the bacterial composition of bulk milk in Japanese farms through the year. We conducted interviews regarding the herds’ general information, milking hygiene, and hygienic management of milking equipment. In addition, the bacterial composition of bulk milk was also examined using 16S rRNA gene full-length amplicon sequence analysis. The results suggest that the bacterial composition of bulk milk may not be constant throughout the year and may be susceptible to bacteria that are tolerant to low temperatures. This finding would be a basis for knowledge regarding the microbial control of bulk milk.

Abstract

Bulk tank milk microbiota (BTMM) is affected by various factors, including the characteristics of raw milk, microflora on teat surfaces, and the milking system. Clarifying the influence of these factors is important for producing high-quality dairy products. This longitudinal study describes the annual changes in BTMM at six dairy farms in northern Kanagawa Prefecture, Japan. Bulk tank milk samples were collected six times a year (in February, April, June, August, October, and December of 2022) to give a total of thirty-six samples. After bulk tank somatic cell counts (BTSCC) had been determined, we performed 16S rRNA gene amplicon sequence analysis to clarify the composition of the BTMM. Although no annual changes were observed in the BTSCC and alpha-diversity index, a significant difference in the beta-diversity index was observed between February and August (p = 0.0315). In February, the proportions of the psychrophilic genera Listeria and Enterococcus were significantly increased (p < 0.05). Similarly, in August, the proportion of commensal milk microbiota in the genera Catenibacterium and Acetobacter were significantly increased (p < 0.05). The results of this study suggest that the composition of BTMM in this region changed throughout the year, which may have been influenced by psychrophilic bacteria in winter.

1. Introduction

The production of high-quality milk and dairy products necessitates the use of high-quality raw milk. One of the crucial factors for evaluating the quality of raw milk is its microbial content, which also serves as a criterion for economic trade. Consequently, it is important to have knowledge of, and to understand, the composition of bacterial flora in raw milk, along with the temporal changes and potential causes of these changes.

Microbial contamination of raw milk in bulk tanks, where milk from different individuals is pooled and stored until shipment, occurs not only via milk from infected individuals, but also from other sources of contamination, such as microflora on the teat surface and milking equipment [1]. As a result, the composition of the bulk tank milk microbiota (BTMM) is typically very diverse. In particular, mastitic udders are known to shed large numbers of pathogenic microorganisms into milk, and amplicon-based high-throughput sequencing analysis has shown that the composition of the BTMM can be frequently affected by bacterial taxa associated with mastitis (e.g., Staphylococcus and Streptococcus) [2]. However, milk from individuals without clinical or subclinical mastitis may also contain mastitis-causing pathogens. And other factors, such as farm management practices and seasonality, have also been shown to significantly influence the microbiota detected in milk [3]. Recently, it has been shown that the composition of BTMM varies not only among farms, but also by geographic region and sampling time [4]. In addition, studies on breeds other than Holsteins have also been reported, showing that the flora of Jersey cows changes depending on whether they are milked in cool (November) or hot (July) seasons [5].

Thus, BTMM is a complex group that includes numerous species from diverse sources, and what we learn about them will help us to produce high-quality milk. However, the causes of the diversity and compositional variations of BTMM in Japan have not been clarified.

The objective of this study was therefore to clarify the annual variation in the composition and diversity of the BTMM. The causes of this variation were examined over the course of a year by continuously observing bulk milk from dairy farms that employ the same feeding pattern and which are located in the same region. The analysis used metagenomic analysis, which does not require culturing and can detect difficult-to-culture bacteria.

2. Materials and Methods

2.1. Farm Selection and Milk Sample Collection

This study was conducted on six farms (A, B, C, D, E, and F) within the same region, located in northern Kanagawa Prefecture. Bulk tank milk from these farms was collected at bi-monthly intervals, six times in total (February, April, June, August, October, and December of 2022). Each farmer was interviewed to ascertain the number of dairy cattle, cattle breeds, feeding system, feeding pattern, origin of feed (i.e., self-supplied or purchased), milking type, milking frequency, milking equipment cleaning methods, bulk tank cleaning methods, elimination of abnormal milk by pre-milking, milking machine manufacturer, and bulk tank inspection frequency. Bulk tank milk was aseptically collected in 10 mL DNA-free tubes by a veterinarian, refrigerated immediately at 4 °C before noon, and brought to the laboratory for testing within 3 h. Bulk tanks used by farmers were kept at around 4 °C.

2.2. Weather Information

The 2022 meteorological data (temperature and rainfall) of six farms located in the same area were investigated using Japan Meteorological Agency data [6] and are shown in Figure 1. Average monthly temperatures ranged from 3.9 °C to 27.4 °C, with the lowest in January and the highest in August. Average monthly precipitation ranged from 19.5 mm to 254.5 mm and tended to be lower in winter (January, February, and December).

2.3. Milk Test

The 10 mL DNA-free tube containing the sample was inverted and mixed well, and the sample was injected into a disposable cassette (DeLaval cassette 928658 80, International AB, Tumba, Sweden) containing small amounts of regents which, when mixed with the milk, react with the nuclei of the somatic cells. Then, a cassette with a DeLaval cell counter (DCC; DeLaval International AB, Tumba, Sweden) was set to measure the bulk tank milk somatic cell count (BTSCC) according to the manufacturer’s instructions.

2.4. DNA Isolation and Pyrosequencing

DNA was extracted from bulk tank milk samples using an ISOSPIN Fecal DNA kit (NIPPON GENE, Tokyo, Japan). Briefly, bulk tank milk samples were homogenized for 10 min at 4350 rpm using a bead mill (Shakeman 6; Bio Medical Science, Tokyo, Japan). DNA was then extracted using a centrifugation-based column purification protocol. Extracted DNA was concentrated using a Quantus^TM Fluorometer (Promega Corporation, Madison, WI, USA) and purity was determined by measuring the absorbance ratios at 260/280 nm and 260/230 nm using a NanoDrop spectrophotometer (NanoDrop Technologies, Inc., Wilmington, DE, USA). Universal primers S-D-bact-1391-a-A-17 and S-D-bact-0008-c-S-20 [7] were used to amplify the V1–V9 region of 16S rRNA by PCR, which was performed using KAPA HiFi^TM HotStart ReadyMix (2×) (Nippon Genetics, Tokyo, Japan). The total reaction volume used for PCR was 20 μL, which consisted of KAPA HiFi^TM HotStart ReadyMix (10 μL), 0.25 μM of each primer, nuclease-free water (7 μL), Haloarcula japonica (2.8 × 10³ molecules/μL), and genomic DNA (100 ng). Haloarcula japonica was spiked in uniformly as an internal control. The PCR was performed in a thermal cycler (ASTEC, Fukuoka, Japan) with the following settings: initial annealing step at 95 °C for 5 min, followed by 98 °C for 20 s, 69 °C for 15 s, and 72 °C for 60 s for 35 cycles, followed by a final extension step of 72 °C for 5 min. Subsequently, for library preparation, a second PCR was performed using a Rapid Barcoding Kit (SQK-RBK004: Oxford Nanopore Technologies, Oxford, UK) for DNA labelling, with incubation at 30 °C for 1 min and 80 °C for 1 min. All labeled PCR products were then purified using AMPure^® XP beads (Beckman Coulter, Brea, CA, USA), and DNA concentration was measured using a Quantus^TM Fluorometer, with purity and quantity further assessed by NanoDrop spectrophotometry.

For the sequencing of library construction, labeled DNA (10 μL) was mixed with Rapid Adapter (1 μL) at room temperature for 5 min. The library mixture DNA (11 μL) was then mixed with sequencing buffer (34 μL), nuclease-free water (4.5 μL), and loading beads (25.5 μL), and sequencing was performed on a Spot-on Flow Cell R9 version (FLO-MIN106D; Oxford Nanopore Technologies) using a MinION^TM Mk1C device (Oxford Nanopore Technologies). Data acquisition and FASTQ file creation were performed using MINKNOW software ver. 21.11.6 (Oxford Nanopore Technologies).

The resulting FASTQ files were trimmed and filtered with NanoFilt software ver. 2.8.0 [8]. Specifically, the minimum average read quality score targeted was less than 10, all sequences shorter than 500 nucleotides were removed, and the first 50 nucleotides of all reads were trimmed. After trimming and size selection, an average of 28,860 reads (4031–169,625) per sample were retained for bacterial identification. Each read was subjected to a minimap2 search against 5850 representative bacterial genome sequences stored in the Genome Sync database [9]. Taxonomic groups were determined based on the National Center for Biotechnology Information classification database [10]. Low-presence taxa (<0.01% of all reads) were excluded from the analysis.

2.5. Statistical Analyses

Annual changes in BTSCC and the alpha-diversity indexes (Richness, Simpson’s, and Shannon’s indexes) of the BTMM across the targeted farms were analyzed by the Friedman test, a nonparametric test. This analysis was performed using EZR plugin (ver. 1.61; Saitama Medical Center, Jichi Medical University) for the R statistical package (The R Foundation for Statistical Computing, Vienna, Austria) [11]. p values less than 0.05 were considered to indicate statistically significant differences. The composition of BTMM was compared monthly using non-parametric tests for similarity analysis (ANOSIM), which used the Bray–Curtis distance scale to measure compositional differences. p values less than 0.05 were considered to indicate statistically significant differences. This analysis was performed using the PAST 4.13 version software package [12]. Linear Discriminant Analysis Effect Size (LEfSe) analysis via the online Galaxy interface was performed to identify the specific bacterial species that showed significant differences in abundance in each month, as indicated by the ANOSIM test results [13].

3. Results

The total head of cattle on the farms in this study ranged from 20 to 51 (average 28.5 head), and all were Holstein Friesian (

Table 1. Characteristics of farms included in the study.

Farm	Number of Milking Cows	Milking Type and Frequency (/Day)	Elimination of Abnormal Milk by Pre-Milking	Milking Equipment Cleaning Methods	Bulk Tank Cleaning Methods	Milking Machine Manufacturer
A	21	Automatic 2	Yes	Alkaline cleaning (once a day) with acid cleaning (once every 3 days)	Automatic cleaning after each milk shipment	SAC
B	26	Automatic 2	No	Alkaline cleaning (once a day) with acid cleaning (once every 3 days)	Automatic cleaning after each milk shipment	DeLaval
C	30	Automatic 2	Unavailable	Unavailable	Automatic cleaning after each milk shipment	Unavailable
D	20	Automatic 2	Yes	Alkaline cleaning (once a day)	Automatic cleaning after each milk shipment	SAC
E	51	Automatic 2	Yes	Alkaline cleaning (once a day)	Automatic cleaning after each milk shipment	ORION
F	23	Automatic 2	Yes	Alkaline cleaning (once a day) with acid cleaning (once every 3 days)	Automatic cleaning after each milk shipment	DeLaval

). The feeding mixture was component feeding, and all were fed using a tie-stall system, with the exception of Farm F, which used purchased feed. Farm F supplemented the feed with dent corn silage at 13 kg/day/head, but only in December. The milking type was an automatic milking system for all farmers, and the milking frequency was two times/day. Interviews also revealed that bulk tanks were automatically cleaned after each milk collection using a disinfectant, and that bulk tank cooling inspections were conducted once a year (Farm A, B, D, E, and F). The milking equipment cleaning methods, pre-milking, and milking equipment manufacturers are shown in Table 1. For Farm C, some items could not be investigated due to the ranchers’ circumstances.

The BTSCC measurement results for each sample are shown in

Table 2. Bulk tank milk somatic cell count (mL⁻¹) measurement results for each sample.

Farm	February	April	June	August	October	December
A	102,000	172,000	143,000	196,000	194,000	329,000
B	72,000	86,000	118,000	80,000	46,000	72,000
C	141,000	103,000	194,000	191,000	47,000	100,000
D	175,000	184,000	123,000	147,000	254,000	216,000
E	152,000	107,000	170,000	171,000	181,000	331,000
F	65,000	76,000	77,000	23,000	53,000	61,000

. BTSCC averaged 137,555 cells/mL overall, with the highest densities observed in December (average 184,833 cells/mL) and the lowest densities observed in February (average 117,833 cells/mL). The results of the Friedman test (Bonferroni correction) showed no significant differences between the months (Figure 2A). The alpha diversity indices (Richness (Figure 2B), Simpson’s (Figure 2C), and Shannon’s (Figure 2D)) showed no significant differences across all groups for any parameters, according to the Friedman test with Bonferroni correction.

A farmer-to-farmer comparison of the BTMM composition on the target farms is shown in Figure 3. The composition ratio of each farm was mainly composed of the phylum Actinobacteria, phylum Firmicutes, and phylum Proteobacteria, and there were no significant differences among farmers, although farmers B and E had a higher proportion of Euryarchaeota in February.

Annual changes in the composition of BTMM on the target farms are shown in Figure 4 (A: 100% stacked bar graph, B: principal coordinate analysis). The BTMM consisted mainly of the phyla Actinobacteria, Firmicutes, and Proteobacteria, but some taxa in the phylum Euryarchaeota and other species were also identified. For example, various species were identified, including Mycolicibacterium spp. and Kocuria spp. in the phylum Actinobacteria, Catenibacterium spp. and Acetanaerobacterium spp. in the phylum Firmicutes, and Acetobacter spp. in the phylum Proteobacteria. According to the results of the ANOSIM test with Bonferroni correction, there was a significant difference (p = 0.0315) in BTMM composition between February and August (Figure 4A). Principal coordinate analysis, employing the Bray–Curtis distance to compare bacterial composition similarity across different months, showed that the BTMM composition varied among farmers in February; however, the composition was similar in August (Figure 4B).

The bar graph (Figure 5A) and cladogram (Figure 5B) show the bacterial species that varied significantly between February (red) and August (green), as indicated by the LEfSe analysis. The genera Listeria and Enterococcus accounted for significantly larger proportions of the BTMM in February, and Catenibacterium and Acetobacter in August.

4. Discussion

No variation was observed in the average BTSCC on the target farms over the course of the year. Somatic cell counts in milk are used as an indicator of intramammary infection [14], and the reasons for variation in BTSCC have been attributed to the contamination of bulk milk with mastitis-causing bacteria [15], as well as herd size, season, and the timing of calving [16]. BTSCC is generally considered to be abnormal milk in excess of 300,000/mL [17], and some samples in this study exceeded that standard. However, because the milking techniques and hygiene and milk line and bulk tank cleaning methods of the target farmers in this study were sanitary, and because they eliminated abnormal milk by performing pre-milking (Table 1), it was unlikely that abnormal milk would enter the bulk milk. However, given that the herds in this study were the same size and had no seasonal breeding, and given that there were no inter-farm differences or annual variations in BTSCC, and in addition that the study was based on bulk milk that was actually shipped, we see no problem in using these samples for a comparative analysis of BTMM between farms.

The analysis of alpha-diversity indexes of the BTMM, including richness, Simpson’s, and Shannon’s indices, did not show any annual variation. However, the richness index, which represents the amount of bacterial species in the flora, exhibited an increasing trend, with greater variation observed after the summer. This trend was noted for both the richness index and the BTSCC, although the differences were not statistically significant. These results differed from those of a previous study [18], which reported a negative correlation between the richness index and BTSCC. The reason for this is that the study by Oliveira et al. [18] observed an increase in BTSCC with an increase in the percentage of Streptococcus spp. and Staphylococcus spp., whereas the present study did not observe an increase in the relative percentage of Streptococcus spp. or Staphylococcus spp. Moreover, Shannon’s index, which is weighted by the percentage of rare species, tended to be higher compared to the other indices (i.e., Simpson’s and Shannon’s indices) that represent the evenness of bacterial flora. This suggests that the number of rare species constituting the flora increased after the summer season.

BTMM composition ratios were similar among farms, but the percentage of Euryarchaeota phylum was higher in February in farmers B and E. This was thought to be due to the higher percentage of archaea (Haloarcula japonica) readings added as an inner control, i.e., the relatively low amount of bacterial DNA in the samples. This may also have an effect on the extremely high percentage of Euryarchaeota in February compared to other months when compared together as a group by month.

The percentage of Pseudomonas spp. in the bulk milk of the farmers included in this study was small and did not fluctuate throughout the year. Since Davide et al. [2] pointed to the inadequate cleaning of milking equipment as the cause of high amounts of Pseudomonas spp. in bulk milk, it was thought that the cleaning of milking equipment was sufficient for the farmers in the study. On the other hand, Streptococcus spp. was found in several samples, and especially in those with BTSSC above 300,000/mL (December, Farm A and Farm E), at a higher rate (19.49%, 0.50%) than in the other samples (0.12–0.21%). Previous studies have attributed the increase in Streptococcus spp. to the contamination of milk from cows with clinical mastitis [2]. Although the cows in this study were clinically observed prior to milking and screened for mastitis milk by pre-milking to ensure that they did not have clinical-type mastitis before milking, the possibility could not be ruled out that milk abnormal enough to affect BTMM was introduced.

Listeira spp. were detected in the bulk milk of the farms included in this study, which is consistent with previous reports [19,20,21]. The origin of Listeria spp. has been linked to fecal contamination due to cattle being fed improperly stored silage [22]. However, the farmers in this study fed almost no silage to their cattle, and no significant increase in Listeria spp. was observed in bulk milk at the only instance of silage feeding (December of Farm F). The detection of Listeria spp. was concentrated in February, the month with the lowest average temperature, suggesting that the origin of Listeria spp. was related to temperature rather than silage feed. We also believe that the detection of Listeria spp. is greatly influenced by temperature, as it has been reported [5] that BTMM in Jersey cattle differs between summer and winter and may be influenced by an increase or decrease in bacteria that is related to temperature.

All of the farms in this study fed their cattle purchased forage, and only on Farm F in December did they use their own feed (dent corn silage), but no bacterial species had a prominent increase in composition percentage in that month alone. The relationship between diet and milk microbiota composition was reported to be different for raw milk microbial communities of cows fed low-starch/high-fiber (LSHF) diets and high-starch/low-fiber (HSLF) diets [23], suggesting that BTMM may be influenced by the diet administered. However, these associations could not be determined because the starch concentration and fiber content of diets were not investigated in this study. Other studies have reported that BTMM varies by breed [24], but the dairy cows in this study were all Holsteins, and the results of this study were not considered to be affected by breed differences.

BTMM composition differed significantly between February and August. The results of the principal coordinate analysis across farms also showed that the BTMM composition in February was highly variable, whereas it was similar in August. It has previously been shown [2] that BTMM composition is influenced by mastitis-causing bacteria, such as Streptococcus spp. and Staphylococcus spp., in the short term, and by climate and feed in the long term (i.e., on an annual basis). Since this study was not conducted over a multi-year period, the effect of climate could not be determined. However, meteorological data show that monthly average temperatures were lowest in January and highest in August, and when combined with the fact that BTMM composition was significantly different in February and August, it is possible that temperature plays a significant role in the change in BTMM composition. Also, given the similarities in BTMM observed across farms in August, it seems likely that the farms may have been exposed to similar climatic conditions.

The difference in the composition of BTMM between February and August was due to the higher proportion of Listeria and Enterococcus spp. in February and Catenibacterium and Acetobacter spp. in August, respectively. Since Listeria spp. and Enterococcus spp., both of which were abundant in February, are psychrophilic [25], it is possible that some herds were affected by these bacteria due to cold conditions, which was one of the reasons for the variation observed in the composition of BTMM among farms. On the other hand, Catenibacterium spp. and Acetobacter spp., which accounted for a large proportion of the microfloral community in August, were reported to be indigenous flora in normal milk [26]. The small variation in BTMM composition among herds in August suggests that there were no significant differences across farms in terms of bacterial exposure during the summer months in this study. These results suggest that although there were no differences among herds in terms of environmental exposure to bacteria in August, some herds may have been exposed to bacteria that were tolerant to the low temperatures in February.

This study has several potential limitations. First, this study examined the relative composition of microbiota in bulk tank milk, and did not consider the absolute number of bacteria in milk. Second, the results of BTMM composition reflect not only the udder flora of each herd, but also the results of the bacterial contamination of bulk tanks. However, the bacterial flora on teat surfaces, milking equipment, and bulk tank hygiene were not examined in this study. Consequently, the origin of the bacteria affecting the composition of the BTMM could not be determined. Third, we could not obtain descriptions of animals, such as average parity, calving date, body weight, body condition score, feeding system by farm (feed ingredients and nutritional value), and average total dry matter intake, due to personal information. Therefore, it was not possible to examine in detail the effects of feed, individual cow differences, and other factors on the BTMM composition ratio.

5. Conclusions

The results of this study suggest that the BTMM composition across the herds examined in this study was not constant throughout the year, and that exposure to psychrophilic bacteria in the winter was a factor in the difference. The results also suggest that temperature is likely to affect the BTMM composition ratio.