Troubleshooting high laboratory pasteurization counts in organic raw milk requires characterization of dominant thermoduric bacteria, which includes non-sporeformers as well as sporeformers

Laboratory Pasteurization Count ( LPC ) enumer-ates thermoduric bacteria and is one parameter used to assess raw milk quality. While there is currently no regulatory limit for LPC, LPC data are used by some dairy processors and cooperatives to designate raw milk quality premiums paid to farmers and may also be used for troubleshooting bacterial contamination issues. Despite occasionally being used as a proxy for levels of bacterial spores in raw milk, there is limited knowledge of the types of organisms that are enumerated by LPC in contemporary raw milk supplies. While historical studies have reported that thermoduric bacteria quantified by LPC may predominantly represent Gram-positive cocci, updated knowledge on microbial populations enumerated by LPC in contemporary organic raw milk supplies is needed. To address this gap, organic raw milk samples from across the United States (n = 94) were assessed using LPC, and bacterial isolates were characterized. LPC ranged from below detection (<0.70 log cfu/mL) to 4.07 log cfu/mL, with a geometric mean of 1.48 log cfu/mL. Among 380 isolates characterized by 16S rDNA sequencing, 52.6%, 44.5%, and 2.4% were identified as Gram-positive sporeformers, Gram-positive non-sporeformers, and Gram-negatives, respectively, and 0.5% that could not be categorized into those groups because they could only be assigned a higher level of taxonomy. Isolates identified as Gram-positive sporeformers were predominantly Bacillus (168/200) and Gram-positive non-sporeformers were predominately Brachybacterium (56/169) and Kocuria (47/169). To elucidate if the LPC level can be an indicator of the type of thermoduric (e.g., sporeforming bacteria) present in raw milk, we evaluated the proportion


INTRODUCTION
Laboratory Pasteurization Count (LPC) is a commonly used raw milk quality parameter that enumerates thermoduric bacteria by heating the product to Troubleshooting high laboratory pasteurization counts in organic raw milk requires characterization of dominant thermoduric bacteria, which includes non-sporeformers as well as sporeformers 63°C for 30 min followed by pour plating with Standard Methods Agar (SMA) and incubation for 48h at 32°C (Wehr and Frank, 2004).The LPC method has been used historically as an indicator of improperly cleaned and sanitized raw milk equipment (Byrne andBishop, 1991, Martin et al., 2023).Although there is currently no regulatory limit for thermoduric bacteria present in raw milk, organic dairy producers are often subjected to the LPC quality parameter.Some dairy processors and cooperatives incentivize organic dairy farmers to maintain high quality milk by assigning premiums to raw milk with low LPC (e.g., < 100 cfu/mL) (Murphy andBoor, 2000, Jayarao et al., 2004), while raw milk with LPC exceeding the threshold of 200 cfu/mL is generally accepted as an indication of poor equipment hygiene (Jayarao and Wolfgang, 2003, Jayarao et al., 2004, Martin et al., 2023).
Several studies throughout the 20th century have found the primary thermoduric bacterial populations in raw milk consist primarily of Gram-positive sporeformers (GPS) (e.g., Bacillus) and non-sporeforming Grampositive cocci and rods (Kikuchi et al., 1996, Ribeiro Júnior et al., 2018).Previously reported GPS found in LPC treated raw milk are predominantly Bacillus and Paenibacillus (Kikuchi et al., 1996, Ribeiro Júnior et al., 2018).GPS are found ubiquitously in dairy farm environments and their primary mode of transmission from these environmental niches into raw milk occurs at the point of milk harvest (Scheldeman et al., 2005, Martin et al., 2019).A recent review by Martin et al. (2023) identified that farm factors associated with increased spore levels in raw milk generally fall within 4 categories including i) housing area and bedding practices, ii) milking routine practices, iii) cow-level factors, and iv) feed factors (Martin et al., 2023).
Typical non-sporeforming Gram-positive cocci and rods that survive the LPC method include Microbacterium, Kocuria, Brachybacterium, Micrococcus, Streptococcus, Enterococcus, and others (Kikuchi et al., 1996, Wehr and Frank, 2004, Gleeson et al., 2013, Ribeiro Júnior et al., 2018).Unlike with GPS, there is little research specifically investigating the sources and farm practices associated with non-GPS thermodurics (e.g., Micrococcus) in raw milk, however, generally several studies have investigated sources and factors associated with high LPC.For example, Elmoslemany et al. (2010) found that elevated LPC was associated with lack of water purification system, presence of a plate cooler (which are often hard to clean adequately and may harbor biofilms) and inadequate frequency of acid washes for equipment.While the authors did not characterize the thermoduric bacteria that were enumerated by the LPC method used in their study, these factors support the historically cited association between cleaning and sanitation practices and levels of thermoduric bacteria in raw milk (Elmoslemany et al., 2010).It is notable that studies focused solely on farm sources and factors associated with GPS have not reported equipment cleaning and sanitation factors as important drivers of GPS levels in raw milk (Martin et al., 2019, Evanowski et al., 2020), and it may therefore be postulated that non-sporeforming Gram-positive cocci and rods were driving the LPC levels in the study by Elmoslemany et al. (2010).Confirmation of this hypothesis would require further investigation into the sources and farm practices associated with non-sporeforming thermoduric Gram-positive cocci and rods.Despite a large volume of historical data on the LPC method and the types of thermoduric bacteria selected for by the LPC test, little information is available on contemporary organic raw milk supplies.Further, the existing literature is lacking key information on the associations between LPC and the subgroups of thermoduric bacteria present in organic raw milk (i.e., GPS and non-sporeforming Grampositive cocci and rods) complicating troubleshooting efforts.Therefore, the goals of this study were to i) characterize the thermoduric bacteria selected for by the LPC method, ii) determine if an association exists between LPC level and a subset of the thermoduric bacteria present in organic raw milk, and iii) compare 2 bacterial characterization methods for identification of thermoduric bacteria in organic raw milk.

Milk sampling and collection
Pilot study sample collection.To determine the impact of temporary freezing of raw milk on the levels and populations of thermoduric bacteria, we evaluated LPC and predominant bacterial populations from 15 raw milk samples before and after freezing at −20°C as described here.Raw milk samples were collected from 15 separate farms (5 certified organic and 10 conventional) in New York State.Each raw milk sample was collected from the bulk tank as previously described by Evanowski et al. (2020) and split into 2 vials: the fresh sample was stored at 4°C for approximately 20 h and the frozen sample was stored at −20°C for approximately 7 d before processing.Sample thawing (frozen samples only), microbiological testing (i.e., LPC), and bacterial characterization (i.e., 16S PCR and sequencing) were conducted as described below for each matched fresh and frozen raw milk sample.
Cross-sectional study.Working with multiregional collaborators on a separate study, a list of 102 US certified organic dairy farms willing to participate was generated.Between the start and completion of the sampling, 6 farms withdrew from the study and 2 samples were excluded from analyses due to lab error, resulting in a total of 94 farms included in the study.Organic raw milk was collected from those 94 organic dairy farms which were located in the following states: PA (n = 28), NY (n = 23), CA (n = 16), WA (n = 6), CO (n = 5), ID (n = 4), WI (n = 4), OR (n = 3), VT (n = 3), MN (n = 1), and IA (n = 1), ranging in herd size from 26 to ~4,000 milking cows.
Organic raw milk samples were collected from the bulk tank in 10 oz vials by the producers, who were instructed to take bulk tank samples as previously performed by Evanowski et al. (2020).The samples were stored in the freezer at the farm until they were shipped on ice to Cornell University's MQIP.Samples arrived frozen and were stored at −20°C until further analysis was conducted, typically within 1 mo; and samples that did not arrive frozen were excluded from the study.

Microbiological analysis
Frozen samples were thawed overnight (~16 h) at 6°C.The thawed samples were shaken 25 times in 7 s in accordance with Standard Methods for the Examination of Dairy Products and 5 mL of the organic raw milk were aseptically transferred into a 10 mL screw cap glass vial (Wehr and Frank, 2004).A temperature control was prepared in the same manner using UHT milk.The samples were lab pasteurized in a water bath at 63°C (±1°C) for 30 min and then promptly placed on ice.Once the temperature of the temperature control reached ≤6°C, a 100 µL aliquot of each sample was dispensed into Petri dishes in duplicate and pour plated with molten SMA tempered to approximately 45°C.The plates were incubated at 32°C for 48 h for LPC, enumerated using an automated counter (Q-count, Advanced Instruments, Norwood, MA), and each colony was marked as either a surface or subsurface colony.Bacterial colonies were selected and isolated based on the following schemes: (1) a representative of all unique surface colony morphologies (typically 1-5 isolates) were selected and (2) up to a total of 5 distinct subsurface colonies were selected for each sample.Each of the selected colonies were streaked for isolation onto Brain Heart Infusion (BHI) agar using sterile loops and the plates were subsequently incubated at 32°C for 24 h.Isolated colonies were inoculated into BHI broth and incubated at 32°C for 24 h and stored at −80°C in 15% glycerol until further characterization.

Bacterial identification
Identification of thermoduric bacterial isolates was performed using PCR amplification of the 16S rRNA gene using the forward primer 16S-PEU7 (Rothman et al., 2002) and the reverse primer16S-DG74 (Greisen et al., 1994).Sanger sequencing was performed at Cornell University's Biotechnology Resource Center.Forward and reverse sequences were generated and subsequently trimmed and aligned using Sequencher (Gene Codes Corporation, Ann Arbor, MI USA).Assembled sequences were compared with sequences in the Ribosomal Database Project (RDP) sequence database (Center for Microbial Ecology, Michigan State University, East Lansing, MI) to assign taxonomy at the highest resolution possible.If ≥7 of the 10 top hits were the same genus with a percent identity >95%, then a genus was assigned.If an isolate could not be assigned a genus, then a less specific taxonomic level was assigned using the same method.While RDP was no longer active, the most up-to-date version downloaded from May 25, 2022 was used for this study.Going forward this version can still be used and the original command line tools are still available to install and use at https: / / jfq3 .gitbook.io/rdptools -docker/ rdptools -docker/ readme.
Isolates were also sent to Cornell Quality Milk Production Services for identification by matrix-assisted laser desorption/ionization-time of flight mass spectrometry (MALDI-TOF MS) using the Autobio Autof ms1000 (Autobio Diagnostics, Zhengzhou, China) per manufacturer instructions.There were 3 sample preparation methods used (i) the direct coating method, (ii) the extended method, and (iii) the extraction method.The direct coating method was used for most identifications, however for the isolates that could not be identified with this sample preparation method, the extended method or extraction method was applied.In the direct coating method, a small amount of an isolated colony was coated in a 1 µL matrix solution (solution containing α-cyano-4-hydroxy-cinnamic acid) on the target plate.Once the sample mixture dried at room temperature, the target plate was placed into the instrument for testing (Park et al., 2021).In the extended method, a small amount of a bacterial colony was placed onto the target plate and 1 µL of a lysis solution (labeled as lysate 1 according to the manufacturer), a solution containing formic acid, was applied to aid in protein release.The mixture was dried and the 1 µL matrix solution was added then dried again and the target plate was placed into the instrument for analysis.In the extraction method, single colonies were mixed with 300 µL of deionized water and 900 µL of anhydrous ethanol in a 1.5 mL centrifuge tubes (Park et al., 2021).This mixture was centrifuged at room temperature for 2-4 min at 8,000-14,800 rpm and the supernatant was decanted.After the supernatant was dried for 2-5 min at 37-40°C, 10 µL of both lysis solutions (labeled as lysate 1 and lysate 2 according to the manufacturer), Lee et al.: Troubleshooting high laboratory… which are solutions containing formic acid and acetonitrile, respectively, were added and mixed with the supernatant.This mixture was centrifuged, as detailed above, and then 1 µL of supernatant was deposited on a target slide followed by drying and adding 1 µL of the matrix solution; after another drying step the sample was placed in the instrument for testing.After the sample preparation, the mass spectra analyses were performed by the Autobio Autof ms1000 and resultant mass spectra were compared with the reference mass spectra in the database (Park et al., 2021).Identification scores range on a scale of 0 to 10 and resulting IDs were considered unreliable with scores between 0 -5.999 and reported here as "No Reliable ID," reliable at the genus level with scores between 6.000 -8.999, and reliable to the species level with scores between 9.000 -10.

Data management and statistical analysis
Data were stored in a Microsoft Access database (Microsoft Access, Redmond, WA).LPC were logtransformed before the analysis and plot generation.The data manipulation was performed using the tidyverse package (Wickham et al., 2019) in R. To calculate the geometric mean, any LPC concentrations that were below detection (<0.70 log 10 cfu/mL) were replaced with a value that was 25% of the detection limit (i.e., samples with a LPC of <5 cfu/mL were replaced with 1.25 cfu/mL or the equivalent log-transformed value for analyses).
Since only a proportion of the colonies were selected and isolated for each sample, the bacterial population for each sample was estimated by adjusting the selected surface and subsurface by isolate counts separately based on the proportion of each taxonomic group and total colony count using the following formula: P = x surface y surface + x subsurface y subsurface , where p is the adjusted proportion of a taxonomic group within a sample, x surface and x subsurface are the proportions of total colonies from a sample that are surface and subsurface, respectively, and y surface and y subsurface are the proportions of isolated surface and subsurface colonies, respectively, that represent the taxonomic group.
Figures were generated with the ggplot2 package (Wickham, 2016) and descriptive statistics were performed with the car package (Fox and Weisberg, 2019) in R. To determine if there was a significant difference between LPC in matched fresh and frozen raw milk samples in our pilot study, we performed a paired Wilcoxon Signed-Rank test using the car package in R (Fox and Weisberg, 2019).To test whether LPC was a predictor variable of proportion of GPS found in a sample, the data was modeled using a logit linear model per-formed with the car package in R. To ensure normality, a value of 0.003 was added to both the numerator and denominator of the logit function so that the 0 and 1 value do not transform to ∞ and -∞ before applying a logit-transformation (Warton and Hui, 2011).All data and codes used to perform analysis can be found at https: / / github .com/FSL -MQIP/ LPC.

Laboratory pasteurization count is not significantly different between fresh and frozen raw milk
To evaluate the impact of freezing milk before LPC analysis, we conducted a pilot study on 15 raw milk samples.In the 15 raw milk samples, the LPC ranged from below detection (<0.70 log 10 cfu/mL) to 2.7 log 10 cfu/mL with a geometric mean of 1.81 log 10 cfu/mL for the fresh samples and ranged from below detection (<0.70 log 10 cfu/mL) to 2.7 log 10 cfu/mL with a geometric mean of 1.86 log 10 cfu/mL for the frozen samples (Figure 1).Using a Wilcoxon Signed-Rank test, it was determined that the difference in LPC between the fresh and frozen samples was not significant (P > 0.05).

Bacterial populations recovered from Laboratory Pasteurization Count plates are predominantly Gram-Positive sporeformers and cocci
Of the organic raw milk samples collected from the 94 US organic dairy farms analyzed in the study, the LPC concentrations ranged from below the detection limit (<0.70 log 10 cfu/mL) to 4.07 log 10 cfu/mL, with 9 samples that had LPC below the detection limit (Figure 3).The geometric mean was 1.48 log 10 cfu/mL with a standard deviation of 0.70 log 10 cfu/mL and a median of 1.54 log 10 cfu/mL.
A total of 382 colonies were isolated from the LPC plates, however 2 of the isolates were excluded from all analyses because (i) one could not be recovered from the frozen stock and (ii) one was a fungal contaminant.In 3 of the samples that had LPC concentrations above the detection limit, no colonies could be isolated, therefore, microbiological analyses were conducted on 380 isolates that came from 82 organic raw milk samples.Based on 16S rDNA sequencing data, 343, 35, and 2 isolates were assigned a genus, family, and order, respectively.Isolates identified at the genus level included Bacillus (n = 168), Brachybacterium (n = 56), Kocuria (n = 47), Streptococcus (n = 30), Microbacterium (n = 20), Paenibacillus (n = 13), Staphylococcus (n = 3), Enterococcus (n = 2), Alkalihalobacillus (n = 1), Lysinibacillus (n = 1), Micrococcus (n = 1), and Neisseria (n = 1).Of the 35 isolates characterized at the family level, 16 were Bacillaceae, 10 were Micrococcaceae, 8 were Moraxellaceae, and 1 was Paenibacillaceae.The 2 isolates that could only be assigned to an order represented Bacillales.Bacillus was the predominant genus identified; 56/82 organic raw milk samples yielded at least 1 isolate classified as Bacillus.
Overall, the thermoduric bacteria isolated from the organic raw milk samples evaluated here can be categorized into 3 groups: GPS, Gram-positive nonsporeformers, and Gram-negative bacteria.the proportion of surface and subsurface colonies from each sample as described above, the proportions of each group were similar, with mean percentage of GPS, Gram-positive non-sporeformers, and Gram-negative bacteria, representing 50.2%, 46.8%, and 1.79%, respectively.The 1.79% representing Gram-negative bacteria is driven almost entirely by one sample.The proportion of GPS, Gram-positive non-sporeformers, and Gramnegatives varied considerably by sample.There were samples consisting of only GPS as well as only Grampositive non-sporeformers (Figure 4).
In relation to generally accepted industry thresholds for LPC, out of the 82 samples that had characterized isolates there were 66 samples with a LPC ≤100 cfu/ mL, 6 samples with a LPC between 100 and 200 cfu/ mL, and 10 samples with a LPC ≥200 cfu/mL.Of the participating farms, a large subset represented below the 100 cfu/mL threshold and not in the higher range, which presents a potential limitation when analyzing the data.The average proportion of sporeformers estimated based on the isolates characterized for each of these categories (i.e., LPC <100 cfu/mL, LPC between 100 and 200 cfu/mL, and LPC ≥200 cfu/mL) was 51%, 67%, and 35%, respectively.A linear logit model was applied and showed that although there was an inverse trend between the proportion of GPS and the LPC log 10 concentration (i.e., samples with higher LPC showed lower proportions of GPS), it was not significant (P = 0.218).
Finally, we assessed the proportion of groups of thermoduric bacteria by their recovery from surface or subsurface of LPC pour plates, as some thermoduric bacteria are known to be microaerophilic and plating method may impact their recovery after heat treatment.The majority of isolates were subsurface (78%) and the others were surface (22%).Of the most abundant thermoduric genera represented in our study (i.e., Bacillus) approximately 32% (54/168) of the isolates were selected from the surface of LPC plates (Figure 5).In fact, Bacillus and other GPS were the predominant thermoduric bacterial genera isolated from the surface of LPC plates -representing ~76% (64/84) of the surface isolates as compared with ~46% (136/296) of subsurface isolates.Conversely, the top 4 most abundant non-sporeforming Gram-positive thermoduric genera, Brachybacterium, Kocuria, Streptococcus, and Microbacterium were nearly all isolated from the subsurface of LPC plates, with 93%, 83%, 90% and 95% of these genera isolated from the subsurface of LPC plates, respectively (Figure 5).

MALDI-TOF MS correctly identifies approximately 66% of bacterial isolates from LPC tests when compared with 16S rDNA identifications
To compare the ability of MALDI-TOF MS to identify thermoduric bacteria from organic raw milk with 16S rDNA sequencing, considered for the purposes of this study to be the gold standard, we evaluated all 380 isolates using this method.Overall, 66.6% (253/380) of the isolates had agreement between the genus identifications from 16S and MALDI-TOF MS.Of the remaining 127 isolates, 92 of isolates could not be assigned a reliable identification using MALDI-TOF MS, and 35 of isolates were mischaracterized by MALDI-TOF MS at the genus level and using a reliability score cutoff of 6.0.

DISCUSSION
Despite LPC data being used as a raw milk quality parameter to designate premiums and identify potential equipment cleaning and sanitation deficits on the farm, there is a gap in knowledge of the thermoduric bacterial populations present in contemporary organic raw milk supplies.Therefore, we tested organic raw milk from certified organic US dairy farms to measure the LPC and characterize the bacterial isolates present.The collected data provides insight into the predominant thermoduric organisms found in organic raw milk, namely we found that 52.6% were Gram-positive sporeforming bacteria, 44.5% were other Gram-positive bacteria, 2.4% were Gram-negative bacteria, and 0.5% could not be categorized into these groups due to identification at the order level.Overall, our data suggest that there The LPC concentrations that were below detection (<5 cfu/mL) were replaced with a value that was 25% of the detection limit, 1.25 cfu/mL, and then log-transformed. is a no significant relationship between the proportion of GPS and the LPC level, which further indicates that when troubleshooting high LPC levels in organic raw milk, identification of the predominant thermoduric population should first be established.To that end we further compared identification of thermoduric bacteria originating from organic raw milk using MALDI-TOF MS to those obtained through 16S rDNA sequencing, we found agreement at the genus level between the 2 microbiological identification tools for approximately 2 thirds of isolates tested.Overall, our study provides contemporary guidance to dairy producers and other community members regarding the levels, populations, and identification of thermoduric bacteria in organic raw milk.

Temporarily freezing milk samples has no considerable effect on LPC levels and the resultant bacterial populations
To transport and test samples that were collected across the US samples were frozen and shipped.We first performed a pilot study to test whether freezing raw milk samples for shipping and storing them temporarily (e.g., up to 7 d) at −20°C had an effect on LPC concentration and thermoduric bacteria populations.No significant difference in the LPC concentration was found, and to our knowledge this is the first study to report on the impact of temporary freezing on thermoduric bacteria in raw milk.However, several other studies have investigated the impact of freezing on other important groups of raw milk bacteria.For example, one study tested the effect of freezing bulk tank raw milk samples at −20°C, −78°C, and −196°C on SPC for various amounts of time up to 28 d and found that all temperatures caused a significant reduction in SPC by d 3 (Read et al., 1969).Schukken et al. (1989) tested the effect of freezing on mastitis pathogens and observed a decrease in number of samples with E. coli isolates and no significant effect on Staphylococcus, which was consistent with the findings from Alrabadi (2015) in which no significant reduction in Staphylococcus spp. was observed after freezing at −20°C up to 8 weeks.However, Bashandy and Heider (1979)  ).The samples are organized in ascending order from lowest LPC (0.70 log 10 cfu/mL) to highest LPC (4.07 log 10 cfu/mL).The 2 isolates that were identified in the Bacillales order did not have a reliable identification at a level specific enough to be represented by one of the 3 groups: Gram-positive sporeformers, Gram-positive non-sporeformers, and Gram-negative, and therefore is displayed as 'No Reliable ID.' reduction in Streptococcus isolates after freezing milk samples at −80°C for 7 and 14 d.Our results suggest that raw milk samples can be frozen temporarily with no significant impact on the level or diversity of thermoduric bacteria.

There is large diversity of thermoduric bacteria present in organic raw milk
Thermoduric bacteria levels in organic raw milk from this study as measured by the LPC method ranged from <0.70 to 4.07 log 10 cfu/mL with a mean of 1.48 log 10 cfu/mL and a median of 1.54 log 10 cfu/mL.Previous studies have reported similar LPC levels.For example, thermoduric bacteria in raw milk from dairy farms in the Midwest US were reported to be 1.71 log 10 cfu/ mL and 1.85 log 10 cfu/mL in the winter and summer, respectively (Buehner et al., 2014).O'Connell et al. ( 2016) conducted a study and reported LPC concentration in bulk tank milk, which ranged from <1 log 10 cfu/mL to 2.68 log 10 cfu/mL with a mean of 0.90 ± 0.47 log 10 cfu/mL (during lactation period 1).Ribeiro Júnior et al. (2018) evaluated thermoduric bacteria in raw milk from 20 Brazil dairy farms and found that the Stacked bar plot of the proportions of characterized surface and subsurface colonies by genus identifications Within each genus, the proportion of surface and subsurface isolates is shown using the red and blue bars, respectively.Isolates that could not be identified at the genus level with a ≥99% identity were identified at the family level and represented by the "Other Bacillaceae," "Other Micrococcaceae," "Other Paenibacillaceae.," and "Family: Moraxellaceae."Two isolates could not be reliably characterized at the genus or family levels and were represented by the order Bacillales.thermoduric count ranged from 0.70 log 10 cfu/mL (5 cfu/mL) to 3.28 log 10 cfu/mL (1.9 × 10 3 cfu/mL) with a mean of 2.51 log 10 cfu/mL (3.2 × 10 2 cfu/mL).The results of our study coincide with those existing literature that most producers fall within the acceptable range for LPC, however there are still a small number of producers that fall above the acceptable industry threshold (≥200 cfu/mL) that may need additional troubleshooting to reduce their thermoduric bacteria load.
The characterization of isolates from this study using 16S rDNA sequencing revealed a great diversity of thermoduric bacteria from organic raw milk.Overall, from the 82 samples of laboratory pasteurized organic raw milk with one or more bacterial isolates, we identified 12 unique genera belonging to 10 bacterial families which were divided into 2 main classifications, GPS and Gram-positive non-sporeforming bacteria.The primary thermoduric genera isolated in our study were identified as Bacillus (44.2%),Brachybacterium (14.7%),Kocuria (12.4%), and Streptococcus (7.9%).Our findings were similar to other studies that have identified thermoduric bacteria in raw milk.For example, Ribeiro Júnior et al. (2018) found the following 8 thermoduric genera in Brazilian bulk tank milk: Bacillus (44.65%),Brachybacterium (14.40%),Streptococcus (14.35%),Enterococcus (9.10%), Kocuria (6.95%), Paenibacillus (5.25%), Micrococcus (4.55%), and Macrococcus (0.75%).Similarly in our study, the isolates found were predominantly Bacillus, followed by Brachybacterium, Kocuria, Streptococcus, Microbacterium, and Paenibacillus.Another study reported the predominant bacteria in laboratory pasteurized raw milk collected in Hokkaido, Japan were represented by Bacillus (30.7%)Microbacterium (28.0%) and Micrococcaceae (17.4%) (Kikuchi et al., 1996).Thus, Bacillus was consistently the predominant genus isolated from laboratory pasteurized milk, however the prevalence of the other predominant genera varied by study.In the study done by Kikuchi et al. (1996), Microbacterium was one of the predominant genera represented in those samples at 33.6% compared with the 5.2% of isolates identified as such in this study.
Kocuria was found at higher levels in our study as compared with Kikuchi et al. (1996), which did not identify Kocuria as a predominant genus.Despite some differences in the prevalence of certain genera found in our study compared with others, which may be attributed to farm or location differences, the overall thermoduric genera represented were very similar.
Finally, the recovery of the thermoduric bacterial populations from the LPC test used here varied by location in the agar plate (i.e., surface or subsurface).While Bacillus and other GPS were recovered from both the surface and subsurface of agar plates, the majority of non-sporeforming Gram-positive bacteria (e.g., Brachybacterium and Kocuria) were recovered subsurface.This preference for subsurface growth may be partially because some thermoduric non-sporeforming bacteria found in milk are known to prefer to grow in low oxygen environments (i.e., are microaerophilic) or at least are able to grow in low oxygen environments.For example, members of the Streptococcaceae family and other lactic acid bacteria that are often thermoduric themselves, are microaerophilic (Bhowmik andMarth, 1990, De Vuyst andVandamme, 1994).The 2 major Gram-positive non-sporeforming genera, Brachybacterium and Kocuria, are aerobic, however some species are facultative anaerobes, which use oxygen when present, however are still able to grow without oxygen (Collins et al., 1988, Savini et al., 2010).These results highlight the importance of using a LPC plating method that allows for recovery of bacteria with different free oxygen requirements (e.g., pour plating).

The Laboratory Pasteurization Count is not a proxy for spore count in organic raw milk
In our study, approximately 53% of the isolated bacteria were GPS and 45% were Gram-positive nonsporeformers, however, from sample to sample there was considerable variation in the proportion of GPS to Gram-positive non-sporeformers.The proportion of GPS found when enumerating thermoduric bacteria is similar to a previous study (Ribeiro Júnior et al., 2018), where the authors found that sporeforming bacteria, specifically Bacillus and Paenibacillus, made up 49.9% of the thermoduric bacteria isolated from the milk samples.A different study found lower percentage (~30%) of GPS, which were predominantly Bacillus and the majority (~69%) were non-sporeforming Gram-positives (Kikuchi et al., 1996).Oftentimes LPC is used as an indicator for spore contamination (Martin et al., 2023), however our data does not support the use of LPC as a proxy for spore level.Overall, our data supported no significant trend between the proportion of GPS and LPC.There is no regulatory limit for LPC, but ≤100 cfu/mL, between 100 and 200 cfu/mL, and ≥200 cfu/mL are commonly used cutoffs to represent low, medium, and high LPC (Jayarao et al., 2004).Milk with LPC >200 cfu/mL, or >2.3 log 10 cfu/mL, has been proposed to indicate inadequate hygiene and sanitation of milking equipment on the farm.This present study found that about 12% of the samples with characterized isolates had LPC ≥200 cfu/mL, which is similar to the 10% of raw milk samples in Saudi Arabia with LPC >200 cfu/mL that Elmoslemany et al. (2016) found.
Our data demonstrate that identification should be included when troubleshooting high LPC as the subpopulations cannot be assessed by the LPC level alone, and the troubleshooting approach will likely vary by the group that is driving the high levels.Elevated GPS levels are typically associated with certain farm factors including housing/bedding practices, milking practices, cow-level factors, and feed factors (Evanowski et al., 2020, Martin et al., 2023).While no studies have specifically investigated farm practices associated with elevated levels of Gram-positive non-sporeforming thermoduric bacteria in bulk tank raw milk, past studies have shown that farm practices, including poor equipment hygiene and improper sanitation, are associated with high LPC levels (Kikuchi et al., 1996, Elmoslemany et al., 2010).Facilitating successful troubleshooting of high LPC levels will likely require characterization of the primary groups of thermoduric bacteria present in the raw milk, and future research is needed to understand where likely source locations of specific thermoduric bacteria may be found in the dairy system.

MALDI-TOF MS can serve as a microbiological identification tool for thermoduric bacteria with further optimization
As our data does not support the use of LPC level alone as an indicator of specific sub-groups of thermoduric bacteria (e.g., sporeformers) in organic raw milk, we evaluated the use of a commonly applied identification tool, MALDI-TOF MS, to provide rapid characterization of thermoduric bacterial populations present in organic raw milk.We chose MALDI-TOF MS as Lee et al.: Troubleshooting high laboratory… it is cost-effective and provides relatively rapid results (e.g., within 1h of when an isolated colony is available) and has been used somewhat widely in the dairy industry to identify mastitis pathogens (Kliem andSauer, 2012, Cameron et al., 2018).MALDI-TOF MS matches peptide mass fingerprints, or mass spectra, of a given sample against the spectra supplied in a given database (Singhal et al., 2015).We found that MALDI-TOF MS correctly identified 66.6% of thermoduric bacterial isolates in this study, could not provide a reliable identification for 24.2% of isolates, and misidentified 9.2% of isolates.The success of MALDI-TOF MS as a bacterial identification tool is dependent upon the database of mass spectra that is used (Seuylemezian et al., 2018).The majority of isolates that were not correctly identified using MALDI-TOF MS was due to a low reliability score, likely due to the lack of thermoduric organisms present in the database.This identification tool has been used more frequently within the dairy industry as a diagnostic tool of bovine mastitis causing organisms, such as Staphylococcus aureus, various Streptococcus species, and Escherichia coli, which has helped build a strong database of peptide mass fingerprints of mastitis-causing organisms (Barreiro et al., 2017, Ali et al., 2021).Despite some overlap between genera containing both thermoduric bacteria and mastitis organisms (i.e., Staphylococcus and Streptococcus), thermoduric bacterial populations primarily consist of Bacillus, Brachybacterium, and Kocuria, which have not been linked to mastitis.Published literature evaluating the agreement between bacterial identifications via 16S rDNA sequencing and MALDI-TOF MS have primarily used Staphylococcus or Escherichia isolates, which may explain why some of the thermoduric bacteria found in this study were not able to be assigned a reliable identification using this method (Wilson et al., 2019).Previous studies indicated that using MALDI-TOF MS can allow for highly accurate bacterial identifications.For example, Cameron et al. (2018) found that coagulase-negative staphylococci were correctly identified with 99.2% (824/831) accuracy using MALDI-TOF MS.In another study investigating mastitis pathogens from milk samples, 7.5% of the isolates obtained either no reliable identification (3.3%) or misidentification at the genus or species level (4.2%) as compared with 16s rDNA sequencing (Jahan et al., 2021).The database for GPS and Gram-positive cocci is insufficient as shown in our study by the number of isolates with no reliable identification using MALDI-TOF MS and the lack of agreement between 16S and MALDI-TOF MS bacterial identifications, indicating that the reliability of MALDI-TOF MS as a microbial identification tool is limited by the reference database (Nonnemann et al., 2019).
The discrepancies in the identification of organisms may also be attributed to the sample preparation method of bacterial colonies for MALDI-TOF MS analysis.The isolates in the present study were prepared primarily using the direct coating method, but if the bacteria could not be identified, the extended and extraction preparation methods were used.A previous study reported that identifying Gram-negative bacteria isolated from various animal sources using MALDI-TOF showed good results using a direct coating method, which was the main method used in the present study (Randall et al., 2015) and found that identifications and reliability scores were significantly better when using an extended method (Randall et al., 2015).
Several studies have also found that the extraction method performed better than the direct coating and extended methods and provided accurate identifications to the species level for various microorganisms, so the mischaracterizations in our study may also be in part explained by the preparation method (Bizzini et al., 2010, Schulthess et al., 2020, Wang et al., 2021).Further improvements and validation of MALDI-TOF MS procedures and databases are needed before it could be considered as a tool to aid in troubleshooting high LPC by providing rapid identification of the groups of thermoduric bacteria driving the LPC outcome, but improvements should be made to increase the number of thermoduric bacteria correctly identified by this method.

CONCLUSION
This study expands the knowledge of thermoduric microbial populations in organic raw milk enumerated by the LPC method.Our study highlights that the use of LPC testing is not appropriate as an indication of the presence of a sub-group of thermoduric bacteria present in organic raw milk, namely sporeforming bacteria.Dairy industry community members using the LPC as an indicator of raw milk quality should use interpret the data as an enumeration technique for thermoduric bacteria, which consists of both sporeforming and nonsporeforming bacteria.Any further troubleshooting that may be needed based on LPC results will depend on the type of thermoduric bacteria that is found in milk supplies.Further development of identification methods for thermoduric bacteria that are rapid and accessible will be critical to support efforts by community members to produce high quality raw milk.The findings of this study are focused on the applications for organic milk, however a similar study to develop a current baseline characterizations of thermoduric bacteria in conventional raw milk is warranted.
Lee et al.: Troubleshooting high laboratory…

Figure 1 .
Figure 1.Box plot of the lab pasteurization counts between fresh (n = 15) and frozen (n = 15) raw milk samples.The interquartile range (IQR) is represented with the boxes with the first quartile (25th percentile) at the bottom and third quartile (75th percentile) at the upper end.The black line within the box represents the median and the red line represents the geometric mean.The whiskers indicate the variability, and the length of the whiskers are 1.5 times the IQR.The outliers, which do not fall within 1.5 times the IQR, are represented by dots.

Figure 2 .
Figure 2. Stacked bar plot displaying proportions of each genus isolated from the fresh and frozen raw milk samples.Isolates from fresh samples (n = 89) and frozen samples (n = 90) were identified using 16S rDNA sequences.

Figure 3 .
Figure 3. Histogram of the distribution of lab pasteurization counts from all samples collected (n = 94).The LPC concentrations that were below detection (<5 cfu/mL) were replaced with a value that was 25% of the detection limit, 1.25 cfu/mL, and then log-transformed.

Figure 4 .
Figure 4. Stacked histogram showing the relative proportions of Gram-positive sporeformers, Gram-positive non-sporeformers, and Gramnegative bacteria in the organic raw milk samples with characterized isolates (n = 82).The samples are organized in ascending order from lowest LPC (0.70 log 10 cfu/mL) to highest LPC (4.07 log 10 cfu/mL).The 2 isolates that were identified in the Bacillales order did not have a reliable identification at a level specific enough to be represented by one of the 3 groups: Gram-positive sporeformers, Gram-positive non-sporeformers, and Gram-negative, and therefore is displayed as 'No Reliable ID.' Figure5.Stacked bar plot of the proportions of characterized surface and subsurface colonies by genus identifications Within each genus, the proportion of surface and subsurface isolates is shown using the red and blue bars, respectively.Isolates that could not be identified at the genus level with a ≥99% identity were identified at the family level and represented by the "Other Bacillaceae," "Other Micrococcaceae," "Other Paenibacillaceae.," and "Family: Moraxellaceae."Two isolates could not be reliably characterized at the genus or family levels and were represented by the order Bacillales.
Lee et al.: Troubleshooting high laboratory…

Table 1 .
Lee et al.:Troubleshooting high laboratory… Comparison of two bacterial identification methods, matrix-assisted laser desorption ionization time of flight mass spectrometry (MALDI-TOF MS) and 16S rDNA sequencing, the gold standard, for 380 thermoduric bacterial isolates collected from 94 certified organic bulk tank raw milk samples