A multifaceted investigation of lactococcal strain diversity in undefined mesophilic starter cultures

ABSTRACT The dairy fermentation industry relies on the activity of lactic acid bacteria in robust starter cultures to accomplish milk acidification. Maintenance of the composition of these starter cultures, whether defined or undefined, is essential to ensure consistent and high-quality fermentation end products. To date, limited information exists regarding the microbial composition of undefined starter culture systems. Here, we describe a culture-based analysis combined with a metagenomics approach to evaluate the composition of two undefined mesophilic starter cultures. In addition, we describe a qPCR-based genotype detection assay, which is capable of discerning nine distinct lactococcal genotypes to characterize these undefined starter cultures, and which can be applied to monitor compositional changes in an undefined starter culture during a fermentation. IMPORTANCE This study reports on the development of a combined culture-based analysis and metagenomics approach to evaluate the composition of two undefined mesophilic starter cultures. In addition, a novel qPCR-based genotype detection assay, capable of discerning nine distinct lactococcal genotypes (based on lactococcal cell wall polysaccharide biosynthesis gene clusters), was used to monitor compositional changes in an undefined starter culture following phage attack. These analytical approaches facilitate a multifaceted assessment of starter culture compositional stability during milk fermentation, which has become an important QC aspect due to the increasing demand for consistent and high-quality dairy products.

undefined starter culture systems, particularly with respect to bacteriophage suscepti bility and robustness (4).Phage attacks can have a significant negative impact on starter cultures resulting in a delayed or even failed fermentation.Undefined starter cultures are typically considered to be more phage-robust when compared to defined starter cultures due to the superior cultural diversity of the former (5).Since the specific composition and relevant technical characteristics of the strains applied in defined starter cultures are established, strains can be more readily replaced with phage-unre lated/insensitive strains without changing the (bio)technological integrity of the culture.Conversely, due to the unknown composition of undefined starter cultures, it is often difficult to mitigate the effects of phage infections or even to determine the impact of phages on the fermentation and ripening process until the final product is graded for its sensory attributes (6).
Lactococcal phages impose serious economic losses on the dairy industry and are therefore among the best studied phage-host systems.Most notably, the phage-host interactions of skunaviruses (formerly termed the 936 group phages) and (to a lesser extent) ceduoviruses (formerly termed the c2 group phages) have been extensively investigated since they are the most problematic virulent lactococcal phages (7)(8)(9)(10).Phage infection typically commences with reversible binding of a phage's tail tip-asso ciated anti-receptor to a receptor present on the host cell surface (11).The phage receptor binding protein (RBP) may recognize and bind to saccharidic, teichoic acid, or proteinaceous receptors (12).The RBPs of skunaviruses recognize and bind irreversibly to a saccharidic receptor on lactococcal cell surfaces, that is, cell wall polysaccharide (CWPS) structures (13).The cwps gene cluster is responsible for the biosynthesis of a given lactococcal CWPS, which is composed of two elements: the peptidoglycan-embed ded rhamnan and surface-exposed polysaccharide pellicle (PSP).There are currently four defined cwps genotypes (A-D) and several C-subtypes (C 1 -C 8 ) recognized among lactococcal strains (14,15).A recent review, of over 400 lactococcal strains, suggests that additional cwps genotypes are likely to exist, indicating the continually evolving genetic composition of this gene cluster among lactococci (16).Recently, RBP phylogeny of skunaviruses has been shown to correspond with lactococcal cwps genotypes, which, in turn, is linked to the biochemical diversity of the associated surface-exposed PSP (i.e., the phage receptor) and thus to host range, demonstrating the industrial relevance of this gene cluster (14)(15)(16)(17)(18).
Starter culture rotation is a common strategy to counteract the effects of phage attacks during various dairy fermentations, by rotating starter cultures that are capable of generating products with reproducible organoleptic properties but have different phage sensitivity profiles.However, to implement an effective culture rotation strategy, it is important to understand both the microbial diversity of the starter cultures, as well as the diversity of phages present in the fermentation environment (19).In the case of undefined mixed starter cultures, typically very little is known about the microbial diversity and complex community dynamics; therefore, this lack of knowledge may make it difficult to choose which starter culture to rotate in after disruptions due to phages (19,20).In the current study, a combination of approaches was used to characterize two undefined mesophilic mixed starter cultures, including culture-based and metagenome analysis combined with the establishment of a qPCR assay (based on lactococcal cwps gene clusters) capable of discerning nine industrially relevant lactococcal genotypes.The metagenome, culture-based, and qPCR-based analyses were successfully used to not only discern the diversity of lactococci within starter cultures but also to monitor changes in the strain composition of an undefined culture in the presence of phages.

Lactococcal strains and bacteriophages
Bacterial strains and phages used in this study are listed in Table S1.Bacterial strains were grown overnight at 30°C in M17 broth (Oxoid, Hampshire, United Kingdom) supplemen ted with 0.5% lactose (Sigma Aldrich, Gillingham, UK) (LM17).Phages were propagated (initially from single plaques, then using 100 µL of phage lysates after initial plaque propagation) in the same medium supplemented with 10 mM CaCl 2 (Sigma Aldrich) and added at the point of host strain inoculation (100 µL of fresh overnight culture).The mixture was incubated overnight at 30°C and resultant phage lysates were then filtered (0.45 µM) and stored at 4°C.

Metagenome DNA extraction
The microbial diversity of two undefined industrial starter cultures, M1 and M2, commonly used in the production of semi-hard cheeses, fermented milk (e.g., viili, kefir, and twaróg), and fresh cheeses (e.g., quark) was investigated.The cultures were activated by first diluting 10 g of lyophilized commercial starter culture in 207 g of sterilized (115°C for 20 minutes) 10% (wt/vol) reconstituted skimmed milk (RSM) powder.Following stomaching, in a Stomacher 400 Circulator (Seward, UK), for 2 minutes at 230 rpm, 20 g of the first dilution was added to another 207 g of sterilized 10% RSM.The second dilution was incubated overnight at 22°C.The following day, 100 mL of pasteurized (90°C for 20 minutes) 10% RSM was inoculated with the overnight culture (at 3%) and incubated overnight again at the same temperature to fully activate the culture.
Metagenomic DNA of the activated starter cultures was extracted following an adapted protocol described previously by Erkus, et al., described as follows: 2.2% trisodium citrate (Sigma Aldrich) solution (heated to 45°C) was added to 2 g of activa ted starter culture to reach a total volume of 25 mL (20).The resulting solution was homogenized in a Stomacher 400 Circulator (Seward) for 5 minutes at 230 rpm.The homogenized solution was centrifuged at 13,800 × g for 10 minutes at room tempera ture.The supernatant and fat layers were removed, and the cell pellet was resuspended in 1 mL of 2% sodium citrate solution (45°C) for washing.The resuspended solution was centrifuged at 10,800 × g for 5 minutes.The supernatant and remaining fat layer were removed and this washing step was repeated 2-3 more times (until full removal of the fat layer).The washed cell pellet was resuspended in 1 mL of lysis buffer [20 mM Tris-HCl buffer (pH = 8), 2 mM EDTA (Sigma Aldrich), 2% polyethylene glycol (Sigma Aldrich)] with the addition of 50 µg/mL of lysozyme (Sigma Aldrich) and 100 U of mutanolysin (Sigma Aldrich).This cell suspension was incubated at 37℃ for 3 h.250 µg/mL of proteinase K (Merck, NJ, USA) was then added and the solution was incubated at 56°C for 1 h. 1 mL of 96% ethanol was added to the lysed solution and mixed gently.The precipitated DNA was further washed and purified using the Qiagen DNeasy Blood and Tissue kit (Qiagen, Hilden, Germany) according to the manufacturer's protocol and eluted in a final volume of 300 µL.DNA concentrations were measured using the Qubit dsDNA BR assay kit with a Qubit 2.0 fluorometer (Invitrogen, Carlsbad, CA; Table S2).

Illumina shotgun metagenome sequencing
Metagenomic DNA preparations from starters M1 and M2, following activation, were sequenced by Macrogen Europe (Amsterdam, The Netherlands) using the NovaSeq 6000 Illumina platform (Illumina, USA), generating paired-end reads with an average read length of 151 bp.The library was constructed using the TruSeq DNA PCR-Free kit and according to the TruSeq DNA PCR-Free Sample Preparation Guide, Part #15036187 Rev. D (Illumina, USA).Quality control (QC) was performed prior to library construction.The raw metagenomic reads of M1 and M2 were deposited in GenBank under Biosample accession numbers, SAMN37487907 (BioProject no.PRJNA1019656) and SAMN37493201 (BioProject no.PRJNA1019710), respectively.

Bioinformatic analysis of metagenomic sequences
Fastq files of paired-end reads were used as the input for genome assemblies using MEGAHIT (version v1.2.9).Assembled contigs were first aligned against the non-redun dant (nr) database maintained by NCBI (https://ncbi.nlm.nih.gov), which includes most sequences available in RefSeq and GenBank, using BLASTn (version 2.11.0+; https:// blast.ncbi.nlm.nih.gov/Blast.cgi) to taxonomically classify each contig using BLAST-based last common ancestor (LCA) taxonomy estimation (21).Reads were aligned against their corresponding assemblies to quantify the number of reads that mapped to each contig using BWA (version 0.7.17-r1188).The relative abundance of the genus and/or species present in the starter cultures was then calculated by dividing the total number of reads mapped to all metagenome assemble contigs by the number of reads mapped to all contigs classified as a particular species or genus.

Strain isolation and CWPS typing
Serial dilutions (prepared in Ringers' solution) of the undefined starter cultures M1 and M2 were directly (i.e., without activation, after initial dilution in RSM) plated on LM17 agar plates and incubated at 30°C for 48 h.After incubation, 181 isolates from M1 and 183 from M2 were picked randomly and incubated in LM17 broth-containing polystyrene 96-well plates (Sarstedt, Nümbrecht, Germany) at 30°C for a further 48 h.Individual colonies were then plated on LM17 agar and after 24 h at 30°C, colonies were typed using colony PCR and previously described cwps genotyping (CWPS types A, B, and C, including C-subtypes C 1 through to C 5 ) multiplex primers (15,17).In addition, the following primer pairs specific to C 6 -type strains were designed: C6-fw 5′-ATATTGGTC AAGGCAAAGCT-3′ and C6-rv 5′-TCAAACCAGTTCGTAAATCCT-3′.Primers were based on llh_1295 (L.cremoris A76, CWPS type C 6 ) with a PCR product of 517 bp.The same PCR conditions were used as previously described (15).16S rRNA gene amplification was performed on any isolates that did not yield a CWPS type, as described previously (24).Sanger sequencing of 16S rRNA PCR products was performed by GENEWIZ (Azenta Life Science, Germany).

Phage screening and phage isolation from whey
A collection of 92 isolates representing the four dominant cwps types (23 isolates of each of the C 1 -C 4 types; Table S3) isolated from two undefined starter cultures (M1 and M2; without activation) were selected for use in phage screening of whey samples.Three whey samples (T1, BD, and BA6) derived from factories that had utilized either one of these two starter cultures and where the potential presence of phages had been reported, were tested for the presence of phages (information on whey samples is presented in Table S1).In a 96-well plate (Sarstedt), 20 µL overnight culture was added to a well containing 150 µL LM17 broth supplemented with 10 mM CaCl 2 (Sigma Aldrich) and 20 µL of a given whey sample diluted 1:20 in SM buffer [50 mM Tris-HCl, 100 mM NaCl (Sigma Aldrich), 10 mM MgSO 4 (Sigma Aldrich), 10 mM CaCl 2 (Sigma Aldrich)].A second plate using 20 µL SM buffer in place of whey was prepared as a positive control for growth.Plates were incubated at 30°C and the OD 600 nm was measured at three time points (0 h, 3 h, and 6 h), incubated overnight, and visually inspected for possible phage-induced lysis (by comparison to the uninfected control).In wells in which lysis of the cultures was observed (either through OD measurements during the first 6 h or complete lysis following overnight incubation), the potential phage was diluted in SM buffer (1:10), filtered (0.45 µM), and spotted on its corresponding host to validate the presence of phages through double layer plaque assays (25).Phages were purified with one round of plaque purification and typed using a previously published 936 (now Skunavirus), c2 (now Ceduovirus), and P335 PCR-based typing system (26).

Host range analysis
Two representative phages were chosen for further studies, that is, a Skunavirus (BA6-A5) and a Ceduovirus (T1-H4).The host range of two of the isolated phages was determined using the same 92 lactococcal strains used in the original whey-based phage screen ing.Host-range analysis was performed in liquid culture, where high titer lysates (i.e., phage lysates with titers of ~10 7 -10 9 PFU/mL) of the two phages were co-incubated with a given lactococcal strain as described above.To confirm the results of the liquid host-range studies, plaque assays were performed using the double-agar method as previously described (25).

Primer design
Primers were designed to target the CWPS biosynthesis gene cluster genotype/subtypes A, B, C 1 -C 6 , and D and specifically using previously defined reference strains of each of the cwps genotypes mentioned (details of the strains, ORFs, and primers designed are available in Tables S1 and S4) (14).For more information on the design and optimi zation of the primer pairs, see Supplementary Text.All GenBank accession numbers associated with these genomes are listed in Table S1.As representative C 7 and C 8 strains are not available in our strain collection, they were not included in the design of this qPCR assay.To design the primers, the target ORFs (listed above) were submitted to Primer3Plus (https://www.bioinformatics.nl/cgi-bin/primer3plus/primer3plus.cgi) with optimal design settings set to select a primer pair with a melting temperature between 50-60°C (with 55°C being optimal) and a product size between 150 and 250 bp (with 200 bp being optimal) (27).The specificity of each primer set was confirmed using the BLASTn search tool (https://blast.ncbi.nlm.nih.gov/Blast.cgi).A control primer set was included based on the conserved groES gene present in all lactococcal strains (28).The groES sequences of nine representative strains (one of each cwps genotype) were retrieved from NCBI and aligned using ClustalW of the Megalign program (Version 5.01, Lasergene software package, DNASTAR Inc., Madison, WI, USA), to identify a conserved region for optimal primer design.The melting temperature of the groES primer set was determined using the OligoCalc tool, which was also used to check for primer self-complementarity (29).Similarly, the specificity of the primer set was confirmed using the BLASTn search tool.Primer sequences, expected amplicon sizes, and melting temperatures are presented in Table S4.Primers were synthesized by Eurofins Genomics (Ebersberg, Germany).

Bacterial enumeration and DNA extraction for standard curves
Representative lactococcal strains (Table S5) of each cwps genotype (A-D) were grown at 30°C overnight in LM17 broth.Following overnight incubation, cells were enumerated on LM17 agar plates once colonies had grown at 30°C under aerobic conditions to a sufficient size to be counted (24-48 h).Enumerations were performed in triplicate using the spot-plating technique.Concurrently, 2 mL of each of these overnight cultures was removed and centrifuged at 14,000 × g for 10 minutes.The supernatant was removed, and the pellets were stored at −20°C until required for DNA extractions, which was performed using the PureLink Genomic DNA Mini Kit (Invitrogen, UK), according to the manufacturer's protocol.All DNA samples were stored at −20°C until required.DNA concentrations were measured using a NanoDrop 2000 spectrophotometer (Thermo Scientific, Waltham, M.A., U.S.A.) or a Qubit dsDNA BR assay kit with Qubit 2.0 fluorometer (Invitrogen, Carlsbad, CA).qPCR protocol qPCR assays were performed using LightCycler 480 SYBR Green I Master Mix (Roche Diagnostics, Mannheim, Germany), at a 1× final concentration, on a LightCycler 480 Instrument II (Roche Diagnostics).Reactions were carried out in a reaction volume of 15 µL containing 5 µl of template DNA (between 50 and 100 ng per reaction).All primers were used at a final concentration of 625 nM.Thermal cycling conditions incorporated an initial denaturing step at 95°C for 10 minutes followed by 45 cycles at 95°C for 10 s, 55°C for 15 s, and 72°C for 20 s.A melting-curve analysis was performed at the end of each qPCR assay, with one cycle at 95°C for 1 min, 40°C for 1 min, 65°C for 1 s, and then gradually raised to 97°C while continuously measuring the fluorescence.All standards and samples were analyzed in triplicate.

Standard curves
DNA was obtained from representative lactococcal strains of each cwps genotype tested (as described above; the strains used are listed in Table S5) was 10-fold serial diluted (10 0 to 10 −7 ) and used in the qPCR assay.The cycle threshold (C T ) values obtained from the dilutions of DNA were plotted against the CFU/ml of the strains, as determined by viable plate counts of corresponding serial dilutions of each strain.Technical triplicate assays were performed for both the viable plate counts and qPCR analyses.The correlation coefficient (R 2 ) and primer efficiency were calculated for each primer set to determine the reliability of the assay (30).Primer efficiencies were calculated according to the equation E = 10 −1 b − 1 × 100%, where b is the slope of the standard curve.The detection limit for each primer set was defined as the lowest CFU/mL where the linearity of the standard curve was maintained.Melting curves were also created for each primer pair.The primer efficiency, R 2 , and limit of detection for each primer pair are presented in Table S5.

Primer specificity
The specificity of each primer set was tested against the DNA of strains (29 in total) harboring target and non-target cwps gene clusters alongside melting analysis of the qPCR products obtained (30).50 ng of DNA was used in each reaction, with ND denoting where a C T value could not be determined at all or due to late C T calls that have a higher uncertainty (within the last five cycles; Table S6).All primer specificity analyses were performed as technical triplicates.

Quantification of cells present in undefined starter cultures M1 and M2 using qPCR
For absolute quantification of cells by qPCR, DNA was extracted from the M1 and M2 undefined starter cultures by first resuspending 5 g of the lyophilized starter cultures (without activation) in 95 g of 10% sterilized RSM and stomaching at 230 rpm for 5 minutes.Total DNA was extracted and purified from 2.5 mL of the homogenized cultures as described above.75 ng of genomic DNA from both M1 and M2 was subjected to qPCR with each cwps primer pair (Table S4).Absolute quantification of cells in the samples was performed by interpolation on a standard regression curve of C T values generated from genomic DNA extracted from reference strains of each cwps type (30).Determination of the slope and y-intercept used in the calculations is described in Table S5.The qPCR results are expressed as Log(CFU/mL) for strains harboring each cwps gene cluster tested (30)(31)(32).The subsequent relative abundance of the cwps types present in the starter cultures was then calculated by dividing the total CFU/mL of all cwps types by the CFU/mL corresponding to each cwps type.

Compositional impact of phage attacks on M1
Protocol 10 g of lyophilized M1 starter culture was resuspended in 90 g of 10% sterile RSM (prepared as described above).20 mL of the first dilution was then transferred to 80 mL of 10% sterilized RSM as the final dilution.To reflect typical fermentation conditions, 20 mL of pasteurized milk (as above) at pH 6.5 was inoculated with 4% of the dilu ted lyophilized M1.Test cultures were additionally inoculated with 1% of a whey or individual phage (at titers of ~10 7 -10 9 PFU/ml).Three whey samples were tested (T1, BD, and BA6).The individual phages (isolated from phage screening of wheys, described in the supplementary text, and chosen at random to represent a Skunavirus and Ceduovirus) tested were BA6-A5 and T1-H4.The samples were incubated in a water bath at 32°C for 6 h and were analyzed in triplicate.After incubation, the pH of the whey-treated samples was measured and compared to an uninfected M1 control to first determine the wheys and/or phages had any impact the acidification of the culture.

Metagenomic DNA extraction of M1 culture following phage exposure
Following pH testing of the samples, 2 mL of each sample was added to 23 mL of 2.2% trisodium citrate (45°C) and homogenized in a Stomacher 400 Circulator (Seward) at 230 bpm for 5 minutes (20).The solution was centrifuged at 13,800 × g for 10 minutes at room temperature.The supernatant was removed, and the cell pellets were stored at −20°C prior to DNA extraction.The DNA extraction was performed using PureLink Genomic DNA Mini Kit (Invitrogen, UK) according to the manufacturer's instructions, after an initial DNase treatment (20 units mL −1 DNase I recombinant, Sigma Aldrich for 30 minutes at room temperature, followed by DNase inactivation at 75°C for 10 minutes).The absolute quantification and relative abundance of cells corresponding to the cwps genotype in each sample were calculated as described above (See Quantification of cells present in undefined starter cultures M1 and M2 using qPCR method).Standard deviations and ordinary one-way ANOVA tests were performed, followed by Dunnett's multiple comparisons test, with a single pooled variance.

Metagenomic assessment of M1 and M2
Genomic DNA of both cultures was extracted following M1/M2 culture activation in pasteurized milk and sequenced (Table S7).The assembled contigs from the sequenced metagenomes of M1 and M2 were taxonomically classified using BLASTn-based LCA taxonomy estimation.Then the relative abundance of the species present in these cultures was determined by mapping the raw metagenomic reads against these now taxonomically classified contigs (Fig. 1; Table S8).Based on this analysis, both cultures were determined to be primarily composed of L. cremoris and/or L. lactis strains (>94%).
Due to the high sequence homology between the two species, it was not possible to accurately assign the reads to L. lactis or L. cremoris through the taxonomic classification of contigs alone.Therefore, the relative proportion of L. lactis/L.cremoris in these starter cultures was determined through reads mapping of three genes (recA, pheS, and rpoB) that have previously been applied in multi-locus sequence typing and differentiation of these two species, although here the 16S rRNA gene was excluded (Fig. S2) (2).In both starter cultures, L. cremoris was found to represent >99% of the L. cremoris/L.lactis population.The next most abundant organisms in the cultures were Leuconostoc garlicum (~1.0%) in M1 and Lactococcus raffinolactis (1.6%) in M2 (Fig. 1; Table S8).To further assess the diversity of the L. cremoris component of the M1/M2 cultures at the strain level, the raw metagenomic reads were mapped against the genes specific to nine cwps gene clusters responsible for CWPS biosynthesis (A, B, C 1 -C 6 , and D; see Materials and Methods) (Fig. 2A and D).The L. cremoris constituent in M1 was assigned to C 2 (70%), C 1 (27%), and C 3 (3%) cwps genotypes (Fig. 2A), while those present in M2 were shown to belong to the C 2 (64%), C 3 (30%), C 4 (5%), C 1 (1%), and C 5 (1%) genotypes (Fig. 2D).

Culture-based analysis of M1 and M2
To investigate the composition of the M1 and M2 cultures through cultivation-based methods, isolates were selected from the original starter cultures (i.e., non-activated culture) following incubation on LM17 plates at 30°C (N.B. there is no notable difference in composition between the activated and non-activated culture, Fig. S3).A total of 364 isolates (181 isolates from M1 and 183 isolates from M2) were collected and typed by multiplex PCR (15,17) to determine the cwps genotype of the individual isolates.Of the 181 isolates retrieved from M1, 171 were C 1 -type and 10 were C 3 -type strains (Fig. 2B).Of the 183 isolates selected from M2, 36 were C 1 -type, seven were C 2 -type, 62 were C 3 -type, 21 were C 4 -type, and 13 C 6 -type strains (Fig. 2E).We were unable to type the remaining 44 isolates using the primers available in the multiplex typing assays; however, 31 of these 44 isolates were determined to be L. lactis/L.cremoris strains (based on the amplification of the lactococcal control, rmlB).The cwps gene clusters of these isolates are potentially novel and will be investigated in future studies.The species of the remaining 13 isolates were determined to be Lactococcus laudensis strains through 16S rRNA gene sequencing.

Setup and validation of the qPCR method
To develop a methodology to monitor the composition of mesophilic starter cultures and their typically dominant L. lactis/cremoris component, we developed a qPCR assay with primer pairs capable of differentiating nine industrially relevant strain genotypes based on distinct cwps gene clusters (A, B, C 1 -C 6 , and D) across two species (L.lactis and L. cremoris; Table S4).Information on the validation and specificity of these primer pairs is found in Supplementary Text.These primers can distinguish between nine distinct genotypes when assessed against a test panel of 29 lactococcal strains (Table S6).In addition, the groES gene was shown to be a highly reliable target for quantifying the total number of lactococcal cells present in a sample.

Application of a qPCR method for M1/M2 cwps-based lactococcal strain monitoring
The qPCR approach to quantify various lactococcal cwps genotypes was employed to assess the lactococcal-based undefined starter cultures M1 and M2 without activation.C T values obtained from the individual cwps primer pairs were determined to be reliable if they were above the individual threshold of the primers (determined by the limit of detection and primer specificity analyses in complex medium LM17) and produced the expected melting curves.Based on the obtained results, both M1 and M2 were established to be composed of C-type strains only (C T values obtained from the A, B, and D-type primer pairs exceeded the designated threshold).Based on the qPCR assay, M1 was shown to be primarily composed of C 2 (82%), C 1 (13%), and C 3 (5%) type strains (Fig. 2C), whereas M2 contains C 2 (67%), C 3 (30%), and C 4 (2%) type strains (Fig. 2F).In addition, the total number of cells based on the cwps genotypes was quantified for M1 and M2 to be 8.65 ± 0.15 and 8.90 ± 0.11 log CFU/mL, respectively (Fig. 3).These values were comparable to the total cell counts based on the groES primers 8.60 ± 0.15 log CFU/mL in M1 and 8.62 ± 0.12 log CFU/mL in M2.Therefore, the total CFU/mL of the combined cwps type strains quantified (based on the described cwps primers) was comparable to the CFU/mL quantified from primers targeting the house-keeping gene groES, indicating that the vast majority of lactococcal strains were typed and quantified (Fig. 3).Total counts calculated using both the cwps and groES primers were also similar to the cell counts obtained when the cultures were plated on LM17: 8.78 ± 0.10 log CFU/mL in the case of starter M1 and 7.18 ± 0.06 log CFU/mL in the case of starter M2.A notable limitation of the current methodology outlined here is that standard curves were based on DNA extracted from strains grown in complex medium LM17 and serial dilutions of the cultures plated on the same medium, whereas the starter cultures were grown in milk.It is recommended that standard curves should be prepared using the same matrix as the samples, where possible, which in this case would be milk (33).However, since some strains used in generating the standard curves were capable of growing well (or in some cases, at all) in milk, this was not possible in this case.

Compositional changes caused by phage attack can be detected by the qPCR assay
The applicability of the qPCR approach to monitor culture compositional changes (of starter culture M1) following phage attack was evaluated using (a) industry-derived phage-containing whey samples and (b) individual phages that are known to infect at least one strain originating from M1.To this end, the composition of M1 following incubation with phage(s) from either whey samples originating from factories where fermentation (milk acidification) had slowed (T1, BA6, BD) or individual phages isolated from phage screening of whey samples (BA6-A5 and T1-H4), results of phage screening of whey can be found in Supplementary Text and Fig. S4) were quantified after 6 h of fermentation using the herein established qPCR assay and compared to the uninfected M1 control, to determine what (if any) compositional changes occurred as a result of phage attack.To determine whether a significant change in composition occurred, ordinary one-way ANOVA tests followed by Dunnett's multiple comparisons tests, with a single pooled variance, were applied to determine if there was a significant difference between the cell counts observed for each primer pair (Fig. 4).
All three whey samples tested, that is, T1, BA6, BD inhibited acidification to some degree, with pH unit differences ranging from 0.44 to 1.07 compared to the uninfected control (Fig. S5).In these factories, the undefined starter culture(s) typically contribute(s) to the acidification of milk ultimately leading to the production of kefir (T1), a Polish curd cheese known as Twaróg (BA6), and a semi-hard cheese (BD).After 6 h of fermentation in pasteurized milk, M1 reached a pH of 4.63 ± 0.06, whereas M1 starter to which the phage-containing whey samples T1, BD, and BA6 had been added only reached pH values of 5.70 ± 0.27, 5.47 ± 0.12, and 5.07 ± 0.06, respectively.This indicates the presence of phages in these whey samples that clearly negatively impacted the fermentation/acidification ability of M1.In addition to whey samples containing unknown phage levels and composition, two individual phages were tested.Phages BA6-A5 and T1-H4 were shown to significantly inhibit the fermentation ability of M1 (P < 0.01), with associated pH values of 5.10 ± 0.00 and 5.00 ± 0.00 after the 6-h incubation period, compared to the uninfected M1 control reaching a pH of 4.63 ± 0.06 (Fig. S5).
Microbial DNA from the uninfected M1 starter culture and all whey/phage-infected cultures was extracted and used as a template in the qPCR assay with all cwps and groES control primer pairs and used to quantify the CFU/mL of strains corresponding to each cwps type present in the samples (Fig. 4).After 6 h of fermentation in pasteurized milk, M1 was established to contain C 1 (6.86 ± 0.26 log CFU/mL), C 2 (7.50 ± 0.19 log CFU/mL), C 3 (6.78± 0.19 log CFU/mL), C 4 (3.86 ± 0.17 log CFU/mL), and C 6 (5.15 ± 0.23 log CFU/ mL)-type strains, and with all other assessed genotypes being below the detection limit.Exposure of the M1 starter culture to phages present in whey sample T1 resulted in a significant decrease in strains with the cwps types C 2 (P < 0.0001) and C 6 (P < 0.0001), with final cell counts after phage infection of 5.78 ± 0.17 and 3.93 ± 0.20 log CFU/mL, respectively.Phages in whey BD resulted in significant decreases of M1-associated starter strains with cwps types C 1 (P < 0.001, 6.03 ± 0.12 log CFU/mL) and C 2 (P < 0.0001, 6.36 ± 0.11 log CFU/mL).Phages in whey BA6 resulted in significant decreases (P < 0.001) of C 2 cwps type strains (6.73 ± 0.13 log CFU/mL).Infection by Skunavirus BA6-A5 and Ceduovirus T1-H4 resulted in a significant (P < 0.0001) decrease of C 2 -type strains (6.53 ± 0.13 log CFU/mL and 6.09 ± 0.10 log CFU/mL, respectively).These changes in the absolute quantities of each cwps type resulted in shifts in the relative abundance of all cwps types present in M1 (Fig. 5).

DISCUSSION
L. lactis and L. cremoris strains are mesophilic LAB widely used in dairy fermentation.To ensure consistent product quality, it is essential that the starter cultures are robust and perform reliably.Consequently, both culture-dependent and -independent approaches have been applied to explore the strain constitution of a number of undefined starter cultures used in the production of Swiss hard cheeses (34), Gouda (20), and continental cheeses (19).In the present study, the microbial composition of two undefined industrial starter cultures, named M1 and M2, commonly used in the production of semi-hard cheeses, fermented milk (e.g., viili, kefir, and twaróg) and fresh cheeses (e.g., quark) was evaluated using both culture-dependent and -independent methods.These microbial explorations led to the establishment of a qPCR assay (based on lactococcal cwps gene clusters) capable of discerning nine industrially relevant lactococcal genotypes, with the potential to be widely applicable, thereby allowing for a better understanding of L. lactis/L.cremoris diversity in mesophilic starter cultures and ultimately aid in the implementation of effective starter culture rotation strategies.
Culture-dependent methods have historically been the most frequently used method to study complex microbial communities (35), although in recent years culture-inde pendent methods such as metagenomics have become a more widespread application in the study of fermented foods (36).The benefits of metagenomics in fermented foods are particularly apparent in the study of traditional/artisanal fermented foods, unde fined starter cultures, or spontaneous fermentation (19,20,34,(37)(38)(39).In these cases, the microbiomes are largely uncharacterized, rendering quality control and product consistency a challenge.Metagenomics has allowed these complex communities to be characterized without the limitations associated with purely culture-dependent characterization studies.In the current study, the culture-based profiles of M1 and M2 appeared to differ greatly from the compositions determined through metagenomic reads mapping of cwps types.For example, C 2 strains were determined in this manner to be the most dominant type in the M1 culture, whereas no C 2 strains were identified from the 181 isolates obtained from starter culture M1.Instead, the vast majority (94%) were C 1 -type strains, and the remaining were C 3 -type (6%).In M2, a similar trend was observed where C 2 -type strains were expected to make up 70% of the culture (based on metagenomics), yet just seven out of the 170 L. cremoris/L.lactis isolates typed were C 2 -type strains.These differences are likely to be due to cultivation conditions (LM17 instead of RSM and plating at 30°C) or the presence of viable but non-culturable (VBNC) cells (30).On the other hand, culture-based analysis of M2 led to the isolation of several L. laudensis isolates as well as lactococcal isolates with a cwps gene cluster that could not be assigned to any of the currently known CWPS types, neither of which were identified by metagenomic analysis.In addition, the composition of a sample can only be reported in terms of relative abundance rather than absolute quantifications in purely metagenomic-based analyses (40).Furthermore, to evaluate the role and functionality of both dominant and minor components of these starter cultures, it is necessary to isolate and characterize individual strains through culture-based methods.Therefore, these differences highlight the importance of a combined approach when evaluating the composition of undefined starter cultures.
In recent years, qPCR assays have been used extensively to track and characterize the microbiota of many complex cultures and food products (30,(41)(42)(43)(44).This method is faster and more sensitive than many culture-dependent methods which can often be time-consuming and limited by the cultivation conditions (medium composition, cultivation temperature, etc.).In addition, as shown above, culture-based analysis may not accurately reflect the true composition of the culture, as it does not take into account VBNC cells and is limited by cultivation conditions.Furthermore, where metagenomics and cultivation-based approaches typically provide a one-dimensional or "snapshot" view of a culture under specific conditions, qPCR assays can be used to actively monitor the composition of a culture as a function of time and under varying condi tions.However, these assays are often limited to defining the species/subspecies-level abundance of cultures and provide little information regarding strain-level diversity or are specific to only a certain culture (20,30,41,42).The majority of currently described LAB-based qPCR assays are limited to quantifying and monitoring certain species (L.lactis, Streptococcus thermophilus, Lacticaseibacillus paracasei, Lactococcus garvieae, etc.) during fermentation (30-32, 41, 45).However, a qPCR assay established by Erkus et al. went beyond the species level to monitor the population dynamics of seven genetic lineages of L. lactis/cremoris and L. mesenteroides strains in an undefined Gouda cheese starter culture (20).One limitation of this assay is that it was found to be quite culturespecific: in a subsequent study on undefined mesophilic mixed (DL) starter cultures used in the production of similar continental cheeses, 19 of their 127 L. cremoris isolates were found to contain gene markers specific to just one of the genetic lineages established by Erkus (19).This indicates that the qPCR assay established by Erkus is highly culturespecific as it is based on genetic lineages present in starter culture Ur and may not be generally applied to other mesophilic starter culture systems.The qPCR-based assay described in the present study utilizes sequence diversity within the variable cwps gene cluster which appears to be present in all known L. lactis and L. cremoris strains and therefore may be more widely applicable to mesophilic starter cultures in general.The cwps gene cluster is an ideal target as CWPS is a technologically relevant attribute of these species as it acts as a receptor for many lactococcal phages.Therefore, this qPCR assay can be useful both as a classification and risk evaluation tool from the perspective of defining what is likely to be phage-unrelated cultures, which is essential when implementing effective starter culture rotation strategies.The applicability of this assay was indeed demonstrated through the characterization and quantification of cwps-type lactococcal strains present in two undefined mesophilic starter cultures (M1 and M2).Overall, the cell numbers observed in both culture-based and qPCR-based analyses were comparable (within one log) for both cultures.It is possible that the subtle differences in cell counts are attributed to culturing conditions, the presence of VBNC cells, or dead cells.While the possible inclusion of dead cells is a limitation of both metagenomic and qPCR-based analysis, since the cells were pelleted and treated with DNase prior to DNA extraction, it is likely that DNA from lysed cells is eliminated before the DNA was extracted (30).Future improvements of these assays could include a live/dead evaluation step.Alternatively, an advantage of culture-independent analyses is that VBNC cells are still likely to be enumerated through both qPCR and metagenomics but not in culture-based assays (33).Also of note, the composition of the two starter cultures as determined through the herein developed qPCR assay is quite similar to that determined by metagenomics, in contrast to the findings of the culture-based analysis.This makes sense as both are culture-independent methods that are not biased by cultivation conditions and capable of enumerating both culturable and non-culturable cells.
It was also demonstrated that this qPCR assay may be used to monitor compositional changes of a mesophilic culture following a phage attack.The compositional changes observed in these phage attacks correspond well to phage screening results from the whey samples and host ranges of the individual phages (Supplementary Text).For example, a significant drop in C 1 and C 2 -type strains was observed in M1 following a phage attack from whey BD.Phage screening identified several C 1 , C 2 , C 3 , and C 4 -type isolates that were sensitive to phage(s) present in the BD whey sample (all of which were typed as Skunaviruses).Since C 3 and C 4 -type strains represent minor components of the M1 starter culture, it seems logical that the C 1 -and C 2 -type strains were most impacted during the phage attacks.In addition, infection from phage(s) present in the T1 and BA6-wheys (as well as from individual phages BA6-A5 and T1-H4) decreases in the absolute quantities of certain cwps-type strains (C 2 -types) following phage attack resulted in the increase in the relative abundance of other cwps type (C 1 and C 3 types).The PSP component of a host's CWPS is a major factor in determining if a phage can bind to a host cell and potentially infect and lyse a strain; therefore, awareness of the CWPS type or subtype of strains within a starter culture provides improved predictions of the risk and potential impact of phage infection (15,17).

Conclusion
By investigating the composition of two undefined mesophilic starter cultures through metagenomics and culture-based analysis, we demonstrated the importance of employing both culture-dependent and -independent methods when defining complex systems.In addition, a qPCR assay was established that is capable of discerning nine cwps genotypes of L. lactis and L. cremoris strains.Both metagenomics and qPCR-based methods were shown to be quite comparable, whereas the compositions of the two cultures as determined by culture-based methods clearly differed from culture-independent methods which can be attributed to differences in culturing conditions and the likely presence of VBNC cells.The added benefit and usefulness of the established qPCR assay to not only characterize but actively monitor meso philic undefined starter cultures was demonstrated by assessing culture composition following various phage attacks (using whey samples and individual phages).The results confirmed the robustness and flexibility of the undefined mesophilic cultures.The developed qPCR assay is expected to be widely applicable to mesophilic starter cultures, allowing for the quantification of lactococcal cells and specific genotypes as well as the identification of phage-susceptible components of a given culture, which aids in efficient starter culture rotation practice.The cwps gene clusters of lactococcal strains are constantly evolving, likely because of continuous pressure from phage predation.Therefore, this qPCR assay is expected to be expanded or further fine-tuned for future purposes.Furthermore, as increasing sequence information becomes available for L. raffinolactis, L. laudensis, and Leuconostoc mesenteroides/pseudomesenteroides strains, it may be possible to develop tailored and species-specific qPCR assays to evaluate (strain) diversity of these (and other) LAB species in mesophilic starter culture systems.

FIG 1
FIG 1 Taxonomic composition of M1 and M2 based on reads mapping of taxonomically classified contigs.Columns indicate the relative abundance of reads corresponding to contigs taxonomically classified as either bacterial (at the genus or species level), eukaryotic, or viral.

FIG 2
FIG 2Comparison of culture-dependent and -independent methods of assessing the composition of undefined starter cultures.Comparison of metagenomics (A and D), culture-based analysis (B and E), and qPCR (C and F) as methods to characterize mesophilic undefined starter cultures M1 (A through C) and M2 (D through F) based on the cwps gene cluster.CWPS types of isolates selected from M1 and M2 were typed using previously described multiplex PCR CWPS typing assays (B and E).Isolates that were unable to be typed by any CWPS type primers (A-C, and C 1 -C 5 -subtypes) but yielded a product from the lactococcal control primers are indicated as unknown.

FIG 3
FIG3 Absolute quantification of lactococcal strains of M1 and M2 using a cwps-based qPCR assay.Quantification of lactococcal strains grouped based on their cwps types expressed as the estimated log CFU/mL (based on standard curves of each primer set against a representative strain).

FIG 4
FIG 4 Absolute quantification of mesophilic undefined starter culture M1 following phage attack compared to an uninfected control using a cwps-based qPCR assay.Absolute abundance of cwps types present in M1 after 6 h of normal fermentation conditions under phage pressure from individual phages (BA6-A5 and T1-H4) or phages present in whey samples (T1, BD, and BA6) compared to an uninfected M1 control.C T values obtained from the primers corresponding to cwps types A, B, D, and C 5 strains were below the detection level and therefore excluded from this graph.A significant change (P < 0.01) in composition in a sample compared to the uninfected M1 control in a sample is indicated by star(s).Two stars indicate a P < 0.01, three stars a P < 0.001, and four stars a P < 0.0001.

FIG 5
FIG 5 Relative abundance of cwps types present in mesophilic undefined starter culture M1 following phage attack compared to an uninfected control using cwps-based qPCR assay.The relative abundance of cwps-type strains present in the samples demonstrates the overall compositional changes in M1 as a result of phage infection.C T values obtained from the primers corresponding to cwps types A, B, D, and C 5 strains were below the detection level and therefore excluded from this graph.