Microbiological and biochemical changes in the curd are crucial factors for the production of raw-milk, ripened cheeses. The microbial ecology of this cheese variety consists of a complex interaction between starter lactic acid bacteria (SLAB, usually deliberately added for curd acidification) and non-starter LAB (NSLAB, adventitious milk contaminants from farm and dairy environments) from milking to ripening (Gatti et al., 2014; Blaya et al., 2018).

NSLAB frequently recovered from cheese are facultative heterofermentative lactobacilli (Beresford et al., 2001). Lacticaseibacillus [formerly Lactobacillus casei group, (Zheng et al., 2020)] is one of the most prevalent genera found in hard cooked, long-ripened cheeses. It includes the species Lacticaseibacillus casei, Lacticaseibacillus paracasei, and Lacticaseibacillus rhamnosus, which are of particular interest because of their proven role in cheese flavor formation during ripening (Bottari et al., 2018), as well as their potential health benefits through the consumption of fermented foods (Hill et al., 2018; De Filippis et al., 2020). Their trend from low abundance in raw milk, to dominance in ripened cheese, involves adaptation to chemical and physical stresses throughout the cheesemaking and ripening processes (i.e., heat-related, acidic, osmotic, and oxidative stresses), besides their ability to grow using energy sources other than lactose (Gatti et al., 2014). These are the physiological criteria used for the delineation of the name of the new genus Lacticaseibacillus from caseus (cheese; Zheng et al., 2020).

The mechanisms responsible for NSLAB survival to stresses encountered during cheesemaking might be attributed to various strategies of adaptation, such as the capability to utilize different energy sources (Lazzi et al., 2014; Papadimitriou et al., 2016) or the activation of strategies that can increase their tolerance to the food manufacturing process, such as the recently described toxin-antitoxin systems (Levante et al., 2019, 2021). Although NSLAB have been extensively studied for their role in cheese ripening (De Dea Lindner et al., 2008; Solieri et al., 2012; Sgarbi et al., 2013, 2014; Levante et al., 2017; Bottari et al., 2020), and reviews are available on this topic (Beresford et al., 2001; Settanni and Moschetti, 2010; Montel et al., 2014; Gobbetti et al., 2015; Bottari et al., 2018), the source of their origin and their relative abundance in raw milk are not well clarified yet.

Further research is necessary to demonstrate whether live microorganisms are present in milk inside a healthy mammary gland (Oikonomou et al., 2020), although milk endogenous bacterial contamination via enteromammary pathway has been hypothesized (Addis et al., 2016). However, it is known that microbial colonization occurs from different sources throughout the route from farm to cheese factory (Gobbetti et al., 2018) and raw milk represents an ideal environment for the growth of many microorganisms (Quigley et al., 2013). To shed light on which and how many of these biotypes will be part of the ripened cheese microbiota, it is of primary importance to know their initial abundance in raw milk, since the number of bacterial cells is known to be one of the most important factors in determining microbial activity in food (Fleet, 1999; Giraffa, 2004). It is expected that in the next future quantitative methods such as flow cytometry will help understanding how microbial loads of certain species in raw milk may affect the microbiome of resulting cheeses (Skeie et al., 2019; Porcellato et al., 2021), as it was also shown for other complex microbiomes (Vandeputte et al., 2017).

Although many studies deal with the description of the complex dairy ecosystem focusing on LAB in raw milk, often reporting isolation, identification, and relative quantification of the most abundant species, they rarely provide absolute quantification. With the aim to answer the initial question: “which and how many lactic acid bacteria are present in cow raw milk used for cheese production?,” this review analyses and comments on the data of the studies available in the literature addressing LAB quantification in raw milk used for the production of different cow’s milk cheeses.

The different analytical techniques used for LAB evaluation in milk can be divided into two methodological approaches: (I) conventional culture-dependent methods which analyze microorganisms after their growth in liquid or solid media; (II) more recent culture-independent methods which directly detect microbial nucleic acids, avoiding the culturing step. A prior careful consideration of numerous factors (e.g., taxonomic resolution, workload) is necessary to define the appropriate methodology for a certain purpose (Temmerman et al., 2004). These different approaches are complementary, thus their combination in a polyphasic approach is suggested to set up more complete and accurate studies (Ercolini et al., 2001; Delbès et al., 2007; Agrimonti et al., 2019) that better describe the dynamic evolution of microbial communities in food ecosystems, in particular in the complex transformation of raw milk to ripened cheese.

In both methods, the qualitative information (i.e., microorganisms taxonomic identification) can be addressed by common molecular analysis (such as PCR-and sequencing-based methods). On the other hand, the quantitative results differ: most of the culture-independent studies on LAB in raw milk report the relative abundance (%) of prevalent taxa; culture-dependent studies mainly based on plate counts, report instead results as Colony Forming Units (CFU/mL).

For this reason, data collected from the studies applying different approaches have been discussed separately in the following two chapters: the first one aims at giving an overview of LAB’s abundance within the complex microbiota of cow raw milk through the review of recent taxonomic studies; the second includes a collection of data about quantification of LAB by culture-dependent analysis in cow raw milk.

Culture-independent techniques available for milk and cheese microbiota analysis are constantly evolving and have been extensively reviewed elsewhere (Randazzo et al., 2009; Ndoye et al., 2011; Quigley et al., 2011; Neviani et al., 2013; Addis et al., 2016; Levante et al., 2020; Tilocca et al., 2020). Common outlines of these reviews are: (I) the usefulness of the culture-independent approach in describing complex ecosystems such as raw milk and cheese, and the ability to overcome culture-dependent limits (e.g., ability to distinguish between dead or viable cells by a combination of DNA-and RNA-based approaches), although a polyphasic approach is recommended; (II) -omics approaches are expected to greatly contribute in shedding light on microbial ecology dynamics.

Methods for absolute quantification at the species level have been implemented, such as Fluorescence in-situ Hybridization [FISH (Bottari et al., 2006)], quantitative PCR [qPCR (Agrimonti et al., 2019)], and Total Bacterial Count computed with the percentage of taxon after 16S rRNA Amplicon Sequencing (Props et al., 2017; Skeie et al., 2019; Porcellato et al., 2021). However, to the authors’ knowledge, no studies applied these methods for the absolute quantification of the Lacticaseibacillus group in cow raw milk used for cheese production, except for Masoud and colleagues who quantified Lacticaseibacillus rhamnosus in raw milk used for Danish cheese production (Masoud et al., 2012).

Different studies are instead available in the literature about the characterization of raw milk microbiota by amplicon-based high-throughput sequencing (HTS) analysis. This method is successful to describe changes in microbiota related to seasonality, geographical origin, and the microbiota evolution at different steps of cheese making, but it suffers from bias because results are usually reported as relative abundance of taxa which are not converted to quantitative values (Skeie et al., 2019), and identification of bacteria beyond the genus level is often not possible (Claesson et al., 2010; Dreier et al., 2022). On the other hand, a fundamental advantage of these studies is the possibility to use the generated raw sequences to perform meta-analysis.

A very useful tool to conduct the meta-analysis is FoodMicrobionet [FMBN (Parente et al., 2016, 2019; De Filippis et al., 2018)], a collection of datasets created by 16S rRNA gene amplicon HTS studies of food bacterial communities. This tool was already used to widely review the microbiota of dairy milk, discussing comparisons between the pasture and feed, farm environments, teat skin, teat milk (from different species, also affected by diseases), bulk tank milk and finally also HTST (high-temperature short time) milk (Parente et al., 2020).

In the present review, the updated version of FMBN 4.1.2 (Parente et al., 2022), integrated with the taxonomy reclassification for the genus Lactobacillus (Zheng et al., 2020) and further recent studies on the raw milk microbiota, has been used. Among the studies available in the database on cow raw milk, those analyzing cow whole raw milk used for cheese production were selected (De Pasquale et al., 2014; Dolci et al., 2014; Calasso et al., 2016; De Filippis et al., 2016; Giello et al., 2017; Falardeau et al., 2019; Cremonesi et al., 2020; Kamimura et al., 2020; Nikoloudaki et al., 2021; Supplementary Tables 1, 2). This dataset included reads from 250 raw milk samples, classified in 1,594 total taxa belonging to 45 different phyla (Chloroplast, Eukaryota, Mithocondria, and unidentified OTUs at the domain level were removed). Nodes and edges tables (.gml file generated by FMBN Shiny app) were then imported into Gephi software [v 0.9.2 (Bastian et al., 2009)] for bipartite network analysis, i.e., network with nodes belonging to OTUs and samples.

Nodes and edge statistics were calculated in Gephi: the degree is the number of connections to a certain node, meaning that for a sample, it represents the number of present OTUs, vice versa for an OTU, it represents the number of milk samples where it was found; the weight is a value assigned to an edge, corresponding to the OTU relative abundance in the sample; the weighted degree is the sum of weights for a node (Supplementary Table 3). Nodes and taxa label size are proportional to their weighted degree, hence the bigger they are, the more the OTU is abundant in raw milk. Only nodes with a weighted degree > 10 were considered for the network (Figure 1). ForceAtlas2 algorithm (Jacomy et al., 2014) was applied for the layout: as a consequence, taxa are closer to samples where they are present with a greater abundance (thus, taxa to the periphery are less abundant); also, taxa close to each other were present more frequently in the same samples.

After the filtering step, 149 taxa were retained (Supplementary Table 3), which belong to 10 different phyla. Part of the OTUs were classified only at the kingdom level (Bacteria, NA dark gray node). Among the most abundant genera, the psychrotrophic bacteria are present in significant proportions: these include Pseudomonas (23.85% of the tot weighted degree in the network), Chryseobacterium (6.56%), and Acinetobacter (6.01%). It is reasonable to assume that their dominance could be due to the refrigerating temperature usually applied for raw milk storage which decreases bacterial diversity (Raats et al., 2011).

According to these data, LAB relevant for cheese production are present as dominant taxa such as Lactococcus (7.24%), or subdominant such as Streptococcus (3.6%), Lacticaseibacillus (2.42%), Lactobacillus (2.35%), Leuconostoc (1.15%), and Enterococcus (0.41%). Despite their relatively low abundance in the raw milk ecosystem, these subdominant taxa can develop in subsequent cheese manufacturing. An example is found in long ripened cheeses, where the Lacticaseibacillus genus is known to become dominant throughout the ripening of raw milk, hard cooked, long-ripened cheeses (Bottari et al., 2018).

The low abundance of some subdominant species makes their isolation from raw milk difficult, limiting thus a potential targeted use. A strategy to overcome this limit could be the enrichment of the autochthonous raw milk microbiota through its spontaneous fermentation. Following this approach, Bancalari and colleagues carried out a spontaneous fermentation of Parmigiano Reggiano raw milk samples to isolate LAB strains potentially usable as adjunctive aromatic starters (Bancalari et al., 2017), while Galli and colleagues incubated the raw milk for 24 and 48 h, at 30°C and 40°C, with the purpose to isolate additional LAB with potentially high GABA (γ-aminobutyric acid) producing capabilities (Galli et al., 2022).

The route of some LAB bearing positive features for cheese production, begins when they reach the raw milk. With the intention of defining which and how many LAB are present in cow raw milk used for cheese production, this review compared and discussed the results of the available studies that addressed this topic through culture-dependent and-independent methods.

The critical review of the literature highlighted an inconsistency in the definition of “raw milk,” a term that can be referred to milk sampled at the farm rather than from the cheese vat after skimming. Together with the sampling procedures, other factors such as different farming systems, milk pre-treatments, and seasonality, take part in the modulation of the raw milk microbiota.

The meta-analysis conducted with the use of FMBN was very effective to highlight the complexity of the microbiota of cow raw milk used for cheese production which is composed of over 45 phyla and showed how the genera belonging to LAB are not the most abundant in this microbial ecosystem. It also indicated how LAB belonging to Lactobacillus and Lacticaseibacillus genera are present in lower relative abundance in comparison with Lactococcus and Streptococcus.

This result is consistent when compared with LAB concentration evaluated by means of plate counts on agar media: lactobacilli were in fact found on average in lower concentration than coccoid LAB in the raw milk used for the production of 24 different cheese types.

This allows once more to conclude that both culture-dependent and-independent techniques complementarily describe the raw milk microbiota, and the combination of the two approaches restores a complete picture of which and how many LAB are present in the raw milk.

A better knowledge of both the amount and the species of LAB present in the raw milk could be useful in the perspective of monitoring and maintaining the biodiversity of the microbiome of this complex substrate through the managing of parameters at the farm and cheese making level and finally predict the expected outcomes in the resulting raw milk cheeses.

LB and MG conceptualized the study. LB performed the investigation, data curation and visualization, and wrote the original draft. MG supervised the study. AL and MG contributed to the methodology of data elaboration and acquired the fundings. EB and MG cured the Section 3.2 of the manuscript. BB critically revised the paper. All authors contributed to manuscript revision, read, and approved the submitted version.

This research has financially been supported by the Programme “FIL-Quota Incentivante” of University of Parma and co-sponsored by Fondazione Cariparma.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.