Dairy lactococcal and streptococcal phage–host interactions: an industrial perspective in an evolving phage landscape

ABSTRACT

Almost a century has elapsed since the discovery of bacteriophages (phages), and 85 years have passed since the emergence of evidence that phages can infect starter cultures, thereby impacting dairy fermentations. Soon afterward, research efforts were undertaken to investigate phage interactions regarding starter strains. Investigations into phage biology and morphology and phage–host relationships have been aimed at mitigating the negative impact phages have on the fermented dairy industry. From the viewpoint of a supplier of dairy starter cultures, this review examines the composition of an industrial phage collection, providing insight into the development of starter strains and cultures and the evolution of phages in the industry. Research advances in the diversity of phages and structural bases for phage–host recognition and an overview of the perpetual arms race between phage virulence and host defense are presented, with a perspective toward the development of improved phage-resistant starter culture systems.

bacteriophage, Lactococcus lactis, Streptococcus thermophilus, dairy fermentation

INTRODUCTION

Like all living organisms, bacteria have predators. Among these predators are bacteriophages (or phages) that are bacterial viruses. To survive phages, bacteria have developed a plethora of strategies that target each step of the phage life cycle, including host recognition, phage DNA injection, DNA replication, protein synthesis, phage assembly and virion release (Sturino and Klaenhammer 2006). Conversely, phages have developed a variety of mechanisms to counter bacterial anti-phage defenses. The consequence of this perpetual arms race between phage and host is the constant evolution of the two biological entities leading to an increased diversity of both.

Lactic acid bacteria (LAB) are fermentative microorganisms that produce lactic acid from the catabolism of various carbohydrates resulting in a rapid reduction in the pH of their environment (Axelsson 2004). This property has long been used by humans to improve the lifespan of perishable foodstuffs. In particular, LAB have been used for millennia to produce a variety of fermented dairy products. The process of fermentation was spontaneous in ancient times, becoming artisanal and then industrial through the development of civilization, the economy and knowledge (Prajapati and Nair 2003). Phages were first identified as responsible for industrial dairy fermentation failures in 1935 (Whitehead and Cox), and it soon became evident that they may have a dramatic economic impact. Therefore, phages from LAB are among the most studied, even if there are no published figures today on their economic impact in the dairy industry. The recent globalization of the economy (Bulletin of the International Dairy Federation 501/2019), with the concentration of production facilities, the ever-increasing size of factories, and the dairy products and by-products traveling across the world have probably impacted the pandemic of LAB phages. Prophylaxis against LAB phages is thus even more critical than ever.

With a specific focus on Lactococcus lactis and Streptococcus thermophilus, the species with the most economic impact on the dairy industry, diversity of phages described in the literature has been reviewed, and an insight into phages affecting starter cultures in the industry based on DuPont field investigations has been reported. The interaction between phages and their hosts has been reviewed, including the relationship between the host receptor and the phage antireceptor as well as the mechanisms that bacteria have developed to fight phages when infected. Finally, the strategies on the composition of starter cultures and their utilization in rotation schemes are described (as recommended by researchers and as applied by starter culture producers).

LACTOCOCCUS LACTIS AND STREPTOCOCCUS THERMOPHILUS PHAGE DIVERSITY

Reducing phage impact in dairy fermentation necessitates knowledge of the phages detected at manufacturing plants: their nature, frequency, number, location, among other epidemiologic elements. This knowledge may not be easy to garner considering that properly performing the analyses requires access to the individual strains utilized during fermentation, and this information is rarely disclosed, remaining proprietary to the starter culture providers.

Lactococcus lactis phages

Phages infecting L. lactis (termed Streptococcus lactis prior to Schleifer et al. 1985) were first described in 1935 (Whitehead and Cox) and were the first phages isolated against a lactic acid bacterium. Phages are traditionally classified based on morphological criteria (head size and shape, tail length, contractile or non-contractile tail, presence of a collar or a base plate) determined by observation using electron microscopy. Early attempts to further discriminate phages infecting L. lactis used their spectrum of virulence (or the spectrum of sensitivity of their hosts; Chopin, Chopin and Roux 1976) or serological studies (Jarvis 1977). In 1984, Jarvis was the first to use DNA–DNA hybridization that allowed a more global analysis of phage relatedness (Jarvis 1984). This methodology was used to classify the L. lactis phages into 12 groups (Jarvis et al. 1991). This classification was further revised in 2006 by Deveau et al. by using more stringent DNA–DNA hybridization and sequence analyses to more reliably define 10 distinct phage groups including 8 of the former groups and 2 new groups. Phages of the 936 group (recently renamed as Skunavirus genus within the Siphoviridae family; Adriaenssens et al. 2018) are the most represented in the publication from Jarvis et al. (1991) (41 isolated individuals) and have been regularly reported as one of the most frequent phage groups responsible for problems in dairy fermentations (for examples: Szczepanska et al. 2007; Raiski and Belyasova 2009; Suarez et al. 2009; Verreault et al. 2011; Murphy et al. 2013). The c2 group (recently renamed as Ceduovirus genus within the Siphoviridae family; Adriaenssens et al. 2018) phages are also problematic, with Jarvis et al. (1991) reporting 29 individuals, and the above-mentioned publications indicating that phages of the c2 group are as frequent as those of the 936 group. However, in several reports it was mentioned that phages of these groups tend to become less prevalent over time (Mahony and van Sinderen 2015; Mahony et al. 2016). Finally, P335 phages (11 isolated phages in Jarvis et al. 1991) compose the third most problematic group in the industry. The P335 group (within the Siphoviridae family) is comprised of phages that are rather diverse and may be lytic or lysogenic (phage integrated into host genome, termed a prophage) (Labrie et al. 2008; Mahony et al. 2013b). The seven other phage groups do not appear to be frequently responsible for issues in dairy fermentations. These groups are 949 (Samson and Moineau 2010), 1358 (Jarvis 1984), P087 (Villion et al. 2009), 1706 and Q54 (Deveau et al. 2006) from the Siphoviridae family, and P034 (Kotsonis et al. 2008) and KSY1 (Chopin et al. 2007) from the Podoviridae family. While the 936 and c2 groups were attributed genus names, the other groups were not.

Streptococcus thermophilus phages

All phages infecting S. thermophilus characterized to date belong to the Siphoviridae family (Quiberoni et al. 2010). Initial attempts to classify them were based on host ranges and serological reaction analyses (Brussow et al. 1994). Le Marrec et al. (1997) refined the classification, identifying two main groups based on their mode of DNA packaging and on major structural protein profiling. This classification is still commonly used and distinguishes the cos and pac groups, named according to their mode of DNA packaging: cos for cohesive ends and pac for head full packaging. Recently, cos and pac groups were attributed a genus name, Moineauvirus and Brussowvirus, respectively (Adriaenssens et al2018). A Polymerase Chain Reaction (PCR) typing method for the phages of these two groups was successfully developed (Binetti et al. 2005; Quiberoni et al. 2006), allowing analyses of S. thermophilus phage distribution. Until recently, all described S. thermophilus phages belonged to one of these two genera with an approximate distribution of two-thirds cos and one-third pac (Quiberoni et al. 2006; Szymczak et al. 2017, 2019; Lavelle et al. 2018a). Three other S. thermophilus phage groups have since been described. The 5093 group was first described in 2011 by Mills et al., and the genome sequence of its representative phage 5093 revealed homologies with genome of phages infecting non-dairy streptococci. Additional phages of this group were then described in 2017 by McDonnell et al. (four phages), in 2018 by Lavelle et al. (two phages) and in 2019 by Szymczak et al. (six phages), allowing the improvement of the previously published PCR typing method (Szymczak et al. 2019). In 2016, McDonnell et al. described a fourth group of S. thermophilus phages, named 987, which exhibits ‘chimeric’ genomic architecture, suggesting genetic transfer from lactococcal phages of the P335 group. Phages of the 987 group were also detected by Lavelle et al. (2018) and Szymczak et al. (2019). Finally, earlier this year a fifth group of phages, named from phage P738, was described (Philippe et al. 2020). The two current representative phages of this new group (P738 and D4446) are genetically distinct from phages belonging to the other groups. Like the 987 group phages, the P738 group displays a chimeric genomic organization, suggesting genetic exchanges with phages from other Streptococcus species.

DuPont collection of phages

Many companies that produce and market starter cultures also offer a service to perform phage monitoring for their customers. This provides an opportunity to access a great number of samples from various dairy technologies and geographical origins over a long period of time, thus offering the possibility to constitute a collection of biodiverse phages. For >40 years, DuPont has isolated phages from customer samples; as of December 2018, its collection comprised 5874 purified phages that were isolated on strains belonging to 19 species from five genera, with >99% being phages isolated on proprietary strains contained in the starter cultures sold over this period. For historical reasons, most of the phages isolated infect L. lactis (3723) or S. thermophilus (1264) and are from samples collected in Europe (80%), the Americas (12%), Middle East and Africa (5%) and the rest of the world (3%). Phages were isolated each year from 35 to >600 distinct industrial dairy samples, with an average of ∼150 isolates per year.

Lactococcus lactis phages

Presently, there are 3723 lactococcal phages in the DuPont collection that were isolated on 514 host strains (282 subsp. lactis including 43 of the biovar diacetylactis, 211 subsp. cremoris and 21 for which the subspecies have not been determined) representing most cheese applications and geographical locations. About half of these phages (1969; see Table 1) have been identified through specific PCR typing (Deveau et al. 2006) and genomic analyses.

Table 1.

Distribution per group (and subgroups for part of the c2 group phages) of 1969 phages from the DuPont collection isolated on L. lactis strains of various subspecies and biovar. Group and subgroup determinations were performed through specific PCR typing (Deveau et al. 2006) and genomic analyses.

Phage group	Subsp. lactis	Subsp. lactis biovar diacetylactis	Subsp. cremoris
c2a	402	88	204
c2-bIL67	243	11	19
c2-c2	183	4	109
936	91	53	210
P335	333	1	5
949	8	0	0
1706	4	0	0
P087	5	0	0
KSY1	2	0	0
P034	4	0	0

a

Phages of the c2 group for which the subgroup was not determined.

The majority of the typed DuPont collection lactococcal phages belongs to the c2 group (63.6%), with 936 and P335 representing the next most dominant groups (18% and 17.2%, respectively). This distribution has been globally stable since 2003 and is reproducible across the various geographic areas. The distribution of phages according to group differs depending on the host subspecies (lactis versus cremoris subspecies; Table 1). Phages of the c2 group were isolated almost equally on both lactococcal subspecies (subsp. lactis 65% and subsp. cremoris 60.7%); however, substantial differences were recorded for phages of the 936 and P335 groups. Phages isolated on subspecies cremoris strains were almost exclusively from the 936 group (38.4%); only five phages (0.9%) belong to the P335 group. On the contrary, there is a higher proportion of P335 group phages (26.1%) relative to 936 group phages (7.1%) isolated on subspecies lactis strains. The distribution of phages isolated on diacetylactis biovar strains (33.8% 936 group and 0.6% P335 group) was similar to subspecies cremoris rather than to the lactis subspecies, although the diacetylactis biovar belongs to the latter. Genomics analyses allow for the discrimination within the c2 group phages, namely the c2-c2 and c2-bIL67 subgroups (Millen and Romero 2016). The c2 group phages are equally distributed over time into these two subgroups; however, there are differences depending on the isolation strain subspecies. For phages isolated on subspecies lactis strains, the distribution is roughly even (43% c2-c2 and 57% c2-bIL67) but is skewed toward the c2-c2 subgroup (86%) for subspecies cremoris. This is likely a consequence of a subset of cremoris strains (∼50%) having an insertion sequence (IS982 family) inserted into their yjaE gene that encodes a protein required for infection by c2-bIL67 subgroup phages (Millen and Romero 2016; see the 'Protein Receptors' section below). As an example, the yjaE homologue L. lactis subsp. cremoris SK11 (Accession No. NC_008527; locus tag LACR_RS04585) is interrupted by a transposase annotated as IS982 that should render SK11 resistant to bIL67-c2 subgroup phages.

Phages belonging to the other groups are much less frequent and were isolated solely on strains of the lactis subspecies. In most cases, detection of these phages ‘peaked’ during a short period. The 1706 group phages were isolated on two different strains in 1994 and 1995 and have not been detected since. The most prevalent of these rare groups in the DuPont collection is the 949 group that was first isolated in 2006 despite being described in 1977 (Jarvis). From 2006 to 2012, seven additional phages from the 949 group were isolated on seven different DuPont starter strains.

As commonly noted in the literature, 936 group phages are the most frequent phages in the dairy environment (Mahony and van Sinderen 2014), which has been reiterated by multiple authors (Deveau et al. 2006; Rousseau and Moineau 2009; Castro-Nallar et al. 2012; Murphy et al. 2013). Mahony and van Sinderen (2014) placed the P335 phages as the second most frequent group. However, in the DuPont collection, the c2 group phages are most prevalent and significantly higher in proportion than the equally represented 936 and P335 group phages. One critical aspect to consider is the diversity of strains that are utilized for phage isolation. Indeed, the diversity of phages obtained is a direct reflection of the diversity of strains used for their isolation. The primary source of phages in the DuPont collection consists of samples collected from dairy plants utilizing DuPont starter culture ranges; moreover, the isolation of phages was almost exclusively performed on strains constituting these starter cultures. As the DuPont mesophilic starter culture range is dominated by lactis subspecies strains, it is thus expected to have a distribution of phages reflecting that of the lactis subspecies. Therefore, it is unsurprising that the DuPont phage collection is dominated by the c2 group phages and the 936 group is underrepresented (phages of the c2 and 936 groups being the main groups of phages isolated on the strains of the lactis subspecies and of the cremoris subspecies, respectively).

Streptococcus thermophilus phages

Since 1985, 1264 isolated phages infecting 243 S. thermophilus strains were added to the DuPont phage collection. A large majority of these phages (99%) have been genotyped using the PCR method developed by Quiberoni et al. (2006) placing phages into the cos and pac groups (Table 2). In addition, further genomic analyses identified phages belonging to other groups. Overall, of the successfully typed streptococcal phages, 69% and 29% belong to the cos and pac groups, respectively, which is in agreement with previous observations by McDonnell et al. (2017) on a series of 40 phages from multiple origins. In contrast, two studies targeting a specific dairy plant found an unbalanced distribution of phages in favor of cos group phages, a distribution generally maintained over time or geographically (Guglielmotti et al. 2009a; Zinno et al. 2010). Among the 243 strains for which phages were isolated in these studies, 113 and 49 strains were only susceptible to phages of the cos or pac group, respectively.

Table 2.

Distribution per group of 1179 phages from the DuPont collection isolated on S. thermophilus strains. Group determination was performed using the PCR method developed by Quiberoni et al. (2006) and genomic analyses.

Phage group	S. thermophilus
cos	812
pac	346
5093	19
987	1
P738	1

Sixty-six phages (5.6%) failed to respond to the PCR typing method. Nineteen of these phages belonged to the 5093 group and were first registered in the DuPont collection in 2009, indicating that this phage group existed in the dairy environment prior to their first description by Mills et al. (2011). The DuPont collection 5093 group phages were isolated on the same strain (or from a phage-resistant variant of that strain), which was first used industrially in 2006. The spectrum of virulence of these phages is very narrow and was limited to their original host (and variants). One of the unidentified phages by the PCR method (D6660) was later classified as belonging to the 987 group by genomic analyses. The host strain of phage D6660 was also sensitive to a series of 10 other phages that are part of the 66 phages that were unidentifiable by the PCR method, suggesting that they may also belong to the 987 group. These 11 phages displayed a spectrum of virulence limited to their isolation host, which was first used industrially in 2011 as part of a mixed starter culture containing lactococci. Investigation into the nature, phylogeny and industrial usage of all the 987 group phage host strains would be beneficial to the understanding on how phages of this group emerged. Finally, genomic investigations of phages that were not identifiable by the PCR method identified D4446 as belonging to the P738 group (Philippe et al.2020) and is the sole representative of the group in the DuPont collection. Phage D4446 was isolated in 2003 from strain S. thermophilus DGCC7891 for which seven other phages (isolated before and after D4446) were typed; all were found to belong to the pac group, suggesting that P738 group phages are rare in the environment.

PHAGE RECEPTORS AND ANTIRECEPTORS

In the first step of the phage infection process, phages must recognize and attach to host cells, representing a first point of host specificity. This attachment occurs through interactions between host receptors and phage antireceptors. Streptococcus thermophilus phages have been shown to recognize a carbohydrate receptor on the host surface (Szymczak et al. 2018). Lactococcus lactis receptors are also found at the cell surface and have been shown to be saccharidic and/or proteinaceous in nature (Monteville, Ardestani and Geller 1994; Ainsworth et al. 2014a). In some cases, such as for c2 group phages, initial binding to saccharidic receptors may be reversible, requiring further interactions for irreversible phage adsorption (Monteville, Ardestani and Geller 1994). Binding to a proteinaceous receptor results in high-affinity protein–protein interaction. While the binding to saccharides is weaker, the avidity of phage receptor-binding proteins (RBPs), due to the large number of RBPs encoded by the phages, can allow for irreversible adsorption (Bebeacua et al. 2013b; Dunne et al. 2018). Phage antireceptors, such as RBPs, are phage-encoded proteins that are primarily responsible for host recognition and binding, and are often found as tail fibers or as part of a baseplate structure (Sciara et al. 2010; Szymczak et al. 2018). The diversity of receptors and antireceptors in L. lactis and S. thermophilus may explain the narrow host ranges of their phages, which typically are only able to infect a small subset of strains (Mahony et al. 2016; Lavelle et al. 2018a; Oliveira et al. 2018).

Host receptors

Cell wall polysaccharides

The receptors for most S. thermophilus and L. lactis phages studied to date, including representatives of 4 streptococcal phage groups (exception is the newly described P738 group) and 6 of the 10 lactococcal phage groups (936, P335, 949, P087, 1358 and 1706), are saccharidic in nature (Mahony et al. 2015; McDonnell et al. 2017; Szymczak et al. 2018a; Marcelli et al. 2019). The lactococcal cell wall is decorated with polysaccharides known as cell wall polysaccharides (CWPS), which are encoded by the rgp (rhamnose-glucose polysaccharide) gene cluster (often referred to as the cwps gene cluster). These polysaccharides were shown to play a role in lactococcal cell wall assembly and cell division, and mutants deficient in CWPS display impaired growth, cell deformities and increased sedimentation in liquid culture (Dupont et al. 2004a; Ainsworth et al. 2014a; Theodorou et al. 2019). Similarly, rgp gene clusters in S. thermophilus are predicted to encode a cell wall-associated rhamnose-glucose polysaccharide that has been shown to affect cell division and morphology in another Streptococcus species (Hols et al. 2005; De et al. 2017). In S. thermophilus, the term CWPS has been used to encompass cell wall-associated polysaccharides encoded by rgp as well as some eps (exopolysaccharide) operons (McDonnell et al. 2020).

RGP

The large, chromosomally encoded rgp gene clusters are composed of highly conserved regions as well as variable regions (Wels et al. 2019). In S. thermophilus, the rgp operon comprises a conserved set of genes (rgpA through rgpF) involved in the assembly of the rhamnose-glucose polysaccharide and a variable region upstream of the rgpAF cluster (Hols et al. 2005). A recent phylogenetic analysis identified five groups of S. thermophilus rgp operons, designated types A to E (Szymczak et al. 2019). We extended this phylogenetic analysis to all currently available rgp sequences, including those present in complete or draft genomes from both public (NCBI) and proprietary (DuPont) databases. rgp gene clusters were extracted as the sequences located between the rpoD (upstream) and radC (downstream) genes, and all the coding sequences (CDS) identified across all rgp gene clusters were grouped into orthologous gene families by using a customized bioinformatics pipeline based on blastn (criteria for inclusion of two CDS into a given gene family: ≥70% identity over at least 70% of the alignment length). The genetic organization of rgp gene clusters was inspected manually following a graphic representation whereby each CDS was attributed a color that is specific to each gene family (Fig. 1; Table S1, Supporting Information). Overall, 167 rgp sequences could be clustered into three major groups (A to C) that are each characterized by a specific rgpF gene (55–64% identity across groups, while the rgpF sequence identity is >95% within each group). Based on gene content, each group could be further divided between five and seven subgroups. Groups A to C correspond to the former types A to C described by Szymczak et al. (2019), while former types D and E now correspond to subgroups A.7 and C.4, respectively. Of note, only four genes are core and shared by all rgp gene clusters: rgpA, rgpB, rgpC and rmlD. Within group A, two distinct rgpD–rgpE gene pairs exist (see Fig. 1), while group B is characterized by the absence of a rgpE ortholog. A large majority of the genes located upstream of rgpA are found in at least two groups, further illustrating the propensity of this locus to be acquired by horizontal gene transfer.

Figure 1.

Overview of the diversity and classification of rgp gene clusters in S. thermophilus. Based on 167 sequences, three groups (A to C) and 18 subgroups (A.1 to C.5) could be distinguished. Genes that are similar (i.e. showing at least 70% identity over at least 70% of the alignment length) across groups and/or subgroups are colored, while the others (unique genes) are shown in white. The group-specific rgpF genes, which could be used as targets for the development of PCR typing tools, are framed in bold. Four genes are core and shared by all rgp gene clusters: rgpA, rgpB, rgpC and rmlD (coding for a dTDP-4-keto-l-rhamnose reductase).

In L. lactis, the rgp gene cluster contains a highly conserved region encoding rhamnan biosynthesis and export genes as well as a variable region of diverse glycosyltransferases that has allowed for the CWPS to be clustered into four main types (A, B, C and D), with C broken down further into eight subtypes (Mahony et al. 2013a, 2020; Ainsworth et al. 2014a). This classification scheme was derived from an analysis of the rgp loci of 107 lactococcal strains (Mahony et al. 2020). Here, based on many more L. lactis genomes, including a majority of unpublished sequences, the rgp genetic diversity can be categorized into at least 8 major types, which can be further divided into at least 21 subtypes. Consequently, we propose that the original CWPS multiplex-PCR tool for the molecular identification of types A, B and C (Mahony et al. 2013a) be updated, as the proposed primer sequences no longer accommodate the existing sequence polymorphisms of each targeted gene, but more importantly because the targeted gene is either found in multiple, distinct types (an ortholog of LLKF_205 is present in up to seven types, and not only in type B) or is not core within the type (case of the llmg_0226 ortholog that is present in many but not all type C strains).

Figure 2 provides an updated overview of the rgp gene cluster diversity and classification. Based on almost 470 L. lactis rgp operon sequences and using an approach similar to the one described above for S. thermophilus rgp gene clusters, eight types including the four former ones (A to D; Mahony et al. 2013a, 2020) along with four novel clusters (named E to H) can currently be distinguished. Detailed information about each type is further provided in Table S2 (Supporting Information), notably strain (genome) names, sequence accession numbers and locus tags. Interestingly, both type strains for subspecies cremoris (e.g. ATCC 19257) and lactis (e.g. ATCC 19435) harbor the same CWPS type (but distinct subtypes, namely C.1 and C.6, respectively), likely explaining that C is also the most represented type within the species (39%; 183 out of 468 sequences). In decreasing frequency, strains harboring types D, A and B each comprise >10% of the available genomes, while strains encoding types E and F are rare, representing <1%. In most subtypes, with the exception of C.3 and C.4, the ratio between cremoris and lactis subspecies is markedly unbalanced, a contrast that is less visible in type C, at type level. Due to the apparent plasticity and polymorphism of the rgp gene cluster in L. lactis, an even sharper distribution across types would be expected, as strains belonging to distinct subspecies should intuitively fall into distinct types. The fact that strains from both subspecies share the same rgp type could be explained by horizontal gene transfer events, a hypothesis that is sustained by complementary genetic analyses such as multi locus sequence analysis (MLSA; data not shown). Since CWPS are involved in phage resistance (Dupont et al. 2004a; Mahony et al. 2013a), the exchange of all or part of this gene cluster could provide a drastic change to the strain sensitivity to phages, possibly explaining the mixed subspecies in subtypes A.1, A.2, C.1, C.2, C.3, C.4, H.1 and H.2. Types B, E, F and G, as well as subtypes C.5, C.6, C.7, D.1, D.2, D.3 and D.5, include strains belonging exclusively to the lactis subspecies, while the cremoris subspecies is dominant in subtypes C.1 and C.2. Of note, the type strains of subspecies hordniae (e.g. DSM 20450) and tructae (e.g. DSM 21502) are of the same type but distinct subtypes (H.2 and H.3, respectively).

Figure 2.

Overview of the diversity and classification of rgp operons in L. lactis. Based on >400 sequences, nine types (A to I) could be distinguished. Genes that are similar (i.e. showing at least 70% identity over 70% of the alignment length) across types are colored, while the others (unique genes) are shown in white. Black stars in genes of types A, B and C indicate the targets of the existing multiplex PCR method, while the blue stars indicate the target (rmlB) of the control PCR amplification (Mahony et al. 2013a).

To date, the RGP structure has been determined only for one S. thermophilus strain, ST64987, and was found to be composed of a backbone tetrasaccharide repeating units, carrying tri- and tetrasaccharide side chains with GlcNAc at branching points (McDonnell et al. 2020). More detailed analyses have been performed on the L. lactis rgp-encoded CWPS. All lactococci studied thus far encode a rhamnose-rich polysaccharide associated with the cell wall, predicted to be encoded by the conserved region of the rgp cluster, although some diversity is still observed in the rhamnan-encoding genes and rhamnan structures (Mahony et al. 2020). The variable region of the gene cluster encodes a poly/oligosaccharidic side chain that is present on the cell surface, termed the polysaccharide pellicle (PSP) (Chapot-Chartier et al. 2010; Mahony et al. 2020). The PSP is responsible for greater structural differences observed among CWPS types, and these differences may help to explain the narrow host ranges of L. lactis phages (Vinogradov et al. 2018a; Mahony et al. 2020). The chemical structures of the CWPS for representatives of all four CWPS types as well as six of the eight C subtypes have been analyzed (Mahony et al. 2020). A rhamnose-rich polysaccharide with branching oligosaccharides was isolated from a representative strain of CWPS type A, and a rhamnose-rich polysaccharide with a complex branched structure and carrying a partial glycerophosphate substitution was isolated from a representative CWPS type B strain (Vinogradov et al. 2018a,b). In contrast, the CWPS structures of C-type and D-type strains analyzed comprise two elements, the rhamnan and the PSP, which are covalently linked (Mahony et al. 2020). A trisaccharide motif identified in C-type PSPs was proposed to act as the core phage receptor, while the remaining components of the PSP were proposed to determine strain specificity (Farenc et al. 2014). In the case of phage 1358 (reference phage for the rare, eponymous lactococcal phage group), the crystal structure of the 1358 receptor-binding protein in complex with this trisaccharide identified the trisaccharide within the phage receptor-binding site (McCabe et al. 2015).

Multiple lines of evidence have confirmed streptococcal and lactococcal RGP as phage receptors. For S. thermophilus, phage adsorption assays on chemically and enzymatically treated purified cell walls ruled out a proteinaceous phage receptor and implicated cell envelope components (Quiberoni, Stiefel and Reinheimer 2000). Phage inhibition experiments with different saccharides found that streptococcal phages are inhibited by carbohydrates that are present in the CWPS (Quiberoni, Stiefel and Reinheimer 2000; Binetti, Quiberoni and Reinheimer 2002; Mahony and van Sinderen 2012). Analysis of phage attachment to different cellular fractions of the host implicated cell wall polysaccharides intercalated with peptidoglycan (Szymczak et al. 2018). Finally, S. thermophilus phage-resistant mutants were found to have mutations in genes encoding glycan biosynthetic pathways including those of the rgp operon (Szymczak et al. 2018). For L. lactis, mutant strains deficient in CWPS have been shown to exhibit increased phage resistance (Chapot-Chartier et al. 2010; Theodorou et al. 2019). Swapping of CWPS type between lactococcal strains resulted in an altered phage sensitivity phenotype (Ainsworth et al. 2014a). In particular, CWPS swapping was used to demonstrate that members of the rare 949 and P087 phage groups use CWPS as receptors (Mahony et al. 2015a). Finally, molecular interactions between phage RBPs and CWPS have been demonstrated in vitro and in silico (Bebeacua et al. 2013b; McCabe et al. 2015).

Exopolysaccharides (EPS)

Chromosomally encoded, Wzx/Wzy-dependent pathway eps operons are commonly found in S. thermophilus strains, and six eps groups (type A to F) were recently identified (Hols et al. 2005; Szymczak et al. 2019). A study by Rodríguez et al. (2008) found S. thermophilus strains that encode capsular exopolysaccharides (CPS) were generally more phage sensitive, and a CPS-negative mutant of strain CRL1190 showed reduced phage adsorption. Additionally, bacteriophage-insensitive mutants (BIMs) of S. thermophilus CSK944 lost the ability to produce EPS concomitantly to becoming phage resistant (Mills et al. 2010). Recent studies on a small number of S. thermophilus mutants that show resistance against respective phages of the cos, pac or 987 groups found mutations in genes belonging to eps operons (Szymczak et al. 2018; McDonnell et al. 2020). Analysis of phage attachment to different cellular fractions of the host also implicated EPS associated with the cell surface as phage receptors for these phage/host pairs (Szymczak et al. 2018). Consistent with streptococcal rgp and eps loci being horizontally exchanged, rgp and eps groups do not fully correlate with each other nor with core genome similarity. However, correlations were observed between eps operon type and phage RBP phylogeny of cos-group streptococcal phages (Szymczak et al. 2019). This correlation, taken together with the resistance to representative phages of the cos, pac and 987 group phages conferred by mutations in streptococcal eps genes, implicates EPS as a receptor for certain streptococcal phages encompassing multiple phage groups.

Protein receptors

Most lactococcal phage groups have been shown to utilize carbohydrate receptors, and, to date, only phages of the c2 group have been shown to require a protein receptor (Valyasevi, Geller and Sandine 1991; Geller et al. 1993). While c2 group phages (representative phages c2 and bIL67) are believed to initially recognize carbohydrate (particularly rhamnose) receptors to bind to the host cell wall, for phage c2 this binding is reversible, and interaction with a 32-kDa cell membrane protein secondary receptor designated Pip (phage infection protein) is required for irreversible binding followed by phage DNA ejection (Monteville, Ardestani and Geller 1994). bIL67 does not utilize the Pip receptor, but instead requires a structurally similar membrane protein, YjaE, for infection (Millen and Romero 2016). Pip and YjaE are orthologous to type VII secretion protein YueB, responsible for irreversible binding of Bacillus subtilis phage SPP1 (Jakutytė et al. 2011).

Phage antireceptors

c2

c2 group phages are unique among lactococcal phages due to their requirement for a proteinaceous receptor. Representative phages c2 and bIL67 were among the earliest lactococcal phage genomes sequenced (Schouler Ehrlich and Chopin 1994; Lubbers et al. 1995). Both genomes are ∼22 kb in size and are 80% identical at the nucleotide level with a region of relatively low conservation consisting of structural genes believed to be involved in host-range determination. This poorly conserved region consists of three late expressed genes, l14, l15 and l16 in c2, which correspond to genes ORF34, ORF35 and ORF36 in bIL67. This region likely encodes the antireceptor responsible for interaction with the host-encoded Pip or YjaE protein, as phages more closely related to c2 in this variable region were shown to require Pip for infection, while phages more closely related to bIL67 in this variable region were shown to require YjaE for infection (Millen and Romero 2016).

Little has been published on the initial carbohydrate recognition of c2 group phages. Carbohydrate-binding modules (CBMs) were predicted in each gene of the variable region, and it was shown that recombinant bIL67 isolates that maintained the 5′ and 3′ ends of orf35 but exchanged the middle section for corresponding gene of phage CHL92 acquired the host range and adsorption patterns of CHL92 (Stuer-Lauridsen et al. 2003; Millen and Romero 2016). However, in another lactococcal phage/host system, this region of l15/ORF35, which includes a CBM, was shown to be dispensable for infection (Millen and Romero 2016). A recent analysis of 10 c2 phages was unable to identify a correlation between amino acid sequence divergence in this variable region to host range or CWPS type of the host, although a possible correlation of the l15/orf35 amino acid sequence to a preference for a subspecies cremoris vs lactis host was noted. Additionally, host spectrum analyses found some of these c2 phages could infect only strains with a specific CWPS, while others could infect strains with different CWPS (Chmielewska-Jeznach Bardowski and Szczepankowska 2020).

It was previously reported that predicted tail adsorption protein L10 of phage c2 (more recently predicted to be a tail tape measure protein) is involved in phage DNA entry (Rakonjac, O'Toole and Lubbers 2005). It may play a role in host recognition as it was shown to correlate to host strain subspecies (lactis vs cremoris) and CWPS type. DuPont collection c2 group phages with the long version of this gene, related to l10 of c2, were found to infect primarily subsp. cremoris hosts, although a study by Chmielewska-Jeznach, Bardowski and Szczepankowska (2020) found phages with the long version were able to infect lactis and cremoris subspecies (CWPS C or U). c2 group phages encoding a shorter version, related to orf31 of bIL67, were found to infect primarily subsp. lactis hosts (CWPS B or U) (Cochu-Blachère et al. 2018; Chmielewska-Jeznach, Bardowski and Szczepankowska 2020).

936

The RBP of lactococcal phages belonging to the 936 group was identified by exchanging the host range of phage bIL170 with that of sk1 through construction of chimeric phages (Dupont et al. 2004b). The RBPs of 936 group phages contain three domains: head, neck and shoulder (Fig. 3), with the receptor-binding site found within the head domain. The shoulder domain (N-terminus) of the RBP is highly conserved among 936 group phages, while neck and head (C-terminal) domains show more diversity that is correlated to host specificity (Tremblay et al. 2006). The RBPs are located at the distal end of the phage tail and are part of a larger structure known as the baseplate (Bebeacua et al. 2013b). Five subgroups of the RBP were identified by sequence analysis, and each group can be correlated to the CWPS type of the hosts they infect (Murphy et al. 2016). A more recent structural analysis of the head domains of these RBPs also identified five groups that are consistent with those previously defined (Hayes et al. 2019).

Figure 3.

Molecular model of a 936 group phage baseplate highlighting the structures of distal tail protein (Dit) and RBP subunits. The specific domains of the Dit and RBP units are highlighted accordingly in the figure. Models were constructed utilizing a threading approach for the ORF 15, ORF 16 and ORF 18 of a selected DuPont 936 group phage using p2 homologs as a scaffold for the modeling within the I-Tasser package. In order to reconstruct a baseplate-like structure, the previously solved p2 phage baseplate was utilized (PDB: 2WZP) to superimpose modeled subunits and the subsequent steric clashes were removed using the Chimera package.

Upon performing a reclassification of phage RBPs derived from an excess of 200 genomes, RBPs belonging to previously defined groups (Hayes et al. 2019) and eight additional RBP types, representing new groups, were identified. The complete classification is provided in Fig. 4 whereby one can observe the relationship of the existing groups alongside those newly classified. Importantly, due to the expansion of, and diversity observed within, the p2-like RBP group, this has been newly defined as two distinct RBP types. Within the proposed classification, this group has been denoted as A1 and A2, which correspond to RBP types 3 and 10 in Fig. 4. It was previously specified that the T4-like group gp12 (group 13 in the proposed classification) constituted a significant divergence from the core p2 RBP types, and it is interesting to note that all DuPont representatives of this group consist of phages infecting diacetylactis strains. When one relates this information to the rgp classification presented above, for those systems whereby both phage and complementary host genomic information exist, it was found that there was a direct correlation between the rgp and RBP classifications. At the rgp subgroup level, some variation exists within a given RBP; however, there is no known instance when the subgroup classification is applied.

Figure 4.

Unrooted phylogenetic tree displaying the diversity of RBPs exhibited by sequenced DuPont 936 group phages. Phylogenetic relationships were inferred through the use of the maximum likelihood method with 500 bootstrap repetitions performed. The resulting Newick file was imported into the Interactive Tree of Life (iTOL) to enable visualization. Clades corresponding to specific RBPs were collapsed and color-coded for clarity purposes. Numbers given in black refer to RBP groups identified among DuPont phages, while letters in red refer to the groups described previously by Hayes et al. (2019) and which have been placed beside the corresponding DuPont groups. Note that the p2-like group described previously has been split into A1 and A2 (DuPont groups 3 & 10, respectively) due to the increased diversity observed meriting independent classification.

The entire virion structure of representative 936 group phage p2 has been resolved by electron microscopy (Bebeacua et al. 2013b), and the structure of its baseplate has also been identified by X-ray crystallography (Sciara et al. 2010). Three proteins comprise the p2 baseplate: distal tail protein (Dit), tail-associated lysin (Tal) and the RBP. Dit proteins form a hexameric ring where each Dit attaches a trimer of RBPs, and a Tal trimer forms a dome closing the central channel at the end of the baseplate. A second Dit hexamer is positioned back to back with the first Dit ring. The head domains of the RBPs, which harbor the receptor-binding sites, are pointed upward, away from the host. The addition of Ca²⁺ ions, known to be required for p2 infection, results in conformational change of the baseplate. In this ‘activated’ state, the RBPs are rotated 200° downward, and the Tal dome opens up forming a channel for DNA passage (Sciara et al. 2010; Bebeacua et al. 2013b). The p2 baseplate structure is believed to be highly conserved among 936 group phages, although exceptions have been recently described. Phages were isolated which were found to encode a second unique RBP (Hayes et al. 2018a). Phages with this additional RBP possess an atypical, less compact baseplate structure that is in the ‘activated’ conformation. One RBP is associated with the Dit ring, while the (less numerous) second RBP protrudes from the baseplate. It was suggested that this atypical baseplate correlates with host affinity, as the host ranges of these phages were all limited to the same single host. Additionally, no other 936 group phages were isolated against this host (Hayes et al. 2018a). In addition to RBPs, CBM decorations on the Dit, major tail protein (MTP) and neck passage structure (NPS) were recently shown to facilitate phage binding in 936 group phages. Many of these CBMs showed structural homology to BppA, which is an accessory baseplate protein involved in host binding in P335 phage Tuc2009 (Collins et al. 2013; Hayes et al. 2019). Two types of Dit proteins have been found to exist in 936 group phages: classical and long (or ‘evolved’). The difference is due to internal insertions in the ‘evolved’ Dits, with the N-terminal and C-terminals aligning with the full length of the classical Dit. These insertions were found to contain CBMs. Dit insertions were previously grouped into four classes. Analyses on these four classes showed the Dit-associated CBMs exhibit the same specificity as the RBPs, and a Dit/RBP/CWPS correlation could be made across the 936 group (Hayes et al. 2018b). A global analysis of the insertions existing within DuPont phages possessing ‘evolved’ Dits has highlighted an impressive level of diversity, with as many as 16 distinct groups being present (Figure S1, Supporting Information).

The EM structure of phage p2 revealed that its tail is covered with protruding decorations consisting of the C-terminal adhesion-like domain of the MTP that was theorized to play a role in host attachment (Bebeacua et al. 2013b). Recently, three groups of MTPs were identified in 936 group phages based on size: a ‘short’ group of ∼215 residues, a second group that is ∼100 residues longer than the ‘short’ group, and a third group whose known members measure between ∼470 and 530 residues. This increased length of the third group is the result of a programmed translational frameshift encoding the Tail protein extension (TpeX). Both the second and third groups are believed to encode the adhesion-like domain noted in p2. More interestingly, each TpeX was found to harbor a CBM domain. Host binding of a representative TpeX was found to be weaker than that of Dit, but, like Dit, binding was host specific. CBM domains were also found in the NPS of 936 group phages. The N-terminus of the NPS is conserved among phages that harbor a NPS, while diversity of the remaining residues allowed for classification into seven groups with varying CBMs (Hayes et al. 2019). A representative NPS CBM domain showed the lowest host-binding affinity compared with host-binding affinities of Dit and TpeX (Hayes et al. 2019).

Research to date indicates that 936 group phages use the lactococcal CWPS as a receptor, and binding to the CWPS is mediated by various phage-encoded CBMs (Bebeacua et al. 2013b; Ainsworth et al. 2014a; Hayes et al. 2018b, 2019). In a proposed two-step mechanism of host binding, reversible binding is performed by the accessory CBMs, which is followed by irreversible binding mediated by the RBP. Due to the host specificity of the accessory CBMs, a second mechanism was proposed in which the accessory CBMs simply enhance the binding capability of the phage by allowing for flexible binding orientations (Hayes et al. 2019).

P335

P335 represents a highly diverse group of lactococcal phages with both lytic and temperate members. These phages can be classified into five morphotypes (I–V) based on nucleotide identity and phage morphology, specifically phage adhesion modules (Mahony et al. 2017; Kelleher et al. 2018). These adhesion modules include the C-terminus of the tail tape measure protein, the Dit, the N-terminal region of the Tal, the RBP and baseplate proteins (BPPs), if present (Kelleher et al. 2018). Phages belonging to P335 morphotypes I and V show some conserved features as both possess a fused Tal/RBP that is likely responsible for the long tail fiber structure observed in members of these morphotypes. Large, ‘evolved’ Dit proteins harboring CBMs can also be found in phages belonging to morphotypes I and V (Kelleher et al. 2018). The adhesion modules of P335 phages belonging to morphotypes III and IV resemble those of 936 group phages. These phages encode a small RBP and are characterized by having ‘stubby’ distal tail regions compared with the large RBP-encoding or multicomponent baseplate structure found in morphotype II phages or long tail fiber structures found in morphotypes I and V phages (Mahony et al. 2017; Kelleher et al. 2018).

Representative phages of P335 morphotype II, TP901-1 and Tuc2009, have been studied in detail. The RBPs of phages TP901-1 and Tuc2009 were identified by exchanging the host range of TP901-1 for that of Tuc2009 through construction of a chimeric phage (Vegge et al. 2006b). Like 936 group phages, the RBPs (also known as BppL) of TP901-1 and Tuc2009 exhibit a modular assembly, comprising a head, neck and shoulder domain, with the receptor-binding site located in the head domain (Spinelli et al. 2006a; Legrand et al. 2016). Similar to 936 group phage p2, the baseplates of TP901-1 and Tuc2009 consist of a central core composed of a Dit hexamer and a Tal trimer. In contrast to p2, a second Dit hexamer is not present, and the TP901-1 and Tuc2009 baseplates also include a double-disk peripheral structure composed of six receptor-binding units. For TP901-1, each of these units, known as tripods, consists of a BppU (upper baseplate) trimer connecting three RBP trimers to the Dit core. The Tuc2009 tripod displays a similar structure, with the exception of an additional trimer of a CBM-encoding accessory protein, BppA, located at the top of the tripod. Unlike p2, the RBPs of TP901-1 were shown to be maintained in an ‘active’ conformation, pointed in the direction of the host (Spinelli et al. 2006; Veesler et al. 2012a; Legrand et al. 2016). Morphotype II phages, including Tuc2009 and TP901-1, display highly conserved classical Dit proteins, and their Tal proteins are highly conserved at the N- and C-termini with a variable mid-region. The Tal proteins encode a peptidoglycan hydrolase (PGNase) domain at the C-terminal, used to degrade the host peptidoglycan for more efficient host penetration. Tals of Tuc2009 and TP901-1 have been shown to undergo proteolytic processing to remove the PGNase, which results in a heterogeneous phage population. Phages with the PGNase are better able to infect stationary phase cells, while phages with truncated Tals have higher adsorption efficiencies (Stockdale et al. 2013; Mahony et al. 2017). The remainder of the morphotype II adhesion module shows great diversity and may comprise a single component, large RBP-encoding gene or a multicomponent peripheral baseplate composed of BppU, BppA and BppL. Some phages displaying the multicomponent baseplate, including TP901-1, lack BppA, and phages that encode a second BppU and/or BppA have also been described, illustrating the diversity with the multicomponent baseplate (Mahony et al. 2017). When present, the N- and C-termini of the accessory baseplate protein, BppA, of morphotype II phages are conserved, while the mid-regions are variable and encode CBMs (Mahony et al. 2017). This domain is likely responsible for the phage adsorption shown to be mediated by the BppA of Tuc2009 (Collins et al. 2013). If BppA is present, the phage's BppU harbors a C-terminal extension that attaches BppA to the baseplate (Mahony et al. 2017). BppU proteins can be divided into two groups based on size. The classical group is represented by TP901-1 and Tuc2009. A second group comprises BppUs whose encoding genes are considerably longer than the classical. In the case of phages encoding a single component, large RBP-encoding gene, a BppU with a CBM, is present while BppA and BppL are absent (Mahony et al. 2017).

CWPS swapping was used to demonstrate that some members of the P335 phage group, including TP901-1, use the host CWPS as a receptor (Ainsworth et al. 2014a). However, a direct correlation between specific RBP subgroups of P335 phages and CWPS types was not observed in a survey of 27 sequenced P335 phages (Mahony et al. 2017). This demonstrates the complexity of phage–host interactions among this diverse group of phages.

cos and pac

Most S. thermophilus phages display a narrow host range and belong to one of the previously mentioned five groups, and phage adhesion modules display synteny across these phage groups. Like lactococcal phages, the adhesion modules include TMP, Dit, Tal and RBP proteins (Lavelle et al. 2018a). Tal and RBP were believed to be fused in cos and pac phages, and this fused Tal-RBP, which encodes CBMs, was believed to play the primary role in host recognition and binding (Duplessis and Moineau 2001; Lavelle et al. 2018b). Very recently, it was proposed that the Tal-RBP may not be the primary RBP in cos and pac group phages. Instead, it was suggested that the CBMs encoded by cos and pac group Tals function similarly to what has been reported in lactococcal 936 phages, where other structural phage genes were found to encode CBMs that facilitate host adhesion (Hayes et al. 2019; Lavelle et al. 2020). The gene directly downstream of Tal was implicated as the primary RBP as it encodes a CBM domain with a fold resembling that of the RBP of lactococcal phages p2 and TP901-1. Furthermore, fluorescence microscopy found the putative RBP CBM of cos group phage STP1 to exhibit higher host-binding affinity than its Tal-associated CBM (Lavelle et al. 2020).

A clear correlation between cos Tal (formerly Tal-RBP) phylogeny and primary host and/or host range has been observed; however, while this correlation has also been observed for pac phages, several pac phages with distinct Tals were found to exhibit a broader host range with incidences of overlapping strains (Lavelle et al. 2018a; Szymczak et al. 2019). The Tals of both cos and pac phages consist of a highly conserved N-terminal gp27-like structural domain; however, a subgroup of pac Tals has two putative CBMs or hydrolases inserted in this domain (Lavelle et al. 2020). This structural domain is followed by a collagen-like domain and then one or two CBMs. The Tal C-terminus consists of a probable collagen-like linker followed by an unknown domain (Lavelle et al. 2020). The CBMs are found in regions that have been referred to as variable regions one and two (VR1 and VR2) due to their diversity (Duplessis and Moineau 2001; Lavelle et al. 2018b, 2020). VR2 is a highly divergent region within Tal, and this region of the encoding gene has been employed as a target in phage detection and characterization (Binetti et al. 2005). Analysis of DuPont cos and pac phage Tal revealed the presence of six superclusters for each variable region. This classification is presented in Fig. 5. In both cases, the diversity within each cluster is extraordinary, rendering precise grouping almost impossible. Indeed, close inspection of VR1/VR2 alignments shows that the resolution of this diversity is apparent at the single phage level in our phages.

Figure 5.

Unrooted phylogenetic tree displaying the diversity of Tal exhibited by sequenced DuPont cos(A) and pac(B) phages. Phylogenetic relationships were inferred through the use of the maximum likelihood method with 500 bootstrap repetitions performed. The resulting Newick file was imported into the iTOL to enable visualization. Six superclusters were identified among both cos and pac phages and are indicated by corresponding branch colors and shaded boxes. The diffuse nature of the clusters renders subgroup classification difficult but highlights an impressive level of diversity.

VR2 has been shown to be involved in host specificity as chimeric cos phages that swapped VR2 displayed altered host range. Additional genetic determinants are believed to influence host specificity since chimeric phages were able to adsorb to both a new host and their original host. Furthermore, while VR2 regions often cluster according to host strain independently of phage group (cos vs pac), phages with similar VR2 but non-overlapping host range have been reported as well as phages with unrelated VR2 but overlapping host range (Duplessis and Moineau 2001; Guglielmotti et al. 2009b; Zinno et al. 2010; McDonnell et al. 2017). In addition to the CBMs found in Tal and the putative RBP, cos and pac phages exhibit ‘evolved’ Dit proteins that harbor additional CBMs. As previously described for lactococcal 936 group phages, these accessory CBMs may enhance the binding capability of the phage (Lavelle et al. 2018b). Mutations in Dit as well as the Tal and/or TMP were observed in phage isolates that were able to infect a phage-resistant derivative of the host (Duplessis, Lévesque and Moineau 2006).

5093 and 987

5093 group phages are unique in that they possess globular structures attached to the base of the tail (Mills et al. 2011; Szymczak et al. 2017). The genes comprising the adhesion module of the 5093 group phages were predicted based on genome synteny with other S. thermophilus phage groups and the presence of conserved domains. The putative RBP was verified by phage adsorption inhibition studies. This RBP is proposed to possess a carbohydrate-binding function, suggestive of a carbohydrate host receptor for the 5093 phage group (McDonnell et al. 2017). In contrast to cos and pac phages, 5093 group Dit and Tal proteins have not been found to encode CBMs (Lavelle et al. 2020).

987 group phages possess shorter tails and a broad baseplate structure resembling that of lactococcal P335 phages (McDonnell et al. 2016; Szymczak et al. 2017). Consistent with this P335-like baseplate, the morphogenesis modules of 987 group phages show nucleotide relatedness to those of the lactococcal P335 group. Similar to the single component, large RBP-encoding gene found in some group II morphotype P335 phages (discussed above), the 987 baseplate appears to be encoded by a single gene that shares N-terminal amino acid identity with the P335 BppU (McDonnell et al. 2016; Mahony et al. 2017). Additionally, this putative RBP encodes a large domain with the fold of the RBP of B. subtilis podophage Phi29 (Lavelle et al. 2020). This gene is proposed to have a carbohydrate-binding function and to be responsible for host binding and specificity. This similarity between the 987 and P335 morphogenesis modules suggests certain phages of both groups may recognize a common host receptor, and indeed, representative 987 group phages were shown to adsorb to certain L. lactis strains in addition to their S. thermophilus host (McDonnell et al. 2016). Adsorption inhibition assays with the purified putative 987 group RBP confirmed its role in host adsorption. The proposed carbohydrate-binding function of the RBP and its homology to P335 phages known to recognize cell surface carbohydrates is suggestive of a carbohydrate host receptor for the 987 group phages (McDonnell et al. 2016). This was confirmed by the analysis of 987 group phage CHPC926 that determined that the phage binds to EPS associated with the cell surface (Szymczak et al. 2018). Additionally, SNPs in the eps gene cluster were identified in 987 group BIMs of ST64987 (McDonnell et al. 2020). Additional determinants may be involved in 987 group host recognition. A highly conserved gene with unknown function can be found immediately downstream of the 987 group RBP. The encoded protein was proposed to play an accessory role in host recognition due to the presence of a sialic acid O-acetyltransferase domain, as O-acetylation has been shown to be present at precise locations of the sialic acid component of the capsular polysaccharide of group B Streptococcus (McDonnell et al. 2016). Like the 5093 group, 987 group Dit and Tal proteins have not been found to encode CBMs (Lavelle et al. 2020).

The diversity of host receptors and phage antireceptors in L. lactis and S. thermophilus shows the complexity of phage–host interactions and illustrates the difficulty of finding broadly applicable industrial phage solutions. However, the understanding of receptors and antireceptors is still of great value with regard to industrial starter cultures. Knowledge of host receptors, including the CWPS and EPS structures and encoding genes, may allow for a prediction of phage sensitivity and for determination of phage-related strains without the need for phage testing (Mahony et al. 2014). This should be taken into consideration for formulation of mixed defined starter cultures. To enhance phage robustness, it may be advantageous to design cultures to contain strains belonging to different CWPS and EPS groups and avoid combining strains with identical CWPS or EPS types (van Sinderen et al. 2014). In addition, this knowledge provides targets for mutation to increase phage robustness of starter strains. Furthermore, awareness of the phage landscape, including the nature of the antireceptors, will allow for a more intelligent use of phage mitigation strategies.

HOST RESPONSES TO PHAGES (RESISTANCE TO PHAGES) AND PHAGE COUNTER RESPONSES

Phage–host interactions with a view toward native anti-phage mechanisms in L. lactis and S. thermophilus have been extensively reviewed in the literature (selected reviews include Allison and Klaehammer 1998; Coffey and Ross 2002; Garneau and Moineau 2011; for more general reviews on bacterial phage resistance, the reader is further referred to Labrie, Samson and Moineau 2010; Dy et al. 2014b; van Houte, Buckling and Westra 2016). Phage sensitivity begins with the recognition of and adsorption to host cell surface determinants by the infecting phage, followed by DNA injection, making cell surface modifications a first line of defense against phage (see below). Should phage DNA gain entry into the host, it is confronted with a diversity of DNA destructive mechanisms in the form of restriction/modification (R/M) and CRISPR (Clustered Regularly Interspaced Short Palindromic Repeats) and Cas proteins (CRISPR-Cas) systems (Makarova, Wolf and Koonin 2013). The final defensive measures consist of abortive infection and toxin–antitoxin systems, which upon infection trigger cell death, thereby abrogating completion of the phage propagation cycle.

Lactococcus lactis response to phages

The native host cell surface structures that serve as discriminating recognition sites for phage RBPs are potential objects of mutations that confer resistance (Chapot-Chartier 2014). In discovering that phage infection of the starter culture caused fermentation failures, it was observed that surviving bacteria were found to be phage resistant (Whitehead 1953). These variants (also termed BIMs) are naturally present in the starter as a minor population that could be selected for upon exposure to phage (Whitehead 1953). The BIMs were often functionally impaired in the cheese-making process (slow acidifying) or quickly became phage sensitive due to reversion or appearance of new phages (Lawrence 1978; King, Collins and Barrett 1983; Klaenhammer 1984; Coffey et al. 1998). The diversity of lactococcal phage types capable of infecting a strain and the propensity for phages to rapidly evolve explain the latter two observations. The slow acidification of BIMs may be explicable, at least in part, in that many are adsorption mutants, which is consistent with CWPS mutation, and such mutations have been shown to result in impaired growth, cell deformities and increased sedimentation in liquid culture (Dupont et al. 2004a; Ainsworth et al2014a; Theodorou et al. 2019). With careful selection and characterization, the generation of phage-resistant variants eventually proved an effective means for replacing a phage-sensitive parental strain and continues as a viable option for starter strain development (Limsowtin and Terzaghi 1976; Marshall and Berridge 1976).

Phage exclusion—lactococcus

Within the DuPont phage collection, c2 group phages are the most frequently detected against dairy starter lactococci. As discussed earlier, the representative Ceduoviruses, phages c2 and bIL67, utilize separate protein receptors encoded by pip and yjaE, respectively. Inactivation of the pip gene in L. lactis C2 provided complete resistance to phage c2 (Garbutt, Kraus and Geller 1997; Kraus and Geller 1998). Subsequently, a second transmembrane protein, YjaE in L. lactis IL1403, was shown to be required for infection by c2 group phage bIL67 (Stuer-Lauridsen and Janzen 2006). The understanding that late-expressed phage genes l14–15–16 (c2) and ORF34–35–36 (bIL67) determine the requirement for either Pip or YjaE, respectively, facilitated a targeted approach toward mutagenesis of pip and yjaE for developing BIMs. In the generation of c2 group resistance, mutations usually resulting in protein truncation or disruption were primarily found in pip or yjaE, when using a c2-c2 or c2-bIL67 type in the challenge, respectively.

In contrast to c2 group phages, a target for mutagenesis conferring broad resistance against 936 and P335 group phages has not yet been identified. Mutations giving resistance to 936 and P335 phages have been found in the genes encoding primary receptor structures categorized as CWPS. Dupont, Janzen and Vogensen (2004a) utilized insertional mutagenesis to search for lactococcal genes involved in phage adsorption. Lactococcuslactis IL1403 and Wg2 integrants, resistant to their respective 936 group phages, were mutated in host genes related to cell wall polysaccharide synthesis and showed reduced phage binding. An investigation of the lactococcal CWPS pellicle found that mutations in L. lactis MG1363 polysaccharide biosynthesis-associated gene llmg_0226 gave resistance to 936 group phage sk1 (Chapot-Chartier et al. 2010). Ainsworth et al. (2014a) exchanged the variable CWPS region between two different C-subtype strains, L. lactis MG1363 (C1) and 3107 (C2), and showed altered phage adsorption. The authors reported that inactivation of CWPS gene LLNZ_01145 in a variant of strain L. lactis NZ9000 (itself a variant of L. lactis MG1363) resulted in resistance to a member of the 936 group. Furthermore, the L. lactis NZ9000 variant now producing the exchanged L. lactis 3107 subtype C2 CWPS became sensitive to L. lactis 3107 P335 group phage φLC3. Lemay et al. (2019) applied a proteomic analysis to study 936 group phage p2 infection of L. lactis MG1363. They identified unique and differentially expressed proteins between infected and uninfected cells. Employing CRISPR-Cas editing, they inactivated gene llmg_0219, which codes for an unknown protein and resides within the rgp gene cluster encoding the polysaccharide pellicle. Inactivation of llmg_0219 conferred phage resistance resulting from reduced phage adsorption. For all of the above examples, the mutations conferring phage resistance were found in the rgp genes, downstream of rgpD (Mahony et al. 2013a; Mahony et al.2020).

Phage exclusion can result from factors other than alteration of cell surface receptors. Several reports of phage resistance via adsorption blocking are found in the literature (see reviews by Allison and Klaenhammer 1998; Forde and Fitzgerald 1999a). Inhibition of phage adsorption is often associated with the production of an extracellular carbohydrate material that may be masking cell surface receptors. LABs are known to produce exopolysaccharides (EPS, which encompasses both extracellular EPS and CPS) that are distinct from CWPS and proposed to have a cellular protective function. EPS-producing strains are used for texture development in fermented dairy products (Zeidan et al. 2017). In lactococci, EPS biosynthesis has been reported to often be plasmid encoded; pNZ4000 and pCI658 are two examples of plasmids that have been extensively studied. pNZ4000 was isolated via conjugal transfer to plasmid-free L. lactis MG1614 and subsequently sequenced (van Kranenburg et al. 1997; van Kranenburg, Kleerebezem and de Vos 2000). Following this, an MG1614 transconjugant containing pNZ4030 (pNZ4000 containing an erythromycin resistance marker; van Kranenburg et al. 1997) was shown to reduce plaque formation (by 50%) and plaque size as compared with the MG1614 parent (Looijesteijn et al. 2001). The reduced phage sensitivity was attributed to an EPS layer protecting the EPS producer, as addition of purified EPS to MG1614 provided no protection. Although only a modest reduction in phage sensitivity was reported, plasmid-encoded EPS production provided significant protection, enabling bacterial growth when the initial phage infection was low. Plasmid pCI658 was also identified via conjugal transfer from a phage insensitive starter to model laboratory strain L. lactis MG1363, which is sensitive to phages c2 and ϕ712 (936 group) (Forde, Daly and Fitzgerald 1999). Transconjugants were observed to generate a ‘fluffy’ pellet upon centrifugation and were resistant to 936 group phage ϕ712 (full resistance) and c2 (partial resistance). Further characterization and sequencing of the EPS biosynthetic genes established pCI658 encodes a hydrophilic exopolysaccharide that masks cell surface receptors, thereby blocking phage adsorption (Forde and Fitzgerald 1999b, 2003). While extracellular EPS may indeed be a mechanism for phage inhibition, virulent phages against known lactococcal EPS producers would indicate that resistance is likely phage- and host dependent. Indeed, this was the conclusion from a phage–host range study of 936 group against EPS-positive and EPS-negative lactococcal strains (Deveau, Van Calsteren and Moineau 2002). No clear correlation was observed between phage susceptibility, and the presence, absence and composition of EPS. Furthermore, there was no difference in phage adsorption by EPS-positive strains and their EPS-negative plasmid-cured derivatives, indicating that EPS was not involved in these instances.

Lastly, resident prophage can also confer host phage resistance (Samson et al. 2013a). Superinfection exclusion (Sie) is a phage-encoded mechanism that protects the host from further infection by specific phages. While the mode of action varies, Sie mechanisms generally prevent phage DNA entry and include altering membrane-associated DNA injection site, preventing peptidoglycan degradation, interfering with host phage binding surface receptors, expression of intracellular repressor genes and anti-sense RNA (Labrie, Samson and Moineau 2010; Bondy-Denomy et al. 2016; van Houte, Buckling and Westra 2016; Owen et al. 2020). First described in L. lactis for P335 prophage Tuc2009, a gene named sie₂₀₀₉ encodes a membrane-associated protein capable of blocking DNA injection of 936 group phages (McGrath, Fitzgerald and van Sinderen 2002). This finding was extended to L. lactis subsp. lactis IL1403 and L. lactis subsp. cremoris MG1363 prophages harboring sie homologs that provided resistance to an additional set of 936 group phages (Mahony et al. 2008).

Lactococcal R/M and CRISPR-Cas

Upon host entry, phage DNA may encounter two notable mechanisms for recognition and subsequent degradation in lactococci: restriction/modification (R/M) (Allison and Klaenhammer 1998; Forde and Fitzgerald 1999a; Coffey et al. 2001) and CRISPR-Cas (Millen et al. 2012). Numerous R/M systems have been described in L. lactis, with strains containing multiple systems encoded chromosomally and/or on plasmids. Although protection afforded by R/M is seldom complete, with escaping phage now methylated and no longer recognized as foreign, inhibition is wide-ranging with respect to phage type and especially effective when combined with other R/Ms or other phage resistance mechanisms. By stacking multiple plasmid-encoded R/M systems of different specificities, the log reduction of phage could be increased several-fold when compared with individual R/Ms (Josephsen and Klaenhammer 1990). Type I R/Ms are characterized by three elements: a restriction endonuclease (HsdR), a methylase (HsdM), and a specificity component (HsdS) that directs the system to its specific nucleotide sequence target. In L. lactis IL1403, Schouler et al. (1998) demonstrated that a lactococcal plasmid encoding only one HsdS subunit could interact with the chromosomal Type I R/M to enhance phage inhibition. Likewise, Seegers, van Sinderen and Fitzgerald (2000) identified native plasmid pCIS3 in starter strain L. lactis UC509.9 that contains an hsdS gene that interacts with a resident R/M to provide a 10^–4 reduction in the efficiency of plaquing (EOP) for phage Tuc2009. When introduced into IL1403, the pCIS3 hsdS also provided EOP reduction similar to that reported by Schouler et al. (1998). In addition to co-opting different HsdS targeting functions, O'Sullivan et al. (2000) reported that two hsdS genes residing on separate plasmids recombined to form cointegrate plasmid pAH90 that exhibited an altered R/M specificity as compared with the two parental plasmids. Therefore, shuffling of hsdS genes can provide a combinatorial route for lactococci to diversify Type I R/Ms in response to phage pressures. Potential R/M synergy is further evidenced in a study of the lactococcal plasmidome, where the authors described unique Type I R/M shufflon systems residing on two megaplasmids and the presence of 77 orphan hsdS specificity subunit genes in the lactococcal genomes was examined (Kelleher et al. 2019).

In contrast to the general presence of R/Ms in lactococci, the occurrence of CRISPR-Cas systems is relatively rare. CRISPR-Cas systems form an adaptive microbial immune system, protecting cells against invading foreign nucleic acids (see corresponding S. thermophilus section for further description; selected reviews: Deveau, Garneau and Moineau 2010; Karginov and Hannon 2010; Marraffini 2015; Rath et al. 2015; Koonin, Makarova and Zhang 2017; Pushnick et al. 2017; Brandt and Barrangou 2019). The system functions in three stages: (i) adaptation, (ii) maturation or processing and (iii) interference. Fundamentally, using phage infection as an example, a Cas protein complex will sample and insert a segment of the infecting phage DNA (spacer) into the CRISPR array, in a step known as adaptation. Survivors of the infection contain the acquired spacer as a vestige of the phage encounter. Subsequently, the transcription of the CRISPR array is processed or matured into small CRISPR RNAs (crRNA) that will serve as a guide to direct the Cas effector complex to any DNA sequence containing the spacer sequence. Upon recognition, the targeted DNA is degraded by the Cas-effector complex, thereby interfering its further propagation.

In Lactococcus, a single CRISPR-Cas system has been found thus far. Lactococcus lactis DGCC7167 contains a self-transmissible plasmid that encodes a Type III-A CRISPR-Cas that confers resistance to various 936, P335 and 949 group phages, and is conjugally transferrable to other strains (Millen et al. 2012). While the cas operon is conserved among a few DuPont collection strains, the CRISPR arrays vary in spacer number and sequence, with a limited number of shared spacers. As of the preparation of this review, in vivo spacer acquisition by Type III-A CRISPR-Cas systems had not been demonstrated. Regardless, introduction and expression of synthesized, phage-targeting spacers, on a cloning vector could direct the system to inhibit any phage. The ‘programmability’ of the lactococcal CRISPR-Cas provides a potentially unlimited capability for combating phage infection.

Lactococcal abortive infection

As a last line of defense, phages evading degradation by R/M or CRISPR-Cas systems may be confronted by a diverse set of anti-phage immunity mechanisms collectively termed abortive infection (Abi) (Lopatina, Tal and Sorek 2020). Abi systems can be viewed as altruistic in that the individual cell, sensing it has been infected, is induced to arrest cellular growth or die before the phage can complete its life cycle, thus protecting the population from further infection (Chopin, Chopin and Bidnenko 2005; van Houte, Buckling and Westra 2016). In L. lactis, over 20 distinct Abi genetic elements have been described (AbiA–AbiZ), which have varying degrees of inhibition against representatives of the commonly problematic phages in the dairy industry (Chopin, Chopin and Bidnenko 2005).

Due to the diversity and complexity of phenotypes, the mechanistic details of many of the lactococcal Abis are not completely understood. According to the review by Labrie, Samson and Moineau (2010), Abis interfere with DNA replication (AbiA, AbiF, AbiK, AbiP, AbiT), RNA transcription (AbiB, AbiG, AbiU), production of phage capsid proteins (AbiC) or packaging of phage DNA (AbiE, AbiI, AbiQ). Select systems have been studied in more depth. AbiZ accelerates cellular lysis and death of infected cells through interaction with the phage-encoded holin and lysin (Durmaz and Klaenhammer 2007). For AbiD1, abiD1 transcription is repressed under normal conditions and activated in the presence of a phage protein following infection. The resulting AbiD1 protein inhibits a phage RuvC-like endonuclease, subsequently impairing phage DNA maturation and packaging (Bidnenko et al. 2009). Of interest is the characterization of AbiE and AbiQ as Type IV and Type III toxin-antitoxin (TA) systems, respectively (Samson et al. 2013a; Dy et al. 2014a). TA systems are composed of a lethal toxin and a cognate protective antitoxin, of which there are six types based on the toxin–antitoxin interaction (Harms et al. 2018). The AbiE anti-phage phenotype was initially attributed to interfering with phage packaging. However, the activation of the AbiE Type IV toxin component halts cellular and phage processes, which may likely be the actual cause for aborted infection (Dy et al. 2014a). Although the exact mechanism for AbiQ abortive infection has yet to be determined, experimental results suggest that phage-encoded components may interact with the antitoxin, enabling the lethal action by the toxin (Samson, Bélanger and Moineau 2013). The diversity of Abi systems reflects the ongoing competitive response to virulent phage. That the majority of Abis in L. lactis are plasmid-encoded illustrates that transferability is a practical attribute for evolving host phage resistance and facilitates their utility for strain construction (Chopin, Chopin and Bidnenko 2005; Ainsworth et al. 2014b).

Lactococcal mechanism stacking

In fact, many lactococci harbor multiple phage resistance mechanisms, either on individual plasmids or clustered onto one element (Allison and Klaenhammer 1998; Forde and Fitzgerald 1999a; Mills et al. 2006; Ainsworth et al. 2014b), indicative of the adaptive measures dairy starter strains employ to counter phage pressure. Two early examples are plasmids pNP40 (McKay and Baldwin 1984) and pTR2030 [Klaenhammer and Sanozky 1985; later independently isolated as pCI829 (Coffey, Fitzgerald and Daly 1989)], both of which possess multiple resistance genes and are naturally self-transferable. pNP40 encodes two abortive infection determinants, AbiE and AbiF (Garvey, Fitzgerald and Hill 1995), a DNA penetration blocking mechanism (Garvey, Hill and Fitzgerald 1996) and R/M LlaJI (O'Driscoll et al. 2004, 2006). pTR2030 was isolated from phage-resistant starter strain L. lactis ME2, which was shown to harbor five unique plasmid-encoded phage resistance determinants (Sanders and Klaenhammer 1983; Higgins, Sanozky-Dawes and Klaenhammer 1988; Hill, Pierce and Klaenhammer 1989; Klaenhammer 1989; Durmaz, Higgins and Klaenhammer 1992). Residing on pTR2030 are AbiA (originally termed Hsp; Hill et al. 1989; Hill, Miller and Klaenhammer 1990), AbiZ (Durmaz and Klaenhammer 2007) and R/M LlaI (Hill, Miller and Klaenhammer 1991; O'Sullivan, Zagula and Klaenhammer 1995). The natural stacking of mechanisms enhances overall resistance to phage and, when coupled with conjugative ability, exemplifies how phage pressures have naturally adapted defensive mechanisms to enable lactococcal starters to survive in the industrial environment, and provides strategies for strain development.

Streptococcus thermophilus response to phages

Phage exclusion—S. thermophilus

As with L. lactis, the diversity of cell wall polysaccharide structures provides an initial barrier against broad-host range phage infection in S. thermophilus (Szymczak et al. 2018, 2019). In reports of spontaneous (phage selected) or induced generation of S. thermophilus BIMs where the mechanism of resistance has been investigated, only a few have been associated with phage blocking, indicative of alterations to cell surface features (Szymczak et al. 2018). For example, phage-resistant mutants of S. thermophilus Sfi1 were generated using lactococcal insertion sequence ISS1 (Lucchini, Sidoti and Brüssow 2000). For one mutant, the insertion was located in a gene that was annotated as a transmembrane protein speculated to be involved in DNA injection, and for another mutant, ISS1 was found to disrupt expression of a Type I R/M. In another example, spontaneous phage-resistant variants were characterized with either reduced adsorption, active R/M systems or lysogeny conferring phage exclusion (Binetti, Bailo and Reinheimer 2007). A study by McDonnell et al. (2020) clearly demonstrated that cell wall associated polysaccharides are involved in phage sensitivity. For one BIM, reduced adsorption by a 987 group phage was due to a mutation in a polysaccharide polymerase resulting in the absence of a hexasaccharide repeating unit in the CWPS structure.

With respect to EPS, selected S. thermophilus strains are known producers and used extensively for natural textural properties in fermented milks (Broadbent et al. 2003; Zeidan et al. 2017). As seen with L. lactis, texturizing S. thermophilus strains are equally as susceptible to phage as non-texturizing strains, indicating that EPS does not de facto confer protection. Rodriguez et al. (2008) studied eight S. thermophilus CPS and slime-EPS (secreted extracellular EPS) producers and found no general correlation between polysaccharide production and sugar composition with phage sensitivity/resistance; producers and non-producers alike were equally susceptible to phage. However, these authors isolated a chemically induced mutant that was unable to produce CPS but maintained secreted slime-EPS production. This mutant showed decreased phage adsorption (67–75% depending on the phage) and a 3–4 log reduction in EOP compared with the parent strain, suggesting a role for CPS in adsorption for this S. thermophilus phage–host pair. Further support for this supposition comes from the fact that the cells used for the plaque assay come from early exponential phase growth when only CPS is produced. In contrast, Broadbent et al. (2003) had previously reported that CPS-producing S. thermophilus MR-1C was as phage sensitive as its CPS-negative derivative DM10, indicating that CPS was not protective against the three phages tested.

Superinfection immunity in S. thermophilus was demonstrated by temperate phage TP-J34, where expression of the ltp gene produces a lipoprotein that interferes with phage DNA injection (Sun et al. 2006). Interestingly, expression of the S. thermophilus ltp in L. lactis conferred protection to a 936 group phage that was extended to P335 and c2 group phages in subsequent studies (Bebeacua et al. 2013a; Ali et al. 2014). Phage resistance derived from a non-coding region resembling PER (phage-encoded resistance—see later in this section) by another S. thermophilus prophage was reported by Da Silva et al. (2018).

Streptococcus thermophilus resistance mechanisms

A notable distinction between S. thermophilus and L. lactis is the scarcity of plasmid-encoded phage resistance in the former. This may be in part due to the general rarity of plasmids in S. thermophilus (Girard and Moineau 2008) in comparison with L. lactis or may reflect an alternative strategy for protection against virulent phage such as CRISPR-Cas systems (discussed below). Chromosomally encoded R/M systems were among the first anti-phage mechanisms reported (see Allison and Klaenhammer 1998). Bioinformatically surveying the National Center for Biotechnology Information (NCBI) database identifies the general presence of R/M-related genes in S. thermophilus genomes, in particular hsdS specificity components, suggesting the relative importance on R/M as a phage defensive mechanism. Streptococcus thermophilus NDI-6 plasmid pCI65st harbors a gene resembling a Type I hsdS specificity subunit (O'Sullivan, van Sinderen and Fitzgerald 1999). Variants cured of pCI65st became phage sensitive, however, the role of an active R/M in phage resistance could not be confirmed. The presence of a plasmid-encoded orphan hsdS raises the potential for expanded R/M-mediated resistance as shown in lactococci (O'Sullivan et al. 2000). Regarding abortive infection, other than one report of Abi-like resistance (Larbi, Decaris and Simonet 1992), an active system has not been confirmed in S. thermophilus. Bioinformatically, a survey of the publicly available S. thermophilus genomes identifies proteins that annotate as belonging to the Abi_2 superfamily and to AbiE/AbiG abortive infection systems. It remains, however, to be demonstrated whether these genes functionally provide phage protection, or as in the case of the AbiE/AbiG toxin/antitoxin systems (Dy et al. 2014a), if they have another role in host physiology. A subcloned lactococcal abiA did confer resistance in S. thermophilus when tested at 30°C but not at 37°C or 42°C, whereas a cloned lactococcal abiG was non-functional at these same temperatures (Tangney and Fitzgerald 2002).

The paucity of phage adsorption variants is explicable by the biological characterization of a then unique prokaryotic immune system widely distributed in S. thermophilus termed CRISPR-Cas (Barrangou et al. 2007; for selected reviews see Deveau, Garneau and Moineau 2010; Karginov and Hannon 2010; Marraffini 2015; Rath et al. 2015; Koonin, Makarova and Zhang 2017; Pushnick et al. 2017; Brandt and Barrangou 2019). There is a diversity of CRISPR-Cas systems in S. thermophilus of which the Type II-A (Cas9-based) is the most active (Horvath et al. 2008). Fundamentally, a short segment of DNA (spacer) is presented in the CRISPR array as a vestige of a prior encounter with a foreign element (e.g. phage). The CRISPR array is transcribed (crRNA) along with a facilitating RNA (tracrRNA for Cas9) that recognize DNA with complementary sequence [termed protospacer when presented on the target molecule, and for Cas9, requiring a unique protospacer adjacent motif (PAM) sequence] and directs the Cas nucleolytic protein(s) to cut the guide-target complex. The Type II-A system in S. thermophilus is one of the most studied and has been adapted (utilizing the orthologous gene from Streptococcus pyogenes) for genome editing (Doudna and Charpentier 2014).

In the initial report (Barrangou et al. 2007), the biologic basis for CRISPR-Cas interference was established through the study of phage–host interaction. Subsequent studies found that the natural generation of S. thermophilus BIMs, for both cos and pac phages, is predominantly achieved through the action of CRISPR-Cas, and through iterative phage challenges, additional spacer acquisition enhanced the resistance phenotype (Deveau et al. 2008). This was further evidenced by Mills et al. (2010) in a follow up to their BIM methodology report (Mills et al. 2007). Importantly, when carried out with consideration for milk fermentation properties, strain physiology in CRISPR-Cas derived variants is unaltered (Barrangou et al. 2013). Characterization of the S. thermophilus CRISPR-Cas systems demonstrated its utility and adaptability toward defense against phage: (i) iterative incorporation of spacers against diverse phages expands range of resistance; (ii) insertion of multiple spacers against a specific phage increases resistance; and (iii) spacer integration into multiple, independent acting CRISPR-Cas systems increases resistance (Barrangou et al. 2013; Deveau et al. 2008; Barrangou and Horvath 2012). Additionally, CRISPR-Cas was shown to function cooperatively with R/M to elevate overall phage resistance (Dupuis et al. 2013). Together, the prevalence of CRISPR-Cas and R/M in S. thermophilus appears to be the primary defense against phage attack in this species.

Novel phage resistance mechanisms

The insights gained from the basic L. lactis and S. thermophilus phage–host interactions have generated several novel engineered approaches to inhibit phage that include: (i) PER based the host's presentation of a phage origin of replication as a molecular decoy titrating phage-specific DNA replication factors; (ii) anti-sense RNA blocking transcription of phage-specific essential proteins; (iii) subunit poisoning as demonstrated by expressing a trans-dominant mutated phage primase protein in S. thermophilus; and (iv) phage-triggered resistance where a lethal gene is conditionally expressed via a phage specific promoter (see review by Sturino and Klaenhammer 2006, for specific references). Under certain regulations, these examples would be defined as genetically modified; however, each could be constructed utilizing only native L. lactis or S. thermophilus genetic elements. Notwithstanding the methodology, stacking phage resistance mechanisms by introducing native R/M system-harboring plasmids (Josephsen and Klaenhammer 1990) also results in strains free of foreign DNA. Based on this logic, it is interesting to consider how one would define the use of phage-derived spacers to construct CRISPR guide RNAs, thereby mimicking the natural adaptive process, to direct phage inhibition by a strain's native CRISPR-Cas (Barrangou et al. 2007; Millen et al. 2012, 2018).

Beyond the native phage resistances described in L. lactis and S. thermophilus, namely R/M, Abi and CRISPR-Cas, the microbial community has evolved numerous other mechanisms to defend against virulent phage (see general reviews by Labrie, Samson and Moineau 2010; Koonin, Makarova and Wolf 2017; Hampton, Watson and Fineran 2020). Recent examples include BREX (Goldfarb et al. 2015), DISARM (Ofir et al. 2018), nine novel phage resistance systems described by Doron et al. (2018), and a chemical-based mechanism involving microbially produced small molecules (Kronheim et al. 2018). Considering the phage pressures in the fermented dairy environment, it will be interesting to see if such mechanisms have also developed in starter bacteria.

Phage countermeasures

Despite the efficacy and flexibility of CRISPR-Cas for phage resistance, the mechanism is not infallible if applied indiscriminately. As observed in our early research, phage can escape CRISPR-Cas owing to point mutations in the protospacer or PAM (Barrangou et al. 2007; Deveau et al. 2008). Although this can somewhat be mitigated by iterative phage challenge with escape or heterologous phages, evolution can invariably overcome biological hurdles given time and opportunity (Deveau et al. 2008). Furthermore, phages have undoubtedly co-evolved with CRISPR-Cas that we theorized would explain instances where adaptive immunity was ineffective for some S. thermophilus host–phage pairs despite confirmation of an active CRISPR-Cas (Horvath and Barrangou 2010). This phenomenon was explained by the identification of anti-CRISPR (Acr) genes residing on phages that did not elicit spacer acquisition in challenge experiments (Hynes et al. 2017, 2018). There is a wide diversity of Acr proteins, which exhibit weak homology to each other and employ varied modes of action, representing an active and ongoing evolution of both CRISPR and anti-CRISPR systems (Stanley and Maxwell 2018; Trasanidou et al. 2019).

Considering Acr activity, further study of phage–host interactions emphasizing non-CRISPR anti-phage mechanisms is imperative. McDonnell et al. (2018) utilized an anti-sense RNA approach to silence CRISPR-Cas to generate S. thermophilus BIMs wherein one of the non-CRISPR BIMs characterized was speculated to be an adsorption-deficient mutant. In an alternative approach, cos phage DT1, which harbors anti-CRISPR protein AcrII6A, was utilized to circumvent the native CRISPR-Cas in phage challenges of S. thermophilus SMQ-301 (Labrie et al. 2019). Phage-resistant variants were isolated that were mutated in the host methionine aminopeptidase (metAP), which impeded a late step in phage DT1 adsorption. MetAP mutations were effective against a broad range of cos phages and in different host strains; however, pac phages were unaffected, suggesting MetAP was not necessary for their replication. The ability to generate non-CRISPR-Cas phage-resistant variants will aid our understanding of essential host-encoded genes for phage propagation. Coupling these mutations with active CRISPR-Cas systems will undoubtedly improve phage protection of starter cultures.

The discovery of phage-encoded anti-CRISPR proteins is only the latest revelation in what is considered to be a ‘biological arms race’. Samson and Moineau (2013) have reviewed the various strategies employed by phages to circumvent host-encoded anti-phage systems. The pervasive presence, diversity and adaptability of 936 group phages to overcome anti-phage systems and persist were discussed by Mahony, Murphy and van Sinderen (2012). The lactococcal P335 group phages are another illustration. As a group, they represent a polythetic collection of diverse virulent and temperate phages (Labrie et al. 2008; Mahony et al. 2017). In response to pressure from abortive infection, an escaping phage ul36 variant had incorporated segments from the host chromosome such that it had a longer tail, different base plate and origin of replication, allowing it escape AbiC (Moineau, Pandian and Klaenhammer 1994). Bouchard and Moineau (2000) found that phage ul36 mutants had acquired host chromosome sequences, likely from prophage remnants, via homologous recombination to circumvent abiK-encoded phage resistance. The authors postulated that this is likely a common occurrence between P335 group phages and their host genome. In another study, ul36 was sequentially passaged through L. lactis isogenic derivatives containing AbiK and AbiT (Labrie and Moineau 2007). Characterization of four mutant phages able to circumvent one or both Abi mechanism(s) found extensive recombination with a host prophage, indicating the collective involvement of resident prophages in evolving lytic phages. Hill, Miller and Klaenhammer (1991) provided direct evidence that P335-group phage ϕ50, which infects a starter strain containing pTR2030, had acquired the LlaI methylase module allowing it to circumvent the pTR2030 R/M. Similarly, lytic P335 phage 4268, which is related to prophages such as BK5-T, contained a methylase that is near identical to one encoded in the host strain (Trotter et al. ). The malleability of the phage genome coupled with the unique nature of the industrial dairy fermentation processes provides ample opportunity for adaptation that must be taken into consideration when designing phage-resistant strains and starter culture programs.

The adaptability of phages is not necessarily limited to intraspecies genetic exchange. Genomic characterization of S. thermophilus phage 5093 found it has a mosaic structure indicating a recombinational path for genetic evolution evident from segments more related to non-dairy streptococcal prophages and limited homology to the cos and pac S. thermophilus phage genomes available at the time (Mills et al. 2011). Similar to some lactococcal phages, an acquired methylase, presumably with a role in circumvention of the host R/M was also identified on phage 5093. A proposed new S. thermophilus phage group, 987, was shown to be a hybrid of lactococcal P335 group phage morphogenesis components and of S. thermophilus replication modules (McDonnell et al. 2016). Two newly characterized phages, P738 and D4446, which share a high degree of identity, represent a new S. thermophilus phage group as they are quite dissimilar to other S. thermophilus phages and most closely related to Streptococcus pyogenes phage T12 (Philippe et al. 2020). The authors propose a common origin among the three phage groups, and that cumulative genetic exchanges between phage and host genomes are responsible for shifts in host infectivity. The recently isolated composite S. thermophilus 987 group phages are likely attributable to the increasing use of L. lactis—S. thermophilus mixed cultures, designed to address industry processing needs for applications traditionally utilizing mono-species starters. The rationale that phage robustness benefits from host genera-based barriers for phage recognition will be exacerbated by intensive use of such mixed starter culture blends.

DEVELOPMENT AND MAINTENANCE OF STARTER CULTURE SYSTEMS

From ‘undefined’ to ‘defined’ starter cultures

Since the discovery that dairy fermentations are negatively impacted by phages (Whitehead and Cox 1935), one objective in developing starter culture systems and the individual strain components is to manage the proliferation of phages and to minimize the negative impact of phage disruption on the dairy fermentation process. Historically, dairy fermentations have relied on so-called ‘undefined-starter cultures’ (also named ‘complex-’, ‘natural-’ or ‘mixed’ starter cultures), composed of an unknown number and diversity of strains. These undefined starters originated from northern Europe from sampling performed in dairy factories renowned for the constant quality of their production, and they have been intensively used for many types of cheese fermentation (cheddar cheese, cottage cheese, yellow cheeses such as Gouda). Specific undefined cultures remain in use today, mainly for yellow cheese and traditional artisanal cheese, despite their inconsistent performance mentioned in 1953 by Whitehead. In industrial large-scale cheese fermentations, undefined cultures have limited application where reproducibility and consistency are required. Therefore, in the mid-20th century, Whitehead (1953) introduced an alternative approach to the design of starter cultures based on individual strain isolates, which sets the foundation for current starter culture development. First applied in New Zealand in the 1930s, this new approach used single-strain starter cultures to produce cheddar cheese. Complete fermentation failure (dead vats) using such single-strain starters led to the discovery of ‘streptococcal’ phages (Whitehead and Cox 1935). The subsequent finding that different species of phages exist and that strains may have distinct spectra of sensitivity to these species led to the idea of using starter cultures containing several strains differing by their profile of phage sensitivity (Whitehead 1953). Indeed, by combining multiple strains with the same functionalities, but differing in their spectrum of phage sensitivity, the impact of an infection on the fermentation by one given phage was limited, the impact being proportional to the ratio of the infected strain relatively to overall culture cell population. All fermentation processes inevitably become susceptible to the proliferation of phages that are contaminating the environment resulting in subsequent issues for the starter cultures. Alternatively, the utilization of multiple starter cultures in rotation based on strains exhibiting distinct phage sensitivities reduces this risk. This rotation-based starter system using multiple-strain combinations was successfully implemented in New Zealand in the 1940s and consisted of the utilization of pairs of strains for a four- or five-day cycle (Whitehead 1953).

The basic principles of the design of starter cultures were applied only a few years later in the US. Collins, in 1962, described the successful use of a combination of four starter cultures, each containing three or more lactococcal strains with different profile of phage sensitivity. The rotational application of additional starter cultures composed of an even greater number of strains was necessary to run mechanized cheese plants where vats were filled multiple times a day, as stated by Lawrence et al. in 1978. The immediate constraint of these new rules on the design of starter cultures was the availability of a sufficient diversity of strains to run the starter culture programs. In 1953, Whitehead could select 10 strains with appropriate cheese functionalities but different phage sensitivity profiles, and ∼10 years later, Collins (1962) had accumulated 20 such strains. In 1978, Lawrence et al. hypothesized that strains displaying desired functional attributes may not be fully distinct and could possibly share sensitivity to some phages, but with a reduced infectivity. Alternatively, BIMs could possibly be selected through the exposure of a strain to its phages and subsequent selection of surviving cells. These basic principles of starter culture design are still applied today, not only for mesophilic cheese application, but for many kinds of applications, both mesophilic and thermophilic, for which phages present a technical and economic challenge.

Maintenance of ‘defined’ starter cultures: phage considerations

Soon after the implementation of defined starter rotations, it became evident that the starter cultures would require maintenance involving the replacement of strains that become phage sensitive (Collins 1962; Lawrence et al. 1978; Richardson et al. 1980). From the perspective of a supplier of commercial starter cultures (culture house) and as noted throughout the review, the prevalence of phage and impracticalities for their complete elimination in an industrial dairy fermentation environment necessitate an approach combining good manufacturing practices and hygiene controls, the development of phage-insensitive strains and culture systems, and a prescribed method of culture use to mitigate phage issues.

Candidate L. lactis and S. thermophilus strains are extensively vetted for defined physiological and technical characteristics related to specific fermented dairy applications. Equally critical, and a principal basis for selection, is a potential starter strain's phage sensitivity profile. For maintaining cultures where a strain component has become routinely phage attacked, phage resistance is the primary consideration. Determination of a strain's spectrum of phage sensitivity is dependent on the depth and breadth of the phage collection employed for the conventional testing method—standard plaque assay. As previously noted, phage–host interactions are exceedingly complex, whereby the diversity of receptor/antireceptor structures, host defenses and phage counter responses can alter the nature of a strain's phage sensitivity spectrum. Nevertheless, the ability of a phage to form plaques on a host strain is a base criterion for determining the phage–host relationships. To this end, most culture houses have regimens for the collection of industrial dairy samples that are tested for phages against a defined set of starter strains. Such programs serve two purposes: first, as a service to dairy processors to monitor the environment for the presence of phage and to manage the starter culture usage; and second, as a research tool for the development and maintenance of starter strains and culture systems. By default, such collections are limited to the available host strains for use as indicators of phage infection. This is a critical consideration as it is typical for dairy processors to employ more than one culture house in structuring a starter program.

New dairy starter strains are typically sourced from traditional ferments or raw milk, the former containing proven example of strains with desired milk fermentation properties (Limsowtin, Powell and Parente 1996). With the demands of modern industrial dairy fermentations and prevalence of commercial starters it is interesting to speculate on the ability to isolate suitable strains. In the most optimistic cases, researchers estimated that ‘20 or more phage-unrelated strains of L. lactis exist’ (Collins 1962; Cogan 1975). While isolating strains from natural sources continues today, the process involves an extensive validation regimen that includes assessments for safety (Koutsoumanis et al. 2020), suitability for various fermented dairy applications, compatibility with other strains (e.g. phage spectrum of sensitivity, production of antagonistic compounds) and the appropriate characteristics to be manufactured at scale. Today, the use of natural strain isolates is also governed by the ‘Rio Convention on Biological Diversity’ (1992) and the Nagoya Protocol on ‘Access to Biodiversity and Benefit Sharing’ [Regulation (EU) No. 511/2014], which oversee the interests of the biological material's origin source. Culture houses are supportive of these treaties; however, addressing the regulatory considerations further prolong the process for potential commercial implementation.

As an alternative to isolation from natural sources, one can take a phage-sensitive strain and develop a phage-resistant variant as described above. The generation (or isolation) of phage insensitive variants using phage as a selective agent remains a viable approach, although this approach is still susceptible to isolation of mutants with altered physiological properties (Lawrence 1978; King, Collins and Barrett 1983; Dupont et al. 2004a; Theodorou et al. 2019). For the examples of mutagenizing pip and yjaE genes in L. lactis that are essential for Ceduoviruses infection, and the utilization of native S. thermophilus CRISPR-Cas systems to immunize against virulent phage, the approach is specific, directed and results in no significant negative impact on strain functionality. Beginning in the 1980s and during the next three decades, a plethora of plasmid-borne phage resistance mechanisms were discovered in L. lactis (see prior section and selected reviews: Allison and Klaenhammer 1998; Forde and Fitzgerald 1999a; Mills et al. 2006; Ainsworth et al. 2014b). These plasmid-encoded mechanisms are frequently associated with conjugative functions that offer a route for constructing improved phage-resistant strains utilizing native lactococcal gene transfer abilities (Sanders et al. 1986; Sing and Klaenhammer 1986; Harrington and Hill 1991; Coakley, Fitzgerald, and Ross 1997; O'Sullivanet al. 1998). This directed approach has the significant advantage of utilizing strains with proven safety, application and industrial technical properties; as the characterization of such strains represents a significant research and development investment for culture houses. The use of proven strains is particularly beneficial for the design of defined multi-strain starter cultures wherein multiple strains with different spectra of phage sensitivity but comparable performance are combined into one culture to mitigate phage proliferation. Combining a set of similarly developed starter cultures in a defined rotation sequence adds an additional hurdle against the spread of phages. A novel twist to defined strain culture rotations was introduced by Sing and Klaenhammer (1993). This rotation is based on known phage resistance mechanisms and employs a set of phage-resistant variants derived from a single strain, where each variant contains a unique plasmid encoding different phage resistances. The availability of phage-resistant strains to develop starter cultures and the ability to replace phage-susceptible strains with phage-resistant variants or alternatives are paramount to maintaining defined strain culture programs.

CONCLUDING REMARKS

Upon the discovery of the first phage infecting LAB in 1935, it soon became evident that both starter culture range management and good manufacturing practices would be necessary to adequately mitigate fermentation failures caused by phages. Initial efforts consisted of isolating and characterizing the phage diversity with a specific focus on phages infecting L. lactis and S. thermophilus. Characterizations focused on the typing of phages, determination of their host spectrum and their virulence. From these studies, it was found that the diversity of strains relative to their spectrum of phage sensitivity was larger than initially thought, thus allowing the development and the maintenance of starter culture ranges based on the routine replacement of strains that became phage susceptible in the industrial environment.

However, because of the ecology and persistence of phage in the environment, starter strains cannot be exploited indefinitely, resulting in a constant need to have access to new strains that are resistant to phages. There are many strains in collections that are sufficiently robust to phages; unfortunately, these strains often lack the appropriate functionalities and are therefore useless as replacements in current starter culture ranges. One may consider improving the functionalities of these strains using natural means such as spontaneous mutation or chemically induced mutagenesis, conjugative plasmid transfer and natural competence that would not result in a ‘genetically modified’ status according to many national and international regulations (e.g. Directive 2001/18/EC from the European Parliament). Alternatively, one could apply these same techniques to improve phage robustness of the phage-sensitive strains. This is one of the main purposes of investigating the interaction of phages with their host receptors and the bacterial mechanisms modulating phage virulence. Despite successful applications of these discoveries (pip and yjaE mutants, CRISPR-based immunization) toward decreasing the pressure of some phages on industrial dairy fermentations, there is still much to accomplish in this field of research. Phages of the P335 and 936 groups remain a major issue for the utilization of strains of the lactis and cremoris subspecies, respectively, and the frequency of other phage groups may increase upon a possible disappearance of the Ceduovirus genus. Similarly, the selective pressure resulting from the CRISPR-based immunization of S. thermophilus strains may give rise to new populations of phages accumulating Acr systems. It is notable that new phage groups have recently emerged in S. thermophilus, some of which appear to result from recombination between phages originating from different host species or genera.

In addition to the selection of new phage-resistant strains (through the selection of new genotypes, or the improvement of already exploited genotypes), successful control of phages in the dairy industry will necessarily rely on an appropriate management of industrial processes (see Garneau and Moineau 2011 for a review). Primarily, the raw materials should not be contaminated by phages. Milk collected from the farms may contain some phages, though likely at a limited level considering there is little if any bacterial propagation in milk at this stage. Whey concentrates and powders, often added to standardize milk for dairy fermentations, may present a higher phage risk. Indeed, whey most often contains phages, and permeation, filtration or spray-drying do not necessarily eliminate or destroy the phages. Also, because whey and milk may be transported to drying facilities using the same tankers, these practices may result in large-scale phage contamination of raw milk. Phages may also originate from the starter culture itself. Strains may harbor prophages that can potentially be induced during the production processes. This is particularly true for mesophilic starter cultures composed of lactococci. Most L. lactis strains, if not all, are known to bear one or more prophages of the P335 group. While the presence of a prophage may provide protection through superinfection immunity against similar phage types, they can also recombine with other lysogenic or lytic phages, resulting in the release of new phages with modified host range and virulence. Lysogenic S. thermophilus (and temperate phages) is much less frequent. Because of the high instability of the prophages, which can possibly mutate to become virulent, it has long been recommended not to industrially exploit lysogenic S. thermophilus, and thus, S. thermophilus lysogens do not seem to represent an issue in dairy fermentations. Once phage contamination occurs in the manufacturing facility, it will become pervasive in the plant environment—tanks, pipes, air and surfaces—making phage eradication very unlikely. It is therefore critical to apply appropriate industrial practices. This includes the application of a rigorous starter culture rotation to avoid recurrent phage amplification, and to allow the washing out of contaminant phages, resulting in the reduction of phage levels in the plant. In parallel, this reduction should be ameliorated by applying sanitation protocols with biocides. The sanitation aims to eliminate the bacterial material that could serve as host for phage propagation as well as the phages themselves. Sanitizer treatments should be both bactericidal and viricidal; indeed, in the absence of bacteria, phages cannot multiply. Most often, phages are more resistant to sanitization than bacteria, and cost-effective treatments may not always be sufficient to decrease the phage population. In addition, the efficiency of the treatment may vary from one sanitizer to another, from one phage group to another and from phage to phages of the same group (Guglielmotti et al. 2012; Campana et al. 2014; Hayes et al. 2017). Among possible treatments (see Giraffa, Zago and Carminati 2018 for a recent review), peracetic acid is often described as the most efficient sanitizer against phages. Sodium hypochlorite and quaternary ammonium-based biocides seem less efficient and phage dependent. It is notable that ethanol and isopropanol, which are commonly used for cleaning laboratories, are poorly efficient on phages. Finally, new disinfectants, such as ethoxylated nonylphenol with phosphoric acid and alkaline chloride foam, proved to be highly efficient (Marcó, Moineau and Quiberoni 2012).

To evaluate the effectiveness of phage control procedures, there is a need to determine the nature and the level of phages in the dairy processing plants. Microbiological investigation has long been the favored methodology for phage detection. Advantageously, plaque assays are highly sensitive and quantitative. However, they are also strain specific and necessitate testing each of the strains for which phages are expected. Therefore, plaque assays may be extremely time consuming. For this reason, alternative approaches should be envisioned. Among the possibilities is the use of described PCR methods allowing for the detection of most (but not all) of the phages infecting L. lactis and S. thermophilus. Constant isolation and genetic investigation of phages are thus strongly recommended to systematically improve these molecular methods. Ultimately, phageomics will expand our understanding of phage–host interactions, phage evolution and could possibly be exploited to detect DNA from unknown or uncultivable phages in the dairy environment.

Despite tremendous research efforts to fight against phages since their discovery, they remain a major economic concern for dairy manufacturers. Efficient solutions were proposed (phage–robust strains, starter culture management, improved industrial design, appropriate sanitizers) that keep phages globally under control. However, these measures have not completely solved the problem, due in part to constant phage evolution and the concentration of the dairy industry with increasingly larger dairy plants managing billions of liters of milk every year. Eradication of dairy phages is likely a utopian objective (and may not be a goal considering the preservation of the biodiversity); one may consider the fight against phages as a never-ending arms race that requires the pursuit of research programs on lactococcal and streptococcal phages. New mechanisms to resist phages would also be valuable to replace those that phages have successfully overcome. A recent example is the discovery of anti-CRISPR-associated proteins (Stanley et al. 2019) that could be exploited to maintain the effectiveness of a CRISPR-Cas system against an active Acr. In the future, the global phage–host ecology should be more deeply investigated with attention to a possible future dominance of a currently minor species or the emergence of new species and the host responses in defense of virulent phages.

ACKNOWLEDGMENTS

We would like to acknowledge the technicians and scientists whose efforts have contributed to the constitution of DuPont collection of phages over the last four decades, and their investigations of this collection, in particular, Isabelle Chavichvily, Bonnie Hiley, Max-Charles Jodeau, Carrie Hanko, Jennifer Seiler and Armelle Cochu-Blachère. A special thanks to Sylvain Moineau and his team for 20 years of fruitful collaboration investigating streptococcal and lactococcal phages and their interaction with their hosts.

SUPPLEMENTARY DATA

Supplementary data are available at FEMSRE online

Conflict of interest

None declared.