Revisiting lactococcal abortive infection-like antiphage systems: diversity, structures, and mechanisms

Abstract

Lactococcus lactis and Lactococcus cremoris are cornerstone bacterial species employed in the production of fermented dairy products such as cheese. However, the non-sterile dairy processing environment regularly exposes these lactococci to phage infection. To counteract this phage threat, bacteria have evolved a wide range of defence mechanisms, including so-called abortive infection Abi(-like) systems, of which many (AbiA to AbiZ) were originally discovered in Lactococcus. Recent discoveries have expanded the list of Abi-like systems in Lactococcus species. In this review, we focus on systems historically, phenotypically, or mechanistically classified as Abi, critically examine and revisit their antiphage activity spectrum and interference with the phage lytic cycle, as well as mechanistic insights of lactococcal Abi-like systems as obtained through the study of phage escape mutants. Furthermore, we group various Abi-like systems based on structural superimposition and use predicted domain information to explore or expand on their mechanism of action. Finally, we show that despite many Abi-like systems being discovered due to their presence on plasmids, most anti-phage defence systems appear to be chromosomally encoded in sequenced Lactococcus strains. Our findings support the notion that co-evolution of Lactococcus species with their phages, possibly accelerated by the extensive application of these bacteria in dairy fermentations, has resulted in the acquisition of a diverse array of phage defence mechanisms, including Abi-like systems.

This review examines the expanding repertoire of experimentally validated Abi-like phage defence systems in industrially relevant Lactococcus species, focusing on their genetic locations, antiphage activity spectra, impacts on the phage lytic cycle, domain and structural similarities, and underlying mechanisms, while also enabling grouping of Abi-like systems through structural superimposition and applying predicted domain information to hypothesize unexplored modes of action.

Introduction

Lactococcus lactis and Lactococcus cremoris [previously two subspecies of L. lactis; (Li et al. 2019)] are members of the lactic acid bacteria (LAB), many of which have a long history of application in the manufacture of various fermented foods, such as cheese, yoghurt and buttermilk (Makarova et al. 2006). Bacteriophages are the most abundant biological entities on the planet and their role in bacterial population dynamics is undeniable (Naureen et al. 2020). The omnipresence of phages has forced bacteria to develop a plethora of antiphage systems for their protection (Georjon and Bernheim 2023). Dairy lactococci are no exception to this, as their vulnerability to phage attack is amplified due to their widespread application as starter cultures in non-sterile dairy fermentation environments (Hayes et al. 2017).

Bacterial defences against phage infection include diverse barriers that are effective at different stages of the phage cycle (Georjon and Bernheim 2023). For instance, bacteria can prevent phage adsorption, disable phage genome injection or degrade the incoming phage genome (Labrie et al. 2010). As one of the last resorts, abortive infection (Abi) systems cause cell death and/or arrest of a critical cell activity upon phage infection, thereby preventing or severely limiting phage proliferation and subsequent infection of neighbouring cells (Bernheim and Sorek 2020, Lopatina et al. 2020).

Due to the recurrent infection of starter cultures and associated risk and financial consequences of failed fermentations, the constant phage threat to L. cremoris/lactis strains during dairy fermentations has had, and continues to have, negative environmental and economic impacts (Ortiz Charneco et al. 2023). Therefore, it is not surprising that many of the first historically classified Abi systems (from here on referred to as Abi-like since the precise mode of action of these systems is not always known) systems were discovered in these species [reviewed by (Chopin et al. 2005, Ainsworth et al. 2014, Philippe et al. 2023)]. Abi-like systems are predicted to be a key element of lactococcal defences against phage infection (Rodríguez-Román et al. 2024). Interestingly, homologues of some of the originally described lactococcal Abi(-like) systems, including AbiD/F group members [also termed Abi2 by Tesson et al. (Tesson et al. 2022)] and AbiEi (Dy et al. 2014), are highly abundant in prokaryotes (Tesson et al. 2022). Notably, most of the lactococcal Abi-like systems have been designated as such (i.e. being an Abi-like system) based on their phenotypic profile rather than on mechanistic characterization, while it is now recognized that mechanistic elucidation is essential to confirm whether a system truly operates via abortive infection (Aframian and Eldar 2023).

Lactococcal antiphage defences further comprise a variety of Type I, Type II, and Type III Restriction-Modification (RM) (Nyengaard et al. 1995, Schouler et al. 1998, Su et al. 1999, McDonnell et al. 2025) and five different Superinfection exclusion (Sie) systems (McGrath et al. 2002, Mahony et al. 2008). Audmula and Fliodhais are two recently discovered antiphage systems that function through a non-Abi [Audmula (Mosterd et al. 2025)], or unknown mechanism [Fliodhais (Grafakou et al. 2024a)]. CRISPR-Cas systems are typically not present in lactococcal genomes and consequently do not appear to play a major role in their overall antiphage defence, although a type III-A CRISPR-Cas system has been described in L. lactis (Millen et al. 2012, Kelleher et al. 2019).

A recent large-scale study of lactococcal plasmids revealed 11 novel Abi-like systems (Grafakou et al. 2024a), yielding a total of 34 distinct Abi-like systems identified in Lactococcus. In addition, examining and comparing functional and structural characteristics of antiphage system proteins has proven useful for predicting their mode of action and grouping them according to sequence/structural/functional similarity. For example, AbiD and AbiF were initially grouped based on the presence of a HEPN domain (Anantharaman et al. 2013). AbiD-like shares 59% identity with AbiD (Grafakou et al. 2024a), while AbiD1 shares 46% identity with AbiF (Bidnenko et al. 2009). Subsequently, AbiD, AbiD1, AbiF and AbiD-like were classified within the proposed AbiD/F group based on structural similarity (Grafakou et al. 2024a). Notably, AbiD-like has previously been referred to as both AbiD-like and AbiD/F (Grafakou et al. 2024a; Mosterd et al. 2025) in different publications, but hereafter termed AbiD-like to distinguish it from the broader AbiD/F group.

With currently 34 Abi-like systems identified in Lactococcus and the last review dedicated to lactococcal Abi systems dating back to 2005 (Chopin et al. 2005), an update incorporating newly identified Abi-like systems as well as the use of recent functional prediction tools is highly desirable. The aim of this review is to focus on systems that have historically, phenotypically, or mechanistically been classified as Abi systems, and to therefore revisit and, where appropriate, (re)classify some of the known and newly identified lactococcal Abi-like systems by predicting, comparing and analyzing their functions, modes of action and structures using current bioinformatic tools, while evaluating their prevalence and genetic locations across L. cremoris/lactis genomes.

The repertoire of known lactococcal Abi-like systems: A 35-year history

Since the discovery of AbiA in the early 1990s (Hill et al. 1990) and up until the identification of AbiZ in the late 2000s (Durmaz and Klaenhammer 2007), 23 Abi-like systems have been described (named AbiA to AbiZ; notably there are no systems designated as AbiM, AbiW, AbiX, and AbiY; Table 1, updated from Philippe et al. 2023). Subsequent to this, the discovery of new Abi-like systems in lactococci seemed to have come to a near stand-still for some 15 years.

Table 1.

Experimentally confirmed lactococcal Abi-like systems, as well as information on their antiphage activity range, affected phage lytic cycle step and phage escape mutants.

System	Active against*			Interference with particular step in phage cycle	Phage escape mutant analysis	Mechanism of cell arrest/death	References
	S	C	P
AbiA (Also Hsp)	+	+	+	DNA replication (P335 group)	Mutation(s) in putative ATP-binding protein; intergenic region; ssDNA annealing protein and Holliday junction resolvase	?	(Hill et al. 1990, Hill et al. 1991, Dinsmore and Klaenhammer 1994, 1997, Dinsmore et al. 1998, Tangney and Fitzgerald 2002, Mosterd et al. 2025)
AbiB (Also abi-416)	+	-	-	RNA transcription	Mutation(s) in major capsid protein and hypothetical protein	?	(Cluzel et al. 1991, Parreira et al. 1996, Grafakou et al. 2024a, Mosterd et al. 2025)
AbiC (Also Prf)	+	-	+	Major capsid production	?	?	(Durmaz et al. 1992, Moineau et al. 1993)
AbiD	+	+	?	After DNA replication	?	?	(McLandsborough et al. 1995)
AbiD-like	+	-	-	?	Mutation(s) in hypothetical proteins	?	(Grafakou et al. 2024a, Mosterd et al. 2025)
AbiD1	+	+	?	Degradation of sk1 RuvC endonuclease	Mutation(s) in hypothetical protein	?	(Gautier and Chopin 1987, Anba et al. 1995, Bidnenko et al. 1995, 1998, 2002, 2009)
AbiE	+	-	-	After DNA replication	?	Likely due to tRNA dysregulation caused by nucleotide transfer to tRNAs (type IV TA system)	(Garvey et al. 1995, Tangney and Fitzgerald 2002, Cai et al. 2020, Hampton et al. 2020)
AbiF	+/−	+/−	-	DNA replication, delayed (Skunavirus)	?	?	(Garvey et al. 1995, Rincé et al. 2000, Tangney and Fitzgerald 2002; Grafakou et al. 2024a; Saunier et al. 2025)
AbiG	+	+/−	+/−	RNA transcription (Ceduovirus and Skunavirus) DNA replication, delayed (P335)	Mutation(s) in major tail protein	?	(O’Connor et al. 1996, 1999, Tangney and Fitzgerald 2002, Grafakou et al. 2024a, Mosterd et al. 2025)
AbiH	+	+/−	?	?	?	?	(Prévots et al. 1996)
AbiI	+	+/−	?	?	?	?	(Su et al. 1997)
AbiJ (Also abi-859)	+/−	+	?	?	Mutation(s) in ssDNA annealing protein	?	(Deng et al. 1997, Grafakou et al. 2024a, Mosterd et al. 2025)
AbiK	+	+/−	+	After DNA replication (Skunavirus), before DNA replication (P335)	Mutation(s) in SaK ssDNA annealing protein; recombination with a prophage	?	(Emond et al. 1997, Bouchard and Moineau 2000, 2004, Boucher et al. 2000, Fortier et al. 2005, Labrie and Moineau 2007, Scaltriti et al. 2011, Wang et al. 2011, Figiel et al. 2022)
AbiL	+	+/−	-	After transcription (Ceduovirus)	?	?	(Deng et al. 1999)
AbiN	+	+	?	?	?	?	(Prévots et al. 1998)
AbiO	+/−	+/−	?	After DNA replication (Ceduovirus)	?	?	(Prévots and Ritzenthaler 1998)
AbiP	+	-	+/−	DNA replication and temporal control (switch-off) of early transcripts	Mutation(s) in major tail protein	?	(Domingues et al. 2004a, b, Grafakou et al. 2024a, Mosterd et al. 2025)
AbiQ	+	+	-	RNA transcription	Mutation(s) in hypothetical proteins	In homologous ToxIN system, cell death is caused by phage infection, while ToxIN system prevents the production of mature virions	(Samson et al. 2013a, Emond et al. 1998, Guegler and Laub 2021)
AbiR	?	+/−	?	DNA replication, reduced	?	?	(McKay et al. 1989, Twomey et al. 2000, Yang et al. 2006)
AbiS	+	?	?	?	?	?	(Holubová and Josephsen 2007)
AbiT	+	-	+	DNA replication (Skunavirus and P335 group) and capsid protein synthesis (Skunavirus)	Mutation(s) in major capsid protein and hypothetical proteins; recombination with prophage genes	?	(Bouchard et al. 2002, Labrie and Moineau 2007, Labrie et al. 2012)
AbiU	+/−	+/−	+/−	RNA transcription, delayed (Skunavirus, Ceduovirus)	?	?	(Dai et al. 2001)
AbiV	+	+	-	Protein synthesis and late gene transcription (Skunavirus)	Mutation(s) in a SaV protein	Likely through host ribosomal RNA degradation	(Haaber et al. 2008, 2009, 2010, Zhu et al. 2024)
AbiZ	+/−	+/−	+	Premature lysis	Recombination with host genome	Likely acts cooperatively with the phage holin to increase membrane permeability, leading to premature lysis of infected cells	(Durmaz and Klaenhammer 2007)
Rhea	+	+/−	+/−	No effect on DNA replication, transcription, translation (Skunavirus)	Mutation(s) in terminase large subunit (ATPase domain)	?	(Grafakou et al. 2024a, Grafakou et al. 2024b, Mosterd et al. 2025)
Aristaios	+	-	-	Transcription (Skunavirus)	?	?	(Grafakou et al. 2024a; Grafakou et al. 2024b; Mosterd et al. 2025)
Kamadhenu	+	-	-	No effect on DNA replication, transcription, translation (Skunavirus)	Mutation(s) in ssDNA annealing protein	?	(Grafakou et al. 2024a, Grafakou et al. 2024b, Mosterd et al. 2025)
Rugutis	+	+/−	+/−	?	Mutation(s) in terminase large subunit (ATPase domain)	?	(Grafakou et al. 2024a, Mosterd et al. 2025)
Hesat	+	-	?	?	?	?	(Grafakou et al. 2024a, Mosterd et al. 2025)
PARIS	+	+/−	?	Unknown in Lactococcus	Mutation(s) in hypothetical protein; intergenic region	Blocks protein translation by cleaving host lysine tRNA (type II TA)	(Burman et al. 2024, Deep et al. 2024, Grafakou et al. 2024a, Mosterd et al. 2025)
Type I CBASS	+	+/−	?	Unknown in Lactococcus	?	Membrane impairment caused by cyclic oligonucleotide signals produced upon phage infection sensing	(Grafakou et al. 2024a, Mosterd et al. 2025)
Type II CBASS	+	+	?	Unknown in Lactococcus	Mutation(s) in hypothetical protein; intergenic region		(Millman et al. 2020b, Grafakou et al. 2024a, Mosterd et al. 2025)
Lamassu	+	-	?	Unknown in Lactococcus	?	?	(Grafakou et al. 2024a, Mosterd et al. 2025)
Septu	+	-	?	Unknown in Lactococcus	Mutation(s) in ssDNA binding protein	?	(Grafakou et al. 2024a, Mosterd et al. 2025)

*S, Skunavirus; C, Ceduovirus, P, P335 phage group; ?, unknown; +/−, indicates either lower/partial activity against phages of this group or activity against specific phages of this genus/group. Updated and adapted from Philippe et al. (Philippe et al. 2023); changes/additions are indicated as bold-face text.

Recent reports on so-called “defence islands” (Makarova et al. 2011, Doron et al. 2018, Gao et al. 2020, Millman et al. 2022)—genomic regions enriched with encoded clustered antiphage systems—have renewed interest in finding novel antiphage systems across various bacterial genera. Additionally, the ongoing issue of phage infection in dairy fermentations (Ortiz Charneco et al. 2023) highlights the need to develop (non-genetically modified) phage-robust starter cultures that also encode other desirable traits, such as the ability to produce novel organoleptic molecules or potent antimicrobial compounds (Mills et al. 2010, Cavanagh et al. 2015). Development of such phage-robust strains may be facilitated by natural transfer of desirable traits through conjugative or mobilizable plasmids (Mills et al. 2006, Derkx et al. 2014, Kohler et al. 2019, Ortiz Charneco et al. 2021, Grafakou et al. 2024b).

The above led to a recent study (Grafakou et al. 2024a) that identified eleven additional plasmid-encoded Abi-like systems in L. lactis/cremoris (Table 1). Five of the identified systems (i.e. Rhea, Aristaios, Kamadhenu, Hesat, Rugutis) were novel Abi-like systems and were named after ancient cheese- and fermented food-associated deities (following the deity-based, antiphage system nomenclature used by Doron et al. (Doron et al. 2018). The remaining six systems [AbiD-like, PARIS, type I and II CBASS (Cyclic oligonucleotide-Based Anti-phage Signaling System), Septu, and Lamassu systems] were homologous to Abi-like systems discovered in other bacteria (except AbiD-like), yet not previously (experimentally) confirmed in dairy lactococci.

Components of lactococcal Abi-like systems

Among the 34 confirmed Abi-like systems in Lactococcus (Table 1), the majority (18 systems) are each represented by single protein-coding genes, while 13 are two-component systems (AbiD, AbiD1, AbiD-like, AbiE, AbiF, AbiG, AbiL, AbiQ, AbiT, AbiV, PARIS, type I CBASS, Septu; of which AbiD, AbiD1, AbiF, AbiD-like, AbiQ, and AbiV comprise a protein and an RNA molecule) and three are three-component systems (AbiR, type II CBASS, Lamassu). Nonetheless, antiphage systems comprising more than two components may have been overlooked, likely due to the prevailing focus on single- or two-gene candidates in recent efforts to identify novel lactococcal antiphage systems (Grafakou et al. 2024a).

Genetic location and prevalence of lactococcal Abi-like systems

Lactococcus strains commonly carry many plasmids [previously reported average of six plasmids per strain (Grafakou et al. 2024a, McDonnell et al. 2025)], which may harbour genes associated with various (bio)technologically desirable traits [such as lactose utilization, casein metabolism, conjugation, bacteriocin production, exopolysaccharide production (Kelleher et al. 2019)]. Since most of the Abi-like systems discovered in lactococci were found on plasmids [aside from AbiH (Prévots et al. 1996), AbiV (Haaber et al. 2008), and AbiN (Prévots et al. 1998)], it might seem as though lactococcal antiphage systems are primarily encoded on plasmids. Being located on plasmids with higher copy numbers than the chromosome may also boost the levels of phage resistance provided by Abi-like systems (Philippe et al. 2023). However, there is a discovery bias towards plasmid-encoded antiphage systems; plasmids are frequently studied for phage-resistance genes due to their potential for conjugative transfer to enhance the phage robustness of industrial production strains (Chopin et al. 2005). The use of conjugation—a natural DNA transfer process—allows the generation of “natural” strains, as opposed to “genetically modified” strains, which are subject to strict regulations and which have poor consumer acceptance (Lucht 2015). To address the knowledge gap of whether lactococcal antiphage defences are primarily plasmid-encoded, we investigated the presence of antiphage defence systems in 584 non-redundant RefSeq lactococcal genomes, including their plasmid sequences, using PADLOC (Payne et al. 2021) (Fig. 1).

Figure 1.

Phage defence systems are primarily located on the chromosomes of L. lactis and L. cremoris. (A) The number of antiphage defence systems identified in each L. lactis or L. lactis genome in Refseq v228; identified using PADLOC. (B) The number of distinct types of antiphage defence systems identified in each genome. (C) The location (chromosome or plasmid) of antiphage defence systems identified in complete L. lactis and L. cremoris (combined) genomes where the chromosome and any plasmids present, where defined.

On average, L. lactis genomes were shown to encode similar numbers and distinct types of systems as L. cremoris genomes (mean = 4.6 vs 4.3 systems and 4.2 vs 4.0 system types) (Fig. 1A and B). Several L. lactis strains encode many more systems than found on average, with > 5% of strains encoding ten or more systems. Whether the apparent overabundance of defence systems in some strains is linked to their selection and use as production cultures remains to be investigated. Additionally, this analysis confirms that RM and Abi-like defences are highly prevalent in L. lactis and L. cremoris, in accordance with a recent study (McDonnell et al. 2025).

To address the question of whether most of these predicted Lactococcus defences are plasmid-encoded, we analyzed the subset of genomes that were fully assembled complete with plasmids (112 strains). Surprisingly, we found that most phage defence systems in these genomes are chromosomally located (Fig. 1C), despite the bias towards discovery of phage defences on plasmids in lactococcal strains. As observed in other genera, it is likely that many of the chromosomally located systems are present in mobile genetic elements. For example, it was recently established that AbiH is largely associated with group 6 integrated conjugative elements in lactococcal chromosomes (van der Els et al. 2024), highlighting the role of mobile genetic elements in the incorporation of genes encoding lactococcal Abi-like systems.

The antiphage activity spectrum of lactococcal Abi-like systems

The activity of antiphage systems in L. lactis/cremoris has mainly been assessed against members of the three most prevalent and problematic phage genera in modern, large-scale dairy fermentation plants, i.e. Skunavirus (formerly 936 phage group), Ceduovirus (formerly c2 phage group), and P335 group phages (Deveau et al. 2006, Mahony et al. 2016). Nonetheless, some systems have also been tested and proven effective against other lactococcal phage genera, such as AbiT [which is effective against members of the Audreyjarvisvirus, Questintvirus, and Fremauxvirus genera (formerly 949, Q54 and 1706 phage groups, respectively) (Fortier et al. 2006, Garneau et al. 2008, Samson and Moineau 2010)]; and AbiQ [which is effective against members of the Audreyjarvisvirus and Questintvirus genera, and coliphages (Fortier et al. 2006, Samson and Moineau 2010, Samson et al. 2013a)].

It is evident that the antiphage activity spectrum of Abi-like systems is diverse (Table 1). Some systems are active only against phages of a single genus [e.g. AbiB (Cluzel et al. 1991, Parreira et al. 1996, Grafakou et al. 2024a)]; while others are active against phages belonging to multiple genera/groups [e.g. AbiA (Dinsmore and Klaenhammer 1994, Tangney and Fitzgerald 2002)]. Additionally, some systems are active against specific members of a given phage genus [e.g. AbiO (Prévots and Ritzenhaler 1998)], and others seem to provide only modest protection against phages [e.g. AbiU, 10⁻¹ to 10⁻² reduction of the efficiency of plaquing of phages from any of the three assessed groups (Dai et al. 2001)]. Interestingly, in all cases, each antiphage system was found to be active against at least one member of the Skunavirus genus, which are the most prevalent lactococcal phages (Mahony et al. 2012), suggesting that antiphage systems targeting the most abundant and prevalent lactococcal phages have been selected for in the studied strains. Notably, these Abi-like systems have been evaluated under different experimental conditions, such as diverse expression vectors and phage panels, complicating the comparison of their activity for practical applications and the assessment of their effectiveness in non-artificial conditions. This highlights the need for uniform testing approaches and a re-evaluation of the activity of the systems against a large panel of phages to establish and compare the activity spectrum of these systems.

Effect of lactococcal Abi-like systems in the phage lytic cycle

Following identification of a given lactococcal antiphage system, which is usually based on a reduced plaquing ability phenotype, characterization of such a system typically involves assessment of the effect of the antiphage system on specific steps of the phage lytic cycle (Table 1). Among the characterized lactococcal Abi-like systems, seven interfere with phage replication [i.e. AbiA (Hill et al. 1991, Tangney and Fitzgerald 2002), AbiF (Garvey et al. 1995), AbiG (Tangney and Fitzgerald 2002), AbiK (Emond et al. 1997), AbiP (Domingues et al. 2004a), AbiR (Twomey et al. 2000), AbiT (Bouchard et al. 2002)], six affect phage transcription [i.e. AbiB (Parreira et al. 1996), AbiG (O’Connor et al. 1999), AbiQ (Samson et al. 2013a), AbiU (Dai et al. 2001), AbiV (Haaber et al. 2010), Aristaios (Grafakou et al. 2024b)], while one was demonstrated to diminish phage protein production [AbiC (Moineau et al. 1993)]. Two systems (Rhea and Kamadhenu) were found not to interfere with any of the above-mentioned steps (Grafakou et al. 2024b), which suggests an (direct or indirect) interference with other (and later) steps of the phage lytic cycle, such as DNA packaging or phage maturation.

Abi-like systems are expected to contain at least one functional module to sense (an aspect of) phage infection, upon which another domain becomes activated to elicit the bacteriostatic/bactericidal effect (Lopatina et al. 2020). Notably, it has become obvious that one of the significant knowledge gaps is our limited understanding of the underlying mechanisms of cell death in many cases (Table 1). Investigating the specific phage infection step affected by activation of a particular Abi-like system may provide mechanistic information on that system. However, this approach fails to differentiate if inhibition of this phage proliferation step is a direct or indirect target for the Abi-like system. Additionally, this approach generally provides less detailed mechanistic information compared to phage escape mutant isolation and analysis, which may point to phage-derived proteins sensed or affected by the Abi-like system.

Phage escape mutant analysis of lactoccocal Abi-like systems

Phage escape mutants, i.e. phage derivatives that bypass the activity of phage defence mechanisms, have been isolated and analyzed for 16 lactococcal Abi-like systems (Table 1), in an attempt to define phage-derived proteins that may circumvent or trigger these antiphage systems. Phages typically acquire insensitivity to these systems through the acquisition of one (or occasionally multiple) point mutation(s). However, in some instances [such as for AbiK (Bouchard and Moineau 2000, Labrie and Moineau 2007), AbiT (Labrie and Moineau 2007), and AbiZ (Durmaz and Klaenhammer 2007)] insensitivity has been achieved via recombination with a prophage or the host genome. Interestingly, escape mutants derived from five systems [AbiA, AbiJ, Kamadhenu, Septu (Mosterd et al. 2025), AbiK (Bouchard and Moineau 2004)] were mutated in genes encoding a single-stranded DNA binding or DNA annealing protein, four systems generated escape mutants with altered structural proteins [major tail or capsid protein; AbiB, AbiG, AbiP (Mosterd et al. 2025), AbiT (Labrie et al. 2012)], two systems provoked escape mutants affecting the ATPase domain of the terminase large subunit [Rhea, Rugutis (Mosterd et al. 2025)], while eight systems resulted in escape mutants with changes in genes encoding hypothetical proteins [AbiB, AbiD-like (Mosterd et al. 2025), AbiD1 (Bidnenko et al. 1995), AbiQ (Samson et al. 2013a), AbiT (Labrie et al. 2012), AbiV (Haaber et al. 2009), PARIS and Type II CBASS (Mosterd et al. 2025)].

Similar to recent reports of identified phage proteins that activate antiphage systems in bacteria other than Lactococcus (Millman et al. 2020a, Depardieu et al. 2016, Rousset et al. 2022, Zhang et al. 2022, Stokar-Avihail et al. 2023), in the case of AbiB, AbiD-like, AbiG, AbiP, and Rugutis, the aforementioned genes mutated in escape mutant phages were also identified as triggers for these systems (Mosterd et al. 2025). Additionally, it was shown that AbiA escape mutants also circumvent AbiJ, which was not observed for other combinations, where the same gene was mutated in escape mutant phages that bypass distinct antiphage systems.

The terminase large subunit, major capsid, and major tail protein have previously been reported to activate antiphage systems present in other bacteria (Gao et al. 2022, Garb et al. 2022, Zhang et al. 2022), showing that despite the diversity, certain antiphage systems seem to be governed by general mechanistic principles regardless of the host origin and their sequence, structural and functional diversity.

Domains and structures of lactococcal Abi-like systems

The functional domains of most lactococcal Abi-like systems were initially explored over a decade ago using PSI-BLAST and HHPred methods (Makarova et al. 2013), but this was prior to the availability of modern structural prediction tools such as AlphaFold (Jumper et al. 2021). Here, we re-examined all lactococcal Abi-like systems using HHpred, Interpro and AlphaFold2/AlphaFold3 [Table 2; updated from (Grafakou et al. 2024a)]. Based on domain predictions and structural resemblance (summarized in Fig. 2), we grouped certain systems and used the predicted domain information to propose possible mechanisms of action. Additionally, we discuss the mode of action for some Abi-like systems where these are already known (summarized in Fig. 2). Structures were predicted by AlphaFold2 (Jumper et al. 2021) or AlphaFold3 (Abramson et al. 2024) on google servers at https://golgi.sandbox.google.com. The figures were made with ChimeraX (Pettersen et al. 2021). Coordinates of predicted structures are accessible on Zenodo (https://doi.org/10.5281/zenodo. 13 911 750).

Table 2.

Functional analysis of lactococcal Abi-like systems.

Figure 2.

Experimentally confirmed Abi-like systems in Lactococcus, grouped by domain prediction and structural superimposition, with schematic mechanistic details provided for select group members where data is available. Created in BioRender.

Defence-associated reverse transcriptase (DRT) systems: AbiA and AbiK

A low level of amino acid sequence identity (23%) was previously detected between AbiA and AbiK (Emond et al. 1997, Chopin et al. 2005). AbiA and AbiK share a reverse transcriptase (RT) catalytic domain (Figiel et al. 2022, Gapińska et al. 2024) (Figs 2 and 3A and B), which superimposes well (Fig. 3C). These defence-associated RT (DRT) Abi-like systems have been classified under the unknown group (UG)/Abi RT group, which also includes AbiP2 from E. coli and other predicted RTs whose antiphage activity and Abi mechanism have not (yet) been experimentally confirmed (Mestre et al. 2022).

Figure 3.

Structures of the defence-associated reverse transcriptase (DRT) systems AbiA and AbiK. (A) Ribbon view, of the crystal structure of AbiA monomer (PDB ID: 8OZ7). Rainbow coloured from N-terminus, blue, to C-terminus, red. (B) Ribbon view, of the crystal structure of AbiK monomer (PDB ID: 7R06). Rainbow coloured from N-terminus, blue, to C-terminus, red. (C) AbiA and AbiK superimposition aligns 256 out of 599 amino acids (aa), with a r.m.s.d of 3.9 Å.

AbiA (originally Hsp) was the first discovered lactococcal Abi system (Hill et al. 1990), and is among those with the broadest antiphage activity spectrum (Dinsmore and Klaenhammer 1994, Tangney and Fitzgerald 2002, Grafakou et al. 2024a). It was found to either interfere with or significantly reduce phage DNA replication (Hill et al. 1991, Tangney and Fitzgerald 2002) (Fig. 2). Recent crystallography studies revealed that AbiA contains three domains: the RT domain, a helical domain unique to Abi polymerases, and a C-terminal higher eukaryotes and prokaryotes nucleotide-binding (HEPN) domain (Mestre et al. 2022, Gapińska et al. 2024). The RT polymerase and HEPN domains are essential for AbiA antiphage activity (Gapińska et al. 2024). While the HEPN domain of AbiA was found to be enzymatically inactive, it was shown to be responsible for dimerization of AbiA (Gapińska et al. 2024). Other proteins containing a HEPN domain have been shown to function as toxins and can be critical components of Abi/toxin-antitoxin systems in prokaryotes (Anantharaman et al. 2013); see “Other HEPN-domain containing systems” paragraph below).

The mode of action of AbiK appears to be phage species-dependent, interfering with the phage cycle during or prior to replication of P335 phages (Emond et al. 1997) or during the DNA packaging stage of skunaviruses (Boucher et al. 2000) (Fig. 2). AbiK was found to consist of two domains, a helical domain and an RT domain (Figiel et al. 2022). Unlike AbiA, AbiK does not contain a HEPN domain and AbiK forms trimers instead of dimers (Figiel et al. 2022, Gapińska et al. 2024). The RT domain has previously been demonstrated to be essential for AbiK activity (Fortier et al. 2005, Wang et al. 2011). AbiK was found to synthesize DNA without a template, generating single stranded DNA sequences without any specificity (Wang et al. 2011). These DNA molecules remain covalently bound to AbiK. A role of the bound DNA in AbiK trimer formation has been proposed, but further investigation is required (Figiel et al. 2022). Nevertheless, the synthesis of random DNA could lead to nucleotide depletion, potentially interfering with phage replication and/or other essential host processes, which in turn would explain the abortive infection phenotype. However, antiphage activity of AbiK and AbiA was described to likely not rely on induced cell death (Gapińska et al. 2024).

Interestingly, phage escape mutants of both AbiA and AbiK have been shown to carry mutations in genes involved in DNA repair or recombination, such as single-strand annealing proteins (SSAPs) among others (Dinsmore and Klaenhammer 1997, Bouchard and Moineau 2000, 2004, Labrie and Moineau 2007, Mosterd et al. 2025). This further highlights the similarity between these two systems. Genes mutated in AbiK escape mutants are structurally diverse SSAPs and have been called sensitivity to AbiK (sak) genes (Bouchard and Moineau 2004, Scaltriti et al. 2011). Notably, phage escape mutants of AbiA with mutations in SSAP- and Holliday junction resolvase-encoding genes were also found to be insensitive to AbiJ. This suggests a mechanistic similarity between AbiA and AbiJ despite the lack of full-protein structural resemblance (Mosterd et al. 2025), possibly attributed to the presence of a HEPN domain in both proteins (see also next paragraph).

Other HEPN-domain containing systems: AbiB, AbiJ, and AbiV

While AbiB and AbiJ show only weak predictions (HHpred hits for HEPN domain containing-proteins with E-values > 0.1) or lack identifiable functional domains based on InterPro and HHpred, AbiV contains a HEPN domain based on InterPro analysis. However, AlphaFold and Dali analysis (Fig. 4A-C) revealed that the three proteins partially superimpose (Fig. 4D) on a bona fide HEPN domain (Fig. 4E-G). Notably, AbiJ and AbiV had previously been described as HEPN-domain-containing proteins (Anantharaman et al. 2013). As mentioned above, HEPN domain-containing proteins are known to function as toxins and can be essential toxin-antitoxin (TA)/Abi components in prokaryotes (Anantharaman et al. 2013). TA systems generally consist of a toxin that interferes with an essential cellular function (for example, transcription or translation) or component (for example, DNA, RNA, or NAD⁺) and an antitoxin that neutralizes the activity or effects of the toxin (Page and Peti 2016). Perturbation of the antitoxin, which can be triggered by phage infection, leads to activation of the toxin and thus abortive infection (Labrie et al. 2010). Based on the nature of the TA system components and their interaction, eight different TA system types (I–VIII) have been proposed (Singh et al. 2021).

Figure 4.

Predicted structures of AbiB, AbiJ, and AbiV. Ribbon views, rainbow coloured, of AbiB (A), AbiJ (B), and AbiV (C) (Rainbow coloured from N-terminus, blue, to C-terminus, red). (D) Superimposition of AbiB, AbiJ, and AbiV. Superimposition of AbiB on AbiJ aligns 118 out of 282 aa, with a r.m.s.d of 3.1 Å Å; superimposition of AbiB on AbiV aligns 112 out of 201 aa, with a r.m.s.d of 4.1 Å; superimposition of AbiJ on AbiV aligns 109 out of 201 aa, with a r.m.s.d of 3.4 Å. (E) Superimposition of AbiB on 4NQF (a bona fide HEPN domain) aligns 118 out of 145 aa, with a r.m.s.d of 2.9 Å. (F) Superimposition of AbiJ on 4NQF aligns 117 out of 145 aa, with a r.m.s.d of 2.9 Å. (G) Superimposition of AbiV on 4NQF aligns 118 out of 145 aa, with a r.m.s.d of 2.9 Å.

AbiV was recently confirmed to be a type III TA system (Fig. 2), while AbiF and AbiQ have likewise been characterized as type III TA systems (see paragraphs below). The AbiV toxin is an RNase, belonging to the HEPN superfamily, which degrades ribosomal RNA. Its antitoxin is an RNA molecule that inhibits the RNase activity of the toxin (Zhu et al. 2024). Since ribosomal RNAs are essential for cellular protein synthesis, degradation of host rRNAs can abort phage infection (Zhu et al. 2024), consistent with previous findings that AbiV inhibits protein synthesis (Haaber et al. 2010).

Activation of AbiB upon phage infection leads to mRNA degradation, thereby causing the arrest of phage development (Parreira et al. 1996). Although no particular mechanistic information is currently available for AbiJ, the presence of the HEPN domain suggests that both AbiB and AbiJ function as TA Abi systems, with AbiB likely operating through a mechanism similar to AbiV (i.e. RNase activity). Notably, the major capsid protein of Skunavirus sv66901 was recently found to trigger AbiB toxicity (Mosterd et al. 2025).

Sequence and structural similarity of AbiC and AbiP

Amino acid sequence alignment of AbiC and AbiP reveals 24% identity with a 57% coverage of AbiC, indicating a relatively modest level of similarity. Based on AlphaFold/Dali analysis, no functional predictions were successfully assigned to AbiC and AbiP; however, the two systems were found to superimpose well on the extra-membrane (or ecto-domain) regions; with the difference of AbiC possessing two transmembrane helices, whereas AbiP contains a single transmembrane helical domain (Fig. 5A and B). This structural similarity justifies grouping these previously distinct systems into a single Abi group, henceforth referred to as AbiC/P.

Figure 5.

Predicted structures of AbiC and AbiP. Ribbon views, rainbow coloured, of AbiC (A) and AbiP (B). Rainbow colours from N-terminus, blue, to C-terminus, red. (C) Superimposition of AbiC and AbiP aligns 93 aa of the ecto-domain out of 228 aa, with a r.m.s.d of 0.91 Å.

Interestingly, AbiC and AbiP exhibit activity against the same phage groups (Table 1). Despite these similarities, their mechanisms seem to differ. AbiC was not found to affect DNA replication (Durmaz et al. 1992), unlike AbiP (Domingues et al. 2004a), while AbiC has been shown to partially inhibit major capsid protein production of a phage of the P335 group (Moineau et al. 1993).

AbiP contains a sequence-independent nucleic acid binding domain, with preferential binding to RNA over ssDNA, and an N-terminal transmembrane domain, both apparently essential for its Abi phenotype (Domingues et al. 2008). Furthermore, the activity of AbiP has been demonstrated to result in higher levels of the early phage transcripts compared to the reference strain, interfering with their “switch off” which occurs 15 min post-infection in the reference strain; correspondingly, lower levels of the middle and late phage transcripts are produced in the presence of AbiP (Domingues et al. 2004a).

AbiP appears to lack specificity for phage-originating nucleic acids, implying that it likely binds non-specifically to cellular nucleic acids (Domingues et al. 2008). Nevertheless, AbiP was not shown to be toxic to the host cell, leading to the hypothesis that its activation requires a phage protein (Domingues et al. 2008). This hypothesis is supported by recent findings, demonstrating that the major tail protein of Skunavirus sk713 triggers AbiP toxicity [Fig. 2, (Mosterd et al. 2025)]. This suggests that AbiP can abort infection by binding cellular nucleic acids once activated by a phage structural component.

Group AbiD/F: sequence and structural similarity of AbiD, AbiD1, AbiF, and AbiD-like

Sequence similarity had previously been reported between AbiD, AbiD1, and AbiF (28–46%) (Garvey et al. 1995, Chopin et al. 2005). The AbiD-like system (also previously referred to as AbiD/F system) was later identified and was shown to share 59% amino acid identity with AbiD (Grafakou et al. 2024a). In this same study, structural analysis of AbiD, AbiD1, AbiF, and AbiD-like revealed high similarity of their protein structures and a resemblance to the Cas13d endonuclease HEPN domain-containing protein leading to their classification into a single Abi group called AbiD/F (Grafakou et al. 2024a). Notably, AbiD and AbiF had also been previously described as HEPN domain-containing proteins (Anantharaman et al. 2013).

The discovery of AbiD, AbiD1, and AbiF as antiphage systems occurred within the same time period. It was determined that AbiD acts after phage DNA replication (McLandsborough et al. 1995). Furthermore, it was demonstrated that transcription of the abiD1 gene is very weak (Anba et al. 1995). Recently it was determined that sequence identity upstream of the AbiD/F group member genes (68–80%) is higher than sequence identity between the genes themselves (46–65%) (Grafakou et al. 2024a). In addition, it was confirmed that this upstream sequence is essential for the associated antiphage effect for AbiF and that abiD1 and abiD-like could only be introduced with their native signals (upstream sequence) due to toxicity issues (Grafakou et al. 2024a). This was a strong indication that members of the AbiD/F group represent two-component systems that consist of a protein that shares a common ancestor with CRISPR-Cas RNA-guided nuclease Cas13d and a short RNA molecule.

This was confirmed for AbiF from the family Prevotellaceae (Zilberzwige-Tal et al. 2025) and AbiF from Clostridioides difficile (Saunier et al. 2025). AbiF was found to function as a type III TA system where the small non-coding RNA (ncRNA) upstream of abiF physically blocks the active site of the AbiF HEPN domain, neutralizing its bacteriostatic activity (Saunier et al. 2025, Zilberzwige-Tal et al. 2025). The structural resemblance of AbiD, AbiD1, AbiD-like, and AbiF and the presence of upstream ncRNAs suggest that the mode of action for the other members of the AbiD/F group possibly relies on a highly similar principle.

Phage escape mutants of AbiD1 (Bidnenko et al. 1995) and AbiD-like (Mosterd et al. 2025) possess mutations in genes encoding hypothetical proteins [orf1 of bIL66 phage for AbiD1; sk1p24 protein of sk1 phage; and sv62601_26 protein of sv62601 phage for AbiD-like (phage accession numbers: L35175.1, PP906167, and PP906171, respectively)]. Spontaneous Skunavirus bIL66 escape mutants of AbiD1 with a single mutation in a gene encoding a hypothetical protein within the so-called “middle-expressed” region believed to be associated with transcriptional regulation, provided partial resistance to AbiD1. Overexpression of another gene in the same operon, which interestingly encodes a RuvC-like endonuclease (Bidnenko et al. 1998), resulted in increased resistance of bIL66 against AbiD1 (Bidnenko et al. 1995). Therefore, AbiD1 was proposed to be involved in degradation of the phage-encoded RuvC-like endonuclease upon activation by the hypothetical protein (Bidnenko et al. 1998, 2002). The hypothetical protein was found to activate translation of AbiD1. In contrast, the escape mutant version (same hypothetical protein with a single amino acid difference) was not found to promote AbiD1 translation, even though both the wild type and the escape mutant hypothetical protein were found to bind to AbiD1 mRNA (Bidnenko et al. 2009). As for AbiD1, AbiD-like was found to be activated by two structurally unrelated proteins of unknown function (Mosterd et al. 2025) (Fig. 2).

AbiF interferes with Skunavirus 712 replication (Garvey et al. 1995) and its activity depends on the host factor RecA. RecA is ordinarily involved in homologous recombination and the SOS response to DNA damage (Garvey et al. 1997), but its mechanism of interaction of AbiF remains to be resolved. The cos site of phage sk1 was found to provide resistance against the action of AbiF in phage 712, a phage normally sensitive to AbiF (Rincé et al. 2000). The cos site of sk1 contains a hairpin structure that inhibits the activity of AbiF, despite the lack of resemblance to the AbiF ncRNA (Zilberzwige-Tal et al. 2025).

CRISPR-Cas systems are rarely found in L. lactis and L. cremoris (Millen et al. 2012, Couvin et al. 2018). However, it has been proposed that type VI CRISPR-Cas system effector Cas13 evolved from AbiF, via gradual changeover of the conserved AbiF ncRNA anti-toxin to function with variable non-coding CRISPR RNAs (crRNAs) (Saunier et al. 2025, Zilberzwige-Tal et al. 2025). The related type II CRISPR-Cas effector Cas9 was found to be indiscriminately activated when introduced into L. cremoris even in the absence of bound crRNAs (Mosterd and Moineau 2025), possibly explaining why Cas9 has only been introduced into L. cremoris through low-copy vectors (Song et al. 2021, Mosterd and Moineau 2025). The toxicity of class 2 CRISPR-Cas effector proteins such as Cas13 and Cas9 possibly relates to their origins as AbiF toxins and the involvement of structurally related proteins in abortive infection.

TOPRIM/OLD family nuclease and Rad50/SMC family ATPase domain containing systems: PARIS and AbiL

AbiL has been included among systems such as PARIS, which feature an ABC ATPase and a TOPRIM nuclease (Burman et al. 2024). The first component of AbiL (AbiLi) shares limited (34%) amino acid similarity to the first component of PARIS (PARIS-i or AriA, 18% query coverage) described in Lactococcus (Grafakou et al. 2024a) and originally identified in E. coli (Rousset et al. 2022). Based on previous analyses, PARIS-i (AriA, Fig. 6A) resembles an ABC-type ATPase related to Rad50 and structural maintenance of chromosomes (SMC) family proteins (Deep et al. 2024, Grafakou et al. 2024a). AbiLi (Fig. 6B) is here predicted to contain a Rad50 ATPase domain as well (Table 2), which confirms previous studies (Makarova et al. 2013, Burman et al. 2024).

Figure 6.

Predicted structures of AbiL and PARIS. Ribbon views, rainbow coloured, of PARIS-i (A), AbiLi (B), PARIS-ii (C), and AbiLii (D). Rainbow colours from N-terminus, blue, to C-terminus, red. (E) Superimposition of PARIS-i on AbiLi aligns 212 out of 453 aa, with a r.m.s.d of 2.7 Å. (F) Superimposition of AbiLii on the structure of an OLD family nuclease (PDB ID 6NJV; alidgned:123/206 aa).

No significant sequence similarity was found between the second components of AbiL and PARIS (AbiLii and PARIS-ii or AriB, respectively). PARIS-ii (AriB, Fig. 6C) contains a topoisomerase-primase (TOPRIM) domain of OLD family nucleases (Rousset et al. 2022, Deep et al. 2024, Grafakou et al. 2024a). AbiLii (Fig. 6D) is here predicted to contain the same functionality as PARIS-ii based on AlphaFold/Dali analysis (Table 2; Fig. 6F), confirming previous studies (Makarova et al. 2013, Burman et al. 2024).

Structural analysis revealed that the central β-barrel and α-helices of AbiLi and PARIS-i (AriA) superimpose well (Fig. 6E), while differences in their structures are observed in the extensions, specifically between residues 250–323 and 195–319, respectively. Despite the predicted shared functions (TOPRIM domain) between AbiLii and PARIS-ii (AriB), the superimposition of their predicted structures showed poor alignment.

It was recently described that PARIS functions as a type II TA system in E. coli (Deep et al. 2024) in which expression of PARIS-ii (or AriB; effector) results in bacterial toxicity, while through protein-protein interactions, PARIS-i (or AriA; infection sensor) acts as an antitoxin. During phage infection, the phage-encoded anti-restriction protein (Ocr) binds to PARIS-i (AriA; sensor and anti-toxin) which undergoes a conformational change. This causes release of the PARIS-ii RNase (AriB; effector/toxin), which aborts infection by inhibiting protein translation through cleavage of a lysine tRNA (Deep et al. 2024) (Fig. 2). However, phages can bypass PARIS activity by expressing lysine tRNA variants that are not cleaved by PARIS (Burman et al. 2024).

Interestingly, to overcome the activity of PARIS, lactococcal phages—where anti-restriction proteins have not been identified—exhibit mutations in a hypothetical protein [c2p02 protein of phage c2, accession number PP906169 supplementary Table S3 (Mosterd et al. 2025)]. Beyond possessing a highly negatively charged cluster, this hypothetical protein bears no sequence or structural resemblance to Ocr (Mosterd et al. 2025), which was shown to activate PARIS Abi system in E. coli (Rousset et al. 2022, Deep et al. 2024). However, this hypothetical protein, at least on its own, has not been found to activate PARIS in Lactococcus (Mosterd et al. 2025).

As previously suggested (Deep et al. 2024), further work will be needed to determine if all ATPase + TOPRIM/OLD phage defence systems, such as the lactococcal AbiL, the Campylobacter RloAB (Miller et al. 2005) and the unnamed two-gene system identified in Nostoc sp. (Makarova et al. 2011), operate through a similar mode of action.

Other TA systems: AbiQ and AbiE

In addition to AbiD/F group members, AbiV and PARIS (see above), some other lactococcal Abi systems, such as AbiE (Garvey et al. 1995, Dy et al. 2014) and ToxIN/AbiQ (Samson et al. 2013b, Fineran et al. 2009) function via TA mechanisms.

Interestingly, AbiE (and in particular the AbiEii component of this two-component system) is widespread in prokaryotic genomes (Beck et al. 2020, Tesson et al. 2022). AbiE functions as a bacteriostatic type IV toxin-antitoxin (TA) system (Dy et al. 2014). In Streptococcus agalactiae, AbiEi antitoxin, which consists of an N-terminal DNA binding domain and C-terminal antitoxin domain, negatively autoregulates abiE expression by binding to inverted repeats within the promoter, thereby repressing its transcription (Dy et al. 2014, Hampton et al. 2018, Beck et al. 2020). AbiEii toxin is a predicted DNA polymerase β-nucleotidyltransferase, that specifically binds GTP (Dy et al. 2014). Notably, the nucleotide specificity of AbiEii can vary among homologues; for instance, the activity of the Mycobacterium tuberculosis MenT3 toxin (AbiEii homologue) involves pyrimidines (Cai et al. 2020). Nevertheless, evidence suggests that AbiEii toxins transfer nucleotide(s) to tRNA(s), impairing translation (Cai et al. 2020, Hampton et al. 2020). Additionally, when plasmid-encoded, AbiE not only provides phage resistance but also prevents loss of the corresponding plasmid (Dy et al. 2014), a property previously demonstrated for other plasmid-encoded TA systems (Guglielmini and Van Melderen 2011).

AbiQ is homologous to the ToxIN Abi system found in Pectobacterium atrosepticum (formely Erwinia carotovora subspecies atroseptica) (Emond et al. 1998, Fineran et al. 2009). Both AbiQ and ToxIN are type III TA (protein-RNA) systems, with AbiQ being the first TA system described and structurally characterzed in Lactococcus (Samson et al. 2013b, Blower et al. 2011); AbiF and AbiV have also been identified as members of this type III TA family. AbiQ/ToxN type III toxins represent endoribonucleases with similar 3D structures and sequence-specific, ribosome-independent ribonuclease activity (Samson et al. 2013b, Blower et al. 2011, Brantl and Jahn 2015). The non-coding RNA antitoxin ToxI (ToxN inhibitor) inhibits toxicity by directly binding to ToxN (Fineran et al. 2009, Blower et al. 2011). Additionally, the toxIN operon is negatively autoregulated, whereby ToxN may repress its own transcription thereby preventing ToxI and ToxN synthesis (Blower et al. 2009). AbiQ/ToxIN are activated upon phage infection which then results in abortive infection (Samson et al. 2013b, Blower et al. 2011, Short et al. 2018). ToxN is activated by phage-induced host transcriptional inhibition in E. coli, leading to loss of the ToxI antitoxin and to the specific cleavage of viral mRNAs by ToxN (Guegler and Laub 2021). This suggests that ToxIN does not function as a typical abortive infection system; rather, phage infection triggers cell death, while ToxIN prevents the formation of new mature virions, thereby protecting neighbouring cells (Guegler and Laub 2021). AbiQ activation affects phage transcription and results in the accumulation of concatenated non-mature forms of viral DNA (Samson et al. 2013a, Emond et al. 1998). Phages can escape ToxIN via RNA antitoxin mimics that inhibit ToxN toxicity (Blower et al. 2012).

Systems with other domains associated with phage defence

Based on HHpred and Alphafold/Dali analysis AbiH resembles sirtuins (Table 2, Fig. 7A and B). Sirtuins (SIRTs) are homologues of the Sir2 protein (silent information regulator 2) found in Saccharomyces cerevisiae and comprise nicotinamide adenine dinucleotide (NAD⁺)-dependent enzymes typically functioning as deacetylases or ADP-ribosyltransferases (Rine et al. 1979, North and Verdin 2004, Gallego-Jara et al. 2021). They are present in prokaryotic and eukaryotic organisms, with seven mammalian homologues of Sir2 identified (SIRT1 to SIRT7) (North and Verdin 2004, Wu et al. 2022). Bacterial sirtuins are involved in transcription, protein translation, carbon and nitrogen metabolism, and virulence (Gallego-Jara et al. 2021). Sir2 domains have been shown to be associated with multiple bacterial phage defence systems, such as the prokaryotic argonautes (pAgo) (Makarova et al. 2009), Thoeris (Doron et al. 2018, Ofir et al. 2021), AVAST, defence-associated sirtuin (DSR) and others (Garb et al. 2022). Several systems, such as Thoeris, DSR2, SIR2/pAgo, and SIR2-HerA, have been shown to rapidly deplete cellular NAD⁺ upon sensing of a phage infection event, thus aborting infection by depleting this essential electron transfer molecule from the cell (Ofir et al. 2021, Garb et al. 2022). Phage defence involving NAD⁺ depletion has been shown to eventually result in cell death, consistent with the concept of Abi systems (Tang et al. 2018, Morehouse et al. 2020, Ofir et al. 2021). Notably, NADase activity of the DSR2 Abi system (DSRs family) has been shown to be activated by a phage tail tube protein (Garb et al. 2022). It is therefore possible that AbiH acts via a similar NAD-depleting pathway.

Figure 7.

Predicted structure of systems with domains previously associated with phage defence. (A) Ribbon view, rainbow coloured, of AbiH. (B) Superimposition of AbiH on the structure of a SIRT3 deacetylase (PDB ID 5YTK; aligned:173/275aa). (C) Ribbon view rainbow coloured, of AbiI. (D) Superimposition of AbiI on the structure of a Class II Ribonuclease H (PDB ID 1EKE; aligned:163/225aa). (E) Ribbon view, rainbow coloured, of AbiO. (F) Up: Superimposition of AbiO C-terminal domain on the structure of an Upf1 helicase (PDB ID 2XZO; aligned: 263/613aa). Down: Superimposition of AbiO N-terminal domain on the structure of restriction endonuclease (PDB ID 3IJM; aligned: 108/146). (G) Ribbon view, rainbow coloured, of AbiTi. (H) Superimposition of AbiTi on the structure of an endo-ribonuclease domain (PDB ID 8VER; aligned:78/286aa) (I) Ribbon view, rainbow coloured, of AbiTii. (J) Ribbon view, rainbow coloured, of AbiU (K) Superimposition of AbiU on the structure of a restriction endonuclease (PDB ID 1FOK; aligned:146/568aa). (L) Ribbon view, rainbow coloured, of AbiZ and its N-terminal domain (right). Dashed boxes highlight the domains. No significant hits have been found in the PDB by Dali or Foldseek. Rainbow colours from N-terminus, blue, to C-terminus, red. The superimposed domains are identified by a red dashed box.

Based on Alphafold2/Dali analysis, AbiI contains a RNase HII domain (Table 2, Fig. 7C and D). RNase H proteins are found in all domains of life including viruses and cleave RNAs in RNA/DNA hybrids (Lai et al. 2000). As with other Abi-like systems (e.g. see above-mentioned AbiV, PARIS, AbiQ), the presence of such a domain in AbiI would be in accordance with the observed abortive infection phenotype (Su et al. 1997).

Based on functional and structural analysis, the AbiO N-terminal domain resembles a restriction endonuclease type II-like domain-containing protein, belonging to the PD-(D/E)XK nuclease superfamily. Its C-terminal domain is predicted to encompass a superfamily 1B (SF1B) helicase (Table 2, Fig. 7E and F). Helicases are known to be involved in antiviral immunity, either as sensors or as effectors (Ahmad and Hur 2015). At least 18 bacterial phage defence systems, including both Abi and non-Abi systems (e.g. Hachiman, Hna, DISARM, Gabija, Zorya, BREX, Nhi) are known to contain helicase domains, which have subsequently been grouped into seven subfamilies (Tuck et al. 2024). For example, Hachiman in E. coli, is a two-component Abi-like antiphage system where the sensor is a DNA helicase (HamB) and the effector is a nuclease (HamA) activated upon DNA damage (Tuck et al. 2024). Similarly, the activation of the Hna single-component antiphage system in Sinorhizobium meliloti, consists of a helicase and a nuclease domain, resulting in abortive infection [Hna; (Sather et al. 2023)]. AbiO was previously shown to be toxic in Lactococcus when its encoding gene was cloned in a high-copy number vector (Prévots and Ritzenthaler 1998), supporting the abortive infection phenotype assigned to this phage defence system. Notably, based on InterPro analysis, AbiO contains a helicase domain found in group 3 of mammalian Schlafen proteins (SLFN_3-like; Table 2), which are associated with cell proliferation regulation, immune responses activation, and viral replication control (Mavrommatis et al. 2013); in contrast, based on HHpred analysis, AbiO resembles Upf1 helicase, which is involved in nonsense-mediated mRNA decay (Chakrabarti et al. 2011).

AbiTi was found to contain an endoribonuclease domain (Table 2, Fig. 7G and H). Notably, although no HEPN domain is predicted for the lactococcal AbiTii protein (Fig. 7I), its homologues in various bacteria are found to have two distinct HEPN domains at their C-terminal end (Anantharaman et al. 2013). The presence of such domains is in accordance with the observed abortive infection phenotype.

AbiU was initially reported to be a two-component antiphage system, shown to delay phage transcription (Dai et al. 2001). As the expression of AbiUi was sufficient to confer phage resistance, while the expression of AbiUii in addition to AbiUi resulted in lower antiphage activity (Dai et al. 2001), AbiU will here be referred to as a single component antiphage system consisting of AbiUi. AbiUii has been predicted to contain a HEPN domain (Dai et al. 2001, Anantharaman et al. 2013) and exhibits sequence similarity to AbiGii (31% similarity; 19% AbiGii query coverage). Based on AlphaFold2 analysis, AbiUi resembles a nuclease (Table 2, Fig. 7J and K).

AbiZ likely acts cooperatively with the phage holin to increase membrane permeability, leading to premature lysis of infected cells (Durmaz and Klaenhammer 2007) (Fig. 2). This early cell lysis results in a reduced phage burst size, as mostly unfinished virions are released (Durmaz and Klaenhammer 2007). Here, although the structure of AbiZ was successfully predicted by Alphafold (Fig. 7L), and its mode of action has been largely described, no functional domain could be identified with confidence for AbiZ.

Aristaios is an Abi-like system recently described to resemble an ADP-ribosyltransferase (Grafakou et al. 2024a). ADP-ribosyltransferases have been associated with phage defence previously and are toxins known to inhibit essential host protein functions (Simon et al. 2014, LeRoux et al. 2022), which is consistent with observations that Aristaios interferes with phage transcription (Grafakou et al. 2024b).

Finally, several Abi-like systems that were initially associated with phage defence in other species and were recently experimentally confirmed in L. lactis/cremoris (i.e. CBASS type I, CBASS type II, Lamassu, and Septu) have been analyzed extensively elsewhere, but are included in Table 2 for completeness (Grafakou et al. 2024a). In short, CBASS systems consist of at least an oligonucleotide cyclase that produces cyclic oligonucleotide signals in response to phage infection, and an effector that, upon activation by these signals, induces cell death before phage replication can be completed (Millman et al. 2020b). Lamassu, is a family of phage defence systems, typically composed of a protein of the structural maintenance of chromosomes (SMC) family (LmuB sensor), a diverse set of enzymatically diverse LmuA effectors (i.e. nuclease, protease), and occasionally a protein of unknown function (type II Lamassu) (Millman et al. 2022). The Septu system has two components: PtuA and PtuB. PtuA contains an ATPase domain, whereas PtuB is a nuclease belonging to the His–Asn–His (HNH) family (Doron et al. 2018, Li et al. 2024). PtuAB inhibits phage proliferation by cleaving phage genomic DNA, and ATP binding to PtuA can regulate the nuclease activity of the PtuAB complex (Li et al. 2024).

Systems with domains lacking phage defence associations

Functional and structural analyses fail to predict any similarity with proteins of known function for some Abi-like systems such as AbiG (Table 2, Fig. 8A and B), which was recently described to be activated by a phage major tail protein (Mosterd et al. 2025).

Figure 8.

Predicted structure of systems with domains previously not associated with phage defence. (A) Ribbon view, rainbow coloured, of AbiGi. (B) Ribbon view, rainbow coloured, of AbiGii. No significant hits have been found in the PDB by Dali or Foldseek for AbiGi and AbiGii. (C) Ribbon view, rainbow coloured, of AbiN. (D) Superimposition of AbiN on the structure of a ligand binding sensor domain (PDB ID 5XUA; aligned: 95/137aa). (E) Ribbon view, rainbow coloured, of AbiRi. (F) Superimposition of AbiRi on the AFDB structure of a Sigma70 RNA polymerase domain (AlphaFold Protein Structure Database, AFDB ID AF-W1VBZ4-F1 model; aligned:364/820aa; however, due to domains flexibility, the real domain identity covers the whole protein). (G) Ribbon view, rainbow coloured, of AbiRii. (H) Superimposition of AbiRii on the structure of a PARB domain nuclease (PDB ID 5K5D; aligned: 196/252aa). (I) Ribbon view, rainbow coloured, of AbiRiii. (J) Up: Superimposition of AbiRiii C-terminal domain on the structure of a helicase ATP binding domain (AFDB ID AF-K0FA99-F1-model_v6; aligned: 470/611aa). Down: Superimposition of AbiRiii N-terminal domain on the structure of an ATP dependant endonuclease domain (PDB ID 7CLG; aligned: 154/167aa). Rainbow colour is from blue (N-terminus) to red (C-terminus). The superimposed domains are identified by a red dashed box.

The structure of AbiN is predicted with high confidence by AlphaFold and resembles a putative periplasmic ligand-binding sensor domain protein (Fig. 8C and D). Periplasmic binding proteins function as bacterial receptors that sense small molecules and undergo conformational changes upon ligand binding (Borrok et al. 2009). This suggests that such a domain serves as the sensor in an Abi-like system, though further research is needed to clarify the mechanism of AbiN and if it functions via abortive infection.

AbiR was initially described to be encoded by two genetic loci, separated by the LlaKR2I RM system, and was found to interfere with phage DNA replication (Twomey et al. 2000). AbiR consists of three genes and its function requires the methylase from the LlaKR2I RM system (Yang et al. 2006). Notably, in the absence of the LlaKR2I methylase, AbiR was shown to be toxic to L. lactis (Yang et al. 2006).

Based on HHpred and Alphafold2/Dali analysis, AbiRi (or AbiRa) resembles an RNA polymerase sigma70 factor (Fig. 8E and F), AbiRii (or AbiRb) resembles partition system ParB (Fig. 8G and H), AbiRiii (or AbiRc) N-terminal domain resembles an ATP-dependent restriction endonuclease, whereas its C-terminal domain shows similarity to a nucleosome-bound INO80 ATPase (Foldseek) as well as to HepA-like Snf2 family ATPases and Dexx-box helicases (AlphaFold Data Base) (Table 2, Fig. 8I and J), with a reliable complex being predicted between these three components. HepA (Hepatitis A virus RNA helicase) and its bacterial homologs (such as HrpA) belong to the Superfamily 2 (SF2) helicases, sharing core ATPase domain features with Snf2 family ATPases and DExx-box proteins. Sigma70 factors are involved not only in directing general transcription but also in activating specific genes in response to stress or other development and ancillary metabolism signals (Paget and Helmann 2003). Therefore, such a protein domain may act as the sensor of the system. ParB is involved in chromosome and low-copy number plasmid segregation during bacterial cell division (Osorio-Valeriano et al. 2021). It is known to bind DNA and proteins such as the nucleoside triphosphatase ParA (Schumacher et al. 2015, Osorio-Valeriano et al. 2021). INO80 is an SNF2 family chromatin-remodelling complex with a core ATPase and has been shown to be involved in regulating transcription, DNA replication and repair (Conaway and Conaway 2009, Wu et al. 2023). Despite the possible interaction of the predicted AbiRii and AbiRiii domains with DNA or their involvement in DNA replication—a process known to be affected by AbiR activity—it remains unclear how the AbiR components interact to confer phage resistance.

The domains, as predicted by AlphaFold2, present in the recently identified Rhea (transposase TnsE), Kamadhenu (YfbU), Rugutis (SEC-C motif), and Hesat (colicin-D) could not be linked to other antiphage defence functions. Rhea has been reported to resemble the DNA-binding protein TnsE, which is believed to recognize replication intermediates typical of conjugative plasmids or of replicating phages (Peters and Craig 2001). Therefore, it was suggested that this domain is responsible for sensing phage infection. Kamadhenu showed structural similarity to YfbU, a protein implicated in cell death triggered by DNA damage (Amitai et al. 2009). Despite a high-confidence AlphaFold2 structural prediction, Rugutis exhibits a fold distinct from any currently available in the Protein Data Bank; however, it contains a SEC-C motif at its N-terminus. Hesat was shown to display strong similarity to colicin-D, a toxin that specifically cleaves the anticodon loop of all four tRNA(Arg) isoacceptors (Graille et al. 2004), suggesting that it inhibits translation, ultimately leading to cell death.

Conclusions

Most lactococcal Abi-like systems have been classified as such based on limited phenotypic observations associated with these systems. Direct evidence that the activity of the system upon recognition of phage infection results in cell death is often absent. Therefore, it is more appropriate to refer to them as Abi-like until further characterization proves their function as bona fide Abi systems. The recent discovery of lactococcal plasmid-encoded antiphage systems (Grafakou et al. 2024a) has substantially increased the number of experimentally confirmed Abi-like systems in Lactococcus, from 23 to 34. These systems appear to be highly diverse, as evidenced by the range of phages they can target, the specific steps of the phage lytic cycle they interfere with, and the diversity of genes that phages can mutate to overcome each system (Table 1). This diversity confirms that the co-evolution of these bacteria and their phages has driven the selection and exchange of a wide array of phage defence mechanisms (Chopin et al. 2005).

The substantial expansion of the (lactococcal) Abi-like system repertoire, the development of strategies to uncover novel defence systems (Millman et al. 2022, Vassallo et al. 2022) and the characterization of some of these systems have significantly advanced our understanding of pan-bacterial immunity (Georjon and Bernheim 2023). These advancements hold the promise of discovering even more defence systems, whether they are unique to Lactococcus or widespread across different bacterial genera.

Most of the known lactococcal antiphage repertoire was originally discovered as plasmid-encoded systems, which led to the notion that plasmids are the main source of phage defence in Lactococcus. However, large-scale sequence analysis of lactococcal genomes and plasmids in this study revealed that most antiphage systems found in sequenced strains (typically strains used in the dairy industry) are chromosomally encoded. These findings suggest that although plasmid acquisition from environmental sources or transfer between strains can increase protection of dairy starter cultures against phage infections, the strains currently in use likely already possess a high baseline level of (less-transferable) phage resistance impeding failed fermentations.

Of the lactococcal Abi-like defence systems, the mechanisms of action of only a few have been extensively characterized and validated as abortive infection: AbiE (Dy et al. 2014, Cai et al. 2020, Hampton et al. 2020), AbiF (Saunier et al. 2025, Zilberzwige-Tal et al. 2025), AbiV (Zhu et al. 2024), AbiQ (Samson et al. 2013b), PARIS (Millman et al. 2020b, Deep et al. 2024), and CBASS (Millman et al. 2020b). These are, with the exception of CBASS, identified as TA Abi systems. Notably, apart from AbiV and AbiQ, the modes of action for these systems have been described in bacteria other than Lactococcus. Therefore, it is intriguing whether future studies will determine if the mechanisms of a given system are consistent across different bacterial phyla.

Additionally, the necessary components responsible for the antiphage activity of each identified system require further investigation, particularly for initially described systems (AbiA to AbiZ) (Ainsworth et al. 2014, Philippe et al. 2023). For example, the component(s) responsible for AbiS activity remain(s) unknown (Holubová and Josephsen 2007), while only recent studies have revealed that AbiV requires an RNA component for its activity (Zhu et al. 2024), in addition to the previously known AbiV protein (Haaber et al. 2008).

Despite the diversity of Abi-like antiphage systems confirmed in Lactococcus, some systems share sequence and structural similarities, allowing classification under the same group (i.e. DRT group for AbiA and AbiK, AbiC/P group, AbiD/F group). However, these similarities and the categorization under the same group do not automatically imply that they are mechanistically identical. This suggests that lactococci have evolved paralogous genes to enlarge the substantial variety of available phage resistance mechanisms.

Furthermore, despite the lack of sequence similarity, some systems share common (functional) domains. For example, quite a few systems contain a HEPN domain (AbiB, AbiD, AbiD1, AbiD/F, AbiF, AbiV, AbiJ, AbiA, AbiTii in other genera than Lactococcus), while others contain a TOPRIM/OLD family nuclease and a Rad50/SMC family ABC ATPase domain (PARIS and AbiL). In addition, certain Abi-like systems are predicted to resemble proteins or contain domains commonly associated with phage defence, such as Sir2 (AbiH), helicase (AbiO), and nucleases (AbiI, AbiU and AbiT). At the same time, particular systems either show similarities to proteins not previously linked to phage defence (AbiN and AbiR) or have no clear similarity to known proteins (AbiG and AbiZ). These findings suggest that while certain functions are common in phage defence, there are also unique or specialized systems that might represent different mechanisms to combat phage infections.

Advancing our understanding of lactococcal phage defence systems has significant implications for the dairy industry. Identifying defences located on transferable plasmids and introducing them via food-grade techniques such as conjugation provides a practical route to generate more phage-resistant strains (Hickey et al. 2001; Grafakou et al. 2024). Moreover, stacking antiphage systems with diverse mechanisms may instill starter cultures with robust, multilayered resistance that is harder for phages to overcome (Mosterd et al. 2025). However, potential interactions—whether antagonistic or synergistic, as observed in other systems (Wu et al. 2024)—along with their in vivo effectiveness and breadth of protection remain poorly understood. Given that lactococcal strains typically harbour approximately four phage defence systems each (see results above), these factors will ultimately determine whether immunity profiles can reliably predict—or guide the design of—the most phage-resistant strains. Nevertheless, future studies on lactococcal and non-lactococcal Abi-like antiphage systems, along with other phage defence mechanisms, will continue to illuminate the landscape of bacterial phage immunity. This knowledge not only advances our understanding but can also guide the development of more robust phage-resistant starter cultures and has the potential to be applicable in other contexts.

Author contributions

A.G, C.M, J.M., and D.v.S. were involved in the conceptualization of the study. A.G. and C.M. drafted the initial manuscript. S.A.J. performed bioinformatic analyses. C.C. performed the AlphaFold2 predictions and analysis. P.d.W. and N.v.P. contributed to meaningful discussion. J.M., D.v.S., and S.A.J. reviewed and edited the manuscript. All authors contributed to the final version of the manuscript.

Conflicts of interest

A.G., P.d.W., and N.v.P. are employees of dsm-firmenich.

Funding

This publication has emanated from research conducted with the financial support of Science Foundation Ireland co-funded by dsm-firmenich (Grant Numbers 12/RC/2273‐P1, 12/RC/2273‐P2, and 17/SP/4678). This work was performed in part using HPC resources from GENCI-IDRIS (Grant number 2023-AD010714075). S.A.J. was supported by a Sir Charles Hercus Fellowship from the Health Research Council of New Zealand and funding from the Marsden Fund, Royal Society of New Zealand Te Apārangi. For the molecular graphics we acknowledge UCSF ChimeraX which is developed by the Resource for Biocomputing, Visualization, and Informatics at the University of California, San Francisco, with support from National Institutes of Health R01-GM129325 and the Office of Cyber Infrastructure and Computational Biology, National Institute of Allergy and Infectious Diseases.