Expanding natural transformation to improve beneficial lactic acid bacteria

Abstract

Nowadays, the growing human population exacerbates the need for sustainable resources. Inspiration and achievements in nutrient production or human/animal health might emanate from microorganisms and their adaptive strategies. Here, we exemplify the benefits of lactic acid bacteria (LAB) for numerous biotechnological applications and showcase their natural transformability as a fast and robust method to hereditarily influence their phenotype/traits in fundamental and applied research contexts. We described the biogenesis of the transformation machinery and we analyzed the genome of hundreds of LAB strains exploitable for human needs to predict their transformation capabilities. Finally, we provide a stepwise rational path to stimulate and optimize natural transformation with standard and synthetic biology techniques. A comprehensive understanding of the molecular mechanisms driving natural transformation will facilitate and accelerate the improvement of bacteria with properties that serve broad societal interests.

The authors review the advantages of using natural DNA transformation to genetically improve lactic acid bacteria for human benefits by providing a step-by-step strategy to predict and trigger this natural gene exchange process.

Introduction

The constant rise of human population puts under pressure our agricultural systems, while, at the medical level, emerging diseases (e.g. due to antibiotic resistant bacteria) and modification of our environment impinge on human health (de Kraker et al. 2016, van Duin, Paterson 2016). Therefore, new complementary strategies and optimization of current bioproduction methods will be required in a near future. In this scope, the huge diversity of beneficial lactic acid bacteria (LAB) turns out to be a promising pool of metabolic factories that we can tap into.

For decades, LAB have assumed a tremendous and growing economic importance in the agro-food sector, thanks to their fermentative lifestyle that converts various sugars into (isomers of) lactic acid (De Filippis et al. 2020). They have a key role in beverages and food maturation/transformation of dairy products and various fishes and vegetables (e.g. cheese, cabbage, yogurt, kefir, sourdough bread, kimchi, sauerkraut, and pickles), and have been used as natural preservatives for millenniums. Currently, LAB participate in a myriad of processes, including improvement of nutrient value, organoleptic diversity, and product texture (Johansen 2018). In addition, their epithelium colonization capabilities and potentially their secretion of antimicrobial compounds such as bacteriocins make several LAB valuable and safe probiotics (Hols et al. 2019, Mathur et al. 2020). Other applications based on LAB are emerging in the industry and could replace the traditional and well-known Escherichia coli and Saccharomyces cerevisiae bio-based factories (Hugenholtz, Smid 2002, Papagianni 2012). The elaboration of fermentation molecules (aroma), biopolymers, vitamins, cofactors, plastic building blocks, solvents, pharmaceuticals, and cosmetics might be expanded to LAB as they neatly adapt to food-related environments. First, LAB could be maintained in a bioreactor at low cost considering that the nutritional media used to grow these LAB cell factories could emanate from raw and nontransformed agricultural products. Second, LAB might be exploited with a recycling scope to valorize biomaterial waste such as whey for bioproduction purposes. In addition, LAB are part of the phyto-microbiome and could be used for biocontrol of agricultural lands as a safe substitute for pesticides (Lamont et al. 2017). Prospectively, LAB gain attention in regard to drug and vaccine delivery, diagnostic, and therapeutics (Van Braeckel-Budimir et al. 2013, Bron, Kleerebezem 2018, Chowdhury et al. 2014) such as fecal transplants and curing oral vectors for specific gastro-intestinal disorders.

Sometimes, beneficial properties required for a specific purpose are borne by different microorganisms. They have to be blended to operationally optimize functionalities. However, some of them could be mutually exclusive or impede growth in industrially relevant conditions. Therefore, these genetically encoded capabilities could be transferred into a unique microorganism adapted to human constraints (e.g. fermentation reactors or food safety issues). Selection of upgraded LAB, thus represents a promising scope to optimize their probiotic or phytosanitary properties, control their antimicrobial behavior, favor their protection against bacteriophages, combine multiple interesting genetic traits in a unique cell factory, and/or enhance the yield of value-added products. From a scientific viewpoint, genetic improvement might also be used to decipher LAB metabolism, stress response, food adaptation, and molecular interactions with the host (plant and animals) immune system (Borner et al. 2019, Cho et al. 2020, Hols et al. 2019).

A comprehensive understanding of how beneficial LAB thrive in relevant environments will be a prerequisite to strain improvement. Interestingly, bacterial natural transformation involving homologous DNA is a highly efficient process for the modification of LAB genomes for research and applied purposes. This intrinsic capacity enables recipient cells to capture extracellular fragments of DNA available in the surrounding medium and promote foreign DNA insertion into their chromosome (Johnsborg et al. 2007). Working with pathogenic LAB strains (i.e. Streptococcus pneumoniae), Frederick Griffith discovered natural transformation in 1928 and directly realized the tremendous potential to transfer genetic traits in between bacteria (Griffith 1928). Nowadays, it is a simple and fast technique to routinely make genetic changes to a significant number of bacteria in laboratory environments.

In this review, we showcase beneficial LAB for their intrinsic property to take up DNA and to modify their genome. We describe the physiological process, the regulation mechanisms, and the protein factors required to generate a productive natural transformation. We use bacterial paradigms, including Bacillus subtilis (non-LAB species) and S. pneumoniae (pathogenic LAB species), to extrapolate some aspects for beneficial LAB strains. We also show how widespread these factors are within industrially relevant LAB. We next elaborate on various practical strategies to augment transformability of LAB strains. Finally, we discuss the benefits of natural transformation to decipher bacterial physiology with concrete examples.

Biogenesis of the competence nanomachine

To be naturally transformable, bacteria have to enter a transient physiological state known as competence, during which they execute a dedicated sequential developmental program, i.e. intricately wired to fundamental processes of bacterial life cycle (Chen, Dubnau 2004, Salvadori et al. 2019). During this stage, they synthesize about 15 proteins that assemble into a sophisticated helical nanomachine at the cellular surface to internalize extracellular naked DNA (Claverys et al. 2009, Dubnau, Blokesch 2019; Fig. 1A). All these proteins that form the so-called transformasome are essential for an effective transformation process and they are typically encoded in half a dozen of operons, including the so-called late competence (com) genes (as opposed to early com genes; see next section; Fig. 1B). They can be clustered into three specific phases.

Figure 1.

The transformasome biogenesis. (A) Schematic representation of the transformasome architecture. The genes involved in the three steps of DNA transformation, namely the DNA docking, transport, and processing, are color-coded in red, green, and blue, respectively. The proteins composing the competence machinery are shaded accordingly. ComC processes the major and ancillary pilin subunits to promote their incorporation into the pseudopilus filament. The two ATPases ComGA and ComFA generate energy via ATP hydrolysis. (B) Operon structure of the late com genes. The red boxes in the promoter region of com genes depict the ComX binding sites (cin-boxes).

DNA docking

The first transformation step requires free stretches of double-stranded DNA floating in the environment. To initiate electrostatic contacts with DNA, bacteria assemble a pseudo-pilus macrostructure reminiscent of type IV pili and the type II secretion system (Laurenceau et al. 2013, Muschiol et al. 2019). They are formed by an ordered heteromer of major (ComGC) and minor (ComGD-GE-GF-GG) pilin subunits that span the cell wall to be exposed at the cellular surface (Chen et al. 2006, Muschiol et al. 2017; Fig. 1A). Synthesized as precursors, the pilin subunits are processed via ComC (pre-pilin peptidase; Chen et al. 2006) to cleave the hydrophilic class III signal sequence prior to incorporation of mature pilins at the base of the filament anchored by the transmembrane platform protein ComGB. Inside the fiber, both major and minor pilins are linked through noncovalent interactions, while major pilins are also bound via covalent disulfide bridges (Chen et al. 2006). The pilus extension is driven by pilin incorporation and requires energy, drawn in the proton-motive force, which is generated by the AAA⁺ (ATPases Associated with various cellular Activities) ATPase ComGA (Laurenceau et al. 2013). The ComG proteins are required to bind DNA even if no such capacity was reported for the pilin polymer itself (Chung, Dubnau 1998, Briley et al. 2011, Laurenceau et al. 2013). Instead, the pseudopilus is supposed to facilitate DNA access to the DNA receptor protein ComEA via the modification of the cell wall and envelope, presumably with the creation of a pore (Provvedi, Dubnau 1999). Nonexclusively, free DNA strands might “slide” on the pilus, or the pilus retraction (depolymerization of pilin subunits) might escort DNA to bring it in intimate contact with ComEA (Lam et al. 2021). Its C-terminus encompasses helix-hairpin-helix motifs responsible for the nonspecific sequence recognition of DNA (Provvedi, Dubnau 1999). It is noteworthy that a comEA deletion mutant still sustains a residual DNA-binding activity, arguing for the presence of other DNA receptors on the cell envelope (Inamine, Dubnau 1995).

DNA translocation

Due to a membrane anchor at its N-terminus, ComEA is in close proximity with the outer leaflet of the cellular membrane. Once ComEA is loaded with DNA, it interacts with and delivers DNA to the channel ComEC (Kramer et al. 2007) and the nuclease EndA (Berge et al. 2013). This large permease ComEC features seven transmembrane helices and homodimerizes to allow the passage of DNA (Draskovic, Dubnau 2005). Simultaneously, the constitutively produced EndA converts double-stranded DNA into single-stranded DNA that translocates in the 3’–5’ orientation (Mejean, Claverys 1988) through ComEC with the concomitant release of soluble nucleotides into the surrounding environment (Midon et al.2011, Moon et al. 2011). The energy required for DNA transport across the membrane is powered by the inner leaflet-associated ComFA that hydrolyses ATP through its ATP-binding cassette domain (Diallo et al. 2017). Its function could be to merely energize the single-stranded DNA passage by itself. However, ComFA, in association with co-factors, might alternatively consume ATP to gate the permease ComEC or to unwind double-stranded DNA.

Genome remodeling

On the cytoplasmic face of the membrane, ComFA forms a complex with ComFC, a competence-essential protein of elusive function, and DprA, that protects the incoming single-stranded DNA from intracellular housekeeping nuclease attack (Diallo et al. 2017). DprA subsequently facilitates homologous recombination by loading RecA onto the foreign DNA (Mortier-Barriere et al. 2007, Quevillon-Cheruel et al. 2012). RecA forms a filament that surrounds DNA and scouts the bacterial chromosome for sequence homology (Berge et al. 2003). In combination with DprA, SsbB (and SsbA), and CoiA, it stabilizes single-stranded DNA in the cytoplasm, increases its survival time, and maximizes the probability of genetic exchange (Desai, Morrison 2007, Attaiech et al. 2011, Morrison et al. 2007).

The central node for competence activation

Competence imposes a dramatic burden on cell physiology (Nester, Stocker 1963, Haijema et al. 2001, Mignolet et al. 2018). First, the production of transformasome subunits and the dynamics of the pseudo-pilus consume abundant building blocks (amino acids) and energy (ATP and PMF). Additionally, competent state engages a series of DNA-related processes (e.g. recombination) barely compatible with a vegetative growth (Berge et al. 2017). This explains why competence is widely regulated in time and space according to internal and environmental conditions (Johnston et al. 2014, Fontaine et al. 2015). In contrast, model LAB strains such as S. pneumoniae R6, Streptococcus mutans UA159, and Streptococcus thermophilus LMD-9, were extensively studied because they tolerate to develop competence in laboratory conditions, sometimes for extended periods of time, and display an intrinsic high rate of transformation (Gardan et al. 2009, Fontaine et al. 2010a, Johnston et al. 2020, Underhill et al. 2019). Owing to disadvantageous collateral effects, competence requires a robust “no-return” commitment at appropriate moments. Therefore, a master competence regulator, part of the early phase of competence, serves as a cellular checkpoint that integrates external and internal cues at the transcriptional and post-translational level to implement the competence developmental program (Fig. 2). In LAB, this bottleneck regulator is an alternative sigma factor named ComX (or SigX; σ^X) in streptococci, lactococci, and Leuconostoc sp. (Fontaine et al. 2015), or SigH in lactobacilli and pediococci (Schmid et al. 2012). Of note, the taxonomy of the Lactobacillus genus was recently revised by its splitting into numerous new genera (Zheng et al. 2020). For clarity and simplicity, we will hereunder refer to them as Lactobacillus, occasionally providing the new name into square brackets. Both sigma factors are phylogenetically related, however, they are genus-specific given that there has been no reshuffling/horizontal transfer in between the two clades (Fig. 3). This presumably means that the sequence-responsive elements present in the promoter of ComX- and SigH-dependent genes are too distant and that the swapping of one sigma factor by the other one would lead to the incapacity to coordinate the activation of the com regulon. In some species, more than one master regulator coexists and can act synergistically to induce competence. For instance, several strains of S. pneumoniae, Streptococcus gordonii, and Streptococcus pyogenes encode two comX genes (comX1 and comX2; Martin et al. 2006), or even three copies in the case of Streptococcus anginosus (Olson et al. 2013), while Staphylococcus aureus (a non-LAB species) combines a sigma factor (SigH) to one or two transcriptional regulators (ComK1 and ComK2) in order to execute competence activation (Fagerlund et al. 2014). Fairly conserved across clades, an asymmetric cin-box (TTNCGAATAA) occurs in each streptococcal promoter region of essential late com genes (at position -10) and is bound by ComX that contacts the RNA polymerase for a coordinated wave of transactivation (Opdyke et al. 2001, Luo, Morrison 2003, Mignolet et al. 2018). Additionally, the ComX regulon includes accessory com genes that are usually species-specific. For example, the competence activation is usually synchronized with the secretion of antimicrobial molecules, including the ribosomally synthesized bacteriocins (Shanker, Federle 2017). In S. mutans, ComX has a direct role on BlpRH, the two-component system that regulates the bacteriocin production and positively influences bacterial predation (Reck et al. 2015).

Figure 2.

Regulation circuitry of master competence activator. Scheme of the global network regulating the abundance of the master competence activator. Pheromones-based (Quorum sensing) or environmental signals activates membrane (two-component systems relaying phosphate moiety) or cytoplasmic (RRNPP) sensors that in turn influence the master competence gene expression. Nonexclusively, specific global regulators can bind the master competence gene promoter to modulate its expression. The Clp bacterial proteasome controls the steady-state level of the master competence activator through a directed proteolysis. Some factors under the master competence activator control ensure a negative feedback loop that inhibits the competence activation and is responsible for the shut-off mechanism.

Figure 3.

Evolutionary relationship between ComX and SigH. Phylogeny of the master competence activator ComX (green) and SigH (blue) in relevant LAB. A total of five LAB genera are color-assigned: Pediococcus (green), Lactobacillus (purple), Lactococcus (beige), Streptococcus (red), and Leuconostoc-Oenococcus (blue). The evolutionary history was inferred by the Neighbor joining algorithm and the distance assessed with the Kimura protein method. A total of 1000 replicates were performed to build the tree and bootstrap values are indicated for each phylogenetic branching. The tree with the highest log-likelihood is shown. The tree is drawn to scale, with branch lengths measured in the number of substitutions per site (scale bar, 2 substitution/site). SigH from B. subtilis was used as the outgroup to root the tree. The multiple alignments, evolutionary analyses and the tree layout were performed with the CLC Main WorkBench software (http://www.clcbio.com/products/clc-main-workbench/).

To protect cells against undesired persisting competence activation, several molecular shut-off mechanisms have evolved (Fig. 2). At the post-transcriptional level, ComX undergoes a rapid turnover under the control of the nonspecific prokaryote proteasome ClpEP or ClpCP (Biornstad, Havarstein 2011, Fontaine et al. 2015). In LAB, this targeted degradation is promoted by an adaptor protein, MecA, that sequesters ComX and plugs it onto the Clp machinery (Boutry et al. 2012, Tian et al. 2013, Wahl et al. 2014). Although the exact role of MecA on the shut-off phase is still under debate, it definitely ensures a hindrance on the ComX steady-state level. This system acts as a locking device to avoid stochastic competence activation under inappropriate environmental conditions. Superimposed, RNA-seq analyses in Streptococcus salivarius and S. thermophilus upon competence induction revealed that the deletion of comX prolongs the activity of early competence phase regulator (ComR, see hereunder; Haustenne et al. 2015, Mignolet et al. 2018). This suggests that genes under ComX control negatively impact upstream control systems governing ComX. Strikingly in S. pneumoniae, ComX targets the polar accumulation of DprA. This essential member of the transformasome plays a dual role and interferes with the regulation of ComX abundance by promoting the dissociation of ComE dimers, an activator of the early competence phase (see hereunder; Mirouze et al. 2013, Johnston et al. 2020). In S. mutans, the regulator of the early competence phase (ComR, see hereunder) is analogously the target of a negative feedback loop mechanism. Indeed, XrpA (comXRegulatory Protein A) interacts with and inhibits ComR to switch off comX expression. However, the xrpA gene is not under the ComX control, but is encoded inside comX and expressed from an alternative reading frame (Kaspar et al. 2018).

Prediction of LAB transformation potential

Microbiologists face a recurring question in bioengineering: without any preconception on transformation-favorable conditions, what is the successful path to demonstrate that a strain has the potentiality to uptake naked DNA? Impairment of the central regulator or one of the late com genes results in a drastic reduction or even a complete abrogation of DNA acquisition (Peterson et al. 2004, Claverys et al. 2009, Mignolet et al. 2018). Therefore, mutation, indel, transposon-insertion, truncation, or absence of at least one com gene in a genome is definitely a tangible hint that a strain cannot be naturally transformed, especially when the first gene of an operon is inactivated. Conversely, an intact set of essential com genes does not provide an absolute certitude. Several bacteria not known to be transformable possess homologs for comX/sigH and all late com genes. Nevertheless, it is certainly a prerequisite. Thence, a careful biocomputational inspection of the genome is the prime rational step to predict natural transformability.

We performed a systematic study (see Supplementary Appendix 1) of industrial LAB genomes to quantitate the proportion of potential naturally transformable strains. We comprehensively browsed the genome of Generally Recognized As Safe strains in streptococci, lactococci, lactobacilli, pedicococci, oenococci, and Leuconostoc sp. and sought for the presence of intact essential competence proteins. In order to avoid ambiguities and false-negative cases (nonsequenced DNA fragments including com genes), we only considered strains for which a complete and closed sequenced genome was publicly available, i.e. 482 strains in this analysis. In each genus, we defined a reference set of valid competence proteins (Claverys et al. 2009, Bron et al. 2019; see Tables S1–S5, Supporting Information) that was used to examine the other strain genomes inside the same genus, estimated the sequence coverage and homology, and reported pseudogenes, disrupted genes (due to small indel, or insertion of prophage, transposon, or any other mobile genetic element), or absent genes in the com gene set. Considering the minor pilin protein (ComGD/GE/GF and GG) that can exhibit a high sequence variability, we looked at the global architecture of the coding operon (gene number and size) to decide whether they are present or absent.

Across LAB species and strains, we noticed a huge variation in the conservation of the com gene content and the presumed capability to perform natural transformation (Fig. 4A). On one side, more than 92% of the S. thermophilus and S. salivarius strains possess a full set of com genes. This high number correlates with their transformability, as natural transformation was empirically demonstrated for a representative handful of strains (Table 1; Fontaine et al. 2010a). This reflects a high selective pressure to maintain such a capacity. In contrast, no more than 25% of lactococci genomes (mainly plant isolates) contain a complete com set. Some Lactococcus lactis strains have been reported to be transformable in the absence of the essential protein DprA (Mulder et al. 2017). Nevertheless, this low percentage is particularly indicative considering that several strains display a massive com gene decay. With the accumulation of up to eight inactive genes, it is unlikely that strains are still proficient for transformation. Lactobacilli, pediococci and Leuconostoc sp. are in an intermediate situation, ranging from 40% to 65% of strains with a potentially intact transformasome. This is highly encouraging to envision natural transformation as a rapid genome-editing tool, even if species-to-species profiles display a strong heterogeneity, with 95% for Lactobacillus [Latilactobacillus] sakei and 0% for Leuconostoc lactis and Oenococcus oeni. However, none of these strains so far have been shown to perform natural transformation, making it impossible to relate the com gene content to the transformability.

Figure 4.

Conservation of competence genes across safe LAB species. (A) Pie charts highlighting the percentage of strains that encode an intact (green) vs. incomplete (red) set of com genes. Small pie charts quantify it for all the strains of a species (number indicated in brackets), while big pie charts summarize the data collected for the five relevant LAB genera. (B) Mutation score and frequency for each com gene. The mutation score (right panel) is the occurrence frequency of mutation for each gene normalized to the gene size (in kb). The single hit frequency (left panel) is defined as the frequency to have two or more mutated com genes for each strain bearing a mutated com gene. Bars are color-coded in agreement to the gene function in the DNA transformation process shown in Fig. 1B (pseudopilus, red; DNA transport, green; and DNA processing, blue). The color-coded vertical bars depict the mean value for the gene set involved in DNA docking, transport, or processing. The recA and ssbB genes have been dismissed from the analysis because recA is not specific for natural transformation and is required for the vegetative growth, while ssbB has a paralog (ssbA) that could bias the analysis. (C) Theoretical and empirical frequency of multiple inactivating mutations in a strain genome (semi-log scale). The observation curve shows the actual frequency for a strain to gather no, a single, or multiple gene mutations (up to eight genes hit) in the com gene set. The theoretical curve is considered as if the mutation events were independent. The inset provides a zoom on values ranging from zero to three gene disruptions/mutations in the com gene set (linear scale).

Table 1.

Avowed transformable LAB.

		Set of late com	Activation of competencea			Reference
Strain	Source	gene	Spontaneous	ComX++	Pheromone
Streptococcus thermophilus b
LMD-9	Dairy	Complete	Moderate	High	High	Gardan et al. (2009), Fontaine et al. (2010a), Fontaine et al. (2010b)
LMG18311	Dairy	Complete	Low	High	High	Blomqvist et al. (2006), Fontaine et al. (2010a), Fontaine et al. (2010b), Gardan et al. (2009)
CNRZ1066	Dairy	Complete	No	High	High	Gardan et al. (2009), Fontaine et al. (2010a), Fontaine et al. (2010b)
DGCC7710	Dairy	Complete	Low	High	High	Fontaine et al. (2010b)
Streptococcus salivarius b
ATCC25975	Human oral cavity	Complete	-	NT	High	Fontaine et al. (2010a)
JIM8777	Human oral cavity	Complete	-	NT	High	Fontaine et al. (2010a)
JIM8780	Human oral cavity	Complete	Low	NT	High	Fontaine et al. (2010a)
HSISS4	Human GI tract	Complete	-	NT	High	Mignolet et al. (2018)
Streptococcus macedonicus
DSM15879T	Dairy	Unknown	Low	NT	High	Morrison et al. (2013)
139C	Dairy	Unknown	No	NT	High	Morrison et al. (2013)
Streptococcus infantarius b
ATCC BAA-102T	Human GI tract	Complete	-	NT	High	Morrison et al. (2013)
11FA-1	Dairy	Complete	-	NT	High	Morrison et al. (2013)
AB2VB1-2	Dairy	Unknown	-	NT	High	Morrison et al. (2013)
AV2A(1)	Dairy	Unknown	-	NT	High	Morrison et al. (2013)
Lactococcus lactis subsp. lactis
IL1403	Dairy	comX*, dprA*	-	Low	-P	Mulder et al. (2017)
KF147	Plant	Complete	-	High	-P	Mulder et al. (2017)
Lactococcus cremoris subsp. cremoris
KW2	Plant	Complete	-	High	-P	David et al. (2017), Mulder et al. (2017)
Lactococcus raffinolactis
LMG13098	Plant	Unknown	Low	NT	-P	Toussaint (unpublished work)
LMG14164	Bird	Unknown	Low	NT	-P	Toussaint (unpublished work)

^aHigh >10⁻⁴ transformation rate; moderate from 10⁻⁴ to 10⁻⁶ ; low <10⁻⁶; - = not identified; NT = not tested; and -P = no identified pheromone.

^bOnly a subset of strains is listed in this table (mainly close genome-strains).

The exact reason why some strains maintain an intact set of com genes while some other tolerate a fragmentary function remains elusive in most cases. Obviously, the selective pressure imposed by the biotope and the microbial partners represent a key factor. In lactococci from the dairy industry, some strains have been domesticated through a process called back-slopping (iteration of strain re-inoculation in new batches of fresh dairy products from a previous fermented product). Such stable environments favor a reductive evolution, where costly functions are rapidly discarded by gene mutation if they cannot confer a substantial advantage for growth (Bolotin et al. 2004, Kelleher et al. 2017, Makarova et al. 2006). This is typically the case for the flagellar operons when bacterial mobility is dispensable for survival (e.g. rich laboratory media under constant shaking; Dragosits, Mattanovich 2013, Knoppel et al. 2018). Akin to the motorized, polysubunited flagellum, the competence pseudopilus necessitates the ATP-dependent production and polymerization of thousands of major and ancillary pilin subunits, while the transport protein ComEC hydrolyses ATP molecules to transport the DNA inward. Moreover, commitment in the competence phase triggers a global metabolic and physiologic switch that disturbs the cell cycle and slows down the growth. Therefore, we can speculate that wild-type lactococcal strains initially encoded a functional transformasome before they underwent a massive com gene decay to adjust to and thrive in their new industrial niches. By contrast, S. salivarius strains had to cope in fiercely competitive ecological niches (e.g. gut and oral cavity) with tremendous and recurrent variations at the physicochemical and biological levels. Maintaining an adequate ability to be naturally transformed can be seen as a strategy to regularly acquire new genetic traits beneficial to outcompete microbial adversaries, and reinforce interactions with symbiotic bacterial consortia and host epithelial cells.

The domestication conditions are pretty similar for L. lactis and S. thermophilus strains, and S. thermophilus exhibits on average a higher number of pseudogenes, presumably due to the cultivation in a dairy niche (Makarova et al. 2006). However, the output is totally conflicting because S. thermophilus strains apparently face a huge selective pressure to keep an intact com gene set. How to explain this discrepancy? Considering the proportion of pseudogenes, S. thermophilus is prone to accelerated genome decay with a high mutation rate. One reason might be that S. thermophilus (in contrast to L. lactis) misses recQ that codes for a helicase important in genome maintenance and DNA repair from UV, alkylation, and oxidative stresses or mere polymerase errors (Bolotin et al. 2004). According to the Muller ratchet theory, asexual reproduction drives the incremental accumulation of mutations to a critical level. Uptake of DNA from kin origin might be, therefore, an opportunity for strains to correct and rescue from detrimental indels and nucleotide polymorphisms. However, S. thermophilus possesses an immune system, known as CRISPR-Cas, that prevents the invasion of double-stranded DNA from bacteriophage and plasmid origin (Garneau et al. 2010). In a collateral manner, this system drastically reduces the transfer of DNA material from transduction or conjugation that could explain why competence and CRISPR systems coexist and coevolves (Jorth, Whiteley 2012, Gonzalez-Torres et al. 2019). Evolvability lies in a delicate trade-off between too low and too high genome plasticity/instability and replication fidelity. While evolutionary trajectories of L. lactis mainly relied on plasmid conjugation and phage transduction, S. thermophilus used natural transformation to maintain sustainable genome flexibility in presence of stringent foreign DNA acceptor (CRISPR) systems. Possibly, competence development in S. thermophilus might be used as an essential nucleotide-based nutrition support that strengthens the energy metabolism during the yogurt acidification process in multispecies environment (e.g. S. thermophilus and Lactobacillus delbrueckii subsp. bulgaricus).

Interestingly, we observed a mutation bias toward genes and phases in competence-deficient strains. Several genes, such as comX/sigH, comGC, comGD, comC, comEA, or coiA, appear as preferential targets of inactivation (Fig. 4B, right panel). Presumably, these genes would be much less favorable for growth in a context where the transformability would be (partially) impaired. For instance, genes could interfere with physiological processes (such as coiA for DNA replication or maintenance) and generate a certain toxicity that would be counterbalanced by adaptability benefits in competence-proficient strains. Alternatively, some mutations merely economize energy and amino acids building block, such as mutations in highly expressed genes (e.g. pilin genes) or mutations in one of the first genes of an operon that usually affects negatively the expression of downstream genes (e.g. comEA), the extreme case being the loss of the regulatory central node (i.e. comX or sigH) that totally annihilates the inducible production of all competence proteins. Nevertheless, compared to the DNA docking and translocation genes (Fig. 4B, right panel), the DNA recombination genes (through coiA and dprA, mainly) proportionally accumulate more mutations. In addition to the colliding effects, these genes could have an impact on the genome integrity; this could also suggest that the two other phases could retain/acquire satellite roles, related or not to competence. As examples, the pseudopilus could be exploited as an adhesive appendage to stick to substrate or capture molecules (DNA and/or other biological constituents), while the DNA translocation system would ensure a nutrition role via the assimilation of nucleotides and phosphate moieties. Interestingly, once the competence gene decay is initiated, there is a selective pressure to massively inactivate com genes (Fig. 4C). Indeed, the actual frequency of mutations for one or two com genes is lower than expected with a theoretical model, while the occurrence of strains possessing four or more mutations is high above predictions where the mutation events would be fully independent (up to a 10 000-fold increase for an octuple inactivation). In this perspective, some genes like comGB, comGD, comGG, or comFA (but not comX/sigH) are rarely inactivated alone, suggesting that they accelerate the com gene decay or compensate toxicity due to other mutations (Fig. 4B, left panel). This reflects their important and conflicting role on competence and cell fitness.

To conclude, the comparative analyses of genomes represent a first hint that indicates whether a strain might be naturally transformable. Nevertheless, even in the scenario of an intact set of com genes, there is a necessity to validate that they do not harbor loss-of-function single-point mutations, are all fully functional, and can sustain an efficient transformation process.

Governing the master competence regulator

The next step to highlight the transformability of a specific strain requires to demonstrate whether the competence machinery might be coordinately produced under the (inducible) overexpression of an extra copy of the master regulator gene (ComX or SigH; Fig. 5, strategy I). Typically, this supplementary gene is borne onto a (multicopy) plasmid, i.e. transferred into the bacterium via an artificial (chemical or electric shocks) transformation (Blomqvist et al. 2006, David et al. 2017, Fontaine et al. 2010b, Mulder et al. 2017). In parallel, the strain can be engineered to carry a reporter fusion (e.g. luciferase or GFP encoding genes) under the control of a late com promoter (e.g. P_comGA or P_dprA). This strategy first demonstrates whether the alloexpression of the competence master regulator impacts late com gene transcription. Moreover, it can also disclose the timing of competence extinction (early/mid/late exponential or stationary phase) that can diverge from species to species (see hereunder). Beside S. thermophilus (Blomqvist et al. 2006, Fontaine et al. 2010b), three recent examples lie in L. lactis/cremoris strains (Table 1) that sometimes encode a full set of com genes, but were non-naturally transformable in human hands until 2017. In contrast to closely related strains partly deficient for com genes (point mutations or indels), David et al. (2017) showed that comX overexpression in strain KW2 activates late com genes, and they were able to induce natural transformation, providing a streptomycin-resistant rpsL or miscellaneous deletion alleles as donor DNA. In parallel, Mulder et al. (2017) showed that two other L. lactis subsp. lactis strains (KF147 and IL1403) were also transformable with similar plasmid-borne inducible extra copy of comX. Additionally, deep sequencing of the transcriptional response in comX/sigH-overexpressing backgrounds or conditions gives a wider picture of the direct and indirect regulon under the master competence regulator control. This strategy was used in L. sakei, where the sigH-overexpression was reported to enhance transcription of the full set of late com genes (Schmid et al. 2012). Unfortunately, L. sakei was non-naturally transformable in this synthetic condition.

Figure 5.

Competence activation strategies and bioediting outputs. Top panel: scheme depicting the six different levers to unlock competence activation and promote natural transformation. (I) Overexpression of comX/sigH. (II) Inhibition of factors that proteolyse or decrease the abundance of the master competence activator. (III) Natural or laboratory conditions that activate comX/sigH expression. (IV) Overexpression of comX/sigH activators (pheromones or sensors). (V) Exogenous supply of natural or synthetic pheromones that activate the comX/sigH expression. (VI) Synthetic screen to identify peptide-based molecules produced intracellularly and able to stimulate the comX/sigH expression. Bottom panel: potentialities of bioediting via natural transformation. Single genes can be swapped (with point-mutated alleles or nonrelated genes) or deleted. Large genomic reorganization or multiple gene exchange can be performed. Libraries of mutant for specific loci/genes can be easily generated. Competence activation increases the rate of spontaneous mutations (soft mutagenesis) that can be selected for a desired phenotype.

Beside gene overexpression, an alternative way to modulate the ComX/SigH steady-state level is to hijack proteins involved in the shut-off/repression mechanisms or the turnover rate (Fig. 5, strategy II). For instance, alteration of the core and accessory proteolysis machinery (i.e. Clp subunits and MecA, respectively) represents an extra control at the post-transcriptional level to increase the intracellular concentration of the master competence regulator. In S. thermophilus, it was demonstrated that the proteolysis-mediated stabilization of ComX can trigger competence in subpermissive conditions (Biornstad, Havarstein 2011, Boutry et al. 2012). This phenomenon is also partially true in L. cremoris subsp. cremoris KW2 (previously known as L. lactis subsp. cremoris KW2; David et al. 2017).

These strategies are highly informative and essential to accelerate/simplify genomic modifications in naturally transformable strains. Nonetheless, the alloproduction of ComX/SigH that drives a permanent or spread competence state is usually toxic and mutagenic, and could lead to undesired mutations or loci rearrangements all over the genome (David et al. 2017, Mulder et al. 2017). This is likely to generate an undesired genome instability or drift that could alter the nature of the microorganism. Thus, extra investigations are required to unveil the “natural” exosignals (e.g. molecules and physicochemical parameters) that stimulate competence activation.

Discovery of the master regulator guardians

Regulatory circuitries that control competence are intricately fine-tuned and differ between species, while the accurate timing of activation is somehow strain-specific. For instance, several bacterial species naturally turn on competence early during the exponential phase (e.g. S. thermophilus; Fontaine et al. 2013), while other clades pace competence according to the stationary phase (e.g. B. subtilis; Hoch 1993). A first glance at comX/sigH promoter might shed light on regulatory sequences, and therefore, the known transcriptional regulators that bind them. Competence is typically triggered in specific conditions encountered in the natural ecosystem of the bacterium. Usually, they deviate from laboratory environments or are difficult to recreate. The signals could range from basic carbohydrates, such as monosaccharides, to complex (branched) biomolecules and media. S. thermophilus LMD-9 becomes naturally transformable in milk, i.e. the starting substrate used in industries for yogurt and cheese production (Gardan et al. 2013). Concerning Vibrio cholerae, Meibom et al. (2005) discovered that competence is induced by chitin, the main biopolymer forming the exoskeleton of crustacean that coexist with V. cholera in aquatic habitats. One-at-a-time testing of conditions that could arise in the natural biotope of a bug is laborious with uncertain chances of success. Instead, phenotypical microarrays allow to screen thousands of defined media elaborated with a range of pH, or media supplemented with nutrients (amino acids, sugars, nucleic acids, lipids, and vitamins), energy sources (carbon, nitrogen, phosphate, sulfur, and glycolysis/Krebs cycle intermediates), oligo-elements, biomolecules, osmolytes, antibiotics, chemicals, or xenobiotics (Fig. 5, strategy III). Growing reporter strains (e.g. harboring P_comX-luxAB constructs) in these diversified conditions is a powerful way to identify inducing molecules (and thereby, the cognate protein sensor or two-component system) and time-monitor competence activation at a high-throughput level.

In Gram-positive bacteria, competence regulation is intimately linked to the metabolic state and relies on a contribution from overlapping layers of quorum sensing, nutritional signals, and other stress responses (Nester, Stocker 1963, Dubnau 1991, Fontaine et al. 2015). Specifically in streptococci, ComX is under the exclusive direct control of either ComE (in mitis and anginosus groups, e.g. S. pneumoniae and S. anginosus, respectively) or ComR (in mutans, bovis, pyogenes, suis, and salivarius groups, e.g. S. mutans, Streptococcus infantarius, S. pyogenes, Streptococcus suis, and S. salivarius, respectively), two pheromone-responsive systems (Johnston et al. 2014). In both cases, the pheromone is a peptide produced, processed/matured and secreted outside of the cell envelope. However, ComE is a response regulator part of the two-component system that extracellularly detects the presence of the pheromone CSP (Competence-Stimulating Peptide; mature form of ComC) through the direct interaction with and the autophosphorylation of the membrane histidine kinase ComD, and the transfer of the phosphate moiety onto ComE (Pestova et al. 1996, Martin et al. 2013). In contrast, XIP (comX-Inducing Peptide, mature form of ComS), the cognate peptide pheromone of ComR, penetrates into the cytoplasmic space to accommodate the peptide-binding pocket of the cytosolic ComR, a protein of the RRNPP (Rgg-Rap-NprR-PrgX-PlcR) family (Fontaine et al. 2010a, Mashburn-Warren et al. 2010). Subsequent subtle conformational changes liberate the ComR DNA-binding domain and expose key residues involved in the selective recognition of specific nucleotides of the comX promoter (Shanker et al. 2016, Ledesma-Garcia et al. 2020, Talagas et al. 2016). Interestingly, this system was discovered in S. thermophilus while performing a transcriptomic profiling of a comX null-mutant. Following a study that highlighted the role of an unidentified pheromone-sensitive system controlling competence activation (Gardan et al. 2009), Fontaine et al. (2010a) realized that the inactivation of comX in conditions favoring competence (i.e. growth in chemically defined medium) bolstered the expression of the ComR direct regulon, namely the bacteriocin regulatory genes as well as the inducing pheromone gene comS itself (positive feedback loop). This data implies that a gene under ComX control negatively influences comX expression through the modulation of the ComR activation level (retro-control). Analogous feedback circuitries were observed for the close relative S. pneumoniae (see DprA bifunctionality, hereinabove) and S. salivarius (Mirouze et al. 2013, Mignolet et al. 2018). Therefore, in such a network configuration, upstream early com genes might be identified via ComX-associated phenotypes and unbiased high-throughput profiling techniques (e.g. deep sequencing or microarrays). In this respect, pheromone-responsive systems, either two-component or RRNPP proteins, and to a lesser extent transcriptional regulators, can be considered as promising candidates that would influence the steady-state level of master competence regulators. Mutations, bioengineering, or environmental conditions that drive the overproduction of the pheromone or the sensor might, therefore, increase the transformation rate and illuminate the process of competence development for fundamental research (Fig. 5, strategy IV). For instance, CovR is an environmental sensor directly repressing comR expression in the salivarius group of streptococci. In S. thermophilus LMD-9 strain, a natural single substitution of the conserved Asp98 (for a Glu) mimics a CovR unphosphorylated state that partially alleviates comR repression. This translates into higher ComR concentration into the cytoplasm and higher rate of transformation compared to other S. thermophilus or S. salivarius strains (Knoops et al. 2022). The identification of external signals modulating the phosphorylation state or activity of CovR will presumably magnify the transformation capabilities of some refractory strains.

Synthetic strategies to hijack regulatory circuitries

Usually, unveiling the “natural” or lab conditions that allow competence development in field strains is costly and often conducted in the long run. Alternatively, hacking the communication systems governing comX/sigH expression is a tantalizing approach that would facilitate competence induction (at the level of costs and time). This is particularly true for streptococci where both pheromones and sensors are identified, and where pheromones are ribosomally synthesized peptides of relatively small size and with no molecular decorations (Fontaine et al. 2015). Strikingly, the exogenous supplementation of synthetic pheromones enables a saturating competence response at extremely low doses (nanomolar range), independently of the sensor cellular location (outside-membrane interface or cytoplasmic compartment; Mignolet et al. 2016, Ledesma-Garcia et al. 2020, Yang et al. 2020; Fig. 5, strategy V). Since the discovery of the ComRS system in S. thermophilus, this pheromone-based strategy appears to be highly efficient. It was used several times in the past to expand competence activation to numerous strains of beneficial streptococci belonging to species S. thermophilus, S. salivarius, S. infantarius, and Streptococcus macedonicus (Table 1; Fontaine et al. 2010a, b, Morrison et al. 2013), as well as pathogenic streptococci including S. mutans (Mashburn-Warren et al. 2010), S. sobrinus (Li et al. 2021), S. suis (Zaccaria et al. 2014), and S. pyogenes (Mashburn-Warren et al. 2012, Marks et al. 2014). The current technologies in chemistry allow a fast and somewhat cheap production of conventional peptides (C- to N-terminus synthesis on solid surface). However, rapid advances in this field will definitively give a similar access in the future for “exotic” peptidomimetics that encompass L stereo-isoform, unnatural, decorated, or bridged amino acids, and even circular and endocyclic peptides (Bedard, Biron 2018). We specifically think about the competence pheromones of Bacillus species, a hexapeptide comprising a modified tryptophan residue (Okada et al. 2005), that might also be chemically synthesized to activate competence when desired. However, some microorganisms are insensitive to their own pheromone in inappropriate conditions. A remarkable example is the S. pyogenes case that requires both ComS and an organized biofilm on epithelial cells to trigger competence, showing how important is the natural environment to activate the competence regulatory cascade (Mashburn-Warren et al. 2012, Marks et al. 2014). Alternatively, several screens and rational design works have been successfully performed to identify linear and cyclic molecules of nonpeptide origin able to modulate the competence pathway in S. pneumoniae (Yang et al. 2020, Milly et al. 2021).

Interestingly, peptide-based signaling exhibits a 2-fold benefit. First, an array of variants can be designed and screened for their capacity to improve peptide stability, protease-resistance, diffusion coefficient, affinity/activability toward their cognate sensor, and ultimately the modulation of the transformation rate. Complementarily, random mutations in the pheromone coding gene to seek for optimized properties might generate a higher diversity of sequences compatible with high-throughput genetic screens (Fig. 5, strategy VI). Moreover, the rise of 3D structures available for regulators, in an apoform or in co-crystal with their cognate pheromone, illuminates the recognition at the atom scale between peptides and regulators (Talagas et al. 2016). It facilitates the rational design of more potent pheromones or pheromones with a revamped selectivity toward noncognate sensors, as was recently performed for streptococci of the salivarius group (Ledesma-Garcia et al. 2020). In a second scope, randomization-based genetic screens can elegantly shine a light on a broad range of peptide sequences, distant from the natural one, and capable of activating signaling regulators. Concerning competence pheromones that impose a high fitness cost and hinder the cell cycle, such strategies can reveal sequences that would be counter-selected in nature because they amplify/prolong the competence state. From another angle, many sensors are only assumed to respond to peptides because their natural peptide has not yet been disclosed. In a recent study, researchers generated a library of short genes for which the last seven codons were randomized, and they managed to unveil a set of peptides that efficiently activate ComR-like regulators unrelated to competence and for which the cognate pheromone stays elusive (Mignolet et al. 2019). Sequence comparison extracted common properties (i.e. amino acid positioning) and may give some hint to fish the natural molecular cue. This synthetic pheromone approach is particularly convenient to stimulate a system on demand, even in absence of known signals (Fig. 5, strategy VI). Albeit designed for cytoplasmic sensors, it could be presumably amenable to extracellular sensing (e.g. two-component system).

Competence benefits for LAB optimization

Nowadays, competence for natural transformation is routinely exploited to genetically modify streptococcal strains in the scope of fundamental research. This includes the engineering of deletion, overexpression, or single-point mutants, the design of genetic libraries for synthetic biology screens, the optimization of cell fitness, and the clustering of interesting traits into a unique microbial chassis. The two main handling advantages of competence over other natural transfer mechanisms such as transduction and conjugation or genome-editing tools such as CRISPR-Cas and λ-red-like recombineering are definitely that it does not require steps of cloning, and that the transferable information can be provided as a mere linear fragment of double-stranded DNA (even though entire circular plasmids can be acquired via natural transformation). The DNA template can either directly originate from genomic extraction or alternatively be native or recombinant genes amplified by PCR, or chemically synthesized in DNA blocks. The only requirement for chromosomal and extrachromosomal insertions is two DNA homology regions (ideally more than 500 bp) that flank the fragment to be inserted, one on each side. In the context of directed evolution and positive selection, competence is particularly efficient and trendy for DNA shuffling. From a linear DNA library of (active) mutated alleles (or functional homologous genes in the case of “family shuffling”), this technique uses the iteration of PCR-random fragmentation cycles to generate novel combinations (hybrids or swaps) and increase the library diversity. Another substantial added value of competence relies on the fact that multiple pieces of DNA can be transferred simultaneously and drive a global genomic rearrangement (Dalia et al. 2017). Therefore, the natural transformation technology is a strain-inherent property that tremendously accelerates and potentiates the genome reshaping, and can produce mutants in a single day.

The transformation rate is definitively strain-specific and relies on intrinsic signaling and regulatory networks. Puzzlingly, it strongly vary across strains of the same species, or even from time to time in a given strain. It can range from 50% in high competence performers such as S. pneumoniae R6 (Johnston et al. 2020), Streptococcus mutans UA159 (Lemme et al. 2011), or Streptococcus thermophilus LMD-9 to very low value (10⁻⁴) in most cases. To select gene acquisition events in poorly transformable strains, a resistance cassette can be linked to the recombinant DNA, and the cognate antibiotic is supplemented in the growth medium to neutralize unmodified cells. On the top of it, counter-selection markers (e.g. oroP, pheS, or sacBR) or site-specific recombinases (Cre) can be combined to introduce or remove alleles in a scarless, or at least a marker-free, output (Fontaine et al. 2010b, Dorrazehi et al. 2020, Mignolet et al. 2018). These methods provide extraordinary opportunities in fundamental research to shed light on bacterial physiological processes. Once robust and convenient stimulating conditions have been developed, natural transformation is the most efficient strategy to select the acquisition of novel phenotypical traits. Particularly convenient for high-throughput screens, hypercompetent LAB can accommodate genetic libraries of a large diversity (up to billion clones) and could be used in the context of bacterial surface display or antigen presentation (Hu et al. 2010, Dorrazehi et al. 2020, Lecomte et al. 2014).

In the context of natural DNA transfers without the use of an antibiotic marker, the major challenges are to use highly transformable strains (Blomqvist et al. 2010) or to associate a desired trait with a selectable (growth improvement, color- or texture-marked) phenotype, and to dispose of a natural donor DNA that exhibits a sufficient homology to sustain recombination in the recipient cells. This strategy, reminiscent of self-cloning, was applied to two examples in S. thermophilus, the prtS locus, and the histidine prototrophy. PrtS is a cell wall-anchored enzyme that fulfills a nutritional role by proteolyzing the milk casein in short importable oligopeptides. Acquisition of the prtS genomic island in prtS-deficient strains speeds up the cellular growth and subsequently the milk acidification process, leading to a reduction in the industrial process costs (Dandoy et al. 2011). The conversion of slow-acidifying strains into fast-acidifying derivatives provides a genuine selective advantage easily visualized on differential agar plate (big colonies surrounded by a yellow halo of acidification). The transfer of prototrophic traits (his⁺) into auxotrophic strains (his⁻) is even more straightforward. Genomic DNA from the prototrophic donor has to be incubated in competence-inducing conditions with the recipient strains disabled for a biosynthetic pathway. Cells are next plated on selective media depleted for the specific essential amino acids and tested for growth (Fontaine et al. 2010b).

As an additional benefit, competence accelerates the evolutionary clock of microorganisms even in the absence of foreign DNA (Grist, Butler 1983, Martin et al. 2020). Indeed, competence induction increases the mutation rate, presumably because proteins involved in the recombination process could display an intrinsic replicative error-prone molecular behavior. In comparison with harsh physical (e.g. UVs) or chemical (e.g. ethyl methanesulfonate) mutagenesis that literally riddle the genome with multiple replicative errors, competence activation could, therefore, be exploited as a soft treatment to promote the emergence of spontaneous mutants.

Concluding remarks

The LAB world is a wide reservoir in which we can tap into to elaborate innovative solutions and face current and future major issues (e.g. soil desertification, multidrug resistance, and food/energy availability; Lamont et al. 2017, Borner et al. 2019, Bron, Kleerebezem 2018, Mathur et al. 2020). The synthetic biology and biotechnologies even expand the possibility landscape because they facilitate (temporally and financially) the fine-tuning of microfactories, either through rational design or randomization-based and evolution strategies. So far, competence for natural transformation is definitely the most convenient tool to expand our knowledge about LAB because it permits to convert bacterial genomes, from single-point mutation to large chromosomic rearrangements. A deep understanding of molecular behaviors that govern the whole process in every transformable microorganism is, thence, the keystone to concretize the full potential of LAB for human need.

The conclusive demonstration of strain transformability is usually a long and demanding path. Even for S. thermophilus and S. salivarius, two species nowadays routinely transformed on lab benches with incredibly simple protocols, more than 10 years of intensive works have been carried out by several research units. With a stepwise strategy based on in silico analyses, genetics, synthetic biology, and phenotypic characterization, future researches will be able to rationally minimize the effort to reach this aim. The use of overexpressing/deletion strains or artificially transformed (i.e. via enzymes, chemicals, or electroshocks) mutants are excellent premises to finely unravel the mechanistic of competence activation. Nevertheless, the identification of the natural signal (molecule or stress) that dictates the transformability is the final objective to fully unleash the potential of natural transformation.

In the future, the use of competence as a natural genome diversification method in bacteria, analogously to plant or animal breeding, will benefit from the microbial diversity, i.e. present in the giant cryostock libraries established by a plethora of academic, medical, and industrial organizations over the world (Ryan et al. 2021) Therefore, there is a flagrant necessity to pursue our effort and widely collect and sequence as possible the highest diversity of microorganisms. They will be the solid basis to combine exciting properties in second generation of improved LAB strains.

Supplementary data

Supplementary Table 1. Conservation of competence genes across the Streptococcus genus.

Supplementary Table 2. Conservation of competence genes across the Lactobacillus genus.

Supplementary Table 3. Conservation of competence genes across the Lactococcus genus.

Supplementary Table 4. Conservation of competence genes across the Leuconostoc and Oenococcus genera.

Supplementary Table 5. Conservation of competence genes across the Pediococcus genus.

Supplementary Appendix. Methods and analysis strategies used to identify the competence gene sets in beneficial LAB genomes, to estimate the mutation frequency in competence genes, and to generate the ComX/SigH phylogenetic tree.

Funding

This work was supported by the Belgian National Fund for Scientific Research (FNRS; PDR T.0110.18 to Pa.H.), the Federation Wallonia-Brussels (ARC 17/22-084 to Pa.H.), and the Belgian Science Policy Office (BELSPO; IAP P17/28 to Pa.H.). J.M. holds a H2020-MSCA-IF-2020 fellowship (grant number 101018461). Pa.H. is Research Director at the FNRS.

Conflict of interest statement. F.T., C.F., Ph.H., and Pa.H., declare that they are listed as inventors on patent(s) or patent application(s) related to transforming bacteria through natural competence.