Cyclic-di-AMP signalling in lactic acid bacteria

Abstract Cyclic dimeric adenosine monophosphate (cyclic-di-AMP) is a nucleotide second messenger present in Gram-positive bacteria, Gram-negative bacteria and some Archaea. The intracellular concentration of cyclic-di-AMP is adjusted in response to environmental and cellular cues, primarily through the activities of synthesis and degradation enzymes. It performs its role by binding to protein and riboswitch receptors, many of which contribute to osmoregulation. Imbalances in cyclic-di-AMP can lead to pleiotropic phenotypes, affecting aspects such as growth, biofilm formation, virulence, and resistance to osmotic, acid, and antibiotic stressors. This review focuses on cyclic-di-AMP signalling in lactic acid bacteria (LAB) incorporating recent experimental discoveries and presenting a genomic analysis of signalling components from a variety of LAB, including those found in food, and commensal, probiotic, and pathogenic species. All LAB possess enzymes for the synthesis and degradation of cyclic-di-AMP, but are highly variable with regards to the receptors they possess. Studies in Lactococcus and Streptococcus have revealed a conserved function for cyclic-di-AMP in inhibiting the transport of potassium and glycine betaine, either through direct binding to transporters or to a transcriptional regulator. Structural analysis of several cyclic-di-AMP receptors from LAB has also provided insights into how this nucleotide exerts its influence.


Introduction
Lactic acid bacteria (LAB) include a diverse range of bacteria commonly found in the environment on vegetal and animal sources and in r aw/fr esh foods. Gener all y r egarded as safe LAB play important roles in food fermentations assisting with the pr eserv ation and transformation of milk, vegetable, meat, legume, and cer eal substr ates . T he desir e to better contr ol fermentations has driven the selection of starter and adjunct cultures with optimal acidification, flavour de v elopment, and textur e alter ation. Certain LAB are also major contributors to food spoilage, particularly under stor a ge conditions with r educed oxygen suc h as modified atmospher e or v acuum pac ka ging. LAB form an important component of the commensal microbial population on various mucosal surfaces of humans and animals, including the oral cavity, gastr ointestinal, and ur ogenital tr acts. In ad dition, a n umber of LAB species are major pathogens causing significant morbidity and mortality.
LAB ar e commonl y exposed to external c hallenges in their ecological niches, during food processing, following consumption or during infection, which may limit their growth or survival. T his ma y include acid stress during food fermentation or transit through the stomach, osmotic stress following salt addition in c heese-making, heat str ess during spr ay drying, nutrient str ess encountered to w ar ds the end of a food fermentation, or immune system attack during infection (van de Guchte et al. 2002 , Tsakalidou andP a padimitriou 2011 ). Like all bacteria, LAB hav e systems that allow them to sense external stimuli and adapt as best as possible . T hese can result in changes at the transcriptional (RNA le v el) and/or post-transcriptional (protein level or activity) levels, whic h r esult in downstr eam physiological or structur al c hanges in the cell.
Second messenger signalling systems in bacteria involve sensing of an external (first) signal, which causes changes in the activity of a synthesis or degradation enzyme that modulates the concentration of an intracellular (second) signal molecule (Yoon and Waters 2021 ). The best c har acterized second messengers include nucleotides cyclic adenosine monophosphate (cAMP) and guanine tetra-or penta-phosphate [(p)ppGpp] that are involved in carbon metabolism and the stringent r esponse, r espectiv el y. Mor e r ecentl y, cyclic dinucleotide monophosphate second messengers have been characterized.
These include cyclic dimeric guanosine monophosphate (c-di-GMP), cyclic-di-AMP (c-di-AMP), and cyclic GMP-AMP (cGAMP), and v ery r ecentl y cyclic trinucleotide monophosphate signalling molecules have also been identified (Yoon and Waters 2021 ). C-di-GMP plays an important role in the motile/sessile lifestyle transition, cell shape, and virulence (Jenal et al. 2017 ). Second messengers cGAMP and other cyclic mono-/di-/tri-nucleotides trigger altruistic cell suicide following bacteriophage infection, thereby pr e v enting vir al r eplication (Duncan-Lo w ey and Kr anzusc h 2022 ). LAB produce c-di-AMP, but have not been found to synthesize other cyclic di-or tri-nucleotides. C-di-AMP was first discov er ed in 2008 (Witte et al. 2008 ) and since then there has been a significant amount of new insight into how it exerts its control in Figure 1. Ov ervie w of the c-di-AMP signalling system. The c-di-AMP synthesis and degradation enzymes respond to external or cellular stim uli r esulting in c hanges in c-di-AMP le v els. Upon r eac hing a certain intr acellular concentr ation, c-di-AMP binds to r eceptors leading to activation or inhibition of activity and a subsequent cellular output and adaptation.
Gr am-positiv e bacteria, Gr am-negativ e bacteria and Archaea. This includes how input signals feed into the system, the receptors that c-di-AMP binds and the resultant cellular outputs/phenotypes (Fig. 1 ). This r e vie w will focus on recent learnings of c-di-AMP signalling systems specifically from work done in LAB species that are of importance in food applications, as colonizers of mucosal surfaces or causes of disease. It will include a bioinformatic analysis of c-di-AMP signalling components in representative LAB genomes, the physiological changes that occur upon c-di-AMP concentration shifts and stimuli that trigger c-di-AMP le v el c hanges. For a br oader understanding of c-di-AMP signalling systems outside of LAB, the reader is referred to several recent revie ws (Commic hau et al. 2019, He et al. 2020, Stülke and Krüger 2020, Zarrella and Bai 2020 ).

C-di-AMP synthesis by the conditionally essential CdaA
The intr acellular le v el of c-di-AMP is regulated by enzymes involved in its synthesis and degradation as well as active export from the cell. LAB, like most Firmicutes , have only one diadenylate cyclase (DAC) enzyme (IPR034701) called CdaA (or DacA) that synthesizes one c-di-AMP from two molecules of ATP (Figs. 2 and 3 ). This contrasts to the more complicated c-di-GMP signalling system where bacteria can contain dozens of synthesis and degradation enzymes (Sondermann et al. 2012 ). CdaA contains an Nterminal domain consisting of thr ee pr edicted tr ansmembr ane spanning regions and a C-terminal cytoplasmic enzymatic domain with Asp-Gly-Ala (DGA) and Arg-His-Arg (RHR) motifs in the activ e site. Structur al anal ysis suggests that CdaA dimers likely need to form oligomeric complexes in order to form a catal yticall y activ e enzyme and pr e v ention of oligomer formation is one way to inhibit c-di-AMP synthesis (Pathania et al. 2021 ). Le v els of intracellular c-di-AMP in Bacillus subtilis have been estimated to be ∼2 μM (Oppenheimer-Shaanan et al. 2011 ).
The first demonstration of c-di-AMP production in a LAB was reported in 2011 (Kamegaya et al. 2011 ). C-di-AMP was identified in cell extracts of Streptococcus pyogenes and reaction products from purified S. pyogenes CdaA using high-pressure liquid chromatogr a phy fr actionation and matrix-assisted laser desor ption ionization-time of flight mass spectrum analysis . T his work also reported that attempts to inactivate the cdaA gene were unsuccessful. Se v er al other studies have since reported the essentiality of cdaA in LAB (Song et al. 2005, Bai et al. 2013, De v aux et al. 2018, Kundra et al. 2021 ) and other bacteria (Whiteley et al. 2015, Gundlach et al. 2017, Zeden et al. 2018 ) under normal laboratory growth conditions, suggesting that cells devoid of c-di-AMP are unable to survive or multiply.
Like in other Firmicutes, the c-di-AMP synthesis gene cdaA is encoded in a highl y conserv ed thr ee-gene oper on in LAB . Downstream of cdaA is cdaR , which encodes a protein of unknown function that contains a tr ansmembr ane domain (TMD) and three or four YbbR domains (IPR012505) located on the external side of the cytoplasmic membr ane. CdaR dir ectl y inter acts with CdaA through interactions between their TMDs (Rismondo et al. 2016, Gibhardt et al. 2020Fig. 2 ). Inter estingl y, in some str eptococcal strains, the CdaA and CdaR proteins are fused together (e.g. A5LQF8, E1M7C3, and V8IIN0). AlphaFold predictions of these large fusion proteins in UniProt indicate that the first TMD of CdaA may be the interacting domain with the TMD of CdaR. The function of CdaR is not known at present, but it appears to play a role in regulating c-di-AMP synthesis by CdaA.
GlmM is an essential phosphoglucosamine mutase that interconverts glucosamine-6-phosphate and glucosamine-1phosphate, forming an early step in the synthesis of cell-wall peptidoglycan and other cell-wall polymers (Barreteau et al. 2008 ). In Lactococcus lactis , GlmM binds dir ectl y to CdaA and inhibits c-di-AMP synthesis activity (Zhu et al. 2016 ;Fig. 2 ). This inhibitory activity of GlmM is shar ed br oadl y in Firmicutes and r ecent structur al anal ysis has indicated that GlmM binding pr e v ents formation of catal yticall y activ e CdaA oligomers (Tosi et al. 2019, Gibhardt et al. 2020, Pathania et al. 2021. In suppressor m utant scr eening studies, a single amino acid c hange (I154F) in L. lactis GlmM resulted in enhanced binding to CdaA and greater inhibition of activity (Zhu et al. 2016 ). Structur al anal ysis of CdaA-GlmM complexes in other Firmicutes identified this residue is in close proximity to the contact site between the two proteins (Tosi et al. 2019, Pathania et al. 2021. The functional significance of the CdaA-GlmM interaction, ho w ever, remains unclear, but does suggest that there is likely coordination between c-di-AMP and peptidogl ycan pr ecursor synthesis . T his hypothesis is supported by the finding of ele v ated intr acellular le v els of the peptidogl ycan pr ecursor UDP-N -acetylglucosamine in L. lactis mutants with high c-di-AMP (Zhu et al. 2016, Pham et al. 2021 ).

C-di-AMP hydrolysis
Degradation of c-di-AMP to phosphoadenylyl adenosine (5 -pApA) and to AMP occurs through the actions of phosphodiesterases (PDEs). In LAB, the most common PDEs are the G GDEF d omain  (Jumper et al. 2021, Bansal et al. 2022, Varadi et al. 2022. Colour coding in protein structures indicates the confidence of the prediction by AlphaFold (dark blue = very high confidence; light blue = confident; y ello w = low confidence; orange = very low confidence). Note that only monomers are shown for simplicity. p rotein containing p hosphodiesterase (GdpP, formerly YybT; IPR014528) and DHH -DHHA1 domain p rotein (DhhP; IPR038763) PDEs . T hese both contain a DHH domain, that contains highly conserv ed activ e site Asp-His-His r esidues, and a DHH-associated (DHHA1) domain at the C-terminus (Rao et al. 2010, Corrigan et al. 2011, Ye et al. 2014, Huynh and Woodw ar d 2016. GdpP is tethered to the membrane via two N-terminal TMDs and contains two potential regulatory domains in the central part of the protein. These include a heme-binding Per-ARNT-Sim (PAS) domain and a region structur all y similar to GGDEF domains, which has weak ATPase activity (Rao et al. 2011, Tan et al. 2013Fig. 2 ). DhhP contains only a DHH-DHHA1 domain and lacks any obvious transmembrane or regulatory domains. GdpP is a highly selective cyclic-dinucleotide hydrolase with high affinity for c-di-AMP (Rao et al. 2010 , Huynh andWoodw ar d 2016 ). DhhP, on the other hand, has promiscuous substrate specificity cleaving both c-di-AMP or 5 -pApA into AMP in vitro (Huynh and Woodw ar d 2016 ). In some cases, such as for Streptococcus pneumoniae DhhP, it can also cleave c-di-AMP directly to AMP in vitro , but with 1000-fold lo w er activity compared with its conversion of pApA to AMP (Bai et al. 2013 ). In se v er al LAB, including S. pneumoniae , Streptococcus mitis , and Enterococcus faecalis , inactivation of both gdpP and dhhP genes result in significantly gr eater c-di-AMP le v els compar ed to m utants with onl y one of these genes inactivated (Bai et al. 2013, Rørvik et al. 2020, Kundr a et al. 2021. Inter estingl y, under biofilm-like gr owth conditions with 5% CO 2 , a S. pneumoniae dhhP mutant, but not a gdpP mutant, contained a very high c-di-AMP le v el (Wooten et al. 2020 ). In other LAB, ho w e v er, inactiv ation of just gdpP results in very high (20-30-fold) c-di-AMP le v els compar ed with the wild-type (Pham et al. 2018(Pham et al. , 2021. It is possible that in some bacteria, pApA may act as a feedback inhibitor of PDE activity, which results in elevated cdi-AMP upon loss of DhhP activity. Stand-alone DHH-DHHA1 homologs are commonly present in organisms that do not produce c-di-AMP, so their role specifically in c-di-AMP hydrolysis in vivo r equir es further investigation (Huynh and Woodw ar d 2016 ).
A few LAB genera, including Enterocococus , Carnobacterium , Tetragenococcus , and Vagococcus contain an additional c-di-AMP PDE called PgpH (PF07698; Fig. 3 ). This protein consists of a Cterminal intracellular HD domain (IPR006674) with a His-Asp motif in the active site that degrades c-di-AMP (Huynh et al. 2015 ). PgpH contains se v en tr ansmembr ane spanning r egions (IPR011621) in the central part and a lar ge extr acellular Nterminal domain (IPR011624) of unknown function. There have been no experimental studies of PgpH homologs in LAB to date.
Like for DACs, there are several regulators of PDE enzyme activity that allow the cell to modulate c-di-AMP le v els in response to intracellular and extracellular stimuli. The stringent response r egulator (p)ppGpp str ongl y inhibits c-di-AMP hydr ol ysis by GdpP and PgpH in se v er al Firmicutes (Rao et al. 2010, Huynh et al. 2015. Limited biochemical studies on LAB GdpP regulation have shown that (p)ppGpp is a very weak inhibitor of GdpP from E. faecalis in vitro (Wang et al. 2017 ). GdpP contains a PAS domain, which has been shown to bind heme (Rao et al. 2011, Tan et al. 2013, and this widespread sensory domain has been found to sense redox potential, light, metal ions, and oxygen in other proteins (Stuffle et al. 2021 ). The PDE activity of the DHH-DHHA1 domain only from . Presence and analysis of c-di-AMP signalling components in re presentati ve LAB. Re presentati ve taxa genomes of LAB were selected from the NCBI database. A phylogenetic tree was constructed using FastME 2.1.6.1 (Lefort et al. 2015 ), which was based on the Genome BLAST Distance Phylogeny (GBDP) distances . T hese distances were calculated from 16S rRNA gene sequences utilizing the TYGS platform (Meier-Kolthoff and Goker 2019 ). T he iT OL softw are (Letunic and Bork 2021 ) w as used to visualize the phylogenetic tr ee and r eceptors. For eac h str ain, w e sear ched for homologs of c-di-AMP synthesis and degradation enzymes, and their receptors using the custom BLAST function in Geneious Prime software version 2022.1.1 (Biomatters Ltd, New Zealand). This search was compared with published c-di-AMP receptors that originated from either LAB or other c-di-AMP-producing bacteria. Specific c-di-AMP binding domains are denoted as follows: RCK_C domain (R), CBS domain (C), and USP domain (U). E. faecalis GdpP was around 13-fold higher than the full-length protein suggesting a role for the PAS and/or GGDEF domains in dampening c-di-AMP hydr ol ysis (Wang et al. 2017 ). Ho w e v er, truncation of the PAS domain in the S. pneumoniae GdpP did not affect PDE activity (Bai et al. 2013 ). The degenerate GGDEF domain from E. faecalis GdpP has been found to exhibit weak ATPase activity (Wang et al. 2017 ), ho w e v er, if and how this affects c-di-AMP hydr ol ysis is not clear. In a screen for heat-resistant mutants in L. lactis , an A285D mutation was identified in gdpP , which is located in the GGDEF domain, suggesting this part of the protein can influence c-di-AMP degrading activity of the DHH-DHHA1 domain (Smith et al. 2012 ). The degenerate GGDEF domain shares structural similarity with diguanylate cyclase enzymes, including DgcR and WspR (De et al. 2009, Teixeira et al. 2021, providing further support for a nucleotide-binding function. Further work is r equir ed to determine what roles the PAS and degenerate GGDEF domains in GdpP and the extracellular domain in PgpH play in sensing signals and adjusting PDE activity.

C-di-AMP export
In addition to enzymatic degradation of c-di-AMP by PDEs, the intr acellular le v el of this signalling molecule can also be lower ed physicall y, by activ e export fr om the cell. Multidrug r esistance (MDRs) transporters of the major facilitator superfamily (IPR011701), whic h ar e widespr ead in LAB hav e been shown to secrete c-di-AMP (Woodw ar d et al. 2010 , Huynh andWoodw ar d 2016 ). Work in Listeria and Bacillus has r e v ealed that mutants with incr eased expr ession of MDRs or inactiv ated MDR genes secr eted higher or lo w er le v els of c-di-AMP, r espectiv el y, as determined by dir ect measur ement or indir ectl y by measuring beta interfer on (IFN-β) induction in infected macr opha ges (Crimmins et al. 2008, Woodw ar d et al. 2010, Schwartz et al. 2012, Yamamoto et al. 2012, Ka plan Zee vi et al. 2013, Townsley et al. 2018 ). In L. lactis , mutations that increased transcription of an EmrB-like MDR gene ( llmg_1210 ) by se v er al hundr ed-fold r esulted in r educed intr acellular c-di-AMP le v els and pr oportionall y higher extr acellular c-di-AMP le v els (Pham et al. 2018 ). This lo w ering of the intracellular cdi-AMP le v el was significant enough to bring about a phenotypic change in salt resistance in L. lactis . It is, ho w ever, currently not known if more subtle (non-mutational) changes in MDR expression or activity can occur in bacteria in response to environmental c hanges, whic h lead to c-di-AMP le v el fluctuations that are significant enough to alter either bacterial physiology or host immune responses.

Phenotypes of LAB mutants with low or high c-di-AMP le v els
A plethora of phenotypic changes in LAB mutants with altered cdi-AMP le v els hav e been r eported (Supplementary Table S1). Mutants with high c-di-AMP have been obtained by inactivation of the gdpP alone or both gdpP and dhhP . Despite its essentiality in some bacteria under normal growth conditions, targeted inactivation of cdaA in wild-type LAB bac kgr ounds or by selection of suppr essor m utants fr om high c-di-AMP m utant str ain bac kgr ounds have allo w ed for the study of cells either devoid of c-di-AMP, or partiall y defectiv e in its synthesis.
Some contrasting phenotypic results for mutants with high or low/no c-di-AMP have been reported in different LAB. This is perhaps due in part to methodological differences, such as growth versus killing assays or different media (osmolarity or solute com-position) being used. It should also be mentioned here that mutants with very high or low c-di-AMP can be unstable and can easil y accum ulate suppr essor m utations (Gundlac h et al. 2015, Whiteley et al. 2015, Whiteley et al. 2017. It is not known what the 'normal' range of c-di-AMP concentration is within cells whereby suppr essor m utations will not occur, but this likel y fluctuates depending upon the growth conditions and specific bacterial species. In our lab, prolonged incubation of a high c-di-AMP gdpP mutant of L. lactis under normal (non-stressed) growth conditions during plasmid transformation and excision experiments commonl y r esulted in the accumulation of suppressor mutations (Choi et al. 2017 ). We ther efor e advise caution when working with mutants with alter ed c-di-AMP le v els . T his can be done by c hec king the cdi-AMP concentration, monitoring phenotypes, and carrying out whole genome sequencing.

Growth and osmoregulation
A commonl y observ ed phenotypic c hange for LAB m utants with no/low c-di-AMP and also high c-di-AMP is growth impairment (Supplementary Table S1). This indicates that a tuned c-di-AMP le v el is necessary for optimal cell multiplication under specific en vironmental conditions . As suc h, c-di-AMP has been r eferr ed to as an 'essential poison' (Gundlach et al. 2015 ). Studies in Streptococcus agalactiae and E. faecalis identified that cdaA , and therefore c-di-AMP, becomes dispensable for growth during anaerobic culturing or on media devoid of osmoprotectants, namely, glycine betaine or carnitine (De v aux et al. 2018, Kundra et al. 2021. In a gr eement with the former condition, inactivation of cdaA was successful in se v er al Streptococcus species ( mutans , sanguinis , mitis , and pyogenes ) under anaerobic or microaerophilic culturing conditions (Xu et al. 2011, Cheng et al. 2016, Fahmi et al. 2019, Rørvik et al. 2020. A thorough understanding of why O 2 is toxic to cdaA mutants of LAB is not yet a ppar ent, but it has also been reported in other Firmicutes, including Staphylococcus aureus (Zeden et al. 2018 ). The toxicity of osmoprotectants to cdaA mutants appears to be due to their uncontrolled accumulation in the absence of cdi-AMP contr ol, whic h triggers osmotic instability. In se v er al LAB, gr owth of cdaA m utants in ric h gr owth media can be impr ov ed by the addition of ele v ated salt concentr ations (Zhu et al. 2016, Kundr a et al. 2021. Conv ersel y, ele v ated salt addition to gr owth media results in poor growth of high c-di-AMP LAB mutants (Smith et al. 2012, De v aux et al. 2018, Zarr ella et al. 2018, Teh et al. 2019. The role of c-di-AMP in osmoregulation has been well established in numer ous bacteria, whic h can be explained by a number of c-di-AMP receptors being involved in potassium and compatible solute transport. A likely consequence of dysregulated osmotic pressure is changes in phenotypes affected by turgor pressure such as cell lysis , cell size , and cell c hain length. CdaA m utants in L. lactis and E. faecalis have been found to be more prone to cell lysis during normal growth (Kundra et al. 2021, Pham et al. 2021. In se v er al LAB, the size of cells was negativ el y influenced by c-di-AMP, with high c-di-AMP mutants having smaller cells (Teh et al. 2019 ) and low/no c-di-AMP cells hav e enlar ged cells (Kundr a et al. 2021 ). This a gr ees with l ysis and cell size c hanges observ ed in other Firmicutes (Corrigan et al. 2011, Gundlach et al. 2017, Zeden et al. 2018, including the striking cell elongation differences r ecentl y r eported for DAC and PDE mutants of Clostridioides difficile (Oberkampf et al. 2022 ). Abnormal changes in cell size may be connected with growth defects of mutants with altered c-di-AMP le v els.

Biofilm forma tion, coloniza tion, and virulence
A number of studies have identified effects of the le v el of c-di-AMP in LAB and phenotypes associated with pathogenicity, including biofilm formation, colonization, and virulence in various model systems (Supplementary Table S1). In se v er al studies, both high or low/no c-di-AMP le v els hav e been shown to r esult in phenotypes that would be predicted to reduce pathogenicity of LAB. Despite this , there is , ho w ever, a lack of understanding of the mechanisms by which c-di-AMP impacts colonization and pathogenicity. The role of c-di-AMP in biofilm formation of LAB is variable between LAB and e v en between studies of the same species (Zarrella and Bai 2020 ). The mechanistic basis for how c-di-AMP affects biofilm formation in S. mutans has been investigated the most. A high cdi-AMP mutant of S. mutans forms greater biofilm biomass and pr oduces mor e exopol ysacc haride (EPS) by ele v ating expr ession of the glucosyltr ansfer ase GtfB (P eng et al. 2016 ). It w as found that the TrkA_C domain containing c-di-AMP binding protein CabPA (a TrkA homolog) binds to the response regulator VicR, which regulates gtfB transcription (Peng et al. 2016 ). Biofilm work in another study in S. mutans reported conflicting results, with a cdaA mutant found to produce greater biofilm biomass, EPS production and GtfB expression (Cheng et al. 2016 ). Furthermore, a study in S. mutans found no difference in biofilm formation of a gdpP mutant (Konno et al. 2018 ). It has been suggested that strain differences and/or culture conditions may account for these variable results (Zarrella and Bai 2020 ). In fermented foods, EPS can enhance the viscosity and mouthfeel, ther efor e it would be of interest to determine if c-di-AMP exerts any influence over EPS enzyme expression in industrial LAB.
In S. pyogenes , a cdaA mutant is unable to secrete the secreted cysteine protease virulence factor SpeB (Fahmi et al. 2019 ). Inter estingl y, a suppr essor m utation r esulting in inactiv ation of the K + transporter KtrB restored SpeB production (Faozia et al. 2021 ). KtrB forms a complex with, and is regulated by the c-di-AMP binding RCK_C gating component KtrA (discussed below). SpeB pr oduction c hanges wer e found to be due to K + imbalance since the addition of sub-growth inhibitory concentrations of ionophor es v alinomycin and gr amicidin r estor ed pr oduction in a cdaA mutant (Faozia et al. 2021 ). A number of other phenotypes (e .g. biofilm formation, acid resistance , and ampicillin resistance) of the S. pyogenes cdaA mutant were either partially or fully restored to wild-type levels upon inactivation of KtrB, demonstrating the significant negative impact of uncontrolled high intracellular K + le v els in cells de void of c-di-AMP (Faozia et al. 2021 ). Ho w e v er, the virulence of the cdaA / ktrB mutant in a subcutaneous mouse model remained similarly attenuated as the cdaA mutant, indicating that other c-di-AMP controlled physiological or structur al c hanges beyond K + balance influence infectivity of S. pyogenes .
Due to its importance in cell growth and virulence, there has been interest in screening of small molecules, which inhibit DAC activity in pathogens (Opoku-Temeng et al. 2016 ). A DAC inhibitor compound called ST056083, which was identified from a 1000 compound library (Zheng et al. 2014 ), was shown to inhibit c-di-AMP le v els, EPS pr oduction, and biofilm formation in E. faecalis (Chen et al. 2018 ). In other work, an inhibitor called IPA-3 was identified from a screen of 1133 compounds using the Streptococcus suis CdaA enzyme (Li et al. 2022 ). This compound was able to inhibit the growth of a number of different c-di-AMP producing Grampositive bacteria, but not E. coli , which does not make c-di-AMP (Li et al. 2022 ).

Antibiotic resistance
Se v er al studies have reported changes in resistance to w ar ds antimicrobials in LAB due to an imbalance of c-di-AMP. This includes mostl y cell env elope tar geting antimicr obials suc h as ampicillin, cefur oxime, v ancomycin, bacitr acin, da ptomycin, and cell-wall hydr ol ytic enzymes (Smith et al. 2012, Fahmi et al. 2019, Kundra et al. 2021, Pham et al. 2021, Rørvik et al. 2021. Ho w e v er, in most cases, the mechanisms by which c-di-AMP affects antimicrobial resistance is unclear. Recent work in L. lactis has explored the mechanisms underlying the sensitivity of a low c-di-AMP cdaA mutant to w ar ds the β-lactam antibiotic cefuroxime (Pham et al. 2021 ). Using a genetic suppressor screen, mutations that either partially or fully inactivated the K + importer KupB or the glutamine importer GlnPQ r estor ed cefur oxime r esistance . T he control of KupB by cdi-AMP (discussed below) and the importance of K + in osmoregulation in L. lactis had been demonstrated before (Pham et al. 2018, Quintana et al. 2019. The connection between c-di-AMP and GlnPQ, ho w e v er, was ne w. It was found that c-di-AMP inhibits GlnPQ activity indir ectl y by its contr ol of K + uptake. Ele v ated ionic str ength activ ation on GlnPQ activity was identified earlier by in vitro experiments using reconstituted proteoliposomes (Schuurman-Wolters and Poolman 2005 ). Imported glutamine is r a pidl y conv erted to glutamate that can be converted to aspartate . T he anionic amino acids aspartate and glutamate are the most abundant free amino acids in L. lactis (Pham et al. 2021 ). Cefur oxime r esistance was found to be inv ersel y r elated with osmoresistance. Together, this work showed that uncontrolled accum ulation of tur gor-inducing intr acellular osmol ytes (K + or anionic amino acids) causes cefuroxime sensitivity in low c-di-AMP cells . T his appears to be an example of where dysregulation of osmotic homeostasis indir ectl y r esults in another phenotypic difference.

C-di-AMP receptors in LAB
C-di-AMP exerts its contr ol thr ough the binding of protein and riboswitc h r eceptors (He et al. 2020 , Stülke andKrüger 2020 ;Figs. 1 and 4 ). A number of protein receptors from LAB have been characterized, but LAB do not appear to possess the c-di-AMP binding riboswitch (Nelson et al. 2013 ). Methods to find c-di-AMP binding partners have included screening of protein expression libraries (Corrigan et al. 2013, Schuster et al. 2016 ) and affinity purification using c-di-AMP coupled beads or resin (Bai et al. 2014, Sureka et al. 2014, Peng et al. 2016. Suppressor mutant screening approaches using high or low c-di-AMP mutants have also identified genes that encode c-di-AMP receptors (Devaux et al. 2018, Pham et al. 2018. Following the identification of c-di-AMP binding domains, targeted testing of proteins from different bacteria has been carried out. The most commonly used method of verifying a c-di-AMP r eceptor involv es the d iffer ential r a dial c apillary a ction of l igand a ssa y (DRaC ALA; Roelofs et al. 2011 ). T his in volves mixing 32 Pradiolabelled c-di-AMP with the protein, then spotting the mixture on a nitrocellulose membrane . T he protein will bind to the membrane at the site of application and c-di-AMP will diffuse beyond this site, unless it is bound by the protein. This method can be used to quantitativ el y determine the dissociation constant (K d ) and can also test binding of unpurified receptors from E. coli whole cell lysates (Roelofs et al. 2011 ). Se v er al differ ent c-di-AMP binding domains have been identified. These include the cystathionine-βsynthase (CBS, [IPR000644]), regulator of K + conductance (RCK_C; [IPR006037]), and universal stress protein (USP; [IPR014729]) domains, whic h ar e pr esent within pr oteins that play r oles in osmoregulation (He et al. 2020 , Stülke andKrüger 2020 ;Fig. 4 ). C-di-AMP binding poc kets hav e also been identified in proteins without designated binding domains, which allow for allosteric inhibition of enzyme activity (Sureka et al. 2014 ).

Potassium transporters
Potassium is the dominant cation in bacteria, present at intracellular concentrations of several hundred millimolars , pla ying a critical role in osmoregulation, pH control, protein expression, and enzyme activity (Stautz et al. 2021 ). A wide variety of differ ent c-di-AMP r eceptors identified in bacteria ar e involv ed in potassium homeostasis (Stülke and Krüger 2020 ). These include K + importers or importer gating components (T rkA, T rkH, KupA, KupB, and KimA), two-component systems that regulate K + importer expression (KdpD), a K + exporter (CpaA), and a riboswitch that regulates K + importer gene transcription. Some of these K + transporters contain distinct c-di-AMP binding domains, including RCK_C and USP. In LAB, various combinations of different cdi-AMP r eceptors involv ed in K + homeostasis are present, with some species containing five ( Lactiplantibacillus plantarum WCFS1 and Limosilactobacillus fermentum 9-4) and others containing just one ( Oenococcus oeni PSU1, Lactobacillus iners C0011D1, Pediococcus acidilactici ATCC8042, Pediococcus pentosaceus ATCC25745, and L. lactis MG1363; Fig. 3 ). There is also variation between strains of the same species, with some strains of L. lactis containing three putativ e K + tr ansporters r egulated by c-di-AMP (Pham and Turner 2019 ).
The first demonstration of a c-di-AMP receptor in LAB was the RCK_C domain containing CabP (KtrA) fr om S. pneumoniae , whic h was identified by passing cell lysates through a c-di-AMP agarose resin (Bai et al. 2014 ). KtrA and TrkA homologs are termed gating components since they bind to and regulate the activity of membrane imbedded K + translocation subunits KtrB and TrkH, r espectiv el y (Stautz et al. 2021 ). KtrA contains one RCK_C domain and forms an octameric ring, whilst TrkA contains two RCK_C domains and forms a tetrameric ring (Stautz et al. 2021 ). Work in S. pneumoniae demonstrated that the interaction between CabP and KtrB and K + uptake from this import system was reduced in the presence of c-di-AMP (Bai et al. 2014 ). Subsequent investigations have confirmed KtrA and TrkA homologs bind c-di-AMP in S. mutans and S. agalactiae (Peng et al. 2016, De v aux et al. 2018. KtrA and TrkA homologs in S. mutans had K d values of 7.8 and 1.2 μM, r espectiv el y (Peng et al. 2016 ). KtrA homologs are more commonly found in LAB than TrkA homologs (Fig. 3 ). KtrA ar e noticeabl y absent in dairy lactococcal species whilst TrkA ar e onl y pr esent in streptococci.
The other K + tr ansporter, whic h has been investigated in LAB is the KUP/HAK/KT family of proteins (IPR003855). Kup homologs are single subunit K + /H + importers containing 12 transmembrane spanning domains with an intracellular ∼200 amino acid domain of unknown function. In a salt r esistance scr een of a high c-di-AMP gdpP mutant of L. lactis , four single amino acid substitutions in K upB (K up2) wer e identified in suppr essors (Pham et al. 2018 ). These m utations wer e all pr esent on the cytosolic side of the protein, either in the C-terminal domain or in an intracellular loop between TMDs . T he K + le v el, whic h was r educed in the high cdi-AMP cells compared to wild-type, was increased in cells containing the mutated KupB, indicating gain-of-function changes. In most L. lactis , there are two Kup proteins (KupA and KupB), which are encoded by neighbouring genes. Some L. lactis strains, including MG1363, have a mutated kupA gene that is unable to generate a functional protein (Fig. 3 ). K upA and K upB from L. lactis IL1403 wer e expr essed as full length pr oteins and shown to bind c-di-AMP using DRaCALA (Quintana et al. 2019 ). Structural studies r e v ealed that despite having low ov er all sequence identity, KUP family proteins hav e conserv ed structur al and k e y residue consistency with KimA proteins, indicating the latter is a sub-family within the KUP family (Tascon et al. 2020 ). An attempt to identify the c-di-AMP binding site in the dimeric KimA structure w as undertaken, ho we v er, no bound nucleotide was identified during cryo-EM analysis (Tascon et al. 2020 ). The suppressor mutations in KupB that restor e salt r esistance in a high c-di-AMP m utant of L. lactis (Pham et al. 2018 ) may affect c-di-AMP binding, ho w e v er, further work is r equir ed to verify this and pinpoint the binding pocket.
Two other c-di-AMP receptors involved in potassium homeostasis are the Kdp system and CpaA exporter. KdpDE is a twocomponent system that regulates expression of the KdpFABC K + importer (Stautz et al. 2021 ; Fig. 4 ). The sensor kinase KdpD of S. aureus contains an N-terminal USP domain, which binds c-di-AMP (Moscoso et al. 2016 ). High c-di-AMP in this species resulted in impair ed activ ation of kdpFABC tr anscription upon osmotic str ess, whic h a gr ees with the inhibitory role c-di-AMP plays on K + accumulation. The K + exporter CpaA is a K + /H + antiporter and contains a c-di-AMP binding RCK_C domain (Corrigan et al. 2013 ). Work in S. aureus has found that high c-di-AMP activates K + export by CpaA (Chin et al. 2015 ), thus lowering intr acellular K + le v els. Experimental examination of LAB homologs of KdpD or CpaA has not been carried out this far, ho w e v er, KdpD is pr esent in v ery fe w LAB (including Carnobacterium , L. fermentum 9-4, and some subspecies of L. lactis ), while CpaA is also absent from most LAB genera, but common in lactobacilli (Fig. 3 ).

Glycine-betaine transporter BusA and carnitine transporter OpuC
A number of different compatible solutes are either synthesized or imported by bacteria in response to osmotic stress, including glycine betaine , carnitine , proline , trehalose , and ectoine (Bremer and Kramer 2019 ). Compatible solutes are preferred over K + as they have less impact on physiological activities within the cell. Most LAB surveyed contain a high affinity ATP-binding cassette (ABC) transporter for glycine betaine termed BusA (OpuA), which consists of two proteins BusAA and BusAB (Fig. 3 ). The BusA complex consists of an extracellular solute-binding domain (SBD) component, which is commonly fused to a TMD (called BusAB), whic h inter acts with an intr acellular ATP ase (called BusAA) (Obis et al. 1999, Bouvier et al. 2000. In L. lactis , pr e vious work identified that under hyperosmotic conditions, glycine betaine uptake is incr eased thr ough both higher BusA expr ession and tr ansporter activity (Obis et al. 1999, Bouvier et al. 2000, van der Heide and P oolman 2000b. T he r epr essor BusR was identified as r egulating the transcription of the BusA complex genes by binding to a motif within the pr omoter r egion of busA in response to ionic strength (Romeo et al. 2003, Romeo et al. 2007 ). BusA activity is also responsive to ionic strength and membrane composition changes (van der Heide and Poolman 2000a , van der Heide et al. 2001, Biemans-Oldehinkel et al. 2006. A connection between BusA and c-di-AMP was proposed when a gdpP mutant of L. lactis was found to have five-fold lo w er busA mRN A le v els compar ed to wild type under non-salt-stressed conditions (Smith et al. 2012 ). Subsequent work in S. agalactiae identified inactivating mutations in busAB , which permitted growth of a cdaA mutant on rich media (Devaux et al. 2018 ). A cdaA mutant of L. lactis was also found to harbour a busA promoter mutation, which inactivated expression of the gl ycine betaine tr ansporter (Pham et al. 2021 ). Together these results suggest that glycine betaine uptake is toxic to low/no c-di-AMP LAB cells . T he BusAA ATPase contains a CBS domain, which in other proteins has been found to bind c-di-AMP (Stülke and Krüger 2020 ). Initial attempts to demonstrate binding between BusAA homologs and c-di-AMP using DRaCALA, ho w e v er, did not pr ov e successful , De v aux et al. 2018.
Recentl y the cryo-electr on micr oscopy (EM) structur al anal ysis of the BusA transport complex r e v ealed c-di-AMP binding to the CBS domain within the ATPase BusAA component (Sikkema et al. 2020 ;Fig. 5 ). This suggests that the DRaCALA method has limitations and that some proteins may need to be in their native membrane complex to exhibit c-di-AMP binding. In addition to confirming the binding of c-di-AMP and glycine betaine by the BusA complex, the cryo-EM study further r e v ealed that BusA is likely to inter convert betw een different conformations in solution. Based on the structural and function studies, a model was proposed to explain how c-di-AMP regulates the transporter function of BusA. The transport cycle begins with the glycine betaine-bound SBD docking to TMD to form a transient outw ar d-facing conformation. The binding of ATP promotes nucleotide-binding domain (NBD) dimerization to shift BusA into an outw ar d-open conformation to allow glycine betaine to diffuse into a hydrophobic pocket in the TMD. The following ATP hydr ol ysis induces an inward-facing conformation with the NBDs shifted a wa y from each other. Dissociation of the NBD domains releases glycine betaine to the cytoplasm to complete the transport cycle. Importantly, the release of glycine betaine from BusA is blocked when c-di-AMP binds to the CBS domain to lock BusA into an inhibitory confirmation while stimulating the ATPase activity of the NBD domain. The model agrees with the observations that glycine betaine uptake by membrane reconstituted BusA complex was strongly inhibited by c-di-AMP and the ATPase activity is stimulated by c-di-AMP. The EM studies also implicated a positive charged motif in the NBD domain as a membr ane-inter acting ionic str ength sensor to r ationalize the ionic strength dependence of the transporter (Sikkema et al. 2020 ;Fig. 5 ).
The ABC transporter for the compatible solute carnitine has also been found to bind c-di-AMP in se v er al LAB ( E. f aecalis and S. agalactiae ) via the CBS domain in the intr acellular ATP ase subunit OpuCA , De v aux et al. 2018. Using DRaCALA, the E. faecalis OpuCA was found to have a K d of 6 μM . From our analysis , OpuC A is present in less than half of the LAB examined, including a subset of streptococci and lactobacilli, but absent in lactococci (Fig. 3 ).

Transcriptional r epr essor BusR
Streptococcus agalactiae and L. lactis BusR bind c-di-AMP via its Cterminal RCK_C domain (IPR006037) and inactivation of busR can r estor e salt resistance to high c-di-AMP gdpP mutants (Devaux et al. 2018, Pham et al. 2018. The BusR RCK_C domain binds c-di-AMP with a K d of ∼10 μM (Pham et al. 2018 ), which is similar to other pr otein r eceptors that ar e mostl y in the low micromolar le v el (He et al. 2020 ). Glycine betaine levels in L. lactis were 10fold lo w er in the high c-di-AMP gdpP m utant compar ed to wildtype, consistent with c-di-AMP driving BusR r epr ession (Pham et al. 2018 ). In a gr eement with this, gl ycine betaine le v els r eturned to wild-type le v els upon inactiv ation of busR . As pr e viousl y mentioned, in vitro BusR DNA binding is also r esponsiv e to ionic strength (Romeo et al. 2007 ) and high c-di-AMP m utants hav e r educed K + uptake, so it is likely that c-di-AMP affects BusR r epr ession by direct binding and indir ectl y by r egulating the intr acellular ion concentration (Pham et al. 2018 ). Recent work with S. agalactiae BusR has demonstrated that c-di-AMP enhances DNA binding (Bandera et al. 2021 ). The cryo-EM and crystal structur al anal ysis of S. agalactiae BusR sho w ed that the oligomeric BusR adopts an autoinhibited conformation with the DNA-binding winged helix-  (Sikkema et al. 2020 ). In BusAB, glycine betaine bound in the solute-binding protein (SBD), which is fused to the TMD. Highlighted here are also the stabilizing scaffold motif located on the periphery of the TMD of BusAB and the ionic strength sensor motif in BusAA with the functionally important Lys16, Arg17, and Lys19 shown in stick representation.
turn-helix motif (wHTH) domain pac ked tightl y a gainst the RCK_C domain (Fig. 6 ). Upon the binding of c-di-AMP to the RCK_C domain, BusR shifts to an activated confirmation with the wHTH domain adopting an extended conformation that enables BusR-DNA inter action (Bander a et al. 2021 ). The DNA-binding specificity is defined by the long coiled-coil region that allow recognition of a 22-bp spaced DNA sequence motif (Fig. 6 ).
BusR is present in around half of the LAB genomes we analysed and all strains that contained BusR also had BusAA (Fig. 3 ). A number of species contained BusAA but not BusR, indicating that c-di-AMP regulation of glycine betaine uptake in these bacteria may be solely at the level of inhibiting transporter activity. Inter estingl y, a number of lactobacilli and streptococci possessed neither the BusR, BusAA, or OpuCA proteins suggesting that there is no c-di-AMP regulation of compatible solute uptake in these species or alternative transporters are present. PC was ca ptur ed in a screen for c-di-AMP receptors in L. monocytogenes using a pull-down assay with c-di-AMP coated beads (Sureka et al. 2014 ). Identification of the c-di-AMP binding pocket fr om crystal structur e anal yses and comparison of PC sequences suggested that it is not well conserved and likely that only a subset of PC enzymes can bind this allosteric inhibitor. PC from LAB E. faecalis and L. lactis were predicted to contain suitable c-di-AMP binding sites and enzyme activity assays r e v ealed up to ∼30% to 60% inhibition by c-di-AMP, r espectiv el y (Sur eka et al. 2014, Choi et al. 2017. No inhibition b y another c yclic dinucleotide c-di-GMP was observed for L. lactis PC, demonstrating the specificity of the inhibition (Choi et al. 2017 ). A crystal structure of the L. lactis PC in complex with c-di-AMP was determined to confirm the binding c-di-AMP site (Fig. 7 A). Lactococcus lactis PC forms a tetramer with c-di-AMP bound at the dimeric interface of two carboxyltr ansfer ase domains. A fr ee L. lactis PC crystal structure was also determined to show the binding of c-di-AMP induces significant conformational changes and imply that an allosteric mechanism is likely to underly the inhibitory effect of c-di-AMP (Choi et al. 2017 ). Se v er al r esidues lining the L. lactis PC binding poc ket, whic h ar e shar ed with L. monocytogenes PC, include a Tyr715 residue that interacts with the adenine base of c-di-AMP and small residues (Ser745 and Gly746) allowing space for c-di-AMP entry into the pocket (Choi et al. 2017 ) (Fig. 7 B). Staphylococcus aureus PC, which is not inhibited by c-di-AMP, contains two lar ger r esidues (Lys and Gln) at positions 745 and 746 (LlPC numbering) putativ el y pr eventing the entry of c-di-AMP (Sureka et al. 2014 ). We analysed r esidues fr om LAB PC that may form a c-di-AMP binding pocket (Fig. 7 C). In addition to Lactococcus and Enterococcus , PC from several other LAB, including Carnobacterium and Vagococcus may be able to bind c-di-AMP. Other LAB contain one or two residues with longer side chains at positions 745-746, thereby potentially preventing cdi-AMP doc king. Ther efor e, it is likel y that onl y a small subset of LAB PCs are potentially regulated c-di-AMP, ho w ever, further w ork would be r equir ed to verify these predictions.

Anionic amino acid biosynthesis
PC and aspartate biosynthesis is essential for acidification of milk by L. lactis since nutritional demands for this amino acid are unable to be met by protein/peptide sources (Wang et al. 1998, Dudley and Steele 2001, Choi et al. 2017. Aspartate is a k e y amino acid in the cell as it is a precursor for five other amino acids as w ell as p yrimidines (Wang et al. 2000b ). In early work with L. lactis PC, it was found that > 90% of the OA generated by PC is converted to aspartate (Hillier and Jago 1978a ). Aspartate was identified as a str ong non-competitiv e inhibitor of L. lactis PC activity, with 2.9 mM causing 50% inhibition, (Hillier and Jago 1978b ). Unlike other PCs, L. lactis PC has high inherent enzyme activity and is insensitive to acetyl-CoA activation (Choi et al. 2017 ). Together this suggests that the primary role of PC in L. lactis is to supply aspartate to the cell.
The finding that c-di-AMP inhibits PC provides an additional layer of regulation of aspartate le v els. It was found that a high c-di-AMP L. lactis mutant contains lo w er aspartate levels, which could be r estor ed following introduction of a variant PC (Y715T) that is insensitive to c-di-AMP inhibition (Choi et al. 2017 ). As well as regulation of biosynthesis in L. lactis , aspartate levels are also dampened via indirect c-di-AMP inhibition of import of glutamine via GlnPQ (Pham et al. 2021 ). Aspartate is the most abundant free amino acid in L. lactis and along with glutamate, which is also regulated by c-di-AMP, make up 55% of the free amino acids (Pham et al. 2021 ). These likel y contribute significantl y to w ar ds the osmotic pr essur e within the cell and form another extension of the osmor egulatory r eac h of c-di-AMP via PC activity modulation.

Stimuli affecting the c-di-AMP pool size
C-di-AMP le v els in bacteria change in response to external and internal signals. In a gr eement with the known tar gets of c-di-AMP, and the effects this nucleotide plays in osmor egulation, c hanges that likely affect cellular osmotic pressure have been shown to trigger adjustments in the c-di-AMP le v el in various LAB. In S. pneumoniae , inactivation of the K + transporter gating component and c-di-AMP receptor CabP (KtrA), which likely lo w ers intracellular K + le v els, r esulted in a significant reduction of the c-di-AMP level (Zarrella et al. 2018 ). This work also r eported that c-di-AMP le v els w ere lo w er and CdaA pr otein le v els wer e r educed in S. pneumoniae cells grown in low K + c hemicall y defined media. A role for K + in affecting c-di-AMP le v els in L. lactis has also been r eported with ov er activ e K + uptake (caused by gain-of-function m utations in kupB or ov er expr ession of kupB ) significantl y incr eased c-di-AMP le v els (Pham et al. 2018 ). Also in this study, ov er accum ulation of the compatible solute gl ycine betaine, fr om inactiv ating busR , led to ele v ated c-di-AMP. Together these findings indicate that intr acellular ionic strength and/or turgor changes are sensed by CdaA and/or GdpP to optimize the c-di-AMP concentr ation. Specificall y, low intracellular K + or compatible solute le v els, leading to low turgor, trigger the cell to lo w er the c-di-AMP le v el in order to activ ate gr eater accum ulation of osmo-activ e substances . Con v ersel y, high intracellular K + or compatible solute le v els, causing high turgor, trigger the cell to increase the c-di-AMP le v el in order to inhibit further accumulation of osmo-active substances. As well as these internal signals, external osmolarity changes in the envir onment hav e been found to trigger c-di-AMP le v el adjustments in se v er al bacteria, including the LAB L. lactis and Lb. plantarum (Pham et al. 2018 ). Non-growing cells suspended in a low osmolarity solution containing glucose needed for synthesis of the CdaA substr ate ATP r a pidl y incr eased their intr acellular c-di-AMP le vels within 5 minutes . T he subsequent addition of salt (0.3 M NaCl or KCl) triggered a rapid depletion or block in c-di-AMP synthesis. T hese studies pro vide a rational connection between the types of stim uli affecting tur gor with the r ole of man y c-di-AMP r eceptors in osmoregulation. Ho w ever, the mechanistic basis of how stimuli is transduced into DAC or PDE enzyme activity changes remains to be solved.
Changes in the bacterial cell envelope of E. faecalis and Enterococcus faecium have also been reported to trigger altered c-di-AMP levels (Wang et al. 2017 ). Inactiv ation of liaR , whic h encodes a component of the LiaFSR three-component cell envelope stress response system, led to an ele v ated c-di-AMP le v el. Whilst the underlying reason for this is unclear, it is known that the LiaFSR system regulates anionic phospholipid distribution in the cytoplasmic membr ane (Tr an et al. 2013 ). This c hange in membr ane composition or c har ge may affect the activities of CdaA or PDE enzymes, which are both membrane-bound.
In non-LAB, a connection between synthesis of the stringent response signalling molecule (p)ppGpp and c-di-AMP le v els thr ough the c-di-AMP binding receptors CbpB (DarB) has been demonstrated (Peterson et al. 2020, Kruger et al. 2021. In turn, (p)ppGpp inhibits the c-di-AMP PDEs GdpP and PgpH (Rao et al. 2010, Huynh et al. 2015, which lead to elevated c-di-AMP levels (Corrigan Figure 7. Structure of c-di-AMP bound PC from L. lactis (A). The c-di-AMP binding pocket (B) and an alignment of key residues in the c-di-AMP binding pocket of PC in LAB (blue highlighted amino acids should have small side chains in order to not block the c-di-AMP pocket) (C). In the close-up view of the c-di-AMP binding site at the dimeric interface (B), the three k e y residues (Q712, Y715, and Q749) that are directly involved in c-di-AMP binding are highlighted. The two residues with short side chains (S745, G746) located deeper in the pocket and permit c-di-AMP binding are also highlighted (Protein Data Bank 5VYZ; Choi et al. 2017Choi et al. ). et al. 2015. Most LAB contain a CbpB homolog (Fig. 3 ), so it is likely that cross-talk between these two second messengers is also widespread in LAB species.
It is curr entl y not known whether there are stimuli that trigger changes in localized or global c-di-AMP signalling networks. For cdi-GMP, a large number of synthesis and degradation enzymes are present and through the formation of multi-protein complexes with their cognate receptors, cells are able to achieve targeted outputs with a diffusible second messenger (Hengge 2021 ). Far fewer enzymes are involved in c-di-AMP level homeostasis, how-e v er, some LAB hav e two PDEs. No inter actions between c-di-AMP receptors and DAC or PDE enzymes have been reported, but maintaining close proximity to c-di-AMP receptors would allow for faster control of the target receptors activity.

Extracellular c-di-AMP roles
C-di-AMP r eleased fr om intr acellular and extr acellular bacterial pathogens triggers a type I interfer on imm une r esponse via the stimulator of interferon genes pathway (Woodward et al. 2010 , Barker et al. 2013, Andrade et al. 2016. In S. agalactiae , extracellular c-di-AMP is hydr ol ysed by a LPXTG-motif containing cell-wall anc hor ed PDE called CdnP, which results in dampening of the host innate immune response and greater infectivity (Andrade et al. 2016 ). From our analysis, CdnP is present in only a few LAB (Fig. 3 ); ho w e v er, it is unclear if the substrate for this enzyme is c-di-AMP since CdnP shares the same domain structure as another ectonucleotidase NudP, which cleaves AMP to adenosine (Andrade et al. 2016 ). Inter estingl y, se v er al of the putative CdnP homologs, including those in E. faecalis , Vagococcus lutrae , and Carnobacterium have undergone a duplication event and contain two 5 -nucleotidase and metallophosphoesterase domains, resulting in a large cellwall anc hor ed enzyme.
As observed for bacterial MDR overexpression mutants, elev ated extr acellular le v els of c-di-AMP hav e also been r eported for high c-di-AMP PDE mutants, including S. pneumoniae and L. lactis (Witte et al. 2013, Pham et al. 2018, Zarrella et al. 2018. Whilst most studies examining the role of c-di-AMP in microbe-host interactions has been in the context of bacterial infections, it is plausible that c-di-AMP released from non-pathogenic food, probiotic, or commensal LAB may also impact host immune responses. Purified c-di-AMP has shown promise as a mucosal adjuvant in vaccine r esearc h (Cheng et al. 2022 ). Work to de v elop L. lactis as a m ucosal v accine deliv ery system has utilized antigen expressing str ains with ele v ated c-di-AMP, ac hie v ed by ov er expr essing cdaA (Quintana et al. 2018 ). The combination of c-di-AMP and antigen led to a gr eater imm une r esponse than when using only c-di-AMP or antigen expressing strains. Further evidence for a role of c-di-AMP r elease fr om non-pathogenic bacteria triggering a host imm une r esponse was r eported in a r ecent gut micr obiome study  ). C-di-AMP r eleased fr om the common gut commensal bacterium Akkermansia muciniphila was identified as a k e y imm une stim ulating molecule for effective cancer immunothera py tr eatment .
As well as impacts on the mammalian host, extracellular c-di-AMP has been shown to influence bacterial physiology, specifically rescuing defects of cdaA mutants. Exogenously added c-di-AMP to L. monocytogenes impr ov ed v ancomycin r esistance (Ka plan Zee vi et al. 2013 ). In other work, exogenously added c-di-AMP, but not c-di-GMP, was able to r estor e gr owth of a cdaA m utant of E. f aecalis in c hemicall y defined media with supplemented peptone, in a dose dependent manner (Kundra et al. 2021 ). This suggests that extracellular c-di-AMP can potentially be imported by bacteria and go on to regulate intracellular targets. As a result, bacterial cell-tocell communication via c-di-AMP signalling may potentially occur in high bacterial cell density en vironments , such as biofilms.

Future perspectives
r A significant gap in knowledge in the c-di-AMP signalling field is a mechanistic understanding of how external and/or internal signals transduce into c-di-AMP level fluctuations via changes in CdaA and GdpP activity. Studies have reported r a pid modulation of c-di-AMP le v els upon c hanging external osmolarity or unbalanced uptake of K + or glycine betaine . T his suggests a post-translational (enzyme activity) response. CdaA and/or GdpP may be sensing membrane curv atur e/str etc hing alter ations or intr acellular ionic str ength c hanges via salt-sensitiv e inter actions with membr ane phospholipids, similarly to that found for osmosensing compatible solute transporters (Wood 2011, Wood 2015. Ionic strength changes may also potentially affect protein-protein interactions leading to r a pid switc hing to and fr om catal yticall y ac-tiv e m ultimeric enzyme forms. Structur al and bioc hemical studies of full-length and membrane-bound DAC and PDE enzymes may r e v eal the mec hanistic understandings of signal transduction. r Cells with an imbalanced c-di-AMP le v el hav e pleiotr opic phenotypes, but are these an indir ect r esult of impair ed osmotic homeostasis? Changes in intracellular osmo-active compound concentrations will cause altered turgor pressure that may affect cell envelope properties, such as resistance to cell-wall acting antimicr obials. C-di-AMP contr ols the le vels of dominant intracellular ions, including K + and anionic amino acids (Asp and Glu), which may have global effects in the cell, affecting the activity of enzymes and transporters, as well as macromolecule interactions (e.g. protein-DNA). Teasing apart the reason for various phenotypes is needed to better understand the broad reaching consequences of imbalanced c-di-AMP le v els.
r Whilst most LAB contain only one synthesis enzyme and one degradation enzyme, it is less likely that c-di-AMP output specificity r equir es localized signalling like that observed for more complicated c-di-GMP signalling networks. Several LAB contain two PDE enzymes or onl y accum ulate significant c-di-AMP le v els when both DhhP and GdpP are inactivated, suggesting there may be both local and global c-di-AMP signalling networks in some LAB.
r Does extracellular c-di-AMP released from LAB from fermented foods, as probiotics or as part of the microbiome impact on the human host during gastr ointestinal tr ansit or colonization? Ther e is e vidence of c-di-AMP pr oduced by bacteria in the intestinal lumen that triggers significant host imm une r esponses. As a potent imm unostim ulatory molecule, future work evaluating c-di-AMP in bacteria-host interactions will be of interest.
r There is the possibility of other as yet unidentified cyclicdinucleotide systems operating in LAB. A recent study reported the c har acterization of a cyclic-oligonucleotide-based antiphage signalling system in L. lactis involving a 3 ,2 -cGAMP-acti vated n uclease (Fatma et al. 2021 ). We found this system is absent in our set of r epr esentativ e genomes and ther efor e unlikel y to be a widespread anti-phage system in LAB. Further exploration of alternative nucleotide signalling systems may r e v eal nov el signalling pathways oper ating in LAB.

Ac kno wledgements
We acknowledge the many students and staff who have worked on c-di-AMP projects in our laboratories, as well as our collaborators.

Supplementary data
Supplementary data are available at FEMSRE online.

Funding
Research funding for our work on c-di-AMP awarded to M.S.T., E.M., and Z-X.L. is from the Australian Research Council (grant DP190100827). Elements of this r esearc h used equipment fr om the Queensland node of Metabolomics Australia funded by Bioplatforms Australia, an NCRIS funded initiative.