A contact-dependent process combined with the presence of CdiA at the cell surface of the attacker raised a question as to whether a cell surface receptor in the recipient cell is involved in docking/recognition of CdiA. If this was the case, then a variant of recipient cells for which the receptor is lacking or altered would become CDI-resistant (CDI^R). Using CDI^EC93 as a model, a transposon (Tn)-based mutagenesis screening led to the identification of such CDI^R mutants that have Tn insertion in either acrB or the promoter region of bamA (Table 1; Aoki et al., 2008).

BamA is an OM protein at the core of the beta-barrel assembly machinery (BAM) complex and required for proper assembly/insertion of other beta-barrel proteins in the OM (Rigel and Silhavy, 2012). As bamA is an essential gene, the bamA-mutated CDI^R mutant is not a null mutant but a knockdown mutant with five-fold less expression (Aoki et al., 2008). The biogenesis-inactive version of BamA, as well as other BAM complex variants, remains capable to mediate CDI, indicating that the presence of BamA but not the function of the BAM complex is required for CDI (Aoki et al., 2008). Treatment of bacterial cultures using anti-BamA antibody that recognizes the recipient BamA on the cell surface disrupted the attacker-recipient cell recognition and thus the CDI-mediated growth inhibition (Aoki et al., 2008). The results strongly support the idea that BamA is an OM receptor of the CdiA^EC93.

Identification of the binding site between the CdiA^EC93 toxin and BamA confirmed BamA as the receptor of CdiA^EC93 (Figure 3). The RBD of CdiA^EC93 (from Arg1358 to Phe1646) binds to BamA’s loop 6/loop 7 variable region that is identical in hundreds of other E. coli strains but shares low-sequence similarity among different CDI-encoding species (Ruhe et al., 2013, 2017). The results correlated well with the observation that CDI^EC93 is unable to inhibit other CDI homologs-harboring species like Salmonella enterica serovar Typhimurium, Citrobacter freundii, Enterobacter aerogenes, Enterobacter cloacae, or Proteus mirabilis but was able to inhibit a variety of E. coli strains (Aoki et al., 2010; Ruhe et al., 2013). To summarize, the CdiA^EC93 uses its RBD domain to bind specifically to the OM protein BamA of the E. coli recipient, demonstrating that CDI is restricted to intraspecies competition in a recipient receptor-dependent manner.

In contrast to CDI^EC93, CdiA of the uropathogenic E. coli (UPEC) strain 536 (CdiA^EC536) was shown to recognize the heterotrimeric OmpC-OmpF complex but not BamA (Figure 3; Beck et al., 2016). More specifically, the RBD region of the CdiA^EC536 interacts with the extracellular loops L4 and L5 of OmpC (Beck et al., 2016; Ruhe et al., 2017). Unlike the binding region of BamA to CdiA^EC93, which is highly conserved in protein sequence in hundreds of E. coli strains, the L4 and L5 of OmpC is highly diverse in protein sequence even among different E. coli strains (Aoki et al., 2008; Beck et al., 2016). Such OmpC polymorphism restricts the range of recipients for CDI^EC536 (Beck et al., 2016). Although OmpF is strictly required for CDI^EC536, using ompF alleles that are highly diverse from that of EC536 does not interfere with the CDI^EC536 delivery process, consistent with the hypothesis that the recognition sites reside in OmpC. The data obtained on receptor preference or specificity of CDI from different E. coli strains suggest that E. coli may use the CDI systems to distinguish “self” from “non-self” cells and promote interactions between siblings.

After identifying the RBD region as the recipient OM receptor binding domain, the RBD region of all identifiable CdiA in the databae resoures of National Center for Biotechnology Information (NCBI) was compared to gain insights in binding specificity and selectivity toward either BamA or OmpC/OmpF (Ruhe et al., 2017). The results indicated that the CdiA RBD region can be divided into four main classes, instead of two, based on their amino acid sequences. Given the variability between the toxin classes, the CdiA^STECO31 from E. coli STEC_O31 was used as a class III effector model to search for its CDI^R mutants. The genetic screen of CdiA^STECO31 CDI^R mutants led to the discovery of a new receptor, namely, Tsx (Figure 3), an OM protein that functions as a monomeric nucleoside-specific porin (Bremer et al., 1990). Although the binding region of the Tsx remains elusive, the RBD region of the CdiA^STECO31 lies in the Gln1385-Tyr1657 (Ruhe et al., 2017).

Besides identification of OM receptors, CDI^R genetic screens also identified several genes encoding IM components (Figure 3; Aoki et al., 2008; Ruhe et al., 2014, 2017; Willett et al., 2015a). In the CDI^R genetic screen using the attacker CDI^EC93, acrB and bamA integrity in the prey cells were found mandatory for the attack to be effective. AcrB is an IM multidrug transport protein belonging to the multidrug/proton antiporter that is composed of AcrB, periplasmic protein AcrA, and OM protein TolC (Tikhonova and Zgurskaya, 2004). Intriguingly, only mutations in acrB but not acrA or tolC conferred resistance to the CDI, suggesting that the AcrB-mediated CDI is independent of its multidrug efflux pump function (Aoki et al., 2008). Of note, cells intoxicated with CdiA^EC93 have reduced proton motive force and steady-state ATP levels, and their AcrB-containing multidrug/proton antiporter function is blocked (Aoki et al., 2009). These results suggested that CdiA-CT^EC93 might interact with AcrB, thus resulting in dissipation of proton motive force (Aoki et al., 2009). Alternatively, AcrB could anchor the incoming CdiA-CT^EC93 in the IM to activate the toxin that forms a pore (Jones et al., 2017).

The recipient’s IM factors required for CdiA^EC536 is the filamenting temperature-sensitive H (FtsH) protein (Figure 3; Ruhe et al., 2014; Willett et al., 2015a). FtsH is an IM-anchored AAA⁺ protease, and its activity is stimulated by the proton motive force (Akiyama, 2002; Langklotz et al., 2012). As CdiA-CT^EC536 is a well-defined tRNase that functions in the cytosol (Aoki et al., 2010; Ruhe et al., 2014), the role of FtsH is suggested to mediate toxin translocation across the IM. It is worth noting that CdiA^EC536 and CdiA^ECL both require OmpC-OmpF heterotrimers and FtsH for toxicity (Willett et al., 2015a; Beck et al., 2016). However, the detailed mechanism of how FtsH is involved in CdiA toxicity remains elusive.

PtsG, the glucose-specific EIICB component of the sugar PTS (sugar phosphoenolpyruvate-dependent phosphotransferase) system, was found to be required for the toxicity of CdiA ^STECO31 (Gabor et al., 2011). As CdiA-CT^STECO31 encodes an EndoU anticodon nuclease that claves tRNA^Glu in the cytosol (Michalska et al., 2018), the IM protein PtsG is believed to enable CdiA-CT^STECO31 translocation into the recipient’s cytosol. Further screening for CDI^R mutants resisting intoxication for CdiA produced by a variety of different bacterial strains discovered that PtsG is also required for toxicity of CdiA-CT^NC101 and CdiA-CT³⁰⁰⁶ (Willett et al., 2015a).

Additional recipient IM factors were also identified by screening for CDI^R mutants resisting intoxication by CdiA produced by a variety of different bacterial strains (Willett et al., 2015a). The screening strategy was designed for identifying entry factors by using chimeric CdiA that harbors the N-terminus from E. coli strain EC93 and the C-terminal-containing toxin domain of other strains. The rationale is that the CdiA C-terminus (CdiA-CT) contains a variable domain that specifies the entry pathway into target bacteria and therefore recognizes and exploits specific proteins on the target cell for entry of the CdiA-CT toxin. Such screen has led to the discovery of six “permissive factors” conferring specific entry of different CDI toxins (Table 1 and Figure 3). Besides identifying known IM factor PtsG that is required for CdiA-CT^NC101 and CdiA-CT³⁰⁰⁶, additional IM proteins including MetI for CdiA-CT^MHI813, YciB for the orphan CdiA-CT of the EC869 (CdiA-CT_o11^EC869), GltK for Photorhabdus luminescens CdiA-CT^TTO1, and RbsC for D. dadantii CdiA-CT^Dd3937 were uncovered (Figure 3 and Table 1; Willett et al., 2015a). Orphan cdiA-CTs encode toxins but have no translation initiation region and therefore are not translated unless grafted with a region encoding an N-terminal CdiA sequence (Poole et al., 2011).

It is worth mentioning that all the identified recipient proteins were IM protein, thus indicating that CdiA-CT is the region recognizing the recipient’s IM receptor but not the OM receptor. This finding is consistent with the evidence that the RBD but not the CdiA-CT region is responsible for binding to the OM receptor of CdiA^EC93 (Figure 2; Ruhe et al., 2017). The authors also used chimeric CdiA-CT^{EC3006–EC869o11} to elucidate which part(s) of the CdiA-CT is responsible for recognition of the cognate IM receptor (Willett et al., 2015a). The CdiA-CT^{EC3006–EC869o11} consists of an N-terminal, CdiA-CT³⁰⁰⁶, and C-terminal fragments, CdiA-CT_o11^EC869, and requires PtsG but not YciB for growth inhibition. The results demonstrate that the IM receptor recognition domain of CdiA lies in the N-terminus of CdiA-CT and was thus designated as the entry domain, while the C-terminus of the CdiA-CT is the toxin domain itself (Willett et al., 2015a).

Cytoplasmic factors were also found to be required for effective CDI mechanism. In contrast to the roles of OM and IM factors involved in recognition and entry, cytoplasmic factors usually participate in enhancing toxin activity. The cytosolic factor of CdiA^EC536 is the O-acetylserine sulfhydrylase A (CysK) (Figure 3; Diner et al., 2012; Beck et al., 2016). The requirement of CysK in antagonizing recipient growth stems from an unexpected result that CdiA-CT^EC536 only displays tRNase activity in the presence of CysK both in vitro and in vivo (Diner et al., 2012). Crystal structure of the CysK/CdiA-CT^EC536 complex revealed that CysK interacts with the C-terminal Gly-Tyr-Gly-Ile (GYGI) motif of CdiA-CT^EC536, and this interaction increases the thermostability and tRNase activity of CdiA-CT^EC536 (Johnson et al., 2016). Intriguingly, CysK also binds to and stabilizes the CdiA-CT^EC536/CdiI^EC536 complex in the attacker cell, and such binding reinforces protection against autoinhibition (Kaundal et al., 2016). The CysK/CdiA-CT^EC536 interaction site mimics the binding site between CysK to its native substrate, CysE. Recent data demonstrated that CdiA-CT^EC536 has a higher affinity to CysK even in the presence of excess CysE (Johnson et al., 2016; Jones et al., 2017). In brief, CdiA- CT^EC536 utilizes CysK of the recipient cell to activate its tRNase activity once in the recipient cytosol.

Other recipient cytosolic factors required for CDI are the Elongation Factor Thermo-Unstable (EF-Tu) and the Elongation Factor Thermo-Stable (EF-Ts) (Figure 3; Jones et al., 2017; Michalska et al., 2017). The CdiA-CT^EC869 toxin interacts with the EF-Tu/GTP/tRNA complex with high affinity. More importantly, the tRNase activity of CdiA-CT^EC869 was only observed in the presence of this complex under in vitro conditions (Jones et al., 2017). Although EF-Ts was dispensable in activating CdiA-CT^EC869 in vitro, it is required in vivo. The role of EF-Ts in vivo was proposed to be promoting the formation of the EF-Tu/GTP/tRNA complex. Aside from EC869, CdiA-CTs from strains NC101 and 96.154 also interact with EF-Tu, and both were unable to intoxicate a tsf mutant that lacks EF-Ts (Jones et al., 2017; Michalska et al., 2017). It is worth noting that CdiA-CT^EC869, CdiA-CT^NC101, and CdiA-CT^96.154 share low-sequence similarity, suggesting that hijacking EF-Tu for activation may be a common strategy used by CDI toxins (Jones et al., 2017; Michalska et al., 2017).

As summarized above, the CDI system requires recipient membrane receptors and cytosolic activators to exert full toxicity. Exemplified by E. coli strains, a wide variety of the RS factors participates in recognition (OM receptor), translocation (IM proteins), and activity (cytoplasmic factors) of CDI toxins. Many other organisms harbor functional CDI, and they differ in gene organization, protein sequence, and cytotoxicity (Aoki et al., 2010; Kiel et al., 2012; De Gregorio et al., 2018; Allen and Hauser, 2019). As such, it is anticipated that novel CDI-dependent recipient receptors and activators would likely be discovered in future studies.

The initial clues for recipient factors affecting the outcome of T6SS killing came from microscopic observations of T6SS firing events (Basler and Mekalanos, 2012; LeRoux et al., 2012). T6SS firing events were monitored by visualization of ClpV-GFP as ClpV is required for disassembly of contracted T6SS sheath, an event subsequent to T6SS firing (Bönemann et al., 2009; Pietrosiuk et al., 2011). The presence of ClpV-GFP foci thus indicates that T6SS firing has just happened (Mougous et al., 2006; Basler and Mekalanos, 2012). It was observed that P. aeruginosa ClpV-GFP foci occurred at the exact place where its neighboring sibling cells also had a ClpV-GFP foci, indicating that one of the activating signals for P. aeruginosa T6SS firing is the T6SS attack from a neighbor sibling cell (Basler and Mekalanos, 2012; LeRoux et al., 2012). This phenomenon was then demonstrated in an interspecies T6SS competition scenario. When punctured by the V. cholerae T6SS, P. aeruginosa fires back using its T6SS at the exact position where it was challenged (Basler et al., 2013). Similarly, Agrobacterium tumefaciens T6SS also triggers a P. aeruginosa counterattack, which led to higher killing of T6SS-active A. tumefaciens as compared to T6SS-inactive strains (Ma et al., 2014). Interestingly, the T6SS counterattack also occurs when sensing the pKM101 T4SS mating pair formation (Mpf) system of E. coli donor cells to resist T4SS-mediated gene transfer of foreign DNA (Ho et al., 2013). Because T6SS firing is also induced by membrane-disrupting compounds such as polymyxin B, the authors concluded that the T6SS counterattack results from Mpf-mediated membrane disruption. Recent studies further showed that the production of two adhesins (TraC and Pep), or the formation of a T4SS channel, but not assembly of conjugative pilus, is capable of activating a T6SS counterattack (Gordon et al., 2017; González-Rivera et al., 2019). Therefore, T6SS firing could be a defensive weapon in response to various assaults challenging membrane integrity (Figure 5).

It has been demonstrated in multiple systems that the T6SS attack could be fine-tuned in response to different recipient cells (Ma et al., 2014; LeRoux et al., 2015; Lazzaro et al., 2017; Wu et al., 2019). For example, in P. aeruginosa, a non-self-recipient cell triggers a stronger T6SS attack than a susceptible sibling (LeRoux et al., 2015). Furthermore, the P. aeruginosa T6SS activity monitored by ClpV1-GFP was significantly elevated when cocultured with B. thailandensis as compared to a monoculture (LeRoux et al., 2012). The authors demonstrated that P. aeruginosa senses a “danger signal” released by lysed sibling cells and activates its T6SS to launch a counterattack (LeRoux et al., 2015). The enhanced T6SS susceptibility triggered by non-self-recipient cells was also demonstrated in A. tumefaciens (Ma et al., 2014). A. tumefaciens only exhibits antibacterial activity against E. coli but not against susceptible siblings in vitro (Ma et al., 2014). Furthermore, A. tumefaciens tends to antagonize other competitive A. tumefaciens strains from different genomospecies but not to the same degree to those within the same genomospecies in planta (Wu et al., 2019). In Serratia marcescens, the transcription level of T6SS is fine-tuned as the T6SS transcript level of S. marcescens varies when challenged by different competitors. Only basal levels of T6SS transcripts were detected when confronted with harmless recipient cells, while upregulation occurs at moderate or higher levels when confronted with contender or aggressive competitors (Lazzaro et al., 2017). Overall, these findings unveil the importance of kin recognition in determining the outcome of the T6SS attack, but future systematic analysis is required to identify the genetic features or determinants governing the fate of a competition (Figure 5).

The first evidence for the involvement of specific T6SS RS factors came from the characterization of the P. aeruginosa effector Tse6-loaded complex, which consists of Tse6, Tsi6 immunity protein, VgrG1, effector-associated gene with tse6 (EagT6), and EF-Tu (Whitney et al., 2015). The presence of EF-Tu in the Tse6-loaded complex was unexpected, and the authors addressed the role of EF-Tu by proposing four possibilities: EF-Tu may be required for (1) stabilizing Tse6, (2) activating Tse6, (3) facilitating Tse6 export from attacker cell, or (4) entering recipient cell. After ruling out the first three, the authors deduced that the interaction of Tse6 with EF-Tu might be required for entering the recipient cell. However, further study on the ability of the Tse6-loaded complex to translocate across membranes using liposome-based in vitro translocation assay showed that Tse6 translocation happened spontaneously in the absence of the inner-face EF-Tu (Quentin et al., 2018). Thus, EF-Tu may not play a role in entering recipient cells across the lipid bilayer, and the exact role of EF-Tu in the interbacterial competition is still to be elucidated.

Another example of RS factors affecting T6SS toxicity is DsbA that functions as a periplasmic disulfide bond-forming protein (Mariano et al., 2018). S. marcescens has two DsbAs, DsbA1 and DsbA2, which are functionally redundant for a proper T6SS functionality. Indeed, in S. marcescens-secreting cells, the presence of either DsbA1 or DsbA2 is sufficient for T6SS activity, but T6SS assembly and secretion levels are significantly compromised in the absence of DsbA1 and DsbA2. Strikingly, the peptidoglycan hydrolase Ssp2 and Ssp4 (English et al., 2012) are able to inhibit Ssp2- and Ssp4-susceptible S. marcescens strains, while a recipient lacking both DsbA1 and DsbA2 was entirely resistant against the activity of these periplasmic-acting effectors (Mariano et al., 2018). The requirement of DsbA for the toxicity of Ssp2 and Ssp4 was also confirmed by artificially expressing and targeting Ssp2 and Ssp4 to the E. coli periplasm, in which their toxicity is relieved if the E. coli strain lacks dsbA. Attacker cells expressing disulfide bond-lacking Ssp2 or Ssp4 did not show T6SS-mediated antibacterial activity. It is generally believed that T6SS delivers effectors from the attacker cell’s cytoplasm directly into the recipient cell, Ssp2 and Ssp4 effectors are unlikely to localize in the attacker cell’s periplasm to form a disulfide bond before its delivery. Thus, it remains unknown how DsbA influences T6SS activity in the attacker cell, but the contribution of DsbA or disulfide bond formation for activity of incoming periplasmic toxins in the recipient cell is likely a widespread mechanism (Figure 5; Mariano et al., 2018).

In A. tumefaciens, a high-throughput screening (HTS) aiming to identify RS factors that affect the T6SS killing outcome was performed (Lin et al., 2020). Using E. coli K12 strain BW25113 as the model recipient cell, several RS factors that enhance E. coli susceptibility to A. tumefaciens T6SS attack were identified. To date, the confirmed RS-encoding genes include clpA, clpP, gltA, ydhS, ydaE, and cbpA, all encoding cytosolic proteins. These results suggest that the RS factors affecting A. tumefaciens T6SS killing outcome are rather involved after injection of T6SS toxins into the recipient cells.

The clpP gene encoding ClpP protease is universal and highly conserved in both prokaryotes and eukaryotic organelles. Its activity depends on other adapter proteins such as ClpA or ClpX AAA⁺ ATPase for substrate recognition (Bhandari et al., 2018; Figaj et al., 2019). The authors showed that clpA but not clpX is required for enhancing susceptibility to A. tumefaciens T6SS killing, suggesting the involvement of ClpAP. ClpP variants deficient in ClpP protease activity or incapable of interacting with its adaptor protein could not restore T6SS effectiveness against a clpP knockout mutant, suggesting that ClpA–ClpP interaction and subsequent proteolysis are critical in enhancing susceptibility to T6SS killing. While the mode of action of recipient ClpAP complex involved in enhancing T6SS killing remains unknown, three hypotheses could be proposed for further testing. First, ClpAP complex may be used to enhance toxin activity, such as the Tde1 and/or Tde2 DNase activity, the major T6SS antibacterial weapons of A. tumefaciens strain C58 (Ma et al., 2014) used for the screen. Second, ClpAP complex could be hijacked by A. tumefaciens to trap or degrade an E. coli defense protein from inhibiting the activity of an incoming toxin. The third hypothesis is that the absence of a ClpAP system may result in substrate accumulation that interferes with T6SS firing or toxin activity of the attacker.

In summary, based on the broad spectrum of recipient cells that T6SS toxins act on, T6SS appears not to require recipient receptor for protein toxin entry. Current evidence suggests that specific RS factors may rather be used for the full activation of T6SS toxins once entering the recipient cells (Figure 5). However, future studies on the mode of action of identified RS factors and more comprehensive genetic screens are required to answer these questions. Besides RS factors, recent studies have revealed the presence of immunity-independent resistance in recipient cell that were nicely reviewed by Robitaille et al. (2020). These recipient defense factors or mechanisms include physical barriers such as exopolysaccharide (Toska et al., 2018), envelope stress responses (Hersch et al., 2020), or peptidoglycan editing (Le et al., 2020). Growing evidence of the involvement of recipient factors in either enhancing T6SS toxicity or defense against T6SS indicates an evolutionary arms race during interbacterial competition, which may play roles in shaping microbiome.

The major roles of RS factors are in recognition, entry, or activation of the toxins. Thus, an approach to identify recipient factors is to search for toxin-interacting proteins. Indeed, EF-Tu, the common RS factor involved in CDI and potentially for T6SS, was identified as one of the components taking part in the toxin–immunity protein complexes (Jones et al., 2017). Thus, co-expression of toxin–immunity complex followed by co-immunoprecipitation or pulldown assay (Brymora et al., 2004; Kaboord and Perr, 2008; Lin and Lai, 2017a) can lead to the discovery of toxin-interacting proteins. This serves as a straightforward method to identify RS factors that may play a role in toxin entry or activation. In addition, the toxin proteins can be used as a bait in well-established protein–protein interaction platforms such as bacterial two-hybrid (BTH) (Battesti and Bouveret, 2012) or yeast two-hybrid (YTH) (Mehla et al., 2015; Lin and Lai, 2017b) to identify potential RS factors by screening a recipient genomic library.

RS factors that are hijacked to activate toxin activity can be identified based on the knowledge of the toxin’s mode of action. For example, periplasmic disulfide bond-forming protein DsbA that is known to be required for folding or stabilization of proteins located in the periplasm could be critical for activity of periplasmic bacterial toxins such as peptidoglycan hydrolases and phospholipase (Kadokura and Beckwith, 2010). Based on this knowledge, the role of DsbA in T6SS-mediated antibacterial activity of S. marcescens was investigated and found to be required for the activity of the peptidoglycan hydrolase Ssp2 and the periplasmic toxin Ssp4 (Mariano et al., 2018). It is possible that DsbA plays a broader role for toxin activation delivered by multiple antibacterial systems. Besides DsbA, involvement of the ClpAP protease in T6SS susceptibility (Lin et al., 2020) also suggested that various types of proteases may be used for activating toxin activity by either cleaving full-length toxin proteins into more active truncated forms or degrading proteins that may inhibit toxin activity. A recent report showed that self-cleavage at both the N- and C- termini of an Rhs-family T6SS toxin TseI is not required for secretion but critical for its toxin activity (Pei et al., 2020). This finding also suggests that protease cleavage could be a strategy used for toxin activation in the recipient cell. Since the mechanism for N-terminal cleavage of TseI remains unknown, it may be mediated by an unknown protease residing in the recipient cell. Future work to test these potential modifying enzymes in activating antibacterial toxins of various systems shall shed light to understand the molecular basis of toxin action once they are translocated into the recipient cells.