GajA is activated upon sensing NTP depletion by transcription and requires assistance from GajB for antiviral defense
The Gabija system consists of two components, GajA and GajB (Fig. 1A), and confers defense against phages (9). To detect the necessary components of the Gabija system, we cloned the Gabija gene cassette sequence (GajAB, located at 94190–97412 of the B. cereus VD045 genome [AHET01000033]), the GajA gene alone, the GajB gene alone, or both genes under the control of separate promoters (GajA + B) into the host strain E. coli B (ATCC® 11303™), which naturally lacks the Gabija system. We then challenged the transformed strain with three coliphages and found that the system conferred protection against all three typical types of lytic phages: T7 (Podoviridae), T4 (Myoviridae), and T5 (Siphoviridae) (Fig. 1B–D). The Gabija system exhibited extensive phage resistance, showing very strong resistance to phages T7 and T4 (> 107 fold decrease in efficiency of plating [EOP]) and weaker resistance to phage T5 (~ 103 fold decrease in EOP) (Fig. 1B–D). In a previous report, the two-gene system together with its flanking intergenic regions (located at 93871–97763 of the B. cereus VD045 genome [AHET01000033]) showed defense against phages (9). Our data further confirmed that only the coding region is essential for defense against phages. Both GajA and GajB genes in the system appear to be essential for its functionality, because deletion of either gene resulted in loss of protection from phage infection (Fig. 1B–D). Surprisingly, we found that GajA + B, which expresses both GajA and GajB but under separate promoters, exhibited no phage resistance (Fig. 1B–D). Therefore, we examined the protein expression of bacteria containing various Gabija plasmids. The results showed that the expression of GajA was significantly higher than that of GajB in the native GajAB construction, while the expression level of GajA and GajB was almost equal in the GajA + B construction (Figure S1), which suggested that the ratio between GajA and GajB is vital for antiviral function. Subsequently, we investigated the cytotoxicity of various Gabija proteins. The bacterial growth curves indicated that Gabija protein has no obvious cytotoxicity to host bacteria in the absence of phage infection (Figure S2).
We examined the role of conserved residues of GajA and GajB in antiviral defense. In previous work, we characterized that GajA is a nucleotide-sensing DNA nicking endonuclease (Fig. 1E), utilizing a TOPRIM domain to cleave DNA and an ATPase-like domain to mediate the regulation by nucleotide concentration (21). Therefore, we constructed three point mutations, E379A, D511A, and K541A, in the TOPRIM domain and two point mutations, K35A and H320A, and a truncation (GajA-CTR, 1–347 amino acid deletion) in the ATPase-like domain, introduced these point mutations into the GajAB gene cassette, and tested their effects on phage defense. We found that all of the above mutations abolished the antiphage function of the Gabija system (Figure S3). E379A, D511A, and K541A have been shown to inactivate the DNA cleavage of GajA and H320A affects the NTP inhibition of GajA (21); however, the K35A mutation, without any observed effect on the DNA cleavage activity of GajA, also abolished the defense function of the whole system. In order to target conserved amino acid residues of GajB, sequence alignment was performed between GajB, DrUvrD, PcrA, EcUvrD, and Rep (Figure S4). Double mutations of the key residues of UvrD helicases in GajB K23A/T24A or D162A/E163A resulted in loss of phage resistance (Figure S5). These results suggest that specific activities of both GajA and GajB are essential for the Gabija defense system.
ATP regulates DNA cleavage activity of GajA by inhibiting its specific DNA binding
Our previous study demonstrated the inhibition of DNA cleavage activity of GajA by NTP (21), but the underlying mechanism was unclear. In this work, we detected the DNA binding of GajA in the absence or presence of ATP or AMP-PNP. In order to avoid substrate cleavage, we replaced Mg2+ with Ca2+ in the reactions to detect the DNA binding of GajA. The electrophoretic mobility shift assay (EMSA) demonstrated the binding of GajA to a DNA substrate containing two overlapped GajA recognition sites AATAACCCGGTTATT (one recognition site in the plus strand shown in bold and the other in the minus strand shown in italic), which was inhibited by ATP and AMP-PNP (Fig. 2A). The DNA binding of GajA was sensitive to the ATP concentration and completely inhibited by 1 mM ATP (Fig. 2B). In the presence of GajA recognition sites, both Mg2+ and Ca2+ supported the specific DNA binding of GajA while only Ca2+ supported a non-specific DNA binding of the GajA in the absence of recognition sites (Fig. 2C). Mg2+ only mediates the recognition site-dependent DNA binding of GajA, which was further verified with a DNA substrate amplified from the pUC19 plasmid containing no GajA recognition sequence (Fig. 2D). When the substrates containing two overlapped recognition sites were cut into two fragments in the presence of Mg2+, GajA was still bound to the cleaved DNA fragments (Fig. 2E), which is consistent with previously reported inefficient turnover of GajA (21). The GajA-bound DNA appears to be a clear gel band shift rather than smear, indicating a specific stoichiometry of GajA/DNA in the complex.
Subsequently, we detected the DNA binding activity of GajA and its mutants in the absence or presence of ATP. K35A, H320A, and D511A mutations decreased the inhibition of GajA DNA binding activity by ATP. K541A mutation enhanced the DNA binding of GajA and changed the gel shift of the GajA/DNA complex, indicating that the mutation altered the binding mode or the stoichiometry in the complex. Like that for the WT, ATP inhibited the DNA binding of all of the GajA mutants, except for E379A (Figure S6A). GajA-E379A was peculiar in that its DNA binding resulted in a wider range of gel shifts and was not inhibited by ATP (Figure S6A). Further, we found that the DNA binding of GajA-E379A was not dependent on recognition sites and metal ions (Figure S6B). The non-specific DNA binding of GajA-E379A indicated that the E379 site is vital to determine the specificity of GajA for recognition sites.
GajB forms a complex with GajA in vitro and in vivo
To investigate the relationship between GajA and GajB, we constructed the predicted coding sequence (AHET01000033.1: 94,190–97,412, sequence listed in Table S1) of GajAB (WT Gabija gene cassette). In this construction, GajA was fused to an N-terminal 6×His-tag. The recombinant proteins were purified on a Ni-NTA agarose column. Interestingly, non-tagged GajB was co-purified along with His-tagged GajA (Figure S7A). We also constructed another vector GajA + B (His-tagged GajA and non-tagged GajB overexpressed under the control of two individual promoters). In this case, GajB was also co-purified with GajA (Figure S7B). To verify the in vivo results, we performed in vitro assembly of the GajA/B complex. The predicted coding sequence for GajA or GajB alone was cloned into pET28a vectors and individual protein was overexpressed, respectively. The supernatants containing His-tagged GajA and non-tagged GajB were mixed before loading onto a Ni-NTA agarose column. GajB was again co-purified with GajA (Figure S7C). Collectively, these data demonstrated a stable binding between GajB and GajA.
Two forms of GajB
By careful comparison of purified proteins from the GajAB and GajA + B constructs as described above, we unexpectedly found that the purified proteins from the GajAB construct contained two forms of GajB of close sizes (Fig. 3A). The smaller form is similar in size to the GajB from the GajA + B construct (that is, GajB encoded by the predicted sequence), while a slightly bigger form of GajB (hereafter named GajBʹ, confirmed by liquid chromatography–mass spectrometry, Figure S8) raised the possibility of a colinear translation starting in the coding sequence of GajA. N-terminal sequencing of GajBʹ (Fig. 3B) demonstrated that GajBʹ possesses a five-amino acid N-terminal extension compared to GajB (indicated by a blue box in Fig. 3B). Together, these results verified that the Gabija system encoded by the B. cereus VD045 gene cassette contains three protein components: a GajA endonuclease and two colinear GajB proteins (hereafter named GajBʹ and GajB, respectively) (Fig. 3C). For further confirmation, we constructed two mutants of the WT GajAB cassette: GajAB-GCG, in which the GajBʹ initiation codon GTG was changed to GCG, and GajAB-rbsATC, in which the GajB initiation codon ATG was mutated to ATC and its ribosome binding site (RBS) region was modified to block the expression of GajB (Figure S9). As expected, the GajAB-GCG mutant expressed GajA and only GajB but not GajBʹ, and the GajAB-rbsATC mutant expressed GajA and only GajBʹ but not GajB (Fig. 3D). The mutations in GajA and GajB investigated in this work did not change the expression of the GajAB gene cassette (Fig. 3D).
GajB senses DNA termini for ATP hydrolysis
Bioinformatics analysis predicted that GajB is a UvrD-like helicase. To elucidate its function, GajB and GajBʹ (> 90% homogeneity) were purified as N-terminal His-tagged proteins (Figure S10A). With ATP as substrate, we first examined the effect of divalent cations on the hydrolytic activity of GajBʹ. Hydrolysis activity was detected and quantified by the PiColorLock™ phosphate detection system kit (Expedeon) by calculating the release of Pi relative to the standard curve. In the pre-experiment, GajBʹ exhibited much stronger ATP hydrolysis activity than GajB. Thus, we used GajBʹ to establish the optimal reaction conditions. With a 56-nt ssDNA (56-nt ssDNA-F in Table S3) in the reaction, GajBʹ exhibited strong hydrolysis activity in the presence of Mg2+ and weak activity in the presence of Mn2+ and Ca2+, while Zn2+, Co2+, and Ni2+ did not support the ATP hydrolysis activity of GajBʹ (Figure S10B). The optimal metal ion concentration for GajBʹ ATPase activity is 10 mM for Mg2+ (Figure S10C) and 2 mM for Mn2+ (Figure S10D). The optimal temperature is 30°C (Figure S10E) and the optimal pH is 7 (Figure S10F). GajBʹ hydrolytic activity was inhibited by NaCl or KCl (Figure S11). Therefore, the optimal reaction condition for GajBʹ ATPase activity is established as 20 mM Tris-HCl pH 7, 10 mM MgCl2, and 1 mM DTT, at 30°C.
The ATPase activity of GajBʹ is DNA-dependent. By comparison of the effects of various types of DNA or RNA (all at the same mass in each reaction) on the ATPase activity of GajBʹ, including short dsDNA (56 bp, linear), short ssDNA (56 nt, linear), T7 DNA (~ 40 kb, linear), pUC19 plasmid DNA, (~ 2.7 kb, circular), M13 ssDNA (~ 8k nt, circular), PCR-amplified DNA (955 bp, linear), ssRNA (600 nt, linear), and dsRNA (600 bp, linear) (sequences are listed in Table S4), it was clear that GajBʹ is only significantly activated by short dsDNA and ssDNA (Fig. 4A), indicating that the ATPase activity of GajBʹ is specifically activated by DNA termini (at the same mass, there are more termini of short DNA than of longer DNA). Further comparison revealed that short ssDNA is more effective than short dsDNA to activate the ATPase activity of GajBʹ (Figure S12). The stimulation is DNA sequence-independent, as complementary ssDNA showed similar stimulation effect (Figure S12).
Among tested NTP and dNTP substrates, GajBʹ showed high preference for ATP, GTP, dATP and dGTP to hydrolyze (Fig. 4B). In contrast, GajB, missing the N-terminal five amino acids compared to GajBʹ, showed only minimal NTP hydrolysis activity (Fig. 4B).
The only difference between GajBʹ and GajB is the five-amino acid N-terminal extension; however, their difference in NTPase activity is dramatic. To explore the role of the N-terminal five amino acids of GajBʹ, we mutated the two glutamine residues to alanine (E3A/E5A) in this region, because the two similar glutamine residues were proved to be the key residues in another hydrolytic enzyme (32). Double mutations in the conserved sites of the UvrD family, K23A/T24A and D162A/E163A, were also constructed. These GajBʹ mutants were purified using the same procedure as that for the WT protein (Figure S13). K23A/T24A and D162A/E163A mutations completely abolished the ATP hydrolysis activity of GajBʹ (Figure S13), indicating a common mechanism between the GajBʹ and UvrD family helicases in ATP hydrolysis. However, the E3A/E5A mutations showed no effect (Figure S13), indicating that the two glutamines are not directly involved in the NTP hydrolysis activity of GajBʹ.
The efficiency of ATP/GTP hydrolysis by GajBʹ was analyzed by TLC. GajBʹ is an efficient ATPase in the presence of ssDNA; 3 µM GajBʹ completely hydrolyzed 4 mM ATP into ADP in 15 min (Figure S14). Similarly, GTP was hydrolyzed by GajBʹ to produce GDP (Fig. 4C). We examined the ssDNA in ATP hydrolysis reactions by Native-PAGE and found that the ssDNA was neither stably bound nor processed by GajBʹ under the optimal conditions for ATP hydrolysis (Figure S15).
We also investigated the ATPase activity of the Gabija protein complex by measuring the ATP hydrolysis activity of the following samples: GajA, GajB, GajBʹ, GajA/B 1:1 complex, GajA/Bʹ 1:1 complex, and native GajAB complex (GajA:B:Bʹ ~4:1:1). The results showed that GajBʹ exhibited the strongest activity and GajA + Bʹ and GajAB showed less activity, likely contributed by GajBʹ in the complexes (Fig. 4D and Figure S16B). The TLC analysis results were similar (Figure S16A, B). The GajA mutations K35A and H320A had no obvious effects on GajAB ATPase activity, while the GajBʹ mutations K23A/T24A and D162A/163A completely abolished the activity of GajAB (Fig. 4E). GajAB-GCG (blocked the expression of GajBʹ) and GajAB-rbsATC (blocked the expression of GajB and introduced mutations E3A/D4A/E5K in GajBʹ, Figure S9B) displayed weak ATPase activity (Fig. 4E).
GajB and GajBʹ were initially suspected to function as helicases based on their homology to UvrD helicases (9). However, in a classic helicase unwinding assay using UvrD as a positive control, GajBʹ displayed no DNA unwinding activity (Fig. 4F). Considering GajBʹ may coordinate with GajA and load onto the DNA nicks introduced by GajA, we further employed a nicked dsDNA substrate by annealing three oligonucleotides (Figure S17 and Table S3). However, still no unwinding activity from GajBʹ was detected (Figure S17).
Composition of the functional Gabija complex
To gain further insight into the regulatory mechanism of the Gabija system, we performed extensive genetic experiments. First, GajBʹ alone exhibited no phage resistance (Figure S18). Under the control of separated promoters and with similar expression levels, GajA and GajB (GajA + B) conferred no resistance to phages (Fig. 5A, B); while GajA plus GajBʹ and GajB (GajA + Bʹ) or GajA plus GajBʹ (GajA + Bʹ-ATC, the GajB initiation codon ATG mutated to ATC) showed slight resistance to phages T7 and T4 (< 102 in EOP) (Fig. 5A, B and Figure S18). If GajBʹ was removed by mutating its initiation codon GTG to GCG, the combination of co-expressed GajA and GajB (GajAB-GCG) almost lost its antiviral function against phage T4 (Fig. 5B), and its resistance to phage T7 decreased significantly (~ 105 in EOP, comparing GajAB-GCG and GajA in Fig. 5A). In contrast, removal of GajB by mutating its initiation codon ATG to ATC from the native Gabija complex (GajAB-ATC) has no obvious effect on its antiviral function (Fig. 5A, B). Modification of the RBS region of GajB in addition to mutating its initiation codon (GajAB-rbsATC) resulted in decreased resistance to phage T4 (Fig. 5A, B), likely due to its effect on GajBʹ expression level. A “T” insertion in front of the GajB initiation codon ATG (GajAB + T), which resulted in a frameshift and an early stop codon in GajBʹ, caused a decrease in phage resistance (Fig. 5A, B). A single “T” deletion in the GajA stop codon TGA (GajAB-T) resulting in the removal of GajBʹ and a fusion protein of GajA and GajB inactivated the Gabija system (Figure S18). The detailed mutation forms of the above mutants are shown in Figure S9 and Figure S19.
The above data indicate that GajBʹ is the functional component in the Gabija system, while GajB is less effective and redundant in supporting Gabija defense, which is consistent with their divergence in DNA terminus-dependent NTPase activity. The indispensability of GajBʹ explains the reason why separately expressed GajA and GajB failed in phage resistance: because of the loss of GajBʹ. However, it was intriguing that even the presence of both GajA and GajBʹ if expressed separately at similar levels (each under the control of the same promoter and RBS, respectively) almost failed in conferring phage resistance (Fig. 5A, B, GajA + Bʹ). In the functional native GajAB gene cassette, the translation of GajBʹ is regulated by the native RBS residing in the GajA coding sequence; therefore, its expression level should be lower than that of GajA, which is regulated by the optimized RBS in the expression vector. Thus, we speculated that the ratio between GajA and GajBʹ is vital for the function of the Gabija system. To verify the hypothesis, we created new constructs to modify the ratio between Gabija components. In constructs GajAB + A, GajAB + B, and GajAB + Bʹ, extra GajA, GajB, and GajBʹ was present (expressed under the control of an additional promoter) to the native GajAB cassette, respectively. GajAB + A with additional GajA was as efficient as GajAB in defense against phages. In contrast, GajAB + B with increased amounts of GajB completely lost phage resistance, and GajAB + Bʹ with increased amounts of GajBʹ almost lost phage resistance (Fig. 5A, B). The K35A mutation abolished the antiviral defense of the Gabjija system (Figure S3B); however, the non-functional K35AGajAB can be rescued by additional WT GajA (Fig. 5A, B). Taken together, these results imply that GajBʹ and GajB levels must be lower than the GajA level for Gabija functionality.
Unlike GajA, GajBʹ and GajB exhibited no significant binding to the pUC19-955 dsDNA containing GajA recognition sites (Figure S20A, B), even in the presence of ATP or AMP-PNP (Figure S20C, D). The binding to the 56-nt ssDNA and 56-bp dsDNA that trigger the ATPase activity by GajBʹ and GajB was also not detectable by EMSA (Figure S20E). Since GajBʹ and GajB form a complex with GajA, we examined the effects of GajBʹ or GajB on the DNA binding by GajA. The results showed that at a molecular ratio of GajA/GajBʹ = 1:0.3, GajBʹ enhanced the DNA binding of GajA (Fig. 5C); however, when level of GajBʹ was equal to or exceeded that of GajA, GajBʹ inhibited the DNA binding of GajA (Fig. 5C). At low molecular ratios, GajBʹ not only enhanced the DNA binding of GajA but also retarded the gel shift of bound DNA, indicating that GajBʹ was incorporated into the GajA/DNA complex (Fig. 5C). GajB showed a similar effect on the DNA binding of GajA as GajBʹ (Figure S20F). ATP or non-hydrolyzable AMP-PNP at 1 mM completely inhibited the DNA binding of GajA, even in the presence of GajBʹ (Fig. 5C), implying that the Gabija system was suppressed under normal conditions (Figure S2). Further, we compared the DNA binding of the GajA and GajAB native complex in the presence of ATP. We found that at low ATP concentrations (0.25 and 0.5 mM), GajAB bound DNA more tightly than GajA alone (Figure S20G), indicating that GajBʹ and GajB in the complex enhanced the DNA binding of GajA and attenuated the inhibition from ATP. The K35A and H320A mutations in the GajAB complex also attenuated the inhibitory effects of ATP on DNA binding (Figure S20H and Figure S21).
GajA with an N-terminal 6×His-tag has a molecular weight of about 69 kDa. In Native-PAGE, GajA runs as an evident band corresponding to about 276 kDa, indicating that GajA forms a tetramer (Fig. 5D). By contrast, GajB forms polymers with high molecular weight while GajBʹ tends to form polymers of various and gradually increasing molecular weight (Fig. 5D). When mixed, both GajB and GajBʹ interact with GajA to form large complexes, as shown near the top of the gel (Fig. 5D).
Finally, we examined the effects of GajBʹ on the DNA cleavage activity of GajA. The robust DNA nicking activity of GajA results in the digestion of T7 genomic DNA (Fig. 5E). In the absence of ATP, addition of GajBʹ at various molecular ratios inhibited the cleavage activity of GajA (Fig. 5E). However, in the presence of 0.5 mM ATP, GajBʹ supplied at a molecular ratio to GajA lower than 1:1 significantly stimulated the cleavage activity of GajA (Fig. 5E), implying the key role of GajBʹ in assisting GajA activation when cellular NTP is not completely depleted by phage invasion.
T7 phage suppressors overcome Gabija defense
To further elucidate the mechanism of the Gabija system, we attempted to find phage mutants that escape Gabija defense. Phage T7 was not able to survive the WT Gabija system. However, after six generations of infection and selection in hosts containing the GajAB + T system with an interrupted GajBʹ (Figure S19B), three suppressors of phage T7 were obtained. These phage mutants could partially overcome (> 100-fold increase in EOP compared to WT T7) the defense from GajAB + T and GajAB-rbsATC, which contain modified GajBʹ, but fail to overcome GajAB and GajAB-ATC, which contain normal GajBʹ (Figure S22). We then sequenced the full genome of each of the phage mutants and compared the resulting sequences to the sequence of the WT T7 phage. The common mutation in these suppressors is a single base insertion resulting in a frameshift in T7 gene 5.5 (Figure S22). These results suggest that disruption of the T7 gene 5.5 protein (gp5.5) enables the phage to partially overcome the Gabija defense when normal GajBʹ is absent.
A previous study showed that T7 gp5.5 binds to tRNA to recruit the E. coli nucleoid protein H-NS (33). H-NS alone inhibits reactions involved in DNA replication, but binding to the gp5.5–tRNA complex abolishes this inhibition (33). Thus, in our selected Gabija suppressors the destruction of T7 gp5.5 should release complexed H-NS, which may affect the function of the Gabija system. Therefore, we tested the in vitro effect of H-NS on GajBʹ ATPase activity and GajA DNA cleavage activity. H-NS had no significant effect on GajBʹ hydrolytic activity (Figure S23A) but significantly inhibited GajA DNA cleavage activity (Figure S23B). Taken together, it is suggested that the binding to the DNA recognition site by GajA is the key to regulation in the Gabija antiviral mechanism, as nucleotides, GajBʹ, and phage suppressors all exert their regulation on the DNA binding of GajA. The inability of these phage suppressors to overcome GajBʹ suggested that the inhibitory effects of H-NS on GajA can be relieved by GajBʹ, further indicating the importance of GajBʹ in assisting GajA function.