Structural modeling reveals viral proteins that manipulate host immune signaling

Abstract

Immune pathways that use intracellular nucleotide signaling are common in animals, plants and bacteria. Viruses can inhibit nucleotide immune signaling by producing proteins that sequester or cleave the immune signals. Here we analyzed evolutionarily unrelated signal-sequestering viral proteins, finding that they share structural and biophysical traits in their genetic organization, ternary structures and binding pocket properties. Based on these traits we developed a structure-guided computational pipeline that can sift through large phage genome databases to unbiasedly predict phage proteins that manipulate bacterial immune signaling. Numerous previously uncharacterized proteins, grouped into three families, were verified to inhibit the bacterial Thoeris and CBASS signaling systems. Proteins of the Sequestin and Lockin families bind and sequester the TIR-produced signaling molecules 3'cADPR and His-ADPR, while proteins of the Acb5 family cleave and inactivate 3'3'-cGAMP and related molecules. X-ray crystallography and structural modeling, combined with mutational analyses, explain the structural basis for sequestration or cleavage of the immune signals. Thousands of these signal-manipulating proteins were detected in phage protein databases, with some instances present in well-studied model phages such as T2, T4 and T6. Our study explains how phages commonly evade bacterial immune signaling, and offers a structure-guided analytical approach for discovery of viral immune-manipulating proteins in any database of choice.

Figure 1.. Proteins with PF21825 domain inhibit type I Thoeris defense.

(A) A fusion protein of Tad2 and Acb4, found in Pasteurella phage Pm86. Each of the Tad2 and Acb4 domains was modeled separately by AlphaFold3 as a homotetramer (Acb4, ipTM = 0.91; Tad2, iPTM = 0.90), and the two models were overlayed. Dashes represent unmodeled linkers between Acb4 and Tad2 domains. (B) A fusion protein of Tad2 and a PF21825 domain, detected in a mobile genetic element in Comamonas sp. 26. Model of a homotetramer, ipTM=0.74. (C) A fusion protein of PF21825 domain and Acb4 from Pseudomonas phage YH6. Model of a homotetramer, ipTM=0.74. (D) Growth curves of B. subtilis cells expressing the type I Thoeris system alone (red), or co-expressing the Thoeris system and a gene with PF21825 pfam domain (A-M represent the 13 tested PF21825 genes, shades of turquoise), or a control strain without a defense system (black), infected by phage SBSphiJ at an MOI of 0.01. Each curve is the average of three replicates, with error bars indicating standard deviation. (E) A representative PF21825 gene (SequestinA, name explained below) that cancels Thoeris defense. Shown are tenfold serial dilution plaque assays, comparing the plating efficiency of phage SBSphiJ on bacteria that express the type I Thoeris defense system alone, or co-expressing the Thoeris system with SequestinA. The control strain lacks the system and expresses GFP instead. Images are representative of three replicates. Data for three replicates for each of the Sequestin genes are presented in Figure S1G. (F) Phages engineered to express PF21825 genes overcome Thoeris defense. Shown are data for wild-type SBSphiJ phage, as well as SBSphiJ knocked-in with members of the PF21825 Sequestin family. Data represent PFUs per milliliter of phages infecting control cells (no system), or cells expressing the type I Thoeris system. Average of three biological replicates with individual data points overlaid.

Figure 2.. Sequestin is a family of viral sponges that bind and sequester 3′cADPR.

(A) Sequestin proteins prevent accumulation of 3′cADPR in Thoeris-expressing infected cells. Cells expressing ThsB from B. cereus MSX-D12, or co-expressing both ThsB and genes from the Sequestin family, were infected with the phage SBSphiJ at a multiplicity of infection of 10. Cell lysates were extracted before infection (t=0) and 105 mins after infection, and were filtered to retain small molecules. Filtered lysates were then incubated with ThsA, and the NADase activity of ThsA was measured using a nicotinamide 1,N6-ethenoadenine dinucleotide (εNAD) cleavage fluorescence assay. Bars represent the mean of three experiments, with individual data points overlaid. Control represents experiments with lysates from cells expressing GFP together with ThsB. (B) Size-exclusion chromatography of SequestinA in apo state or following incubation with 3′cADPR. Light absorbance values at 260 nm and 280 nm are shown. (C) HPLC analysis of 3′cADPR incubated with either buffer (control, top lane) or with purified SequestinA (middle lane) at 1:1 ratio. Bottom lane shows the molecule release after the molecule-bound SequestinA was denatured by chloroform. (D) Structure of the AlphaFold3-predicted complex formed by a homodimer of SequestinA, together with 3′cADPR. Cartoon and surface representations are shown. (E) Close-up view of the predicted interactions of 3′cADPR with SequestinA from the model shown in panel E, highlighting key interacting residues. (F) Effect of point mutations in SequestinA and SequestinB on the ability of the sponge to cancel the defensive activity of type I Thoeris. Data represent PFU per milliter of SBSphiJ phage infecting negative control cells, cells expressing Thoeris, or cells co-expressing Thoeris and a sponge variant. Average of three replicates with individual data points overlaid.

Figure 3.. A structure-guided computational and experimental pipeline for the discovery of viral proteins that manipulate immune signaling.

(A) Crystal structures of Tad1, Tad2, and Acb2 (PDB: 7UAW, 8SMG, 8IY2) and AlphaFold3-predicted structures for Acb4 (uniport accession: A0A0B3RSW4, ipTM = 0.93), and SequestinB (ipTM = 0.94), together with their respective ligands. Cross-sections into the proteins are presented, with surface as an electrostatic map in a blue-red scale. Blue blebs represent positively charged pockets within the proteins. A dimer unit of the Tad1 hexamer is presented for clarity. (B). A computational pipeline to predict viral proteins that interact with host immune signals. The phage protein space is clustered, and representative proteins are selected from clusters of short proteins of unknown function. High-scoring AlphaFold-Multimer predictions for homo-oligomers are analyzed by CastP and Autosite to identify pockets with sizes typical to those found in known sponge proteins. Proteins with positively charged pockets are further evaluated as possible new anti-defense proteins. (C) Schematic representation of the experimental set up. Liquid cultures of bacteria harboring the defense system are grown in a 96 well plate format. Each well is transformed with a different plasmid that expresses a pair of anti-defense candidates, with antibiotics added to select for cells that acquired the plasmid. Bacteria are subsequently infected with phages in a low MOI, and OD is measured to identify successful phage infections. (D) An example of an infection assay described in (C). Cells expressing the type I Thoeris system from B. cereus MSX-D12 were infected with phage SBSphiJ at MOI 0.01. Each curve represents growth of bacteria transformed with a pair of candidate anti-defense genes, with negative control cells transformed with a plasmid expressing GFP (black), and positive control cells transformed with a plasmid expressing Tad1 (green). Purple curves represent cases where the transformed plasmid allowed the phage to propagate and cause eventual culture collapse, suggesting that one of the genes in the transformed plasmid inhibited Thoeris defense.

Figure 4.. Lockin is a family of phage proteins that inhibit Thoeris defense.

(A) Structure of the LockinA homo-hexamer, predicted by AlphaFold2-Multimer (ipTM = 0.78) in cartoon representation. (B) Cross sections in the LockinA structure model showing the void space pockets predicted to be formed within the protein. Top and side cross sections are shown on the lea and right, respectively. Surface representation is displayed as an electrostatic map (blue-red scale for positive and negative charge, respectively). Blue blebs are positively charged pockets within the protein. (C) A close-up view on one of the positively charged pockets, showing that it is formed in the interface between protomers. (D) Growth curves of B. subtilis cells expressing type I Thoeris from B cereus MSX-D12 (red), or co-expressing type I Thoeris and a Lockin protein (shades of purple), or control cells expressing GFP instead of the Thoeris system (black). Cells were infected with SBSphiJ phage at an MOI 0.01. Results of three experiments are presented as individual curves. (E) A representative Lockin protein (LockinA) capable of overcoming type I Thoeris defense. Shown are tenfold serial dilution plaque assays, comparing the plating efficiency of phage SBSphiJ on bacteria that express the type I Thoeris alone or with LockinA. Images are representative of three replicates. (F) Plating efficiency of phage SBSphiJ on negative control cells, cells expressing type I Thoeris, and cells expressing both type I Thoeris and a Lockin gene. Data represent PFUs per milliliter, and bars show the average of three replicates with individual data points overlaid. (G) Same experiment as in panel E, but with type II Thoeris from B. amyloliquefaciens Y2. (H) Phages engineered to express Lockin genes overcome defense of types I and II Thoeris. Shown are data for wild-type SBSphiJ phage, as well as SBSphiJ knocked-in with members of the Lockin family. Data represent PFUs per milliliter of phages infecting control cells (no system), or cells expressing a type I or II Thoeris system. Average of three biological replicates with individual data points overlaid. (I) Lockin proteins prevent accumulation of 3′cADPR in Thoeris-expressing infected cells. Legend is as in Figure 2A.

Figure 5.. Lockin is a sponge protein that sequesters Thoeris signaling molecules.

(A) Size-exclusion chromatography of apo state or 3′cADPR-bound LockinA. 3′cADPR-bound LockinA shows a substantial shift compared to LockinA in the apo state. (B) HPLC analysis of 3′cADPR incubated with either buffer (control, top lane) or with purified LockinA (middle lane) at 1:1 ratio. Bottom lane shows the molecule release after the molecule-bound Lockin was denatured in 98°C for 10 minutes. (C) Crystal structure of LockinA in complex with 3′cADPR. (D) Detailed view of LockinA residues that interact with the 3′cADPR adenine base and diphosphate backbone. Green dashed lines indicate hydrogen bonding interactions. Subscript denotes individual protomers within the Lockin hexamer. (E) A 2D map presenting a detailed view of 3′cADPR adenine base and diphosphate backbone and their LockinA residues interactions. (F) Effect of point mutations in LockinA on the ability of the sponge to cancel the defensive activity of type I Thoeris. Data represent PFU per milliter of SBSphiJ phage infecting negative control cells, cells expressing Thoeris, or cells co-expressing Thoeris and a sponge variant. Average of three replicates with individual data points overlaid.

Figure 6.. Acb5 proteins inhibit CBASS signaling.

(A) Growth curves of E. coli cells expressing either a genomically-integrated CBASS system from E. albertii MOD1-EC1698 (red), co-expressing CBASS and Acb5 proteins (shades of orange), or negative control cells lacking the system (black), infected with T2 phage at an MOI 0.001. Results of three experiments are presented as individual curves. (B) Acb5 is capable of overcoming CBASS defense. Shown are tenfold serial dilution plaque assays, comparing the plating efficiency of T2 on bacteria that express the CBASS defense system, or bacteria co-expressing CBASS with Acb5 proteins, or a control strain that lacks the system. Images are representative of three replicates. (C) Plating efficiency of phage T2 and phage Bas60 on negative control cells, cells expressing CBASS, and cells expressing both CBASS and an Acb5 gene. Data represent PFUs per milliliter, and bars show the average of three replicates with individual data points overlaid. (D) Acb5 prevents cGAMP accumulation in infected cells. Cells expressing the CD-NTase enzyme of the E. albertii CBASS, or co-expressing the CD-NTase and Acb5A, were infected with the phage Bas60 at a multiplicity of infection of 10. Cell lysates were obtained before infection (0 mins) and during infection (30 min), and filtered to retain small molecules. cGAMP concentration in filtered lysates was measured using a 3′3′-cGAMP ELISA kit, and cGAMP concentration per cell was calculated based on estimated cell counts. Cells co-expressing CBASS and Acb2 were used as positive control. Bars represent the mean of three experiments, with individual data points overlaid. Limit of detection is presented as a dashed line. (E-F) AlphaFold3-predicted tetramer complex for Acb5A (ipTM = 0.89) and Acb5B (ipTM=0.88). Protomers are presented in different shades of orange. (G) Cross section in the AlphaFold3-predicted tetramer complex for Acb5A with surface displayed as an electrostatic map (blue-red scale for positive-negative charges). Blue blebs represent positively charged pockets within the protein.

Figure 7.. Acb5 is an enzyme that cleaves CBASS signaling molecules.

(A) Thin-layer chromatography (TLC) analysis of Acb5A nuclease activity. Recombinant Acb5 was incubated with α³²P-radiolabeled 3′3′-cGAMP, and degradation products were visualized by TLC. Purified Acb1 (an enzyme that cleaves 3′3′-cGAMP) and Acb4 (a 3′3′-cGAMP sponge) are included as controls. Data are representative of at least three independent experiments. (B) HPLC analysis of synthetic 3′3′-cGAMP. Y axis represents light absorbance at 260nm. (C) HPLC analysis of degradation products of 3′3′-cGAMP following 30 minutes incubation with purified Acb5A. M/Z values, as measured in mass spectrometry, are indicated for each product. Data are representative of 3 independent experiments. (D) HPLC analysis of synthetic 2′3′-cAMP and 2′3′-cGMP standards. (E) HPLC analysis of synthetic cUA. (F) Degradation products of 3′3′-cUA following 60 minutes incubation with purified Acb5A. (G) AlphaFold3-predicted tetramer complex for Acb5A with 2′3′-cAMP and 2′3′-cGMP (ipTM = 0.87, pTM=0.90), colors represent pLDDT scores. (H) Close-up view of the predicted interactions of 2′3′-cAMP and 2′3′-cGMP with Acb5A from the model shown in panel G, highlighting key interacting residues. (I) Effect of point mutations in Acb5A on the ability of the enzyme to cancel the defensive activity of CBASS. Data represent PFU per milliter of T2 phage infecting negative control cells, cells expressing CBASS, or cells co-expressing CBASS and an Acb5A variant. Average of three replicates with individual data points overlaid.