A phage-encoded anti-CRISPR protein cooperates with host enolase to prevent type III CRISPR immunity

Abstract

CRISPR systems provide powerful adaptive immunity against phage infection. In response, phages can evolve anti-CRISPR (Acr) proteins to evade CRISPR immunity. The few type III Acrs identified so far exhibit conditional effectiveness in countering type III immunity or rely on unknown or poorly understood inhibitory mechanisms. Here we report the discovery of AcrIIIA2, a type III-A Acr encoded by Streptococcus thermophilus (Sth) phages. Biochemical and structural analyses reveal that AcrIIIA2 co-opts host enolase to form a specific ternary complex with the Sth type III-A (Csm) crRNP (CRISPR ribonucleoprotein complex), obstructing its immune responses. Enolase is a highly abundant glycolysis enzyme known to exhibit additional, moonlighting functions. The enolase-chaperoned AcrIIIA2 blocks the initial step of phage RNA binding, thereby preventing downstream type III immune responses including cOA (cyclic oligoadenylate) signaling and degradation of phage RNA and DNA. Mechanistically, AcrIIIA2 disrupts target RNA recognition by directly engaging the 3’ seed region of the crRNA while also interacting with Csm2 and Csm5 subunits of the Csm complex. Enolase participates in the anti-immune response by serving as an essential structural scaffold, stabilizing the Acr-CRISPR interactions. These findings uncover a new anti-defense strategy that exploits a well-conserved host factor to block CRISPR immunity.

INTRODUCTION

Bacteria and archaea have evolved diverse defense mechanisms to protect themselves against mobile genetic elements (MGEs) such as phages and plasmids^1,2. Among these, CRISPR systems are unique in providing adaptive immunity through the specific recognition and degradation of foreign nucleic acids^3,4.

Of the seven CRISPR system types identified^4,5, type III systems are the most complex and coordinate multiple enzymatic activities during immune responses⁶. Type III CRISPR systems are widespread in both bacteria and archaea and consist of multiple subtypes (III A–F) with III-A (Csm) and III-B (Cmr) being the most abundant and well characterized^7,8. A key feature of type III systems is their activation by crRNA-guided recognition of MGE RNA^9,10, which triggers a cascade of immune processes. These processes include target RNA cleavage by the Csm3/Cmr4 endoribonucleases¹¹, DNA degradation via the HD domain of Cas10 (Csm1/Cmr2)^12,13, and the synthesis of cyclic oligoadenylates (cOA) by the Palm domain of Cas10¹⁴. The cOA molecules then act as secondary messengers, activating auxiliary ribonucleases including Csm6 (III-A)¹⁵ and Csx1 (III-B)¹⁶, amplifying the immune response.

Structural and biochemical studies of type III-A and III-B systems have highlighted the importance of short sequences at the 3′ ends of crRNAs (referred to as “seed sequences”), which are essential for initiating interactions with target RNAs^17,18. Type III crRNA seed regions facilitate initial crRNA–target RNA interactions, exposing otherwise buried crRNA segments within the crRNP complex and enabling full base-pairing with target RNAs¹⁷. The strength and stability of this 3’ crRNA-target RNA interaction can influence the degree of target RNA interaction^19,20, ultimately affecting the activation of downstream nucleases and cOA signaling. The specific requirements for type III nucleases vary across type III systems and different conditions. For example, in some cases, phage immunity depends on both Cas10 DNase activity and cOA-mediated Csm6 RNase activity, particularly when targeting late-expressed phage genes^21,22. In other systems, Csm6 RNase activity alone is sufficient to confer immunity, regardless of the timing of target phage gene expression²³.

Despite their complex coordination of nuclease activities, type III CRISPR systems are under constant evolutionary pressure from MGEs that deploy anti-CRISPR (Acr) proteins to neutralize CRISPR immunity. While numerous Acrs targeting types I, II and VI CRISPR systems have been identified, and in some cases, extensively characterized^24,25, those targeting type III systems remain largely unexplored²⁶. Currently, only one type III-A specific Acr (AcrIIIA1)²⁷, two type III-B Acrs (AcrIIIB1 and AcrIIIB2)^22,28,29, and a type III Acr (AcrIII-1)³⁰ predicted to inhibit either type III-A or III-B systems have been identified.

Studies of type III Acrs have primarily focused on non-lytic viruses infecting archaeal hyperthermophiles of the order Sulfolobales^22,28–30. AcrIII-1, a viral protein, degrades cOA signaling molecules to prevent activation of Csm6/Csx1 RNases but does not inhibit Cas10 DNase activity, which remains sufficient in some type III systems to defend against viral infections³⁰. Similarly, AcrIIIB1 appears to target Csx1 RNase activity through uncharacterized, direct interactions with Cmr crRNP complexes, likely modulating cOA synthesis²². AcrIIIB2 interferes with the activity of Cmr crRNP complexes with conflicting evidence of effectiveness against early expressed phage genes^28,29. Interestingly, these findings suggest that the identified type III Acrs exhibit partial or conditional effectiveness, preventing immunity only under specific circumstances.

In contrast, AcrIIIA1, the only known type III-A Acr is encoded by certain staphylococcus phages and prevents type III immunity regardless of early or late phage gene expression²⁷. AcrIIIA1 inhibits the Csm crRNP complex through an unclear mechanism involving interactions with Csm2, t(m)RNA fragments, and the host RNase R nuclease²⁷. Of note, house-keeping ribonucleases including RNases J1, J2, R, E and PNPase, are important for full crRNA maturation and for maintaining robust anti-plasmid and anti-phage defenses, at least in staphylococcus bacterial species^31,32. The involvement of these non-Cas host proteins in type III immunity adds further complexity to the evolving effectiveness of type III Acrs, as their presence and function may vary across different organisms. Therefore, studying type III Acrs within their native hosts and exploring additional protein interactions may be essential for fully understanding mechanisms of type III CRISPR-Cas inhibition by Acrs.

In this study, we report the discovery and characterization of AcrIIIA2, a novel Acr protein encoded by Streptococcus thermophilus (Sth) phages. We demonstrate that AcrIIIA2 specifically inhibits the Sth type III-A CRISPR system, independent of phage gene expression timing. Remarkably, AcrIIIA2 selectively associates with Sth enolase, a highly abundant and conserved glycolysis enzyme responsible for conversion of 2-phosphoglycerate to phosphoenolpyruvate³³. Beyond the primary metabolic role, certain bacterial enolases are known to exhibit diverse moonlighting functions^34–38. In vivo purification and biochemical assays demonstrate that AcrIIIA2 and enolase form a stable subcomplex that inhibits the type III-A system by preventing target RNA binding, effectively shutting down all associated type III-A activities. The high-resolution cryo-EM structure of the ternary complex formed between the AcrIIIA2/enolase subcomplex and the Csm crRNP complex provides key insights into how AcrIIIA2 co-opts the chaperone function of enolase in blocking target RNA recognition through the prevention of the 3′ crRNA seed-target RNA interaction.

RESULTS

Discovery of a type III-A anti-CRISPR protein

Streptococcus thermophilus (Sth) phages encode multiple anti-CRISPR (Acr) proteins, including AcrIIA3³⁹, AcrIIA5⁴⁰, AcrIIA6⁴⁰, AcrIIA24, and AcrIIA25⁴¹. All previously identified Acrs in Sth phages specifically inhibit type II-A (Cas9-containing) CRISPR systems. However, type III-A systems are highly prevalent in Sth with 88% (232/263) of sequenced genomes containing a type III-A system, compared to 100% and only 9% harboring type II-A and type I-E systems, respectively⁴². We hypothesized that the sole identification of type II-A Acrs reflect the biased nature of the functional screens performed rather than absence of phage Acrs capable of blocking type III-A CRISPR immunity. As previously reported, our model Sth organism contains four functional, yet independent CRISPR systems: two type II-A, one III-A and one I-E^23,43. Using a guilt-by-association approach, we identified a gene downstream of the known AcrIIA6 in Sth phage SW3 as a potential type III Acr candidate (Fig. 1a). To test for Acr activity, we performed a phage-based CRISPR interference assay (Fig. 1b). In this assay, we used Sth CRISPR bacteriophage insensitive mutants (BIMs), each containing a naturally-acquired, functional spacer effectively targeting Sth phage 2972. These BIMs correspond to the four CRISPR systems: CR1 (type II-A), CR2 (type III-A), CR3 (type II-A), and CR4 (type I-E)⁴⁴. During infection with Sth phage 2972, BIMs are expected to survive due to CRISPR-mediated immunity. However, if the candidate Acr protein inhibits a specific CRISPR system, the corresponding BIM will undergo host cell lysis despite carrying a functional spacer (Fig. 1b). This assay allowed us to pinpoint which CRISPR system was targeted by the Acr candidate.

Figure 1. AcrIIIA2 was discovered from Sth phages.

a, Sth phage SW3 encodes a known anti-CRISPR (Acr) protein, AcrIIA6. Using a guilt-by-association approach, the gene located directly downstream of AcrIIA6 was identified as an Acr candidate. b, A phage-based interference assay was used to test the candidate’s Acr activity. The diagram illustrates the assay outcomes when a type III-A BIM is reinfected with Sth phage 2972. In the absence of an Acr, the type III-A crRNP complex recognizes and cleaves the invading phage genome, leading to normal cell growth. Conversely, if a type III-A Acr is present, the CRISPR system is inhibited, allowing phage infection to persist resulting in host cell lysis. This figure was created with Biorender.com. c, Either an empty vector or a plasmid encoding the candidate Acr was introduced into BIMs that had acquired a spacer against Sth phage 2972 within the CRISPR arrays CR1 (II-A), CR2 (III-A), CR3 (II-A), or CR4 (I-E). The BIMs were then challenged with phage at an MOI of 1, and growth was monitored by OD₆₀₀ over six hours. The graphs represent a single biological replicate for each, with the line indicating the mean of three technical replicates. Cell lysis occurred only in the type III-A BIM, identifying the candidate as a type III-A-specific Acr (AcrIIIA2). d, The phage-based interference assay was repeated with additional type III-A BIMs targeting phage genes expressed at early, middle, and late stages of the infection cycle. In all cases, AcrIIIA2 inhibited CRISPR activity, resulting in host cell lysis.

When challenged with phage 2972 at a multiplicity of infection (MOI) of 1, the wildtype (WT) Sth strain rapidly lysed upon infection, while the CR1-CR4 BIMs maintained phage immunity as expected (Fig. 1c). When the Acr candidate was expressed, BIMs corresponding to CR1 (type II-A), CR3 (type II-A), and CR4 (type I-E) maintained immunity. In contrast, the CR2 (type III-A) BIM targeting an early expressed phage gene (Orf34) fully lysed, albeit with a slight delay compared to the WT strain, indicating a novel and specific type III-A Acr (Fig. 1c). To determine whether Acr activity depends on the timing of phage gene expression, we tested additional CR2 BIMs with spacers targeting early (Orf30), middle (Orf6), and late (Orf18) phage genes²³. Regardless of the gene target, expression of the Acr candidate resulted in loss of type III immunity and host cell lysis (Fig. 1d). Additionally, expression of the Acr candidate in WT Sth cells had no effect on growth, reinforcing that the observed lysis in CR2 BIMs was Acr-dependent and CRISPR-specific (Supplementary Fig. 1a). To further validate our assay results, we expressed AcrIIIA2 from an Sth phage. In the absence of a naturally occurring example, we replaced the early-expressed, nonessential gene Orf33 in phage 2972⁴⁵ with the candidate acr gene. The resulting AcrIIIA2-expressing phage retained the ability to lyse the WT strain and, notably, also lysed the CR2 (type III-A) BIM, consistent with type III-A Acr activity (Supplementary Fig. 1b). Moreover, this lytic activity was specific to the CR2 (type III-A) BIM (Supplementary Fig. 1b). Together, these results demonstrate that the candidate phage protein specifically inhibits the type III-A CRISPR system, independent of the expression timing of phage gene transcripts, leading to its designation as AcrIIIA2, the second type III-A specific Acr.

AcrIIIA2 contains 105 amino acids with an isoelectric point of 9.91 and no predicted functional domains. To identify homologs of AcrIIIA2, we conducted a BLASTp search against the NCBI protein database. This analysis uncovered 12 additional homologs in Sth phages. Amino acid alignment revealed that 8 of these homologs were identical to the AcrIIIA2 protein from Sth phage SW3, while the other four alleles each contain a single nucleotide variation (Supplementary Fig. 1c). Additional AcrIIIA2 homologs were identified exclusively in Streptococcus species including salivarius, sanguinis, and oralis, but incomplete genome assemblies make it difficult to confirm their integration, though they are likely phage-associated (Supplementary Table 1).

Previous analyses of Sth type III BIMs showed that cOA-activated, Csm6 target RNA cleavage is the primary activity driving type III-A CRISPR immunity in vivo²³. To further investigate how AcrIIIA2 functions, the HD DNase domain of Csm1/Cas10 was inactivated through mutation (BIM dHD, HD changed to HA) in the type III-A BIM targeting the early phage gene Orf34, preserving only Csm6 RNase activity. At an MOI of 1, AcrIIIA2 expression resulted in host cell lysis, confirming the inhibition of Csm6 activity (Supplementary Fig. 1d). Our prior work showed that Sth Csm1 HD DNase activity becomes detectable only under low MOI conditions and when csm6 is either deleted (BIM ΔCsm6) or when the Csm1 Palm active site is inactivated through mutation (BIM dPalm; GGDD changed to GGAA)²³. When challenged with an MOI of 0.01, both strains underwent lysis upon expression of AcrIIIA2, indicating that AcrIIIA2 also inhibits Csm1 HD DNase activity in vivo (Supplementary Fig. 1d). Together, these results demonstrate that AcrIIIA2 is a robust and specific anti-CRISPR protein that effectively disables the major RNase and DNase activities associated with type III-A CRISPR immunity.

AcrIIIA2 interacts with the Sth type III-A CRISPR complex and host enolase

To further investigate the mechanism of Acr inhibition, we tested for potential interactions between AcrIIIA2 and the Sth type III-A crRNP (Csm complex) using Sth host cells. Purifications of the Csm complex with an N-terminal histidine-tagged Csm2 subunit were performed from Sth under native Csm expression levels. Sth Csm complexes were purified via Ni²⁺ affinity chromatography followed by size-exclusion chromatography (SEC). In the presence of AcrIIIA2 with a C-terminal strep-tag expressed from a plasmid, a notable SEC elution shift indicated purification of a larger Csm complex (Fig. 2a), exceeding the expected size from monomeric, AcrIIIA2 binding alone. Sodium dodecyl-sulfate polyacrylamide gel electrophoretic (SDS-PAGE) analysis confirmed the presence of core Cas proteins (Csm1-Csm5) in both complexes. However, an additional ~17 kDa band corresponding to AcrIIIA2 and an unexpected ~50 kDa band was observed. Mass spectrometry confirmed the ~17 kDa band as AcrIIIA2 and identified the additional ~50 kDa band as Sth enolase (Supplementary Fig. 2a), a highly conserved glycolysis enzyme³³. RNA analysis on a denaturing PAGE gel revealed identical crRNA profiles in both complexes, indicating that AcrIIIA2 does not affect crRNA processing or assembly into Csm complexes or co-purify with additional nucleic acids (Fig. 2a). The observed SEC peak shift can therefore be attributed to the predicted octameric structure of Sth enolase³⁶ in combination with AcrIIIA2. Together, these results demonstrate a stable interaction between AcrIIIA2 and the native Sth Csm complex, while also revealing an unexpected interaction partner, Sth enolase.

Figure 2. AcrIIIA2 interacts with the Sth Csm and host enolase.

a, Csm2-tagged Sth crRNP complex (Csm) was purified via Ni²⁺ affinity and size-exclusion chromatography (SEC) from Sth. The Csm complex eluted at 10.94 mL, but co-expression with AcrIIIA2 (Csm+Acr) caused earlier elution at 9.44 mL. SDS-PAGE analysis confirmed the presence of Csm proteins and additional bands corresponding to AcrIIIA2 and Sth enolase. Nucleic acids from the SEC peak fractions were analyzed on a UREA denaturing PAGE gel, showing no changes in crRNA profiles or evidence of additional nucleotide binding. b, AcrIIIA2 with a C-terminal strep-tag was purified from Sth using strep-tactin affinity chromatography. SDS-PAGE analysis revealed a ~17 kDa band corresponding to AcrIIIA2 and a ~50 kDa band corresponding to Sth enolase. Enolase co-purification occurred regardless of whether Cas proteins Csm1–6 were present (WT) or absent (ΔCsm) in the cells. c, Purification of the Sth Csm complex from E. coli showed similar early elution when AcrIIIA2 and enolase were co-expressed, confirmed by SDS-PAGE analysis. d, AcrIIIA2 was purified either alone or co-expressed with Sth enolase from E. coli. The enolase co-purified with AcrIIIA2, as indicated by an additional band matching the size of individually purified enolase. e, Binding assays with purified recombinant components. Ni²⁺ affinity purified Sth Csm complex incubated with either purified AcrIIIA2 alone, Sth enolase alone, or both together revealed that enolase and AcrIIIA2 co-purified with the Csm complex only when both AcrIIIA2 and enolase proteins were present. The input samples are shown in Figure S2C.

To assess whether AcrIIIA2 could reciprocally co-purify Cas proteins and Sth enolase, we purified strep-tagged AcrIIIA2 using Strep-tactin affinity chromatography. SDS-PAGE analysis confirmed successful purification of AcrIIIA2 along with the copurification of Sth enolase (Fig. 2b). Mass spectrometry further validated the presence of Sth enolase and Cas proteins (Csm1-Csm5) (Supplementary Fig. 2b), though the Cas proteins were below the detection threshold of the gel stain. To test if the AcrIIIA2-enolase interaction occurs independently of Csm complex, AcrIIIA2 was purified from a Sth strain lacking type III-A Cas proteins (ΔCsm1–6). The AcrIIIA2/enolase complex still formed in the absence of Csm complexes, indicating a stable, direct interaction between AcrIIIA2 and enolase, independent of Cas proteins (Fig. 2b).

Reconstitution of AcrIIIA2 interactions in E. coli

To better facilitate downstream biochemical and structural analyses, we tested whether the interactions that we observed in Sth between AcrIIIA2, Sth enolase, and the Sth Csm complex would also be observed in a heterologous E. coli expression system. The Sth Csm complex, with an N-terminal, histidine-tagged Csm2, along with crRNA, were co-expressed from a single plasmid as previously reported^46,47. C-terminal, strep-tagged AcrIIIA2 and untagged Sth enolase were expressed from individual plasmids and co-introduced into E. coli along with the Csm complex expression vector. The Csm complex was purified via Ni²⁺ affinity chromatography followed by SEC both in the presence and absence of AcrIIIA2 and Sth enolase. In both conditions, intact Csm complexes were successfully isolated. Notably, a major SEC peak shift was observed in the presence of AcrIIIA2 and Sth enolase, indicating successful reconstitution of their interactions with the Csm complex (Fig. 2c). SDS-PAGE analysis confirmed the presence of both with the purified complexes. Importantly, these results closely mirrored the outcomes from the native Sth purifications.

To further validate the formation of the AcrIIIA2/Sth enolase subcomplex, we expressed C-terminal, strep-tagged AcrIIIA2 in E. coli both in the presence and absence of untagged Sth enolase. AcrIIIA2 was then purified using Strep-tactin affinity chromatography. Additionally, Sth enolase was independently purified via Ni²⁺ affinity chromatography as a control. SDS-PAGE analysis verified the successful purification of both AcrIIIA2 and enolase proteins (Fig. 2d). Notably, Sth enolase copurified with AcrIIIA2 when expressed in the same cell, consistent with observations from native Sth purifications. Interestingly, AcrIIIA2 exhibited improved integrity when complexed with Sth enolase compared to its individual purification indicating that binding of AcrIIIA2 to Sth enolase is critical for the stability of AcrIIIA2 (Fig. 2d).

The essential nature of Sth enolase in its native host prevented the success of our enolase gene deletion experiments, so we leveraged our E. coli-based expression system to further investigate AcrIIIA2 interaction with the Sth Csm complex. The Csm complex was purified via Ni²⁺ affinity chromatography then incubated with purified AcrIIIA2 alone, Sth enolase alone, or both proteins together. SDS-PAGE analysis revealed that Sth enolase formed a stable interaction with the Csm complex only in the presence of AcrIIIA2, and reciprocally, AcrIIIA2 co-purified only when enolase was present (Fig. 2d and Supplementary Fig. 2c). The successful recapitulation of AcrIIIA2 and enolase interactions in E. coli confirmed the native purification results and validated the use of the E. coli purified products for subsequent in vitro assays to characterize AcrIIIA2 inhibition.

AcrIIIA2, with host enolase, impairs target RNA binding and cleavage by the Csm complex in vitro

We next tested the effects of purified AcrIIIA2 and enolase on the activities of isolated type III-A Csm crRNP complexes in vitro (Fig. 3). Our in vivo phage interference assays indicated AcrIIIA2-dependent inhibition of both Csm1- and Csm6-mediated DNA and RNA destruction, respectively (Supplementary Fig. 1c). To further explore these effects, we investigated whether target RNA binding by the Csm complex, an essential upstream step, was disrupted by the AcrIIIA2/enolase proteins. For this assay, the E. coli purified Csm complex in the absence or presence of co-expressed AcrIIIA2/enolase, was titrated into the reaction and incubated with 3’-end fluorescently labeled target RNA (complementary to the crRNA) or non-target RNA (lacking crRNA complementarity) (Supplementary Table 2). Binding was visualized using native PAGE mobility assays. As expected, the Csm complex selectively interacted with target RNA containing sequences matching the crRNA in a concentration-dependent manner (Fig. 3a and Supplementary Fig. 3a). In contrast, target RNA binding was substantially reduced when the Csm complexes were associated with the Acr-enolase components (Csm/Acr/Eno). A minor super-shifted band was observed suggesting weak target RNA binding by the Csm/Acr/Eno complex (Fig. 3a). These findings reveal that Csm crRNPs with associated AcrIIIA2 and enolase results in a significant impairment in target RNA binding.

Figure 3. AcrIIIA2 with Sth enolase reduces target RNA binding and downstream nuclease activity.

a, Fluorescent labeled target RNA was incubated with the Sth Csm complex (0.2–200 nM), and RNA binding was assessed using a native PAGE gel. A distinct upward shift indicated near complete binding at the highest Csm concentration (Csm+RNA), with no shift in the corresponding non-target (N) lane. Co-purified AcrIIIA2 and enolase (Csm/Acr/Eno) reduced RNA binding, with ~50% of remaining bound RNA forming a supershifted complex. b, Similarly, target RNA cleavage was evaluated using denaturing UREA PAGE gels. The Csm complexes alone showed near complete, regular 6-nucleotide cleavage at the highest concentration, while cleavage was significantly reduced in the presence of AcrIIIA2 and enolase (Csm/Acr/Eno). Non-target RNA (N) control lanes confirmed specificity. c, The Csm complex (200 nM) was pre-incubated with AcrIIIA2, enolase, or the AcrIIIA2/enolase subcomplex (2 μM) before the addition of labeled target RNA. Target RNA binding (left) and cleavage (right) were analyzed on native and denaturing PAGE gels, respectively. Pre-incubation with the AcrIIIA2/enolase subcomplex reduced target RNA binding and cleavage, mirroring results from the copurified Csm/Acr/Eno complex. d, Csm1-mediated DNA cleavage was tested by combining the (dCsm3) Sth Csm or copurified (dCsm3) Csm/Acr/Eno complexes with M13mp18 ssDNA and either target (T) or non-target (N) RNA. Cleavage products were visualized on a SYBR Gold-stained agarose gel. DNA cleavage was observed in the presence of target RNA and the Csm complex only. The addition of AcrIIIA2 and enolase inhibited cleavage activity. e, Synthesis of cOA was quantified using three different Csm preparations: Csm alone (top), Csm co-purified with the Acr and enolase (Csm/Acr/Eno, middle), and Csm preincubated with the Acr/enolase subcomplex (bottom). After a four hour reaction, the products were separated using an HPLC column gradient and detected at 254 nm. In the absence of target RNA (left), no cA₆ was detected. Upon addition of target RNA (right), cA₆ was produced in the Csm reaction. However, when Acr and enolase were associated with the complex, cA₆ production was significantly reduced.

We further investigated whether the Acr/enolase-mediated inhibition of target RNA interaction extended to target RNA cleavage. We performed the same target RNA binding assay, but with the addition of MgCl²⁺ (required for catalytic activity of Csm3 RNase) and visualization of the fluorescently-labeled substrate RNAs after denaturing PAGE. Titrations of Sth Csm complex showed the expected 6-nucleotide target RNA cleavage pattern, with one cut per Csm3^48,49, specific to target RNA (Fig. 3b and Supplementary Fig. 3b). In the presence of bound AcrIIIA2 and enolase, the Csm complexes exhibited a significant reduction in target RNA cleavage (Fig. 3b). These findings were consistent with the reduced, but not completely abolished, target RNA binding results (Fig. 3a) and together further support the notion that AcrIIIA2 in combination with enolase greatly impairs the ability of the Csm complexes to recognize and stably bind target RNA.

We next aimed to determine whether the addition of independently isolated AcrIIIA2/Sth enolase subcomplexes to purified Csm complexes would produce similar reductions in target RNA binding and cleavage as Csm/AcrIIIA2/enolase complexes formed in the cell. We repeated the target RNA binding and cleavage assays using a fixed amount of Csm complexes that were preincubated with AcrIIIA2, enolase, or the AcrIIIA2/enolase subcomplex before adding target or non-target RNA. Csm complex binding was unchanged when either AcrIIIA2 or enolase was added individually (Fig. 3c and Supplementary Fig. 3c). However, upon addition of the AcrIIIA2/enolase subcomplex, target RNA binding was reduced to the same level as the copurified Csm/AcrIIIA2/enolase complex (Fig. 3c), indicating that inhibition occurred only when both Acr and enolase were present. The target RNA cleavage results mirrored the binding assay, showing a reduction only after the AcrIIIA2/enolase subcomplex was added (Fig. 3c and Supplementary Fig. 3c). These findings underscore the requirement for both AcrIIIA2 and enolase in mediating the inhibition of type III-A CRISPR activity.

Inhibition of downstream DNA cleavage and cOA synthesis mediated by AcrIIIA2-enolase in vitro

We next investigated the effects of AcrIIIA2/enolase on the two activities of the Csm1 subunit: HD domain-mediated DNase activity and Palm domain-driven cyclic oligoadenylate (cOA) synthesis. For the Csm1 DNA cleavage assay, we used Sth Csm complexes containing catalytically inactive Csm3 (dCsm3) to prevent any target RNA cleavage following target RNA binding^49,50. We examined the Csm complexes with and without associated AcrIIIA2 and Sth enolase. These complexes were incubated with unlabeled synthetic target or non-target RNA (Supplementary Table 2) and single-stranded, M13 phage substrate DNA (ssDNA). DNA cleavage products were resolved by agarose gel electrophoresis. As expected, the Csm complex efficiently cleaved the majority of ssDNA in the presence of target RNA (Fig. 3d). However, when AcrIIIA2 and enolase were bound to the Csm complex, the ssDNA remained intact (Fig. 3d). The results demonstrate a clear inhibition of Csm1-mediated DNA cleavage, consistent with our in vivo phage interference assays (Supplementary Fig. 1c).

We also evaluated the impact of AcrIIIA2/enolase/Csm complex interactions on the cOA synthesis activity of the Csm1 Palm domain. Cyclic hexa-adenylate (cA₆) production was quantified using high-performance liquid chromatography (HPLC). When the reaction contained only ATP, the apo Sth Csm complex produced no detectable cA₆. However, the addition of target RNA shifted approximately half of the ATP peak to the expected cA₆ peak, confirming robust synthesis (Fig. 3e and Supplementary Fig. 3d). In contrast, the copurified Csm/AcrIIIA2/enolase complex completely abolished cA₆ synthesis (Fig. 3e), corroborating the results of the upstream binding and DNA cleavage assays. Notably, preincubation of the Csm complex with the Acr/enolase subcomplex also resulted in a significant reduction in cA₆ production (Fig. 3e), whereas individual additions of either AcrIIIA2 or enolase had no measurable effect on cOA synthesis (Supplementary Fig. 3d). In conclusion, AcrIIIA2 and enolase suppress upstream target RNA binding and cleavage that extends to inhibition of downstream Csm1-mediated DNA cleavage and cOA synthesis.

Structure overview of the Sth Csm/AcrIIIA2/Enolase complex

We isolated the crRNA-bound Sth Csm, AcrIIIA2, and Sth enolase as a ternary assembly by co-expressing all components in E. coli and obtained their cryo-EM structures (Fig. 4a and Supplementary Table 3 and 4). The host enolase forms an octameric ring that serves as a structural scaffold for the Acr assembly (Fig. 4b and Supplementary Video 1). On one side of the enolase ring, the AcrIIIA2 protein is positioned within its interior, burying 2,041 Ų of the solvent-accessible surface area. The enolase-supported AcrIIIA2 then captures the elongated Sth Csm complex at the end associated with Csm5, opposite to the catalytic subunit Csm1 (Fig. 4b). The AcrIIIA2 protein also forms an extensive interface with Sth Csm, interacting with Csm5, the 3′ end of the crRNA, and Csm2, burying 1,342 Ų, 436 Ų, 202 Ų, respectively, of the solvent-accessible surface area. This structural arrangement demonstrates that AcrIIIA2 inhibits the defense function of Sth Csm with the assistance of the host enolase.

Figure 4. Cryo-EM structure of AcrIIIA2/enolase bound Sth Csm-crRNA complex.

a, Schematic of the Streptococcus thermophilus (Sth) type III-A CRISPR-Cas operon. b, Cryo-EM density map of Sth Csm-crRNA complex (class 3:2) bound to AcrIIIA2 and enolase at 2.72 Å resolution (top); and its corresponding model in cartoon representation (bottom) in two orthogonal views. Protein subunits and RNAs are colored as follows: AcrIIIA2, crimson; enolase, burlywood; Csm1, skyblue; Csm2, dark gray; Csm3, cornflower blue; Csm4, cadet blue; Csm5 slate blue; crRNA, green; enolase, burlywood. See also Supplementary Video 1. c, Surface rendering showing AcrIIIA2 (red) is buried in octameric enolase complex and contacts the seed region at the 3’ end of the crRNA; d, Surface rendering showing contacts of AcrIIIA2 (red) with Csm2 and Csm5. The enolase catalytic residue (Asp242) is highlighted as golden blob on each monomer.

Like the structures of several Csm systems in the absence of Acr^50,51, two major stoichiometric assemblies are observed, one containing three Csm3 and two Csm2 (3:2) while the other containing four Csm3 and three Csm2 (4:3) subunits. In both assemblies, no difference in binding to AcrIIIA2 and enolase is observed (Supplementary Fig. 4). We, therefore, focused on structural analysis on the 3:2 assembly due to its better overall resolution (Supplementary Table 3 and 4).

The distant placement of AcrIIIA2 with respect to the catalytic activity of Csm1 suggests that AcrIIIA2 does not directly obstruct the catalytic functions of Csm1. Instead, AcrIIIA2 is expected to block target RNA from accessing the seed region of the crRNA (Fig. 4c and 4d), consistent with the target RNA binding studies (Fig. 2a).

There is evidence suggesting that the AcrIIIA2/Csm complex can simultaneously bind to the two structurally equivalent sides of the enolase octameric ring (Supplementary Fig. 4a). However, refinement of the double-decker assemblies reveals an asymmetry, predominantly favoring binding on one side. This observation suggests that the binding of the AcrIIIA2/Csm complex to one side of the enolase ring weakens its ability to bind to the opposite side (Supplementary Fig. 4). AcrIIIA2 does not directly obstruct the conserved catalytic residue (Asp242)³³ of enolase monomers, indicating that enolase may retain its functionality even when bound to the Acr (Fig. 4d).

Binding specificity suggests co-evolution of AcrIIIA2 to bind to Csm and Enolase

The observed binding mode of AcrIIIA2 to Sth enolase suggests a critical requirement for the octameric assembly. The largely unstructured AcrIIIA2 binds at the concaved center of the Sth enolase octamer, resembling the interior of a bowl, and forms several specific contacts. Acr interacts asymmetrically with five of the eight enolase subunits and forms polar interactions with chains E1, E3, E4, E5 and E7 (Fig. 5a). Chain E1 residues Asp 86, Ser2 and Glu77 contact Asn21, Lys15 and Lys50 of Acr, respectively, while chain E4 residue Asn74 contacts the hydroxyl group of Tyr131 of Acr and chain E7 residue Glu28 forms hydrogen bonds with the hydroxyl group of Ser8 of AcrIIIA2 (Fig. 5a). Notably, the hydroxyl and carboxyl oxygen of Ser36 of AcrIIIA2 contacts the carboxyl oxygen of Ala96 and amino group of Lys105 of enolase, and the natural variation of Ser36 to Ile36 from Sth phage SW7 abolished its Acr activity in vivo (Fig. 5a and Supplementary Fig. 5a). We introduced additional mutations in AcrIIIA2 residues that interact with the enolase octamer (Tyr79 or Lys15 to Glu) which resulted in either complete loss or a marked reduction of Acr activity (Supplementary Fig. 5b). Likewise, introducing the corresponding double mutations into the native Sth enolase (Glu28Lys/Lys105Ala or Asp83Lys/Glu77Lys) had no effect on normal cell growth but also led to a complete loss of Acr activity (Supplementary Fig. 5c), further underscoring the importance of these residues and confirming that enolase binding is essential for Acr inhibition. Most enolase residues involved in AcrIIIA2 binding are highly conserved (Supplementary Fig. 6) regardless of their predicted quaternary form. Thus, it is the octameric form rather than the specific residues that better serves Acr binding. The E. coli enolase exists as a homodimer³⁵, which explains the lack of AcrIIIA2 function in E. coli.

Figure 5. Structure and interactions of AcrIIIA2 with Sth enolase, crRNA and Csm complex.

a, Top-view of enolase-Acr complex (center); enolase chain E7 interactions with AcrIIIA2 (top-left); enolase chain E3, E4 and E5 interactions with AcrIIIA2 (bottom-left); chain E1 interactions with AcrIIIA2 (right). b, Sth Csm/AcrIIIA2/enolase complex (center). AcrIIIA2 is shown as surface, colored in red while Csm2, Csm5 and crRNA are shown as cartoon, colored in dark gray, slate blue, and green, respectively. Enolase, Csm1, Csm3, Csm4 and Csm5 are show as transparent surface. AcrIIIA2 C-terminus interaction with the positively charged β-hairpin loop (C-clamp) of Csm5 (top-left); electrostatic surface potential of AcrIIIA2 C-terminus and Csm5 β-hairpin loop in the same orientation as top-left (bottom-left). Red, white and blue indicate negative, neutral and positive electrostatic potential surfaces, respectively; Acr N-terminus interactions with Csm2 M-loop and Csm5 N-loop (top-right); Acr interactions with the terminal four nucleotides from crRNA 3’ end (bottom-right).

The most extensive interactions formed between AcrIIIA2 and the Sth Csm complex are with the Csm5 subunit. Largely unstructured, the long and meandering extension of AcrIIIA2 spanning nearly 60 Å wraps around Csm5, engaging three of its structural elements. The N-terminal segment of AcrIIIA2 (residues 26–32) contacts the random coil (residues 303–313) of Csm5, an insertion to its core Rossmann fold between α2 and β4 (N-loop for interacting with the N-terminus of AcrIIIA2) (Fig. 5b, top-right). The fully extended, C-terminal segment of AcrIIIA2 (residues 99–102) lies flat on a β-hairpin (residues 324–337, C-clamp) immediately following the random coil (Fig. 5b, top-left and bottom-left). Consistently, the C-terminal truncation of AcrIIIA2, retaining only residues 1–83, abolished its inhibitory activity in vivo (Supplementary Fig. 5). Similarly, the naturally occurring AcrIIIA2 variant from Sth phage CHPC663, which carries a single Glu90 to Lys90 substitution, also failed to suppress CRISPR immunity in vivo (Supplementary Fig. 5). To a lesser extent, the midsection of AcrIIIA2 (residues 54–64), also extended, interacts with the hairpin loop (residues 203–213) of Csm5 inserted between α1 and α3 of its Rossmann fold core (L13) (Fig. 5b, bottom-right).

In addition, the loop following α2 of AcrIIIA2 (residues 44–50) forms electrostatic interactions with the loop in Csm2 connecting α3 and α4 (M-loop for interacting with the mid-region of AcrIIIA2) (Fig. 5b, top-right). Finally, the N-terminal segment of AcrIIIA2 including the first methionine hold the last four nucleotides of the crRNA in place by both electrostatic and hydrophobic stacking interactions (Fig. 5b, bottom-right).

Sequence comparison of Sth Csm to the homologous Csm systems indicates that AcrIIIA2 explores the unique elements of Sth Csm5 and Sth Csm2 to establish its specific anti-CRISPR activity. Though both Csm5 and Csm2 are well conserved among Sth species, specific protein elements involved in binding AcrIIIA2 are unique to Sth species (Supplementary Fig. 7). The majority of the organisms have the Csm5 C-clamp element without the Csm2 M-loop, although one organism, Abiotrophia defective, instead retains the Csm2 M-loop without the Csm5 C-clamp. Streptococcus suis (Ssu) contains both the Csm2 M-loop as well as the Csm5 C-clamp (Supplementary Fig. 7), which coincides with its enolase being octameric⁵² suggesting the Acr inhibition mechanism will also apply to this organism.

AcrIIIA2 binding blocks target RNA access to the seed region of crRNA

Our Sth Csm/Acr/enolase structure suggests a potential mechanism for AcrIIIA2 to obstruct target RNA binding. Comparison of the Sth Csm structure when bound with Acr to a homologous Sth Csm structure when bound with target RNA show that the 3’-end of the crRNA that contains the seed region is significantly bent by the bound AcrIIIA2 (Supplementary Fig. 8), which compromises its required trajectory to base pair the incoming target RNA. In addition, the C-terminal extension of AcrIIIA2 contains largely negatively charged residues and is in the open space where the target RNA likely approaches the seed region of crRNA. Therefore, the electrostatic repulsion of the AcrIIIA2 C-terminus could also exclude the target RNA from binding.

AcrIIIA2 does not disrupt enolase activity in vivo or in vitro

In vivo testing was conducted to determine whether AcrIIIA2 impacts Sth enolase canonical activity or phage infectivity. Enolase inhibition in Bacillus subtilis led to growth defects and cell wall disruption that stimulated phage infectivity⁵³. No growth defects were observed in the presence of AcrIIIA2 (Supplementary Fig. 9a). Furthermore, similar phage lysis curves (Supplementary Fig. 9a) and efficiency of plaquing results (Supplementary Fig. 9b) were observed in the absence and presence of AcrIIIA2, suggesting no enhancement of phage infectivity. Moreover, addition of AcrIIIA2 did not significantly impair the in vitro enzymatic activity of purified enolase to convert 2-PGA (2-phosphoglycerate) to PEP (phosphoenolpyruvate) (Supplementary Fig. 9c). Collectively, these results indicate that binding of AcrIIIA2 to enolase does not interfere with the canonical enzymatic activity of enolase.

DISCUSSION

We report the discovery of AcrIIIA2, a specific and potent inhibitor of type III-A CRISPR immunity, encoded by Streptococcus thermophilus phages. Through a combination of cell-based assays and detailed biochemical and structural studies, we demonstrate that AcrIIIA2 employs a unique mechanism of action requiring both interactions with the Csm complex and the highly-conserved host protein enolase (Figs. 2, 4 and 5). Our structural analysis reveals that AcrIIIA2 inhibits Csm function by directly interacting with the 3’ seed region of the crRNA (Figs. 4c, 5b and Supplementary Fig. 8) as well as nearby Csm5 and Csm2 subunits (Figs. 4b–d and 5b). Together, our results reveal that AcrIIIA2 is highly effective at preventing of type III immunity by blocking the initial recognition of phage target RNA needed for activation of all downstream type III immune activities including cOA signaling, as well as RNA and DNA cleavage (Fig. 3).

Most known Acr proteins function by interfering with CRISPR immunity through two primary mechanisms: preventing nucleic acid target recognition or inhibiting target cleavage. These mechanisms typically involve direct binding of the Acr to CRISPR effector complexes, either obstructing key functional sites or inducing allosteric inhibition⁵⁴. However, our current understanding of Acr mechanisms is largely derived from studies on type I and type II CRISPR systems, where Acrs have been shown to block DNA binding, inhibit DNA cleavage, or disrupt effector protein assembly^25,26. In contrast, knowledge of type III CRISPR inhibiting Acrs remains limited. To date, only four Acrs targeting type III systems have been identified, and their mechanisms are far less characterized^22,27–30. Unlike type I and II systems, type III CRISPR complexes degrade both RNA and DNA and produce cyclic oligoadenylates (cOA) that activate auxiliary nucleases, adding layers of complexity to potential Acr inhibition strategies.

Of the four previously identified type III Acrs, three (AcrIIIB1, AcrIIIB2, and AcrIII-1) were discovered in Sulfolobus archaeal viruses. AcrIII-1³⁰ and AcrIIIB1²² selectively inhibit certain components of the type III-B Sulfolobus system, blocking Csx1-mediated RNA cleavage while leaving Cas10 DNase activity intact. Similarly, AcrIIIB2 provides conditional inhibition when expressed from the phage genome, specifically suppressing immunity when targeting late-expressed genes^28,29. In contrast, our newly identified AcrIIIA2 exhibits robust inhibition of both Cas10 DNA cleavage and Csm6 RNA degradation in vivo, independent of gene expression timing (Fig. 1c–d and Supplementary Fig. 1c). The only other known type III-A Acr (AcrIIIA1), which inhibits the Staphylococcus epidermidis type III-A system, also blocks immunity when targeting either early or late-expressed phage genes²⁷. AcrIIIA1 interacts with Csm2 and t(m)RNA fragments, which are associated with host RNase R, though the precise consequences of these interactions remain unclear²⁷. AcrIIIA2, however, lacks comparable t(m)RNA binding properties (Fig. 2a), and the Sth type III-A system is not known to engage host factors for immunity. Beyond AcrIII-1, a cOA ring nuclease that blocks Csx1 RNase activation³⁰, all known type III Acrs remain poorly characterized, lacking both a detailed understanding of their inhibition mechanisms and structural insights into their interactions^27–29. AcrIIIB2 inhibition appears to act through preventing turnover of cleaved target RNA^28,29. Here, we not only identify a novel type III-A Acr but also present the most comprehensive characterization of a type III Acr to date, elucidating its inhibition mechanism, interaction partners, and key residues needed for Acr function.

Our structural analysis reveals that the AcrIIIA2 N-terminus interacts with key nucleotides within the crRNA seed region (Figs. 4c, 5b and Supplementary Fig. 8), a region previously identified as essential for target RNA recognition and binding^17,19,20. This interaction likely is responsible for the observed prevention of crRNA-target RNA pairing, thereby inhibiting CRISPR interference. Additionally, AcrIIIA2 forms extensive interactions with Csm5 and Csm2 subunits of the Csm complex (Fig. 4b–d and 5b), highlighting its multipronged inhibitory strategy. Evolutionary analysis of homologous Csm complexes across species suggests that the structural elements required for AcrIIIA2 binding to Csm complexes, such as the Csm2 M-loop and Csm5 N-loop elements, are highly restricted to certain Sth species (Supplementary Fig. 7), suggesting the importance of co-evolving, phage-host interactions in shaping the ‘arms race’ between bacteria and their associated phages.

The discovery that enolase, a well-known glycolysis enzyme, plays a crucial role in AcrIIIA2-mediated inhibition adds a novel layer of complexity to our understanding of the Acr mechanism. Enolase is traditionally known for its central role in glycolysis, where it catalyzes the conversion of 2-phosphoglycerate to phosphoenolpyruvate, making it an essential enzyme in bacterial fermentation and energy production³³. Consequently, enolase is one of the most abundant cytoplasmic proteins in bacteria⁵⁵. Our assays suggest that this canonical glycolytic function remains largely intact, as neither cell growth nor phage infectivity is significantly altered in vivo (Supplementary Figs. 9a–b). Furthermore, the catalytic activity of enolase in vitro is also not significantly affected by interaction of enolase with AcrIIIA2 (Supplementary Fig. 9c). Indeed, the active site of enolase within each monomeric unit, are quite remote from sites where AcrIIIA2 binds the enolase octamer (Fig. 4d), suggesting a separation of its chaperone from its catalytic function. Therefore, AcrIIIA2 interaction with enolase is critical for type III inhibition but does not substantially impact the glycolytic function of the highly abundant enolase enzyme.

Interestingly, a naturally occurring enolase inhibitor was discovered in Bacillus subtilis phages revealing another instance where phages have previously recognized and targeted enolase as a means of host manipulation⁵³. In this case, PEIP (the phage-encoded enolase inhibitor protein) was found to support phage propagation through inhibiting enolase function. PEIP acts by disassembling the octameric structure of enolase which results in defects in cell wall integrity, impairment of host cell growth and enhancement of phage proliferation⁵³.

Beyond its metabolic role, enolase is known for its moonlighting functions in various pathogenic Streptococcus species, influencing host-pathogen interactions and contributing to bacterial virulence^37,38. However, given that Sth is non-pathogenic, it is unlikely that enolase serves a similar role in this context. Instead, structural analysis of the Csm/Acr/enolase complex (Fig. 4 and 5) suggests the enolase proteins act as a key structural cofactor. The octameric ring of enolase subunits³⁶ provides a stabilizing base for AcrIIIA2 and plays a key role in binding of AcrIIIA2 to the Csm crRNP complex (Fig. 5). Intriguingly, enolase in E. coli is recruited to the RNA degradosome through interactions with RNase E³⁴, while the type III-A system of Staphylococcus epidermidis has been shown to recruit ribonucleases from the degradosome for crRNA maturation and anti-phage and plasmid immunity^31,32. While we do not know if Sth type III-A system utilizes additional host RNases for its crRNA biogenesis or function, our results indicate a direct interaction of AcrIIIA2 with enolase in vivo and in vitro (Figs. 2, 4 and 5). Overall, our findings reveal a strategy in which AcrIIIA2 hijacks host enolase for structural stability and function, further underscoring the versatility of the enolase enzyme beyond its classical metabolic role.

The identification of AcrIIIA2 opens possibilities for discovering additional type III Acrs. Both AcrIIIA2 and the previously discovered AcrIIIA1²⁷ appear to interact with the Csm complex as well as host proteins (Fig. 2, 4 and 5), suggesting that other type III Acrs may act similarly. To uncover further Acrs, investigating native host systems for their ability to withstand CRISPR interference, followed by native co-purification may be required. In conclusion, AcrIIIA2 defines a new class of Acr proteins that requires a host protein for its interaction with a CRISPR-Cas complex to inhibit CRISPR-Cas immunity. Our work also provides the first structural insight into the mechanism of action of a type III Acr.

METHODS

Strain culturing

Streptococcus thermophilus (Sth) cultures were grown overnight in M17 medium (HiMedia) at 37°C under static conditions or on M17 agar plates at 42°C. When pTRK882 plasmids were present, M17 was supplemented with 10 μg/mL erythromycin. For pTRK882 plasmid maintenance, E. coli strain MC1061 cultures were grown in Luria Broth (RPI) with erythromycin 200 μg/mL at 37°C. E. coli TOP10 strains, used for plasmid maintenance, and BL21-AI strains, used for protein purification, were grown with the following antibiotics: 50 μg/mL kanamycin for pRSFDuet1, 100 μg/mL ampicillin for pET, and 100 μg/mL chloramphenicol for pACYC vectors.

Identification of candidate anti-CRISPR genes

To identify potential acr genes, we used a guilt-by-association approach, focusing on genomic regions adjacent to known Sth Acrs. Specifically, the region between the lysis and replication models of Sth phages is known to be variable and has previously been shown to contain accessory genes, including Acrs, in related phages^39–41. Candidate open reading frames (Orfs) in this variable region were examined. One such gene in Sth phage SW3, located immediately downstream of AcrIIA6, was selected as a potential type III Acr candidate for further experimental testing.

S. thermophilus phage 2972 lysis curves

Overnight Sth cultures were diluted in fresh M17 media (HiMedia) to an optical density (OD₆₀₀) of 0.2 and supplemented with 10 mM CaCl₂. Phage 2972 lysate was added to a 96-well microplate along with 200 μL of the freshly diluted culture to achieve the desired multiplicity of infection (MOI). The plate was then placed in a BioTek Epoch2 Microplate Spectrophotometer, maintained at 42°C with double orbital shaking. OD₆₀₀ readings were collected every 2 minutes over 6 hours. Each assay was performed in technical triplicates with data collected from three biological replicates. Growth curves were built using GraphPad Prism 10.4.1.

S. thermophilus phage 2972 genome editing

Plasmid constructs of pTRK882 were engineered to contain a CR1 type II-A repeat-spacer-repeat array under the P_pgm promoter, using the Orf33 spacer described in a previous phage 2972 study⁴⁵. The plasmid also carried 500 bp of homology to the phage 2972 genome flanking Orf33, with the acrIIIA2 gene inserted between these regions to serve as a recombination template. The plasmid was first maintained in E. coli MC1061 and subsequently introduced into wild-type Sth using an established natural transformation protocol²³.

Overnight Sth cultures (500 μL) containing the pTRK882 construct were diluted into 0.75% M17 (NutriBact) top agar supplemented with 10 mM CaCl₂ and mixed with 10-fold serial dilutions of 100 μL of phage 2972 lysate. The top agar was then plated on 1% M17 (NutriBact) agar plates and incubated overnight at 42°C. Resulting plaques were screened by PCR. Positive plaques were scraped into 100 μL of phage buffer and subjected to a first and second round of viral amplification by infecting WT Sth cultures grown at 42°C to an OD₆₀₀ of 0.2 in M17 (NutriBact). Cultures (10 mL) were supplemented with 10 mM CaCl₂, inoculated with 50 μL of the engineered phage and grown at 42°C for 6 hours before filtration with a 0.2 μm filter.

AcrIIIA2 homologs alignments

Homologs of AcrIIIA2 were identified in January 2025 using BLASTp (protein-protein BLAST) in NCBI (https://blast.ncbi.nlm.nih.gov) with an e-value cutoff of ⩽1×10⁻⁵. The identified homologs were aligned and organized using the Clustal Omega multiple sequence alignment (MSA) plugin in Geneious Prime^® 2025.0.2.

Native protein purification from S. thermophilus

A 6x Histidine tag was introduced at the N-terminal region of Csm2 in the genome of Sth DGCC7710 via natural transformation, as previously described²³. This strain harbors the active type III-A system from Sth JIM8232, replacing its native type III system²³. When specified, AcrIIIA2 was constitutively expressed from the P_pgm promoter on the pTRK882 plasmid, with a C-terminal Strep-tag^®II (IBA Lifesciences) added to the Acr protein.

For protein purification, 50 mL of overnight Sth culture was inoculated into 5 liters of fresh M17 media and grown standing at 42°C for ~18 hours. Cultures were centrifuged at 6,000 × g for 20 minutes at 4°C, and the resulting pellets were frozen at −80°C. Pellets were resuspended in buffer (40mM Tris, 500mM NaCl, pH8), and homogenized with a CF1 Cell Disrupter (Constant Systems Ltd.): 3 rounds at 20 kpsi. The lysate was centrifuged at 20,000 × g for 30 minutes at 4°C and the soluble fraction collected. The Sth Csm complex or AcrIIIA2 proteins were isolated via Ni²⁺ or Strep-Tactin^® XT affinity chromatography. Further purification was performed using size exclusion chromatography (SEC) on a Superdex^™ 200 Increase column (Cytiva). Purified proteins were concentrated and stored at −80°C in 40mM Tris, 500mM NaCl, pH8. Protein bands were identified via mass spectrometry following affinity purification.

Nucleic acid isolation

Nucleic acids that copurified with native Sth Csm and Sth Csm bound to AcrIIIA2 and Sth enolase were identified by incubating 1 μg of each complex with proteinase K for 15 minutes at 37°C. The reactions were then denatured at 95°C for 5 minutes before the products were resolved on 15% denaturing UREA PAGE gels. Electrophoresis was carried out at 200 V for 40 minutes. The gels were stained with SYBR Gold and imaged using a ChemiDoc MP imager (BioRad).

Proteolysis

Each sample was resolved by SDS-PAGE and stained by Coomassie Blue. Gel bands were excised, reduced by incubating with 5 mM of dithiothreitol (Sigma) at 56 °C, alkylated by 13.75 mM of iodoacetamide (Sigma) at room temperature in dark, and digested using trypsin/LysC mix (Promega) overnight at 37 °C. Following digestion, the peptides were extracted and dried down.

LC-MS analysis

The peptides were separated on an Acclaim^™ PepMap^™ 100 C18 column (75 μm × 15 cm) and eluted into the nano-electrospray ion source of an Orbitrap Eclipse^™ Tribrid^™ mass spectrometer (Thermo Scientific) at a flow rate of 200 nL/min. The spray voltage was set to 2.2 kV and the temperature of the heated capillary was set to 275 °C. Full MS scans were acquired from m/z 300 to 2000 at 60k resolution in the orbitrap, and MS/MS scans following collision-induced dissociation (CID) were collected in the iontrap. The spectra were analyzed using SEQUEST (Proteome Discoverer 2.5, Thermo Fisher Scientific) with mass tolerance set as 20 ppm for precursors, and 0.5 Da for fragments. The search output was filtered to reach a 1% false discovery rate at the protein level and 10% at the peptide level. The quantitation was performed based on spectral counts.

Recombinant protein purification from E. coli

The Cas proteins and crRNA of the Sth JIM8232 type III-A system were expressed under individual T7 promoters, as previously described⁴⁶. A 6x Histidine tag was added to the N-terminus of Csm2 using an established natural transformation with homology dependent recombination procedure²³. AcrIIIA2 and the enolase protein from Sth DGCC7710 were expressed under T7 promoters using pRSFduet1 and pET vectors, respectively. When specified, a C-terminal Strep-tag^®II (IBA Lifesciences) was added to AcrIIIA2, while enolase contained an N-terminal 6x Histidine tag. Plasmids were introduced into E. coli BL21-AI cells in the indicated combinations. Overnight cultures (10 mL) were grown in Terrific Broth (TB) (RPI) at 37°C. Large-scale cultures were initiated by inoculating 1 liter of fresh TB with 10 mL of the overnight culture and grown at 37°C. Once the cultures reached an OD₆₀₀ of 1, protein expression was induced with 0.5% arabinose and 0.5 mM IPTG. The temperature was reduced to 20°C, and the cultures were incubated for an additional ~18 hours. Cells were harvested by centrifugation at 6,000 × g for 20 minutes at 4°C, and the resulting pellets were frozen at −80°C. Pellets were resuspended in buffer (40 mM Tris, 500 mM NaCl, pH 8) and lysed by high-frequency sonication. The lysate was centrifuged at 20,000 × g for 30 minutes at 4°C to collect the soluble fraction. Sth Csm and enolase proteins were purified via Ni²⁺ affinity chromatography, while AcrIIIA2 was purified using Strep-Tactin^® XT affinity chromatography (Cytiva). Final purification of the Csm complex, either alone or in complex with AcrIIIA2 and Sth enolase, was performed using size-exclusion chromatography (SEC) on a Superdex^™ 200 Increase column (Cytiva). Peak fractions from SEC were concentrated and stored at −80°C in buffer (40 mM Tris, 500 mM NaCl, pH 8).

In vitro binding using recombinant purified proteins

To test in vitro co-purification of AcrIIIA2 and enolase with the CRISPR complex, the Sth Csm complex was first purified from E. coli via Ni²⁺-affinity chromatography as described above. AcrIIIA2 and Sth enolase were individually purified using either a C-terminal Strep-tag^® II (AcrIIIA2) or an N-terminal Strep-tag^® II (enolase) followed by Strep-Tactin^® XT affinity chromatography (Cytiva), as described above. The purified proteins were then combined in storage buffer (40 mM Tris, 500 mM NaCl, pH 8) at a ~1:2 ratio of Csm to Acr/enolase and incubated for 30 minutes at 4 °C. After incubation, the mixture was subjected to a second round of Ni²⁺-affinity chromatography. Co-purification was evaluated by resolving the elution fractions on SDS-PAGE gels, staining with Coomassie Blue, and imaging the gels using a ChemiDoc^™ MP imager (Bio-Rad).

In vitro target RNA binding

To assess target RNA binding to the Sth Csm complex, an Alexa Fluor^® 647 fluorescent label was attached to the 3’ end of synthetic RNA. These RNAs were either complementary (target) or non-complementary (non-target) to the expressed crRNA (Supplementary Table 2). Labeling was performed using pCp-AZDye647 (Cytidine-5’-phosphate-3’-(6-aminohexyl)phosphate, labeled with AZDye 647, Triethylammonium salt) from Jenna Bioscience. For the binding assay, 100 nM of labeled target or non-target RNA was mixed with 0.2 to 200 nM of SEC-purified Csm complex, with or without bound AcrIIIA2 and Sth enolase, in a reaction buffer containing 20 mM Tris, 250 mM NaCl, pH 8. Additionally, 25 nM EDTA and 5 μg E. coli tRNA competitor were added, and the reaction was incubated at 37°C for 15 minutes. The samples were then placed on ice, and half of each reaction was loaded onto pre-chilled 5% native PAGE gels. Electrophoresis was performed at 120 V for 80 minutes at 4°C, and the gels were imaged using a ChemiDoc MP imager (BioRad).

To investigate the effects of the AcrIIIA2/Sth enolase subcomplex on target RNA binding, 200 nM of Sth Csm was preincubated with 2 μM AcrIIIA2, Sth enolase, or the AcrIIIA2/Sth enolase subcomplex for 15 minutes at 37°C. Target or non-target RNA was then introduced under identical buffer conditions and reagent concentrations as described above. The samples were also processed and analyzed following the same native PAGE protocol. Each assay was conducted in biological triplicates.

In vitro target RNA cleavage

To assess target RNA cleavage, SEC-purified Sth Csm complexes, with or without bound AcrIIIA2 and Sth enolase, were incubated in a reaction buffer containing 20 mM Tris, 250 mM NaCl (pH 8), 10 mM MgCl₂, and 5 μg E. coli tRNA competitor. Alexa Fluor^® 647-labeled target or non-target RNA (Supplementary Table 2) was added at a final concentration of 100 nM, and the reaction was incubated at 37°C for 15 minutes. An equal volume of gel loading buffer (95% formamide, 18 mM EDTA, 0.025% SDS) was added, followed by denaturation at 95°C for 5 minutes. Half of the reaction mixture was loaded onto pre-warmed 15% UREA PAGE gels and ran at 200 V for 40 minutes. Fluorescent RNA bands were visualized using a ChemiDoc MP imager (BioRad).

For AcrIIIA2/Sth enolase subcomplex experiments, Sth Csm (200 nM) was incubated with 2 μM AcrIIIA2, Sth enolase, or the AcrIIIA2/Sth enolase subcomplex for 15 minutes at 37°C. Fluorescently labeled target or non-target RNA (100 nM) was then added, and the reaction was further incubated for 15 minutes at 37°C. The same buffer conditions, reagents, and UREA PAGE analysis protocol described above were followed. Biological triplicates were collected for each assay.

In vitro Csm1 ssDNA cleavage

The Sth Csm complex containing catalytically inactive Csm3 (dCsm3), either alone or co-purified with AcrIIIA2 and Sth enolase at a final concentration of 20 nM, was incubated with 50 ng of M13mp18 single-stranded DNA (New England Biolabs, Inc.), 1 mM DTT, and 200 nM unlabeled target or non-target RNA (Supplementary Table 2) in a reaction buffer containing 20 mM Tris, 40 mM KCl (pH 8.5). Reactions were incubated at 37°C for 1 hour. Following incubation, purple 6x gel loading dye (New England Biolabs, Inc.) was added, and samples were denatured at 95°C for 5 minutes. Half of each reaction was loaded onto a 1.5% agarose gel stained with SYBR Gold dye and electrophoresed at 90 V for 1 hour. DNA bands were visualized using a ChemiDoc MP imager (BioRad). The assays were performed until three biological replicates obtained.

In vitro Csm1 cOA synthesis

The cOA inhibition assay was performed by incubating a 250 nM Csm–crRNA binary complex, 200 nM target RNA, 1 mM ATP, 750nM Acr/enolase complex and 10 mM MgCl₂ in a reaction buffer (33 mM Tris acetate, pH 7.6, 66 mM potassium acetate) at 37°C for durations of 4 hours in a total volume of 30 μl. The reaction was terminated by heating at 95°C for 10 minutes to inactivate the enzyme, followed by centrifugation to remove denatured protein subunits.

The resulting products and control samples were analyzed using a Shimadzu Prominence LC-20 HPLC system equipped with a SunFire C18 column (4.6 mm × 150 mm, 3.5 μm particle size). Samples of 5 μl were injected, and a linear gradient method was employed with eluent A (20 mM ammonium bicarbonate) and eluent B (100% acetonitrile) at a flow rate of 0.3 ml/min over 22 minutes. The gradient conditions included: 2–30% B from 0 to 12 minutes, a transition to 95% B from 12.1 to 16 minutes for column washing, a constant 95% B from 16.1 to 17 minutes, and re-equilibration to 2% B from 17.1 to 22 minutes. Product detection was carried out at a wavelength of 254 nm. The experiments were performed in triplicates.

Efficiency of plaquing

Cultures of Sth carrying either an empty pTRK882 vector or one expressing AcrIIIA2 were grown overnight at 37°C in M17 medium (HiMedia) supplemented with 10 μg/mL erythromycin. Fresh M17 medium containing 0.75% agar and 10 mM CaCl₂ was inoculated with 500 μL of the respective bacterial strain and spread over M17 plates containing 1% agar. Sth phage 2972 lysate was serially diluted in phage buffer (10⁻² to 10⁻⁸), and 2 μL drops of each dilution, along with undiluted lysate, were spotted onto the bacterial lawn. After drying, plates were incubated overnight at 42°C to allow lysis zones to form. Images were captured using a ChemiDoc MP imager (BioRad). The assay was performed in technical triplicates, with data collected from three independent biological replicates.

Cryo-EM sample preparation, data collection, and 3D reconstruction

4μL of 3 mg/ml wildtype Sth Csm/crRNA/Acr/Enolase sample was applied to glow-discharged UltrAuFoil 300 mesh R1.2/1.3 grids (Quantifoil). The grids were blotted with blotting paper for 3 s at 95% humidity and flash-frozen in liquid ethane using FEI Vitrobot Mark IV. The grids were stored in liquid nitrogen before being used for imaging.

A total of 13,537 raw micrographs were collected using a Titan Krios cryo-transmission electron microscope equipped with a Gatan K3 direct electron detector (Thermo Fisher Scientific). Movies were recorded at a magnification of 105,000× in super-resolution mode, with an energy filter set to 15 eV. The data corresponded to a corrected physical pixel size of 0.53 Å/pixel. A total dose of 50 electrons per Ų was applied over 50 frames, with the defocus range set between −0.5 μm and −2.0 μm.

Motion correction and contrast transfer function (CTF) estimation were carried out in CryoSPARC v.4⁵⁶. Initial particle picking was performed using blob-based picking, followed by iterative rounds of 2D and 3D classification with C1 symmetry. Non-uniform 3D refinement was applied to generate high-resolution maps. A solvent mask was generated specifically for the Csm/Acr complex and was employed during subsequent local refinement steps. CTF post-refinement was conducted to correct for beam-induced particle motion, yielding the final maps. Data processing and refinement statistics are summarized in Supplementary Fig. 4.

Structural models were built with manual adjustment in COOT⁵⁷ and refined in PHENIX⁵⁸ to satisfactory stereochemistry and real space map correlation parameters. Model validation and refinement statistics are provided in Supplementary Fig. 4.

Native S. thermophilus enolase mutations

S. thermophilus CR2 type III-A BIM enolase mutant strains were constructed using an established natural transformation procedure with homology dependent recombination²³. The pUC19 plasmid constructs containing the double mutations were cloned using Gibson assembly (NEBuilder^® HiFi DNA Assembly Master Mix) then amplified from to act as the recombinant PCR products. Two kb homology arms were utilized for the recombination event then mutations were verified via Sanger sequencing (Eurofins Genomics LLC).

Enolase activity assay in vitro

Enolase activity was measured using an enolase activity assay (Abcam, cat# ab241024), which works through a coupled enzyme reaction. In this assay, D-2-phosphoglycerate is converted into phosphoenolpyruvate (PEP), which then reacts with a peroxidase-based substrate to generate an intermediate product. This intermediate product stoichiometrically reacts with the OxiRed Probe to generate color detectable at 570 nm.

For the experiment, a 50 μL reaction mixture was prepared, containing enolase, Acr/enolase or Sth Csm/AcrIIIA2/enolase complex with 5 nM enolase concentration. The reaction was initiated by adding 50 μL of the assay reaction mix and the absorbance was measured at 570 nm for 60 minutes at 25°C to determine enolase activity.

Figure 6. Model of AcrIIIA2-mediated inhibition of the Sth type III-A CRISPR complex.

In the absence of AcrIIIA2 (top), Sth crRNP complexes recognize and bind complementary target RNA, triggering a cascade of immune activities. Csm3 cleaves the target RNA, while Csm1 mediates DNA cleavage and cyclic oligoadenylate (cOA) synthesis, which in turn activates Csm6 for further RNA degradation. These combined nuclease activities establish robust anti-phage immunity. Upon phage expression of AcrIIIA2 (bottom), the Acr forms a subcomplex with the Sth enolase octamer, which then associates with Csm5, Csm2, and the 3’ crRNA region of preassembled Sth crRNPs. This interaction occludes the 3’ crRNA seed region, preventing target RNA recognition and suppressing crRNP nuclease functions. As a result, CRISPR immunity is blocked, enabling phage propagation. This figure was created with Biorender.com.

DATA AVAILABILITY

The atomic coordinates of the cryo-EM structures of Sth Csm class3:2 and class 4:3 have been deposited in the Protein Data Bank under the identifiers 9NO4 and 9NQ7 in the Electron Microscopy Data Bank under the entries EMD-49593 and EMD-49645 respectively. The mass spectrometry data have been deposited to the MassIVE database (https://massive.ucsd.edu/ProteoSAFe/static/massive.jsp) with the identifier MSV000097041 (ftp://MSV000097041@massive.ucsd.edu/).