Archaeal Kink-Turn Binding Protein Mediates Inhibition of Orthomyxovirus Splicing Biology

lab-led

Archaeal members of the L30 protein family, but not their orthologues from any other domain of life, block production of the spliced influenza A virus, influenza B virus and isavirus transcripts without measurably disturbing host splicing, identifying a shared and apparently noncanonical feature of orthomyxovirus splicing.

Kohei Oishi; Daniel Blanco-Melo; Andrew P. Kurland; Jeffrey R. Johnson; Benjamin R. tenOever

2023 · Journal of Virology · primary research

Senior authors

Benjamin R. tenOever

Correspondence

Benjamin R. tenOever

Research areas & themes

Citation

Oishi K, Blanco-Melo D, Kurland AP, Johnson JR, tenOever BR. Archaeal Kink-Turn Binding Protein Mediates Inhibition of Orthomyxovirus Splicing Biology. Journal of Virology. 2023. Volume 97, issue 4, article e01813-22. DOI 10.1128/jvi.01813-22. PMID 36943134. PMCID PMC10134859.

One-sentence contribution

Archaeal members of the L30 protein family, but not their orthologues from any other domain of life, block production of the spliced influenza A virus, influenza B virus and isavirus transcripts without measurably disturbing host splicing, identifying a shared and apparently noncanonical feature of orthomyxovirus splicing.

Executive summary

Orthomyxoviruses replicate in the nucleus and expand their coding capacity by splicing segment 7 and segment 8 transcripts, producing M2 and NS2 respectively. The slow accumulation of these spliced products acts as a timer that separates early and late events of infection. How this splicing is executed has remained unclear, and it has never been reconstituted in vitro, which has led to proposals that it is noncanonical.

The authors screened a synthesized library of codon-optimized archaeal RNA binding proteins for effects on influenza A virus protein accumulation. One member, the kink-turn binding protein L7Ae, selectively eliminated M2 and NS2 while M1 and NS1 accumulated. The effect reproduced without infection and without the viral polymerase, required nuclear localization of L7Ae, and was lost against a recombinant virus in which M2 and NS2 are made from 2A peptide constructs rather than by splicing. In inducible canine kidney cells, L7Ae reduced titers of several influenza A virus subtypes by two to three logs, influenza B virus by one to two logs, and blocked splice product formation in infectious salmon anemia virus, while leaving vesicular stomatitis virus untouched.

Specificity was assessed against an intron-containing reporter, host RNA sequencing and proteomics, all of which showed little change. Mapping localized the sensitive element to the 3 prime splice acceptor region and a short stretch of adjoining open reading frame, but crosslinking immunoprecipitation did not resolve a footprint. Twenty passages under selection produced no true escape, only a segment 8 variant that raises splicing efficiency at a severe fitness cost.

Scientific context

Orthomyxoviruses carry segmented negative sense genomes and complete replication in the nucleus, using splicing and ribosomal frameshifting to expand coding capacity. M2, from segment 7, is the ion channel required for uncoating. NS2, also called the nuclear export protein, exports viral ribonucleoproteins, modulates polymerase activity through enhanced complementary RNA production, and contributes to small viral RNA generation. Slow accumulation of the spliced products has been described as a molecular timer coordinating the infection cycle, a concept developed in part in earlier work from this laboratory. Splicing is ubiquitous in eukaryotes, absent in bacteria, and present in unusual forms in archaea, and influenza virus splicing has been argued to be noncanonical and possibly post-transcriptional, having never been reconstituted in vitro. The paper's stated entry point is that screening archaeal RNA binding proteins might reveal something about that mechanism.

Central question

Do any archaeal RNA binding proteins interfere with the splicing of influenza A virus transcripts, and if so, what does the specificity of that interference reveal about how orthomyxoviruses execute splicing and whether the feature is shared across the family?

Experimental strategy

The screen is the conceptual core. Rather than perturbing host factors, the authors introduce proteins from a domain of life whose RNA biology diverges from the host, on the reasoning that an archaeal protein might recognize a structural feature of viral RNA that the host machinery treats differently. Reading out four viral proteins at once, the unspliced M1 and NS1 alongside the spliced M2 and NS2, makes a splicing-specific hit distinguishable from a general replication defect from the outset.

The hit was then decomposed. Reconstitution with polymerase and viral RNA, and then with polymerase II driven plasmids, removes infection and then the viral polymerase from the requirement. Intronless M2 and NS2 constructs separate an RNA processing effect from protein instability. A recombinant virus in which the same products are made through a 2A peptide provides the decisive genetic test, since a splicing-directed inhibitor should have no effect on it. Orthologue panels across bacteria, protists, archaea, plants, arthropods and vertebrates convert the finding into a statement about lineage specificity, and truncation and alanine substitution map the requirement onto the protein.

Specificity for the virus was tested at three levels, an intron-containing fluorescent reporter, host transcriptome by RNA sequencing, and host proteome by mass spectrometry. Mapping on the RNA side used scanning 30 nucleotide scrambles across segment 7, chimeras between the L7Ae-sensitive viral minigene and the L7Ae-insensitive beta-globin reporter, and crosslinking immunoprecipitation sequencing. Serial passage under selection tests whether the virus can evolve around the inhibitor and at what cost.

Key findings

1. In a screen of codon-optimized archaeal RNA binding proteins expressed in fibroblasts and challenged with A/Puerto Rico/8/34, RBP14 and RBP16 of the RNase III family reduced viral proteins generally, while RBP11 reduced M2 and abolished NS2 despite accumulation of M1 and NS1 (Figure 1A). RBP11 is the L30 family member known in archaea as L7Ae.

2. Reconstitution with the polymerase components, nucleoprotein and individual viral RNA segments showed that L7Ae from Archaeoglobus fulgidus blocked M2 and NS2 with little effect on neuraminidase, and raised M1 and NS1 correspondingly (Figure 1B). Polymerase II driven expression gave the same result, and intronless M2 and NS2 constructs were unaffected, which excludes protein instability as the explanation (Supplementary Figure 1A and 1B).

3. Among L30 orthologues cloned from bacteria, protists, archaea, plants, arthropods and vertebrates, only the archaeal members inhibited M2 and NS2, and five additional archaeal L7Ae proteins all did so (Figure 1C and 1D, Supplementary Figure 1C to 1E).

4. In doxycycline-inducible canine kidney cells, L7Ae induction caused specific loss of M2 and NS2 during H1N1 infection, and reduced wild-type virus titers by approximately three logs, while a recombinant virus expressing M2 and NS2 through 2A peptides was unaffected in both protein expression and titer (Figure 2A to 2D). This is the central genetic demonstration that the target is splicing.

5. L7Ae reduced titers of H1N1 A/California/04/2009, H1N1 A/Texas/36/91, H3N2 A/Panama/99/2007 and live-attenuated H5N1 A/Vietnam/1203/2004 by two to three logs, influenza B virus by one to two logs, and suppressed splice product formation by infectious salmon anemia virus, while having no effect on vesicular stomatitis virus (Figure 2E to 2J and Supplementary Figure 2D).

6. L7Ae did not affect expression from an intron-containing DsRed reporter, changed only 18 host genes with a Pearson correlation of 0.985 across normalized read counts, and produced no significant proteome changes by mass spectrometry (Figure 3B to 3D). Among the few enriched transcripts were small Cajal body-specific RNAs carrying box C and D elements, which the authors suggest were stabilized by L7Ae binding. That suggestion is an interpretation.

7. During infection, M2 and NS2 messenger RNA measured across the spliced junction fell by about 50 percent with L7Ae, while M1 and NS1 rose modestly at 6 hours (Figure 3E). Inhibition required nuclear L7Ae, since a nuclear export sequence tagged version lost activity (Figure 3F and 3G).

8. Truncation showed the flexible alpha 1 and alpha 6 terminal helices are dispensable. Substitutions at lysines 37 and 79, which are known to be required for binding box C and D elements, did not abolish inhibition, while the archaea-defining residues isoleucine 88, glutamate 89 and valine 90 contributed most (Figure 4A to 4E). The authors read the retention of activity by the K37A and K79A mutant as evidence that the viral substrate is not a canonical kink turn.

9. Scanning 30 nucleotide scrambles across segment 7 identified four constructs that abolished M2 on their own by destroying splice sites, the branch point or the pyrimidine tract, while all other scrambles remained L7Ae-sensitive (Figure 5A and 5B). Chimeras with the beta-globin reporter restored splicing in the presence of L7Ae only when the 3 prime splice site of segment 7 was replaced, not the 5 prime site, and deletion series narrowed the requirement to roughly 20 nucleotides of intron upstream of the 3 prime splice site plus about 9 nucleotides of the M2 open reading frame, with a similar result for segment 8 (Figure 5C to 5I).

10. Crosslinking immunoprecipitation sequencing enriched viral RNA relative to control but produced no clear footprint, although the 3 prime splice sites of segments 7 and 8 were captured (Supplementary Figure 4). This is a negative result and the authors attribute it to a dynamic or transient substrate.

11. Twenty passages under L7Ae selection yielded two segment 8 variants and no true escape. The C508A and G576A changes truncated NS1 prematurely. The G60A variant permitted low-level NS2 by increasing splicing efficiency rather than by losing L7Ae sensitivity (Figure 6A). An engineered virus with an optimal vertebrate 5 prime donor site in segment 8 also remained sensitive, and L7Ae still raised NS1 from that construct, which the authors take as further evidence that the 3 prime acceptor region is the point of engagement (Supplementary Figure 5A).

12. In competition, the G60A variant was outcompeted by wild-type virus by passage 2 without L7Ae and dominated with L7Ae, and its titers did not exceed 100,000 plaque forming units per milliliter under either condition (Figure 6B and 6C). Resistance therefore carries a substantial fitness cost.

Mechanistic model

The study does not establish the molecular mechanism, and the authors state their account as a postulate. What the data establish is that an archaeal L30 family protein, acting in the nucleus, prevents maturation of the spliced orthomyxovirus transcripts without requiring the viral polymerase, without acting on the proteins themselves, and without detectable effect on host splicing, and that sensitivity maps to the 3 prime splice acceptor region plus a short adjoining coding stretch in both segment 7 and segment 8.

The proposed model is that orthomyxoviruses form a kink-turn-like structure during lariat formation, used to recruit host splicing components, and that archaeal L7Ae binds that structure and occludes it. Several observations complicate a simple canonical kink turn account, and the authors say so. No canonical kink turn was found in the viral genomes. Mutations that abolish canonical box C and D binding leave the inhibition intact, which points to a substrate that deviates from the canonical fold. Crosslinking did not yield a footprint, and published SHAPE-seq work has also failed to detect the M2 or NS2 lariat, so the authors argue the relevant structure is transient and forms only during splicing. The residues that matter most are the ones that distinguish archaeal L7Ae from its orthologues, which is why the specificity exists, but the RNA counterpart of those contacts is unidentified. The mapping data are consistent with direct engagement near the 3 prime acceptor site, and also with an indirect effect in which L7Ae occupancy elsewhere prevents assembly of the splicing complex at that site. The data do not distinguish these.

Conceptual or technical advance

A single heterologous protein now separates orthomyxovirus splicing from host splicing experimentally, which makes the long-standing claim that influenza virus splicing is noncanonical testable rather than inferential. The same tool works across three genera of the family, including a fish pathogen, so it also converts a feature described for influenza A virus into a family-level property. The finding that the virus cannot escape without either truncating NS1 or driving splicing efficiency up at a large fitness cost is informative about how tightly the timer is constrained. The authors propose therapeutic and transgenic animal applications. Those are prospects, and the reader should note the declared conflict of interest, since tenOever is a co-founder of Archean Biologics and an author of a patent covering commercialization of L7Ae.

Relationship to the broader research program

The paper builds on the laboratory's earlier work on segment 7 and segment 8 processing, in particular the finding that influenza A virus uses suboptimal splice sites to coordinate the timing of infection, published by Chua and colleagues in 2013, and on the splicing-independent recombinant virus constructed in that line of work, which supplies the decisive control here. The related question of what structural element in the NS segment governs this processing is pursued further in ongoing work in the laboratory on a conserved hairpin in segment 8. Category 3 synthesis, visible when this paper is set beside the earlier splicing work, is a sustained argument that viral splice site suboptimality is not incidental but is the mechanism by which the virus schedules its own replication cycle.

Related publications

• Chua and colleagues 2013, Cell Reports, on suboptimal splicing and the timing of influenza A virus infection. Predecessor and methodological foundation. Source of the timer concept and of the splicing-independent 2A recombinant virus used here as the key control.
• Oishi and colleagues on the segment 8 conserved hairpin. Follow-up. Continues the search for the RNA element responsible for the processing behavior probed here.

Limitations and boundaries

Everything is done in cultured cells, predominantly HEK293T, A549 and MDCK lines, with no animal infection, so in vivo efficacy and tolerability of L7Ae expression are untested. L7Ae is delivered by transfection or by doxycycline-induced lentiviral expression, which is a strong and non-physiological expression regime rather than a therapeutic one. Host impact was assessed as transcriptome and proteome under induction in one cell line and a single intron-containing reporter, which does not exclude effects on specific splicing events, on small nucleolar and small Cajal body RNA function, or in other cell types, and the authors note it is unclear how the observed enrichment of those RNAs might affect host biology. The RNA element is localized to a region rather than identified, and the key crosslinking experiment returned no footprint, so direct binding to the proposed structure is not demonstrated. The proposed kink-turn-like structure is inferred from the identity of the protein and from mutational data, not observed. Escape selection was 20 passages from a single starting strain in one cell line, which bounds but does not exclude the emergence of resistance under other conditions. The potency measurements cover a small panel of laboratory strains, and the influenza B virus effect is smaller than the influenza A virus effect. The isavirus experiment reports loss of splice products and not titers. Finally, the authors declare a competing interest tied to commercialization of the protein studied here.

Audience summaries

25 words

A protein borrowed from archaea blocks the splicing step that influenza and related viruses depend on, without disturbing the same process in human cells.

75 words

Influenza viruses make two of their proteins by splicing, and the pace of that splicing schedules their replication. Screening archaeal RNA binding proteins identified L7Ae, which eliminated both spliced products of influenza A virus while leaving the unspliced ones intact. Orthologues from every other domain of life failed to do this. L7Ae also suppressed influenza B virus and a salmon orthomyxovirus, and virus passaged under selection could not escape without a severe fitness penalty.

150 words

Orthomyxoviruses splice segment 7 and segment 8 transcripts to produce M2 and NS2, and the slow accumulation of these products separates early from late infection. A screen of codon-optimized archaeal RNA binding proteins identified the kink-turn binding protein L7Ae as eliminating M2 and NS2 while M1 and NS1 accumulated. The effect reproduced without infection or viral polymerase, required nuclear L7Ae, and disappeared against a virus engineered to make both products through 2A peptides rather than splicing. Only archaeal orthologues had this activity, with the archaea-defining residues contributing most, while substitutions that abolish canonical box C and D binding did not. Host transcriptome, proteome and an intron-containing reporter were largely unaffected. Chimeric minigenes localized sensitivity to the 3 prime splice acceptor region and a short adjoining coding stretch in both segments, but crosslinking yielded no footprint. Twenty passages under selection produced only a variant that raises splicing efficiency at substantial fitness cost.

Discoveries supported by this paper

Documented publication relationships

• Influenza A Virus Utilizes Suboptimal Splicing to Coordinate the Timing of Infection — methodological foundation. Direction: this paper → linked paper. stated in the Related publications section of 2023-oishi-archaeal-kink-turn-binding-protein

Pathogens

Technologies