Surface immune signaling unlocks NLR activation through mRNA alternative splicing
The potato late blight receptor Rpi-vnt1.1 is held inactive by an N-terminal extension, and surface immune signalling splices that extension out of the mRNA, so pattern recognition licenses the intracellular receptor before an effector ever arrives.
PTI and ETI are usually described as parallel layers. This paper wires them in series. Rpi-vnt1.1 is transcribed as two isoforms: one retaining an intron that adds an N-terminal extension ahead of the MADA motif, one spliced to give a conventional MADA-NLR. The retained form barely oligomerises, barely conducts calcium, and barely kills cells. PAMP treatment shifts the balance towards the spliced form, at both transcript and protein level, without changing total transcript. Lock the isoform either way and PTI can no longer enhance the response, placing surface signalling upstream of splicing. The requirement is BAK1-dependent, and transgenic potato locked in the retained isoform loses late blight resistance. Extensions longer than 16 residues are inhibitory generally, and 619 Solanaceae MADA-NLRs carry one.
Key findings (11)
PTI is required for full Rpi-vnt1.1-mediated immunity. INF1 and flg22, but not the non-recognised nlp20, accelerated Dex-induced ion leakage, enhanced hypersensitive response and increased H2O2 accumulation on AVRvnt1 recognition. The effect was not explained by altered Rpi-vnt1.1 transcript or protein levels, nor by AVRvnt1 accumulation.
Evidence dexamethasone-inducible AVRvnt1 expression in Rpi-vnt1.1 transgenic N. benthamiana; ion leakage, HR and DAB staining; controls in bak1 and eds1 backgrounds
Source Figures 1A-1C, 1D-1F; figures S1, S2
Rpi-vnt1.1 produces two isoforms differing at the 5' end: an intron-retained form (Rpi-vnt1.1-IR) carrying an N-terminal extension, and an intron-spliced form (Rpi-vnt1.1-IS) beginning at a second start codon with the canonical MADA motif leader.
Evidence transcriptome sequencing of Solanum okadae treated with flg22; RNA-seq read coverage across the 5' coding region
Source Figures 2A and 2B; figure S3
PAMP treatment shifts the isoform balance without changing total transcript. INF1 and flg22, but not nlp20 or mock, promoted splicing in S. okadae, S. venturii, S. tuberosum C1848 and the Desiree Rpi-vnt1.1 transgenic line.
Evidence RT-PCR with separate primers for total and alternatively spliced transcript, ELF1a reference
Source Figures 2C and 2D; figure S6
The shift is real at protein level. Parallel reaction monitoring mass spectrometry with an IR-specific peptide and a shared IR/IS peptide showed flg22 markedly reduced the IR-specific signal and significantly decreased the IR/total ratio (Student's t test, P < 0.001).
Evidence isoform-resolved PRM mass spectrometry in S. okadae
Source Figures 2E-2G; figure S8
A YFP splicing reporter reproduces the switch. YSR, fusing the Rpi-vnt1.1 5' region (UTR, intron and MADA coding sequence) to YFP, predominantly gave the 35 kDa IR variant; INF1 or flg22 treatment raised the 31 kDa IS variant above it.
Evidence YSR reporter with YSR-IR and YSR-IS controls, transiently expressed in N. benthamiana; western blot
Source Figures 2H-2J
Locking the isoform in either direction makes the receptor insensitive to PTI. Mutating U12 splicing acceptor sites (IR1) or the second start codon (IR2) preserves retention; truncating the intron (IS1) or mutating the first start codon (IS2) enforces splicing. In all four, PAMP treatment no longer altered isoform ratio, placing PTI upstream of splicing.
Evidence engineered YSR mutants; western blot after mock, INF1, flg22 or nlp20 treatment
Source Figures 3A and 3B; figure S10
Only the spliced isoform is immune-competent. In full-length Rpi-vnt1.1, IR1 gave a weak response and IR2 failed entirely, phenocopying the P-loop mutant Rpi-vnt1.1-KN, while both IS variants behaved like wild type. All accumulated comparably. BN-PAGE showed wild-type and both IS variants oligomerise strongly on AVRvnt1 activation, IR1 markedly less, and IR2 and KN not at all. Calcium influx followed the same pattern.
Evidence ion leakage, HR, DAB staining, BN-PAGE, co-immunoprecipitation and cell-based calcium influx assays
Source Figures 3C-3F; figures S9B, S12, S13
The inhibition is a property of length, not sequence. Fusing N-terminal extensions of varying length from the Rpi-vnt1 family and from MADA-NLRs of other species onto Rpi-vnt1.1-IS1 showed extensions shorter than 16 amino acids had little effect, while those longer than 16 markedly suppressed HR.
Evidence HR assays with a graded series of heterologous N-terminal extensions
Source Figure S11
AlphaFold3 Multimer models the difference. Rpi-vnt1.1-IS forms a stable resistosome with a well-ordered funnel; Rpi-vnt1.1-IR fails to assemble one. Across 85 curated N-extended MADA-NLRs, full-length models consistently predicted the N-terminal residues as low-confidence flexible segments that did not stabilise in the pore-forming region and made misplaced lipid engagements, frequently displacing or occluding the MADA alpha1 pore helix.
Evidence AlphaFold3 Multimer resistosome prediction across 85 N-extended MADA-NLRs
Source Figure S14; data S9
The splicing requirement holds during real infection and depends on BAK1. P. infestans strains T30-4, JH19 and P13527 (all expressing AVRvnt1) and the AVRvnt1-silenced P13626 all enhanced splicing without changing total transcript, so induction does not require the effector. Silencing BAK1 greatly suppressed P. infestans-induced splicing, reproduced in N. benthamiana bak1 mutants. Transgenic Desiree lines locked in IR1 showed significantly reduced late blight resistance against the full resistance conferred by wild-type Rpi-vnt1.1.
Evidence P. infestans infection of S. okadae and transgenic Desiree potato; PVX-mediated gene silencing; bak1 mutants; lesion area quantification
Source Figures 4D-4H; figures S15-S18
N-terminal extensions are widespread. Screening 4,370 MADA-NLRs from 133 angiosperm genomes found 218 with similar N-terminal extensions, 87 of which carry a second methionine at the start of the MADA motif. In 8,530 MADA-NLRs from 346 reannotated Solanaceae genomes, 619 have the feature, and these are phylogenetically closer to singleton NLRs such as Rpi-vnt1.1.
Evidence comparative genomic screen across angiosperm and Solanaceae genomes; phylogenetic placement
Source Figures S4A, S4B, S5
NLRs
| Name | Class | Role in this paper |
|---|---|---|
| Rpi-vnt1.1 Rpi-vnt1.1-IR, Rpi-vnt1.1-IS, Rpi-vnt1.1-IR1, Rpi-vnt1.1-IR2, Rpi-vnt1.1-IS1, Rpi-vnt1.1-IS2, Rpi-vnt1.1-KN (P-loop mutant) | singleton subject | central subject; the receptor whose activation is gated by alternative splicing of its own mRNA |
| RB | singleton comparator | unrelated NLR used as a negative control in co-immunoprecipitation |
| N-extended MADA-NLRs | singleton subject | the wider class that makes the mechanism potentially general; phylogenetically closer to singleton NLRs |
Effectors
| Name | Pathogen | Role |
|---|---|---|
| AVRvnt1 | Phytophthora infestans | the activating effector; notably, PTI-induced splicing occurs whether or not AVRvnt1 is expressed, so the priming step is effector-independent |
| INF1 | Phytophthora infestans | one of two PAMPs that trigger splicing |
| flg22 | bacterial flagellin epitope | the second PAMP that triggers splicing; used for RNA-seq and mass spectrometry |
| nlp20 | various | negative control PAMP throughout, and the reason the effect can be attributed to surface recognition rather than infiltration |
| PehC | Ralstonia solanacearum | used in a published tomato RNA-seq dataset to test whether PTI-induced splicing of immune receptor genes is general |
Structures
| Entry | State | Description |
|---|---|---|
| AlphaFold3 Multimer predicted | resistosome (predicted); protomer count not stated in the extracted text | Rpi-vnt1.1-IS resistosome with a well-ordered funnel-shaped architecture; Rpi-vnt1.1-IR fails to assemble an equivalent structure |
| AlphaFold3 Multimer predicted | resistosome (predicted) | 85 curated N-extended MADA-NLRs modelled as full-length resistosomes; N-terminal residues consistently low-pLDDT and flexible, with misplaced lipid engagement and displacement or occlusion of the MADA alpha1 pore helix |
Open questions
- The splicing machinery itself is unidentified. The paper states the precise components and mechanism linking PTI signalling to U12-type splicing remain undetermined.
- Whether the 619 Solanaceae and 218 angiosperm N-extended MADA-NLRs are regulated the same way is untested; only Rpi-vnt1.1 was examined functionally.
- Why extensions longer than 16 amino acids are inhibitory while shorter ones are not is described but not explained mechanistically.
- There is no experimental structure of either isoform. The structural account rests entirely on AlphaFold3 Multimer models.
- Transgenic lines locked in the spliced isoform showed uniformly low expression and only quantitative resistance, which the authors attribute to counterselection against autoimmunity. That interpretation is not directly tested.
- How the isoform balance is set in the absence of infection, and whether it varies by tissue or developmental stage, is not addressed.
Connections in this corpus
| Paper | Why |
|---|---|
ibrahim-2026-nrg1-chloroplast | Six co-authors in common (Ibrahim, Toghani, Yuen, Eilmann, King, Wang) and a shared Bozkurt-lab background, so the pairing is institutional as well as intellectual. Both concern how and where NLR activation is controlled rather than how effectors are recognised: Ibrahim by subcellular compartment, Gao by transcript isoform. |
gomezdelacruz-2026-mla3-mimicry | The two papers in the corpus without Kamoun authorship, and complementary halves of the recognition problem. Gomez De La Cruz asks how a receptor acquires the ability to bind an effector; Gao asks what has to happen before a receptor that already can bind one is allowed to fire. Both then engineer the controlling element. |
marques-2026-legume-tissue | Both concern regulation of NLR availability rather than recognition specificity. Gao finds an NLR held inactive by an mRNA isoform until surface signalling removes the block; Marques finds NLRs restricted to the tissue where they are needed. Transcript-level and organ-level answers to the same problem of keeping immune receptors off until and where they are wanted. |
guo-2026-ccg10-resistosome | Both bear on what the N terminus does to resistosome assembly. Guo resolves a CC domain rotated away from the LRR in a receptor lacking the EDVID motif; Gao shows an N-terminal extension ahead of the MADA motif that prevents funnel formation altogether. Both also use BN-PAGE oligomerisation as the readout that separates assembly-competent from assembly-blocked variants. |
seager-2026-avrcap1b-nrc3 | Two ways to stop a resistosome forming, one endogenous and one imposed. Gao has an autoinhibitory N-terminal extension that prevents assembly from the inside; Seager has an effector that jams assembly from the outside. Both use BN-PAGE to distinguish partial from absent oligomerisation, and both find intermediate states rather than a binary switch. |
toghani-2026-sni-undecamer | Both use AlphaFold3 across a large curated set of MADA-NLRs to reason about resistosome assembly, and both treat low-confidence regions as informative rather than discardable. Gao reads low pLDDT in N-terminal extensions as evidence they cannot stabilise in the pore; Toghani reads low ipTM across a clade as evidence of an atypical architecture. |
Provenance and caveats
Not a Kamoun-lab paper. Kamoun is not an author and the lab is not acknowledged. Six Bozkurt-lab and TSL-affiliated co-authors overlap with ibrahim-2026, which is the real connection. The About page should be clear that inclusion reflects topical scope, not authorship.
No experimental structure of either isoform exists. The claim that the N-terminal extension prevents funnel formation rests on AlphaFold3 Multimer models across 85 N-extended MADA-NLRs, supported by BN-PAGE oligomerisation and calcium data. The functional data carry the claim; the models illustrate it.
The predicted resistosome protomer count for Rpi-vnt1.1 is not stated in the text extracted. [UNVERIFIED - do not assert a stoichiometry for the Rpi-vnt1.1 resistosome without checking figure S14.]
Transgenic lines locked in the spliced isoform all showed low expression and only quantitative resistance. The authors attribute this to counterselection against autoimmunity, supported by untagged IS1 overexpression triggering HR without AVRvnt1, but the interpretation is inferred rather than demonstrated.
The isoform naming is easy to garble: IR is the intron-retained, N-extended, immune-incompetent form; IS is the intron-spliced, conventional, competent form. IR1 and IR2 are two different engineering routes to the same locked-retained state, as are IS1 and IS2 for the locked-spliced state. They must not be presented as four distinct phenotypes.
Sources
- Full citation: Gao C, Meng X, Chen X, Yang L, Ibrahim T, Toghani A, Yuen ELH, Eilmann N, King F, Li K, Wang L, Sun B, Wang Y, Bozkurt TO, Dong S (2026) Surface immune signaling unlocks NLR activation through mRNA alternative splicing. Science 393:65-70.
- DOI: 10.1126/science.adx9929
- PDF on file:
papers/Science_2026b.pdf - Extraction:
data/papers/gao-2026-vnt1-splicing.json(schema v1.0) - References: 58 works, retrieved from OpenAlex
- Licence: [UNVERIFIED] - not stated on PDF