Can genetically engineered RNA inducers of proximity regulate protein activity?

A two-motif RNA that recruits enzymes

GENRIP stands for genetically engineered RNA inducers of proximity. The design is a single RNA molecule with two recognition motifs joined by a short linker: a variable motif that binds the protein of interest (POI), and a motif that binds an effector enzyme able to modify that protein. Because the enzyme is brought next to its substrate, the RNA does not need to catalyse anything itself; it only has to hold the two proteins close enough for modification to occur.

The effector side is the unusual part. Instead of recruiting a general enzyme, the authors chose enzymes that already bind RNA and modify proteins, so the RNA motif can dock the enzyme directly. They name this class RNA-binding enzymes and use two examples: the E3 ubiquitin ligase MEX3A, which carries KH domains for RNA binding plus a RING domain for ubiquitination, and the kinase CamK. The POI side is more flexible: its motif can come from an endogenous RNA-binding protein, from an in vitro-selected aptamer, or from an engineered tag such as GFP fused to the target.

The RNA must form a stable ternary complex

The first test asked whether a GENRIP built for QKI and MEX3A, called RNA-AQ, actually binds both proteins at once. Molecular dynamics simulation indicated that RNA-AQ, MEX3A and QKI can form a stable ternary complex with a root mean square deviation below 5.0 Å. In RNA pull-down assays, biotin-labelled RNA-AQ enriched both QKI and MEX3A from cell lysates, whereas control RNAs carrying only one of the two motifs could not enrich both proteins simultaneously.

A FLIM-FRET assay then measured whether the RNA changes the distance between the two proteins inside living cells. Cells expressing mNeonGreen-QKI and mScarlet-MEX3A showed a significantly shortened mNeonGreen fluorescence lifetime when RNA-AQ was present, while control RNA sequences did not produce a significant effect. This is a proximity measurement, not a degradation measurement, and it supports the intended mechanism: the RNA shortens the QKI–MEX3A distance rather than merely binding one partner.

QKI loss depends on MEX3A and the proteasome

In KYSE150 cells, which have relatively high levels of both QKI and MEX3A, RNA-AQ reduced QKI protein in a time-dependent and concentration-dependent manner, and the degree of degradation correlated positively with the RNA-AQ mRNA level. The effect was specific to the protein: RNA-AQ had minimal impact on QKI and MEX3A RNA levels, and GENRIPs lacking either the QKI motif or the MEX3A motif failed to down-regulate QKI. The same QKI degradation was reproduced in SKOV3, PC9, PANC1 and 4T1 cells.

Three experiments tie the loss to the intended pathway. Knocking down MEX3A with lentiviral shRNA reduced RNA-AQ-driven QKI degradation from 55% to 19%. The proteasome inhibitor MG132 reversed QKI loss, whereas chloroquine, which blocks lysosomal acidification, did not. Immunoprecipitation showed that RNA-AQ increased QKI ubiquitination. Together these results indicate that the QKI reduction is MEX3A- and proteasome-dependent, which is the causal chain the design predicts.

Endogenous MEX3A targets stay largely unchanged

Recruiting a native E3 ligase raises an obvious concern: the RNA might redirect MEX3A away from its normal substrates or perturb them indirectly. The authors addressed this with RIP-seq to identify mRNAs bound by MEX3A and RNA-seq to measure transcript changes after RNA-AQ transfection. RNA-AQ did not alter the mRNA levels of reported MEX3A targets such as CDKN1A, CDKN2B and E2F7, nor did it affect the vast majority of newly identified MEX3A-bound transcripts.

This is transcript-level evidence, so it speaks to mRNA abundance rather than to protein stability of those other substrates. Within that scope, the result supports the claim that RNA-AQ has minimal impact on endogenous MEX3A substrates, which matters because the platform's selectivity argument rests on the RNA motif directing the enzyme rather than on the enzyme being inert.

Motif order and folding set the potency

The authors compared variants that differ in motif order, motif number and linker length. RNA-QA, with the motifs in the opposite order to RNA-AQ, degraded QKI less efficiently. RNAfold secondary-structure prediction offered an explanation: RNA-AQ retained the complete structures needed for both MEX3A and QKI binding, whereas RNA-QA lacked a structure that can bind MEX3A. Adding an extra MEX3A-binding module to RNA-QA restored the dual structural features and improved degradation relative to RNA-QA and RNA-QAQ, while adding extra modules to RNA-AQ did not further increase correctly folded structures or improve degradation.

The practical rule is that sequence optimisation should be guided by predicted RNA folding, because a motif that is present in the sequence may not be available in the folded structure. The same section shows the platform's reach: an RNA aptamer against Tau, a protein without a natural RNA-binding motif, produced degradation of Tau, and a three-motif RNA-TAQ reduced QKI and Tau protein levels simultaneously.

Promoters add cell-type and compartment control

Because GENRIP is expressed from a plasmid, its production can be placed under existing genetic control elements. The authors tested the liver-specific TBG promoter: GFP driven by TBG was detected in L02 liver cells but not in 293T kidney cells, while a ubiquitous CMV promoter gave comparable RFP expression in both. TBG-driven RNA-AQ substantially decreased QKI in L02 cells but had minimal influence on QKI in 293T cells.

Promoter choice also shifted where the RNA acted within the cell. In KYSE150 cells, CMV-AQ and U6-AQ were compared for QKI protein levels in the cytoplasmic and nuclear fractions, linking the expression element to subcellular localisation of the effect. This is the basis for the paper's claim of cell-type selectivity and precise subcellular targeting, and it is a property inherited from the promoter rather than engineered into the RNA itself.

CamK pairing activates rather than destroys

Degradation is only one output. Using the RNA-binding kinase CamK as the effector, the authors report phosphorylation-induced activation of the POI, with Tau as the target and CamK-mediated phosphorylation contributing to Tau aggregation. This is the opposite direction of regulation from the MEX3A case: the same two-motif architecture is claimed to support both ubiquitination-mediated degradation and phosphorylation-induced activation, which is the core of the modularity argument.

The paper also tests whether the QKI-directed RNA has phenotypic consequences. RNA-AQ reduced viability of KYSE150 cells relative to a mix of the single-motif RNAs, altered colony formation and Annexin-V/PI apoptosis readouts across transfection doses, and, when delivered by adenoviral vector or by liposome-mediated in vitro transcribed RNA, reduced tumour growth in mice bearing subcutaneous tumours, with QKI protein measured in the treated tumours. These are efficacy experiments in specific models, not evidence that the approach works across tumour types or in patients.

What the evidence supports and what it does not

The strongest claim the data support is mechanistic: for the QKI–MEX3A pair, RNA-AQ forms a ternary complex, shortens the intermolecular distance, increases QKI ubiquitination, and lowers QKI protein in a MEX3A- and proteasome-dependent way, without changing QKI mRNA. The CamK–Tau arm and the multi-target and aptamer variants extend the design space, but they are proof-of-concept demonstrations rather than a systematic comparison of efficiency across effectors.

Two boundaries follow from the design. First, GENRIP depends on an endogenous RNA-binding enzyme being present and usable in the target cell, which is why the authors argue that building systems on different RNA-binding E3 ligases matters given heterogeneity in E3 ligase expression across cell types and tissues. Second, the selectivity evidence is transcript-level for bystander MEX3A substrates, and the in vivo work uses specific tumour models and delivery routes. The paper's own framing is that this bridges direct genetic control at the protein level; the reader's takeaway should be that the QKI degradation arm is the best-supported case, while activation, multiplexing and therapeutic use remain demonstrations whose generality is still to be established.

Sources used in this answer