miRNA does both, but the split is unequal and condition-dependent. In animals, most miRNA-mRNA pairs are repressed at translation with the transcript left largely intact; a smaller, highly consequential subset is degraded through deadenylation and decapping. The dominant outcome depends on base-pairing complementarity, the Argonaute protein loaded, the number of target sites, and the cell state. A search for a single answer returns a half-truth. The accurate answer is that degradation and translational repression are two ends of one continuum, and most endogenous animal targets sit nearer the repression end.

One RISC, Two Outputs

The microRNA-induced silencing complex is built around an Argonaute (AGO) protein. The miRNA guides AGO to a target mRNA, usually through a short seed match at the 5′ end of the miRNA. What happens next is decided by how firmly AGO engages the transcript. In animals, most sites pair through only 6–8 nucleotides. That partial pairing is enough to recruit repressors but not enough to trigger the slicer active site inside AGO2. The result is repression without cutting.

The repression itself runs through two separable arms. The first arm blocks translation initiation. AGO2 recruits the scaffold protein GW182 (TNRC6 in humans), which in turn recruits the CCR4-NOT deadenylase complex and the decapping machinery. CCR4-NOT strips the poly(A) tail, and the mRNA loses the poly(A)-binding protein that helps recruit ribosomes. Decapping follows, and the unprotected transcript is destroyed by XRN1. This is genuine degradation, not just stalling.

The second arm acts on translation directly. GW182 recruitment interferes with the cap-binding initiation factor eIF4F and with the 60S ribosomal subunit joining. Ribosomes fall off or fail to load efficiently. The transcript persists but produces less protein per copy. These two arms are not mutually exclusive; the same complex can throttle translation first and dismantle the transcript later.

The deciding variable is complementarity. When a miRNA pairs with near-perfect complementarity across the coding region, plant-style, AGO2's slicer domain cleaves the mRNA between positions 10 and 11 of the miRNA. The cut transcript is then degraded by cellular exonucleases. In animals, such near-perfect sites are rare, which is why slicing is the exception rather than the rule. A plain-language way to hold this: the miRNA can either jam the assembly line or send the blueprint to the shredder, and the seed match decides which.

How Much Degradation, How Much Repression

Quantifying the split has occupied the field for over a decade, and the numbers are worth stating carefully because they are often misquoted. Ribosome-profiling and mRNA-seq studies in mammalian cells found that for many miRNA targets, the drop in protein output exceeds the drop in mRNA level. In several datasets, mRNA abundance fell modestly while translation efficiency fell more sharply, implying repression dominates. The ratio varies by target, by cell type, and by the specific miRNA.

The reason the ratio drifts is mechanistic. A target with a single weak seed site tends to be repressed translationally with little decay. A target with multiple seed sites, or with a site in a favorable 3′ UTR context, recruits more CCR4-NOT and tips toward deadenylation and decay. Cooperativity matters: two nearby sites can produce more than additive repression. This is why the same miRNA can behave as a pure repressor at one target and a near-silencer at another.

A counterintuitive detail worth flagging: the poly(A) tail shortening that precedes decay is often the first measurable event, and it can be detected before any change in mRNA abundance. Translational repression and deadenylation are kinetically intertwined rather than sequential. A transcript can be translationally silenced while its tail is being trimmed, then decay once the tail falls below a threshold length. Reporting "miRNA degrades mRNA" or "miRNA only represses translation" both flatten this timeline.

Timeline of a miRNA–Target Encounter

Time after miRNA loading Molecular event mRNA level Protein output
Seconds to minutes AGO–GW182 engagement, eIF4F interference Unchanged Begins to fall
Minutes CCR4-NOT recruitment, poly(A) tail shortening Unchanged or slightly lower Falling
Minutes to hours Decapping, XRN1 attack on the body Detectable decline Low
Hours Transcript cleared or held at low steady state Reduced Sustained low
Hours to days Target site occupancy, feedback on miRNA levels Depends on feedback Depends on feedback

The table is a qualitative schematic. It shows the ordering the field broadly agrees on, not precise kinetics, which differ by system. What it makes visible is that translational repression and degradation are not alternatives chosen once; they unfold in sequence, with repression often leading.

Edge Cases That Break the Simple Rule

Perfect complementarity in animals does occur, chiefly in a few developmental and viral contexts, and it switches AGO2 into slicer mode. When that happens, the mRNA is cut and degraded, and the miRNA behaves like a small interfering RNA. This is the clearest case of genuine miRNA-directed degradation.

Non-canonical sites complicate the picture further. Some miRNA-target interactions rely on 3′ compensatory pairing rather than the seed, producing weak repression with minimal decay. Others involve sites in the coding sequence or the 5′ UTR, where the effector machinery behaves differently. A site in the coding region can be repressed translationally while the transcript stays stable, because the decay machinery is less effectively recruited there.

Cell state changes the output. During stress, neurons, and certain developmental transitions, miRNA-mediated repression can be reversed, and stored mRNAs are reactivated. The transcript was never degraded; it was held. Conversely, in rapidly dividing cells, decay can dominate because the deadenylation machinery is abundant and active. The same miRNA-mRNA pair can therefore be a degradation target in one tissue and a translation-repression target in another.

A professional caveat: many "miRNA targets" reported in databases rest on computational seed prediction alone. Predicted sites vastly outnumber validated ones, so claims about whether a given pair is degraded or merely repressed require experimental confirmation, typically ribosome profiling plus mRNA half-life measurement. Do not treat a predicted site as an established mechanism.

Misconceptions Worth Retiring

"miRNA always degrades its targets." False. In animals, translational repression with transcript stability is common, sometimes the majority outcome for a given pair. "Translational repression and degradation are separate pathways." They share the same core machinery, GW182 and CCR4-NOT, and often occur in sequence. "A seed match guarantees repression." Seed matches are necessary but not sufficient; site number, spacing, and 3′ UTR context set the strength. "Perfect complementarity is the animal norm." It is the exception; most animal sites are partial and do not trigger slicing. "If mRNA levels are unchanged, the miRNA is inactive." A transcript can be strongly repressed at the protein level with stable abundance.

These errors propagate because mRNA-seq alone cannot distinguish repression from decay. If you measure only transcript abundance, you will systematically miss translational repression, and you may wrongly conclude a miRNA is inactive. Pair abundance data with translation measurements or half-life assays before drawing conclusions.

Practical Actions for Researchers and Readers

If you are evaluating a claim about miRNA mechanism, work through a fixed sequence.

Confirm the target site: seed match, position in the UTR, number of sites, and conservation. A single predicted site is weak evidence. Measure both layers: mRNA abundance and translation (ribosome profiling or polysome fractionation). One layer alone cannot separate the two mechanisms. Test transcript stability directly with transcription shutoff (actinomycin D or a comparable approach) and compare half-lives with and without the miRNA. Check AGO loading: confirm the miRNA actually associates with AGO in the relevant cell type, ideally by immunoprecipitation. Look for slicing: if near-perfect complementarity is present, assay for the cleavage product before assuming repression. Reconcile thresholds explicitly. If one assay suggests decay and another suggests stable transcripts, report both and state the discrepancy rather than choosing the convenient one.

For readers of review articles, the actionable filter is simple: a claim that a miRNA "degrades" a target should be backed by a measured drop in transcript half-life, not just lower abundance. A claim of "repression only" should be backed by stable transcript levels plus reduced translation. Absent both, treat the mechanism as unresolved.

Limits of What the Evidence Supports

Several boundaries deserve honest statement. Most quantitative splits between degradation and repression come from a limited set of cell lines and reporter systems, so generalizing to primary human tissue is not guaranteed. Reporter assays with bulged or perfect sites measure what the site can do, not always what the endogenous transcript does in its native context.

The kinetic ordering is broadly accepted but not precisely parameterized across systems. Numbers for how fast poly(A) tails shorten or how quickly decapping follows vary substantially, and no single threshold applies universally. Presentations that quote one conversion or one half-life as universal should be treated with caution.

Finally, distinguishing repression from decay in vivo remains technically demanding. Ribosome profiling captures a snapshot, and transcription shutoff has its own artifacts. The strongest conclusions combine multiple orthogonal measurements. Where such data are absent, the correct stance is that miRNA can degrade mRNA and can repress translation, and that the balance for any specific pair is an empirical question, not a default.