Nature Biotechnology
(2026) Cite this article
Knockdown efficiency of RNA-targeting CRISPR systems is commonly measured by quantitative polymerase chain reaction with reverse transcription (RT–qPCR). Here we discovered that guide RNAs copurify during RNA extraction and inhibit RT–qPCR for amplicons spanning or upstream of the guide RNA binding site, resulting in overestimation of knockdown efficiency across all CRISPR systems tested. We recommend using a processive reverse transcriptase with strong strand-displacing activity, with orthogonal methods, to ensure accurate quantification when using RT–qPCR.
RNA knockdown (KD) avoids genome alteration and is used widely in biological studies. KD can be achieved by small interfering RNAs (siRNAs) and short hairpin RNAs (shRNAs), antisense oligonucleotides (ASOs) or RNA-targeting CRISPR systems1,2,3,4, the latter being especially useful for nuclear-localized transcripts5,6. Single-effector systems, including PspCas13b and RfxCas13d (CasRx), achieve strong on-target KD but may cause collateral RNA cleavage6. In contrast, the multisubunit Csm system mediates efficient KD without detectable collateral activity7. Here, while comparing Csm with other systems, we identified a pervasive reverse transcription (RT)-based quantification artifact, defined its mechanism and propose strategies to avoid it.
To reproduce published Csm-mediated KD, we targeted nuclear-localized XIST and cytoplasmic BRCA1 using reported guide RNAs (gRNAs), primer sets and a nuclear-localized all-in-one Csm construct7. We observed KD efficiencies similar to those reported7. Unexpectedly, the RNase-dead (RD) mutant (Csm3-D33A) exhibited apparent KD comparable to wild-type (WT) Csm (Fig. 1a–c), despite no protein-level KD (Extended Data Fig. 1a). DNase-dead (DD; Csm1-D16A)8 and a double mutant (DM; Csm1-D16A, Csm3-D33A) Csm also yielded strong apparent KD, independent of the RNA extraction method (Extended Data Fig. 1b).
a, Schematic of RNA-targeting CRISPR-mediated RNA cleavage and decay. b, Experimental workflow. c, RT–qPCR with amplicons spanning XIST or BRCA1 gRNAs shows apparent KD by both WT and RD Csm. d, Flow cytometry shows mCherry protein KD by WT, but not RD Csm. e, RT–qPCR from samples in d shows apparent KD by both WT and RD Csm with a spanning, but not an upstream, amplicon. nt, nucleotide. f,g, RT–qPCR of BRCA1 (f) or XIST (g). For RD Csm, spanning and upstream amplicons report apparent KD, whereas downstream do not. For WT Csm, the spanning amplicons overestimate KD. Spanning amplicon data are duplicated from c for comparison. h, Flow cytometry shows mCherry protein KD by WT PspCas13b and CasRx, but not RD mutants. i, RT–qPCR from samples in h shows apparent RD-mutant KD with a spanning amplicon, whereas a downstream amplicon detects KD only for WT. In e–g and i, green arrows indicate gRNA binding sites; amplicon positions and distances are shown above; bar colors correspond to amplicons. Two-tailed one-sample t-test and two-tailed two-sample Welch’s t-test (c,f,g,i). Two-tailed one-sample t-test (d,e,h). Biological replicates (c–i): n = 3. Data are mean ± s.d.
Apparent KD by RD Csm has been attributed to ribonucleoprotein binding within open reading frames, potentially interfering with translation or nuclear export and promoting transcript decay9. However, RD Csm produced apparent KD of XIST (Fig. 1c), which is neither translated nor exported. To test whether binding alone induces RNA decay, we constructed an mCherry reporter enabling RNA and protein measurement from the same sample. WT Csm with an open reading frame-targeting gRNA reduced mCherry protein robustly, whereas RD Csm did not (Fig. 1d). DD Csm reduced protein similarly to WT, whereas RD and DM did not (Extended Data Fig. 1c), indicating that Csm3 RNase activity drives true KD and Csm1 DNase activity is dispensable. Thus, binding alone does not explain apparent KD by RD Csm.
We next measured mCherry RNA by quantitative polymerase chain reaction with RT (RT–qPCR) using an amplicon spanning the cleavage site—a common strategy to avoid amplifying cleaved fragments2,7,9. All RD mutants showed strong apparent KD for this amplicon (Fig. 1e and Extended Data Fig. 1d). In contrast, an upstream amplicon showed KD for WT but not RD Csm (Fig. 1e), consistent with protein measurements. Thus, spanning amplicons falsely report KD.
Because the original XIST and BRCA1 amplicons also spanned the gRNA binding sites, reported WT Csm KD efficiencies may have been inflated. We therefore designed upstream and downstream amplicons. For RD Csm targeting BRCA1, the downstream amplicon correctly showed no KD, whereas spanning (78%) and upstream (42%) amplicons exhibited apparent KD (Fig. 1f). For WT Csm, all amplicons showed KD, but the spanning amplicon consistently yielded higher KD (92%) than the downstream one (64%) (Fig. 1f). This effect was stronger for XIST: the spanning amplicon reported strong KD for both WT and RD Csm (90%), whereas the downstream amplicon showed substantially lower KD for WT (35%) and minimal KD for RD (14%; Fig. 1g). Together, downstream and distal upstream amplicons more accurately report KD, whereas spanning amplicons likely overestimate it.
We then tested whether other RNA-targeting CRISPR systems exhibit this limitation2,3. PspCas13b or CasRx were cotransfected with the mCherry reporter and mCherry-targeting gRNAs. WT enzymes efficiently reduced mCherry protein, whereas RD mutants did not (Fig. 1h). At the RNA level, however, spanning amplicons reported significant KD for both WT and RD enzymes (Fig. 1i). In contrast, downstream amplicons showed strong KD only for WT enzymes, whereas RD mutants matched the nontargeting (NT) control (Fig. 1i). Thus, RT–qPCR with spanning amplicons overestimates KD across all systems tested.
To determine the mechanism underlying RT–qPCR misquantification, we first tested whether Csm ribonucleoprotein remains bound to target RNA after extraction. Proteinase K treatment of extracted RNA did not rescue apparent KD measured with the spanning amplicon, ruling out residual protein as the cause (Fig. 2a). Next we hypothesized that gRNAs copurify with cellular RNAs and directly inhibit RT–qPCR. A synthetic mature Csm gRNA10 spiked into the RT reaction caused concentration-dependent apparent KD for upstream and spanning amplicons, whereas downstream was unaffected (Fig. 2b). Thus, gRNA alone is sufficient to inhibit RT–qPCR. Transfection of a gRNA plasmid alone did not produce apparent KD9 (Extended Data Fig. 1e), suggesting that Csm proteins are required to stabilize gRNAs in cells, enabling accumulation and copurification. To separate RT from qPCR, synthetic gRNAs were spiked into qPCR reactions. Spanning amplicons were inhibited much less than RT spike-in (Extended Data Fig. 1f). Furthermore, RNase treatment of cDNA did not rescue the affected amplicons (Extended Data Fig. 1g), indicating that inhibition occurs primarily during RT.
a, RT–qPCR of BRCA1 after treatment with Proteinase K. Residual protein does not cause inhibition. Two-tailed one-sample t-test. b, Synthetic BRCA1-targeting gRNA spiked into RT inhibits spanning and upstream, but not downstream, amplicons. c, Complementary ddASO in RT sequesters gRNA and rescues upstream and spanning amplicons, whereas downstream amplicons are unaffected; one-way ANOVA (b,c). d, Relative expression of all RT–qPCR amplicons plotted by distances from gRNAs. Only spanning and proximal upstream amplicons are inhibited. Upstream inhibition was fit to a one-component association model: decay constant 170 nt, 95% confidence interval, 73–390 nt. e, uMRT RT–qPCR is not inhibited by gRNA and enables accurate spanning amplicon quantification. RD samples: two-tailed one-sample t-test. WT samples: Welch’s ANOVA. Amplicon color (a–e): magenta, spanning; blue, upstream; cyan, downstream. f, Model: copurified gRNA binds uncleaved RNA and inhibits RT–qPCR; ddASO sequestration or uMRT displacement of gRNAs rescues quantification. Biological replicates: n = 3 (a–d), NT (e) and BRCA1 RD n = 3; otherwise, n = 2. Data are mean ± s.d.
To confirm gRNA-mediated inhibition, we designed a 3′ dideoxycytidine-capped ASO complementary to the mature gRNA (ddASO). Adding the ddASO before RT rescued the spanning and upstream amplicons in RD samples in a concentration-dependent manner, whereas the downstream amplicon was unaffected (Fig. 2c). In WT samples, ddASO shifted the spanning amplicon toward the downstream measurement, indicating that the spanning amplicon had been depressed artificially (Fig. 2c). Thus, downstream amplicons provide a more accurate KD measurement under standard RT conditions.
The RT-inhibition by gRNA depends on amplicon position. With random hexamer priming, RT can initiate throughout the transcript. RT initiated downstream of the gRNA may be blocked by the guide, whereas RT initiated upstream will not. Accordingly, proximal upstream amplicons should be inhibited, distal upstream less so and downstream unaffected. Consistent with this model, upstream inhibition decreased with distance from the gRNA and disappeared beyond ~500 nt (Fig. 2d), whereas downstream amplicons showed no distance dependence (Fig. 2d).
Although downstream amplicons avoid the artifact, spanning amplicons are preferable because cleaved fragments could contribute to quantification. We therefore tested ultraMarathonRT (uMRT), a processive reverse transcriptase with strong strand-displacement activity11. With uMRT, RD Csm and RD PspCas13b produced no apparent KD with any amplicon (Fig. 2e and Extended Data Fig. 1h), indicating that uMRT is not inhibited by bound gRNAs. For WT enzymes, uMRT yielded similar KD across amplicons and matched downstream measurements from conventional RT (Fig. 2e), suggesting that fragment stability affects quantification minimally.
In summary, gRNA-target binding can inhibit RT–qPCR for spanning- or proximal upstream- amplicons, inflating measured KD (Fig. 2f). This artifact affects all tested RNA-targeting CRISPR systems and may explain reported apparent KD by catalytically inactive enzymes9. Related artifacts may also influence siRNA, shRNA and ASO KD measurements12,13. Artifact magnitude probably depends on gRNA abundance, binding affinity, extraction method and RT priming strategy. Phenol–chloroform extraction, which enriches small RNAs, may exacerbate gRNA carryover. Oligo(dT) priming may inhibit all upstream amplicons because RT initiates from the poly(A) tail. RNA sequencing may also be affected when cDNA synthesis occurs without previous RNA fragmentation.
