Answers are grounded in this paper and cite their sources.
In their letter entitled “The action of Arabidopsis DICER-LIKE2 in plant growth inhibition”, authors from the Hongwei Guo laboratory (Liu et al. 2025) bring up several points to question the validity of our evidence for a new understanding of the function of the DICER-LIKE ribonuclease DCL2. Two broad matters are brought up. First, our evidence for an RNA interference (RNAi)-independent role of DICER-LIKE2 (DCL2) in conferring growth restriction in mutant backgrounds lacking the major cytoplasmic double-stranded RNA-directed ribonuclease DCL4, described in a paper published in The Plant Cell last year (Nielsen et al. 2024). Second, our evidence that this growth restriction is in fact a manifestation of DCL2-mediated sensing of double-stranded RNA that switches on innate immunity dependent on specific nucleotide-binding leucine-rich repeat receptors (NLRs), described in a preprint available to the community (Nielsen et al. 2023). We welcome the initiative by Liu et al. to further debate the evidence for these RNAi-independent roles of DCL2 and thank the editors for allocating space in the pages of The Plant Cell for this debate. We assume that readers of this response will have read the letter of Liu et al. and, therefore, we do not introduce basic notions on RNAi and DCL ribonucleases again. Nonetheless, we remind the reader of three points, key to the debate that follows: RNAi is understood to be the most important element of basal antiviral immunity in plants (Ding and Voinnet 2007). Antiviral RNAi is understood to involve two redundantly acting DCL enzymes: DCL4 and DCL2 (Ding and Voinnet 2007). Often, DCL2 is described as a back-up of DCL4, because in the vast majority of cases, DCL2-produced siRNAs only become detectable when DCL4 is inactivated, either genetically or through viral anti-RNAi virulence factors (Deleris et al. 2006; Wang et al. 2011; Andika et al. 2015). In addition, we hope the reader will forgive us for taking them on a slight detour whose purpose is to provide the necessary background to understand our direct answers to the points raised by Liu et al. This detour involves three elements: The basic observations at the core of the debate. How they have been explained prior to our contributions. What the main problem with these explanations is. In the absence of DCL4, Arabidopsis plants display a growth phenotype with incomplete penetrance (Gasciolli et al. 2005; Parent et al. 2015; Wu et al. 2017). This phenotype is exacerbated by partial loss of DCL1 function (Bouche et al. 2006), or, as shown by the Guo laboratory, by inactivation of RNA decay factors (Zhang et al. 2015), or, as we show in our papers, by inactivation of the HSP90 cochaperone SGT1B (Nielsen et al. 2023; Nielsen et al. 2024). In these genetic backgrounds, the penetrance is 100% and the phenotype is more severe than in dcl4 single mutants. In all cases, the growth phenotypes can be completely suppressed by inactivation of the DCL2 gene (Bouche et al. 2006; Zhang et al. 2015; Nielsen et al. 2023; Nielsen et al. 2024). The Guo laboratory and others have advocated the following explanation for the ability of DCL2 to confer growth phenotypes in the absence of DCL4: DCL2 makes ectopic siRNAs, 22 nucleotides (nt) in size. Others have shown that the 22-nt size of small RNA is important because in contrast to 21-nt siRNAs, 22-nt siRNAs efficiently trigger siRNA amplification through the RNA-dependent RNA polymerase, RDR6 (Chen et al. 2010; Cuperus et al. 2010; Yoshikawa et al. 2013; Fei et al. 2018; Sakurai et al. 2021; Yoshikawa et al. 2021). The ectopic siRNAs resulting from DCL2 activity and uncontrolled amplification, particularly those targeting the mRNAs SMXL4/5—encoding chaperonin-like factors—and NIA1/2—encoding nitrate reductases—are then proposed to cause growth phenotypes via silencing through the RNAi pathway (Zhang et al. 2015; Wu et al. 2017, 2020). This explanation is pleasing at first sight and clearly consistent with established knowledge. After all, siRNAs can silence gene expression through RNAi, and the genes in question are related to growth and metabolic activity. There is one major problem with the explanation described above, however: there is no proof for it. Clearly, the siRNA populations in question coincide with the growth phenotype, but so do many other transcriptomic changes that are not suggested to cause the phenotypes. Why this difference in treating similar evidence? Simple guesswork? Not quite, but almost. The authors rely on the fact that double knockout mutations in the genes concerned can produce phenotypes “similar” to those obtained in dcl4 mutants. Unfortunately, this is not a strong argument. First, it assumes comparable molecular effects obtained by genetic knockouts and ectopic siRNA populations, but in contrast to the knockouts, it is unknown when, where, and to what extent the siRNAs silence the targets in question. Second, it clearly does not prove that the siRNAs cause the phenotype, because such proof of causality would require the specific reversal of the gene silencing proposed to underlie the phenotype. There is one experiment used by the RNAi field for more than two decades to resolve this crucial question of causality: engineer a target resistant to siRNA/miRNA-guided silencing and test whether expression of this target is sufficient to suppress the phenotypes observed. As explained in our paper (Nielsen et al. 2024), this is not straightforward to achieve experimentally for at least two reasons. First, the implication of an entire siRNA population, not just a single small RNA species, necessitates extensive re-engineering of the target in question. Second, we do not know exactly which properties of the NIA and SMXL genes cause RNAi initiation, and if engineered versions retain those properties, the experiment would be compromised. Thus, the experiment may be cumbersome, although not outright impossible, to carry out. Unfortunately, results of such an experiment have not been reported thus far. Our work provides evidence for a completely different model of DCL2 function that explains the growth restriction phenotype of dcl4 (and other RNA metabolism mutants) as a manifestation of autoimmunity. In this model, DCL4 and DCL2 both perform crucial, yet different functions in basal antiviral immunity. DCL4 mediates classical RNAi, but DCL2 provides a second layer of immunity that only comes into play when RNAi has been defeated, and cytoplasmic double-stranded RNA (dsRNA) concentrations are abnormally high. In that situation, the very processing of dsRNA by DCL2 is a signal that switches on intracellular NLRs to activate an innate immune state that includes growth restriction. Hence, in dcl4 mutants, DCL2 processes endogenous dsRNA, switching on NLR-mediated immunity that explains the DCL2-dependent growth phenotypes. Clearly, this model predicts: that DCL2 causes growth restriction independently of a silencing function of the siRNAs that it produces, that activation of NLRs actually causes the growth restriction seen in dcl4 mutants, that those NLRs have a role comparable to that of DCL2 in basal resistance to at least one virus rendered incapable of RNAi suppression. The letter by Liu et al. raises critical points on selected pieces of the evidence in support of this model but omits other crucial points of evidence from our papers. Below, we provide answers to what we consider to be the five main points raised. Wherever relevant, we also include mention of evidence from our studies omitted in the letter by Liu et al. Liu et al. point out that our small RNA-sequencing performed to get siRNA profiles of Col-0 WT, dcl4, dcl4 DCL2/dcl2 (heterozygous for the dcl2 knockout allele), and dcl4 dcl2 has much less sequencing depth than that of other studies addressing DCL2-dependent phenotypes in dcl4 mutant backgrounds. This is undoubtedly true, but is it relevant? The data clearly show detection of several DCL2-dependent siRNAs upon loss of DCL4, including the ones believed by Liu et al. to cause developmental phenotypes (from NIA1/2 and SMXL4/5 loci). The data also show that these and other DCL2-dependent siRNAs (some confirmed by small RNA blots) accumulate to comparable levels in dcl4 and dcl4 DCL2/dcl2 plants, and their DCL2-dependence is confirmed in the dcl4 dcl2 double mutants (Fig. 1). Regardless of sequencing depth, the results are clear. If we were claiming differential expression of siRNAs from loci with low read counts, we would understand the concern. But in this case, there are ample reads to safely conclude that there is no decrease in siRNA levels from NIA1/2 or SMXL4/5 in dcl4 DCL2/dcl2 (heterozygous for the dcl2 knockout allele) compared to dcl4 DCL2/DCL2 (with two wild type DCL2 copies). This is the key point, because we show that reducing the DCL2 protein dosage by half in dcl4 DCL2/dcl2 is sufficient to significantly reduce the penetrance of the growth restriction phenotype (Nielsen et al. 2024). Full DCL2 dosage is not required for siRNA production. Analyses of small RNA-Seq data from 21-day-old plants of the indicated genotypes, mapped onto TAS1B, INVERTED REPEAT 71 (IR71), NIA1, NIA2, SMXL4, and SMXL5. rptm, reads per ten million. The results shown in this figure were first published as elements of Figure 4 in our 2024 Plant Cell paper (Nielsen et al. 2024). Liu et al. further argue that the way we decided to collect the material for small RNA-sequencing compromises the utility of the result. This point concerns the comparison of small RNA profiles in the following mutants, whose phenotypes are noted between parentheses. Col-0 wild type (all plants green and healthy) dcl4 DCL2 (two DCL2 wild type alleles) (a percentage of the population shows growth restriction, chlorosis, and anthocyanin accumulation) dcl4 DCL2/dcl2 (heterozygous) (a smaller percentage than in dcl4-2 shows growth restriction, chlorosis, and anthocyanin accumulation) dcl4 dcl2 (homozygous) (all plants green and healthy) We have taken care to explain the reasoning behind the way the experiment is done in the paper (Nielsen et al. 2024). The difficulty relates to the incompletely penetrant phenotypes. Our objective was to assess whether reducing the dosage of DCL2 by half affects DCL2-dependent siRNA levels as may be expected if silencing by some of these siRNAs caused the phenotype. But if we grew the plants under conditions in which the phenotypes penetrate and selected the ones with phenotypes, we would be likely to see the same siRNAs in the same phenotypic categories of the different mutant backgrounds, simply because the generation of the siRNAs coincides with, but does not necessarily cause, the growth phenotype. In contrast, conducting the experiment the way we did, with tissue harvest prior to appearance of visible symptoms, addresses the question we wanted to address: whether reducing the dosage of DCL2 by half affects DCL2-dependent siRNA levels transcriptome-wide. The result is clear, it does not: The mRNA-derived DCL2-dependent siRNAs differentially expressed between dcl4 and dcl4 DCL2/dcl2 split into three categories: some unaffected, some less abundant, some more abundant, likely reflecting the transcriptomic differences that arise as a consequence of the effect of DCL2 dosage on activation of immune responses, even in presymptomatic seedlings. We recognize that this is a choice with some imperfection. Doing the experiment in the way suggested by Liu et al. could have merit if indeed high SMXL4/5 siRNA levels were found in nonsymptomatic dcl4 DCL2/dcl2 heterozygous plants. We considered this possibility to be unlikely, simply because all the available evidence suggests that growth restriction coincides with (but again, is not necessarily caused by) high expression of SMXL4/5 siRNAs. Therefore, we did not embark on yet another experiment whose results would not distinguish between causality and correlation. We stress that the results we report are useful, because they establish that reducing the DCL2 dosage by half does not appreciably affect the ability to produce steady state levels of many DCL2-dependent siRNAs compared to the full DCL2 dosage. At this point, it becomes relevant to dig a bit deeper into the dependence of growth restriction phenotypes on DCL2 protein dosage. The letter by Liu et al. touches on this, but the weight of the argument does not come through completely clearly in their Hence, we what we actually report in the published in The Plant Cell last The penetrance of the growth phenotype in dcl4 mutants is not just at least a expression of DCL2 protein (Nielsen et al. 2024), by heterozygous for dcl2 knockout have the of DCL2 protein in wild and the penetrance as the of the growth phenotype in the population observations to of DCL2 and (Fig. and contrast, levels of different DCL2-dependent siRNAs do not show this dependence on DCL2 protein as clearly noted in the paper (Nielsen et al. (Fig. This that the growth restriction phenotype is related to DCL2 protein it is with a silencing effect of that involves several into an protein to produce an RNA the silencing by The penetrance of the DCL2-dependent phenotype is to DCL2 protein in of plants of the indicated were in first on and at were to and for to of the DCL2 by in the indicated compared to Col-0 wild type plants. the between the of of the phenotypes described in and DCL2 protein to endogenous DCL2 levels in as by in In both the are as and to The a model with a strong correlation. RNA of with no phenotypes, of in the background compared to The same was to to endogenous siRNAs from REPEAT and to (a of the of the is shown as The results shown in this figure were first published as elements of and in our 2024 Plant Cell paper (Nielsen et al. 2024). for of the as for the steady state levels of many small is particularly has been in et al. and is from many observations in Arabidopsis and observations also include effects on steady state of several even in strong mutants in DCL1 (Liu et al. clearly the of a for small RNA of the and the results on siRNA levels in show that the silencing is not a function of DCL in in the specific of DCL2 in a dcl4 mutant But in this case, the of growth restriction is in fact dependent on DCL2 at least a all but a silencing effect by DCL2-mediated siRNAs as the cause of growth restriction. Thus, we are with two the DCL2 protein does completely to small RNA processing to bring growth restriction, or the direct the levels of resulting to in is the relevant that growth restriction. we show that growth restriction and basal antiviral completely on and on required for in the our results point to the as the that the signal to bring growth restriction through activation of innate immunity (Nielsen et al. 2024). Our of dcl2 point mutants by is by Liu et because of them full between siRNA and of growth restriction. We and, exactly for that we care not to on the of these mutants in our report (Nielsen et al. 2024). Nonetheless, as noted above, the results show that it is to dcl2 mutants that have activity in DCL2-dependent siRNAs in wild type are as in the mutants, the fact that the mutants completely suppress the growth phenotype (Fig. This does siRNA generation for silencing as the role of DCL2 to a the other it is explained by our model that the acting as a the growth phenotypes. of dcl2 point mutants that suppress growth in of the of the in the DCL2 knockout and and the of the mutations in the dcl2 by genetic for mutations resulting in at least partial or mutations resulting in loss of activity. and of the indicated under RNA to a to the siRNA of the of the is shown as The results shown in this figure were first published as elements of and in our 2024 Plant Cell paper (Nielsen et al. 2024). Liu et al. further on our evidence in the preprint that dcl4 mutants are by an and that the growth phenotypes are caused by by the two NLRs and (Nielsen et al. 2023). that inactivation of these NLRs does indeed suppress the growth restriction in dcl4 and dcl4 Hence, we can all that these NLRs are necessary for the growth restriction phenotype, a major The main argument becomes the How do we know whether silencing of some by DCL2-dependent siRNAs, SMXL4/5 or causes activation This is a NLRs as they to changes in those factors that are by virulence factors to establish and Hence, mutants caused by activation through genetic inactivation of virulence targets are and have indeed caused much in of phenotypes the et al. 2017). But Liu et al. only mention our that and mutants suppress growth phenotypes and activation in dcl4 and dcl4 mutant backgrounds, and to mention several other to this of all the relevant results would be the dcl4 mutants not only in Arabidopsis but also in and, the growth and a gene expression with of the a element the in of differentially expressed in similar to other does not the DCL2 this state is suppressed by inactivation of the gene (Nielsen et al. 2023). would be a if silencing of more or less genes in both and Arabidopsis would to But with the this becomes impossible, not least because a 22-nt DCL2 that could amplification, does it the The growth phenotype of dcl4 or dcl4 can be suppressed by inactivation of either one of two nucleotide-binding repeat intracellular immune and (Nielsen et al. 2023). of or has no effect on of a RNA virus rendered on wild type plants by of anti-RNAi virulence with inactivation of dcl4, knockout of either or this basal antiviral in the of if in the of with an effect size that of inactivation of DCL2 viral are similar in dcl4 dcl2 and dcl4 dcl2 consistent with DCL2 and acting in the same antiviral pathway (Nielsen et al. 2023). This key result provides strong genetic evidence for the of and in basal antiviral a result that one would not if their activation in dcl4 mutants were a result of ectopic silencing of endogenous genes by DCL2-dependent siRNAs. the other this is what is from our model of DCL2-mediated dsRNA How that sensing including exactly and are dependent on dsRNA processing by DCL2 are of and we of to these in a In with the of as the cause of DCL2-dependent growth phenotypes, Liu et al. three further points that we to The first relates to of DCL2-dependent growth phenotypes in dcl4, dcl4 and dcl4 mutants. Liu et al. are concerned that explain such phenotypic as they this through DCL2 processing of is to explain the and of growth of dcl4 as as dcl4 and dcl4 We are by this because in many Arabidopsis mutants show phenotypic and dependence on growth conditions et al. et al. Zhang et al. et al. 2010; et al. 2017). the and the are major of cytoplasmic RNA decay et al. et al. et al. and 2021). In single mutants of these many siRNAs are et al. et al. 2015), so it is not to dsRNA in cytoplasmic in dcl4 and dcl4 mutants. This could to different of activation through DCL2 even NLRs on the of such are unknown at this point, but the phenotypic and the of growth in dcl4 by and mutations do not the model of by DCL2-mediated dsRNA The second point concerns the of the of growth phenotypes, related to the first point of Liu et al. to be particularly concerned that we report growth phenotypes of dcl4 in to There is no As reported in the (Nielsen et al. 2024), when plants are under phenotypes are to of We this for of growth phenotypes in dcl4 because we that the of phenotypes was a bit more under In addition, genetic backgrounds with suppressed growth at the at the all dcl4 mutants show the phenotype. these properties that plants to by Nonetheless, phenotypes are also visible under to 4 of Liu et al. are concerned that of two and are required for DCL2-dependent activation and that this require DCL2 to either activate different or trigger through the same of the of by knockout of either or is that they in not in is an of the is an of the on plant function has shown that are that nucleotides whose receptors may as of et al. 2023). such a unknown at this it to that a model action of a and a not the on plant innate immune the of their Liu et al. that evidence could be by what the effect of mutations and on dcl4 may the of the of all of plant siRNAs and et al. 2005; et al. and silencing is significantly in mutants et al. et al. because of of small et al. is the key protein of the of several of small gene silencing et al. and 2005; et al. If one silencing as a as it is in the of this would DCL4 or DCL2 siRNAs. of DCL siRNAs are by then into that then Hence, with this of if inactivation of and the DCL2-dependent growth phenotype, it would prove that not DCL2 processing is the Guo laboratory has shown that of suppress growth phenotypes in dcl4 mutants (Zhang et al. 2015), and that a mutant growth phenotypes of dcl4 and dcl4 mutants et al. 2020). this in Liu et al. conclude that genetic evidence by Wu et al. that 22-nt siRNAs, but not DCL2 protein per cause plant growth But does it The problem is that this a of silencing in plants. The plant RNAi is not an important by and the RNA-dependent RNA RDR6 2006), key to understand the DCL2-dependent growth phenotype, because this phenotype on as in work from both the Guo laboratory and from us (Zhang et al. 2015; Nielsen et al. 2024). can to of dsRNA from siRNA Hence, with in this not only silencing per but also the of intracellular dsRNA available for processing by DCL2. Thus, results of genetic on inactivation of or distinguish between the because these mutations affect both silencing activity and generation of dsRNA for processing by DCL2. and mutants both out of a genetic for mutants in In this these mutations not only to silencing but to a of the siRNA population, the of the for dsRNA and siRNA generation et al. et al. we welcome the debate by Liu et we the in their letter the weight of our evidence for RNAi-independent functions of DCL2 in dsRNA sensing to be in some evidence for the by the Guo laboratory that silencing of key endogenous such as SMXL4/5 and causes the DCL2-dependent phenotypes in dcl4 knockout mutants, has simply not been Thus, this model be as a which there is Our that RNAi-independent functions of DCL2 in cytoplasmic sensing of double-stranded RNA provide the in fact the available model that explains all available We stress that this does not that siRNAs by DCL2 be used in We very much support the that in has been defeated, by DCL2 activation and siRNAs that to through As it is not that DCL2-dependent siRNAs in some mutant backgrounds may cause silencing under some But all we can for at this point is that the to DCL2-dependent growth in dcl4 mutants involves the NLRs and that are also required for basal antiviral We by that we do not question the validity of the genetic results from the Guo laboratory on the of of mutants in dcl4 and RNA decay factors by inactivation of the and genes (Zhang et al. 2015). we are by the of that genetic and the results to be an important to the This debate is concerned with the of the This work was by a from to There are no new data with this
No takes yet. Share an insight, caveat, or question.
Nielsen et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: