Transgene-free, virus-based gene silencing in plants by artificial microRNAs derived from minimal precursors

Abstract Artificial microRNAs (amiRNAs) are highly specific, 21-nucleotide (nt) small RNAs designed to silence target transcripts. In plants, their application as biotechnological tools for functional genomics or crop improvement is limited by the need of transgenically expressing long primary miRNA (pri-miRNA) precursors to produce the amiRNAs in vivo. Here, we analyzed the minimal structural and sequence requirements for producing effective amiRNAs from the widely used, 521-nt long AtMIR390a pri-miRNA from Arabidopsis thaliana. We functionally screened in Nicotiana benthamiana a large collection of constructs transiently expressing amiRNAs against endogenous genes and from artificially shortened MIR390-based precursors and concluded that highly effective and accurately processed amiRNAs can be produced from a chimeric precursor of only 89 nt. This minimal precursor was further validated in A. thaliana transgenic plants expressing amiRNAs against endogenous genes. Remarkably, minimal but not full-length precursors produce authentic amiRNAs and induce widespread gene silencing in N. benthamiana when expressed from an RNA virus, which can be applied into leaves by spraying infectious crude extracts. Our results reveal that the length of amiRNA precursors can be shortened without affecting silencing efficacy, and that viral vectors including minimal amiRNA precursors can be applied in a transgene-free manner to induce whole-plant gene silencing.


INTRODUCTION
MicroRN As (miRN As) are a class of 20-24-nucleotide (nt) endo genous small RN As (sRN As) that r epr ess gene expression post-transcriptionally in plants and animals.Most plant miRNAs are 21-nt long and regulate the expression of genes encoding transcription factors and proteins that impact key biological processes such as growth, de v elopment, stress adaptation or hormone regulation ( 1 , 2 ).MiRNA biogenesis in plants [re vie wed recently in (3)(4)(5)(6)(7)] occurs entirely in the cell nucleus, and initiates with the transcription of MIR genes by RN A pol ymerase II, resulting in the synthesis of large primary miRNA precursors (or pri-miRNAs) that are highly variable in length (from hundreds to thousands of nucleotides) and in secondary structure.Pri-miRNAs are capped, spliced and polyadenylated, and harbor a characteristic foldback (or hairpin) structure that also varies in length (from 50 to 900 nt).MiRNA foldbacks have a common structure consisting of a ∼15-17 base-pair (bp) basal stem (BS), a miRN A / miRN A* duplex, and a distal stemloop (DSL) region variable in size and shape and flanked by single-stranded RN A (ssRN A) segments ( 8 , 9 ).Processing of most pri-miRNAs occurs through a 'base-to-loop' mechanism in which the nuclear RNase III DICER-LIKE1 (DCL1) associated with its accessory proteins SERRATE (SE) and HYPONASTIC LEAVES1 (HYL1) first cleaves at the BS of the foldback to release a shorter precursor miRN A (pre-miRN A).Next, a second cleavage at a 21-nt distance from the first cut releases the miRN A / miRN A* duplex including two-nucleotide 3 overhangs that are 2 -Omethylated by the methyltr ansfer ase HUA ENHANCER1 (HEN1) ( 3 , 10 ).Alternati v ely, for loop-to-base processed precursors, DCL1 produces a first cut that releases the terminal loop and then continues towards the base of the precursor to release the miRN A / miRN A* duplex ( 8 , 11 ).Typically, one of the strands of the duplex (the guide strand) is sorted out into an ARGONAUTE (AGO) protein (AGO1 for the majority of plant miRNAs) resulting in an RNAinduced silencing complex (RISC) that recognizes and silences complementary target RN As mostl y through their endonucleolytic cleavage, and less frequently through their translational r epr ession ( 12 ).
Artificial miRNAs (amiRNAs) are 21-nt miRNAs computationally designed for r edir ecting the endogenous miRN A bio genesis and silencing machineries to selecti v ely suppress specific target RNAs with high efficacy and specificity (with no predicted off-targets) (13)(14)(15).AmiRNAs are produced in planta by expressing a modified MIR gene producing a pri-miRNA precursor in which the sequence of the nati v e miRN A / miRN A* duplex is substituted by the sequence of the amiRN A / amiRN A* duplex.The selection of an appropriate pri-miRNA backbone is critical for ensuring an efficient and accurate precursor processing and produce high amounts of amiRNAs necessary for effecti v e target silencing.The 521-nt long pri -miRNA precursor deri v ed from Arabidopsis thaliana (Arabidopsis) MIR390a ( pri-AtMIR390a ) is a highly efficient and accurately processed pri-miRNA compared to other plant pri-miRNAs used for producing amiRNAs ( 16 , 17 ).Moreover, pri-AtMIR390a has been used for expressing amiR-N As in m ultiple plant species including model and crop plants, to effecti v el y silence both endo genous genes and viral RN As (17)(18)(19)(20).Interestingl y, pri-AtMIR390a has a relati v ely short DSL region of only 31 nt, allowing the synthesis of the complete foldback with just two oligonucleotides spanning this structure.This feature has facilitated the de v elopment of a cost-effecti v e, high-throughput methodolo gy to directl y introduce amiRN A inserts into a series of 'B / c' vectors including the pri-AtMIR390a sequence ( 16 ).
AmiRNAs are used as biotechnological tools for both functional genomics and cr op impr ovement ( 19 , 21 ).Howe v er, a current limitation of the amiRN A technolo gy is the use of relati v ely long precursors, mostly based on fulllength pri-miRNA sequences, to produce the amiRNAs in vivo .Reducing the size of amiRNA precursors should have se v eral biotechnological advantages such as a more costeffecti v e production of amiRNA constructs or a more efficient in vitro or in bacteria synthesis of amiRNA precursors for topical application to plants.Another limitation is the need to genetically modify plant genomes with amiRNA transgenes.The use of plant DNA viruses to successfully express authentic amiRNAs in Nicotiana benthamiana (22)(23)(24) is promising, although infections are typically triggered by the agroinfiltration of plant tissues with plasmids including the amiRNA-expressing viral vector and limited to a narrow list of host species.Ther efor e, this so-called amiRNA-based virus-induced gene silencing (amiR-VIGS) technology r equir es further optimization for its broader application in multiple plant species and to avoid the generation of genetically modified plant tissues.
Here, we further expanded our recently de v eloped silencing sensor system in N. benthamiana ( 25 ) and used it to systematically analyze the minimal structural and sequence r equir ements for producing effecti v e amiRNAs from the broadly used, 521-nt long pri-AtMIR390a precursor .We functionally screened a large collection of constructs expressing amiRNAs against the N. benthamiana ma gnesium chelatase sub unit CHLI-encoding SULPHUR ( NbSu ) or the 1-DEOXY-D-XYLULOSE-5-PHOSPHATE SYNTHASE ( NbDXS ) genes from AtMIR390a -based precursors with artificially shortened BS or DSL regions.By combining phenotypic, biochemical and molecular assays with high-throughput sequencing-based analysis of precursor processing we show that highly effecti v e, accurately processed amiRNAs can be produced in N. benthamiana from a shortened chimeric MIR390 -based amiRNA precursor of only 89 nt.This 'minimal' precursor includes the complete BS region of pri-AtMIR390a (with no additional ssRNA segments) and the DSL region deri v ed from Oryza sativa MIR390 ( OsMIR390 ) with a 2-nt deletion.The precursor was further validated in Arabidopsis transgenic plants expressing amiRNAs for silencing CHLORINE42 ( AtCH42 ) or FL OWERING L OCUS T ( AtFT ) w hich resulted in bleaching and late flowering phenotypes, respecti v el y.Finall y, we show that the minimal precursor has the unique ability to produce authentic amiRNAs and trigger widespread gene silencing in N. benthamiana when expressed from an RNA virus vector spr ay ed to leaves in a non-transgenic, DNA-free manner.

Plant materials and growth conditions
N. benthamiana and Arabidopsis thaliana (Arabidopsis) Col-0 plants were grown in a growth chamber at 25 • C with a 12 h-light / 12 h-dark photoperiod or at 22 • C with a 16 h-light / 8 h-dark photoperiod, respectively.DCL1i and DCL4i N. benthamiana knockdown plants were described previously ( 26 ).Arabidopsis Col-0 plants were genetically transformed using the floral dip method ( 27 ) with the Agr obacterium tumef aciens GV3101 str ain.T1 tr ansgenic Arabidopsis were grown as previousl y ( 28 ).Photo gra phs were taken with a Nikon D3000 digital camera with AF-S DX NIKKOR 18-55 mm f / 3.5-5.6GVR lens.

Plant phenotyping
Plant phenotyping analyses were conducted in blind as described ( 28 ).Briefly, the flowering time of each independent line is the number of days elapsed from seed plating to first bud opening (or 'days to flowering').The 'CH42' phenotype was scored in 10 days old seedlings and was reported as 'weak', 'intermediate' or 'se v ere' if seedlings had more than two leav es, e xactly two leav es or no leav es at all (only two cotyledons), respecti v ely.
AmiRNA constructs 35S:pri-amiR-GUS Nb and 35S:pri-amiR-NbSu wer e r eported pr eviousl y ( 25 ).All DN A oligonucleotides used for generating the constructs described above are listed in Table S1.The sequences of all amiRNA pr ecursors ar e listed in Appendix S3.The sequences of the BS-AtMIR390a -based B / c amiRNA vectors are listed in Appendix S4.

T r ansient expr ession of constructs and mechanical inoculation of viruses
DNA constructs were agroinfiltrated in N. benthamiana leaves as described ( 39 , 40 ), using A. tumefaciens GV3101 strain.Infection assays with TSWV LL-N.05 isolate were done as described ( 41 , 42 ).Mechanical inoculation with crude extracts from PVX infected plants was done as described for TSWV.

Chlorophyll extraction and analysis
Pigments from N. benthamiana infiltrated leaf tissues were extracted as described ( 43 ).Absorbance measurements and content in chlorophyll a were calculated as described ( 28 ).

RNA pr epar ation
Total RNA from N. benthamiana leaves or from Arabidopsis seedlings was isolated as follows.Tissue was ground in liquid nitrogen to fine po w der and homogenized in extraction buffer (1 M guanidinium thiocyanate, 1 M ammonium thiocyanate, 0.1 M sodium acetate, 5% glycerol, 38% watersa tura ted phenol).Ne xt, RNA was e xtracted by chloroform extraction and precipitated after adding 0.5 volumes of isopropanol for 20 min.Triplicate samples from pools of two N. benthamiana leaves or 9-12 Arabidopsis seedlings were analyzed.

R eal-time R T-qPCR
cDNA was obtained from 500 ng of DNaseI-treated total RNA from apical leaves at 14 days-post agroinoculation (dpa), using the PrimeScript RT Reagent Kit (Perfect Real Time) (Takara) according to the manufacturer's instructions.
Real time RT-qPCR was done as described in a QuantStudio 3 Real-Time PCR system (Thermo Fisher Scientific).NbSu / NbDXS / PVX and AtCH42 target RNA e xpression le v els were calculated relati v e to N. benthamiana PROTEIN PHOSPHATASE 2A ( NbPP2A ) and Arabidopsis ACTIN 2 ( AtACT2 ) r efer ence genes, r especti v ely, using the delta delta cycle threshold comparati v e method of the QuantStudio Design and Analysis Software (version 1.5.1.;Thermo Fisher Scientific).Oligonucleotides used for RT-qPCR are listed in Table S1.

Small RNA blot assays
Small RNA blot assays and band quantification from radioacti v e membranes were done as described ( 25 ).Oligonucleotides used as probes for sRNA blots are listed in Table S1.

Small RNA sequencing and data analysis
Total RNA quantity, purity and integrity was analyzed with a 2100 Bioanalyzer (RNA 6000 Nano kit, Agilent) and submitted to BGI (Hong Kong, China) for sRNA library preparation and SE50 sequencing in a DNBSEQ-G-400 sequencer (BGI).After reception of quality-trimmed, adaptor removed clean reads from BGI, fastx collapser ( http:// hannonlab.cshl.edu/fastxtoolkit ) ( 44 ) was used to collapse identical reads into a single sequence, while maintaining read counts.A custom Python script was then used to map each clean, unique read against the forward strand of the amiRNA pr ecursor over expr essed in each sample (Data S2) not allowing mismatches or gaps, and also to calculate the counts and RPMs (reads per million mapped reads) for each mapping position.Processing accuracy of amiRNA foldbacks was assessed by quantifying the proportion of 19-24 nt sRNA (+) reads that mapped within ± 4 nt of the 5 end of the amiRNA guide as r eported befor e ( 39 , 43 ).Phasing register tables were built by calculating the proportion of 21-nucleotide sRNA (+) reads in each register relati v e to the corresponding amiRNA cleavage site for all 21-nucleotide positions downstream of the cleavage site, as described previously ( 16 ).List of 21-nucleotide sRNA (+) reads of target RNAs and species-specific tasiRNA-generating controls ( AtTAS1c in Arabidopsis and AtTAS3 in N. benthamiana ) are shown in Data S3.

Stability and sequence analyses of amiRNA precursors during viral infections
Total RN A from a pical leaves of each of the three biological replicates was pooled before cDN A synthesis, w hich was performed as explained above.PCR to detect amiRNA precursors, PVX and NPP2A was performed using oligonucleotide pairs A C-654 / A C-655, A C-657 / A C-658 and A C-365 / A C-366 (Ta b le S1), respecti v el y, and Phusion DN A polymerase (ThermoFisher Scientific).PCR products were analyzed by agarose gel electrophoresis, and products of the expected sized were excised from the gel and sequenced.

Pr epar ation and spr aying of crude extr acts obtained from virus infected plants
For each plant, 1 g from 1-2 apical leaves was ground to fine po w der, homogenized in 5 ml of buffer (50 mM potassium phosphate pH 8.0, 1% polyvinylpyrrolidone 10 and 1% polyethylene glycol 6000), filtered with sterile Miracloth (Millipore) and transferred to a 30-ml high-density pol yethylene va porizer (Yizhao).All volume ( ∼5 ml) was spr ay ed to two different leaves (third and fourth leaves counting from the bottom) of a 3-week old N. benthamiana plant, at a 5-10 cm distance (Supplementary Figure S1).

A silencing sensor system in N. benthamiana for easily visualizing and quantifying amiRNA-induced silencing of endogenous genes
We recently described a silencing sensor system for easily visualizing and quantifying silencing of the magnesium chelatase subunit CHLI [hereafter SULPHUR gene ( NbSu )] in N. benthamiana by the amiRNA amiR-NbSu-1 ( 25 ).To further expand the use of the sensor system, we tested if we could trigger a similar visible bleaching with amiRNAs directed against the 1-DEOXY-D-XYLULOSE-5-PHOSPHATE SYNTHASE ( NbDXS ) gene.Briefly, thr ee differ ent constructs expr essing amiR-NAs targeting NbDXS at unique sites and with no predicted off-targets (Supplementary Figure S2A, Data S1) were independently agroinfiltrated in two areas of two leaves from thr ee differ ent N. benthamiana plants, together with a negati v e control construct expressing an amiRNA against Esc heric hia coli b-glucuronidase GUS.At 7 dpa, areas agroinfiltrated with anti-NbDXS amiRNA constructs displayed a visually obvious bleaching predicted from NbDXS knockdown while areas expressing the control construct did not (Supplementary Figure S2B).The chlorophyll analysis of agroinfiltrated areas showed a similar decrease in chlorophyll content in bleached areas expressing anti-NbDXS amiRNAs compared to areas expressing the control construct (Supplementary Figure S2C).Next, two leaves of thr ee differ ent plants wer e independently agroinfiltrated in the whole leaf surface with each of the amiRNA constructs described abov e. sRNA b lot assays of RNA preparations fr om agr oinfiltra ted leaves showed tha t all anti-NbDXS amiRN As accum ulated as single-size sRN A species, with amiR-NbDXS-1 accumulating to significantly higher le v els than the other two amiRNAs (Supplementary Figure S2D).RT-qPCR analysis re v ealed that all samples e xpressing anti-NbDXS amiRN As accum ulated significantl y lower le v els of NbDXS mRNA than control samples (Supplementary Fig- ure S2E).These results indicate that anti-NbDXS amiRNAs are highly acti v e and induce a strong bleaching phenotype in the tissue where they are expressed.Thus, the methodology described in ( 25 ) and outlined in Figure 1 was used in this work to test the effect on NbSu and NbDXS silencing efficacy of amiR-NbSu-1 (hereafter amiR-NbSu) and amiR-NbDXS-1 (hereafter amiR-NbDXS), respecti v ely (Figure 2 A), pr oduced fr om pri-AtMIR390a precursors with reduced BS or DSL regions (see next four sections).

Functional analysis of pri-AtMIR390a -based amiRNA precursors with shortened DSL regions
First, we analyzed the silencing activity of amiR-NbSuand amiR-NbDXS-expressing constructs including pri-AtMIR390a -based precursors with progressi v ely shorter DSL regions ranging from 31 nt of the full-length pri-AtMIR390a (or simply ' pri ') precursor to only 6 nt (Figure 2 B).As expected, at 7 dpa areas expressing amiR-NbSu or amiR-NbDXS from the pri precursor displayed the reduced pigmentation characteristic of NbSu or NbDXS knockdown, while areas expressing amiR-GUS Nb did not (Figure 2 C).Interestingly, a similar bleaching was also observed in ar eas expr essing amiR-NbSu or amiR-NbDXS from AtDSL-Δ6 and AtDSL-Δ13 precursors, while a less intense bleaching was noticed in areas expressing amiR-NbSu (but not amiR-NbDXS) from the AtDSL-Δ21 precursor (Figure 2 C).No bleaching was observed in areas expressing amiRNAs from the AtDSL-Δ25 precursor (Figure 2 C) .Visual silencing phenotypes were confirmed by the chlorophyll analysis of agroinfiltrated areas, with bleached areas displaying lower chlorophyll a content than darker green ar eas (Figur e 2 D).RNA blot assays of RNA preparations fr om agr oinfiltrated leaves showed that amiR-NbSu and amiR-NbDXS accumulated to high le v els when expressed from pri or from shorter AtDSL-Δ6 and AtDSL-Δ13 precursors (Figure 2 E).In ov ere xposed b lots both amiRNAs were also detected although to low le v els when expressed from AtDSL-Δ21 and also from AtDSL-Δ25 in the case of amiR-NbDXS (Figure 2 E).Small RNA species of different sizes were also detected in some cases, most likely reflecting that the position of DCL1 cleavage is not always uniform, as proposed before ( 28 ).Twenty-four-nt sRNAs present in the ov ere xposed b lots presumab ly arise from transgene-deri v ed dsRN As, w hich typicall y occurs in N. benthamiana w hen agroinfiltr ating ectopic tr ansgene constructs, as observed before ( 45 ).Finally, RT-qPCR analysis re v ealed that target mRNA e xpression was considerab ly reduced in samples expressing amiRNAs from pri or from shortened precursors compared to control samples, except for samples expressing amiR-NbSu or amiR-NbDXS from AtDSL-Δ25 (Figure 2 F).Notably, samples expressing amiRNAs from pri or from AtDSL-Δ6 accumulated similar lo w tar get mRNA le v els, while the expression of shorter AtDSL-Δ13 and AtDSL-Δ21 precursors generally induced a more moderate reduction in target mRNA le v els (Figure 2 F).In contrast, no significant reduction in target mRNA le v els was observed when amiRNAs were expressed from AtDSL-Δ25 (Figure 2 F).These results indicate that the DSL region of pri-AtMIR390a can be shortened to some extent without significantly affecting amiRNA accumulation and target mRNA silencing.

AmiRNAs produced from pri-AtMIR390a -based precursors including shortened OsMIR390 -derived DSL regions trigger gene silencing with high efficacy
The OsMIR390 precursor has a DSL region (OsDSL) consisting of only 16 nt, which r epr esents one of the shortest DSL regions from all conserved plant MIRNA precursors ( 43 ).Her e, we r easoned that the OsDSL r egion might also tolerate nucleotide deletions while preserving efficient amiRNA production and target mRNA silencing.To test this, se v eral constructs were generated for expressing amiR-NbSu and amiR-NbDXS from chimeric constructs including the complete (16-nt long) OsDSL region or progressi v ely shortened versions of it (Figure 3 A).In addition, a construct in which the 8-nt terminal loop of OsMIR390 was replaced by the shorter 4-nt long terminal loop of At-MIR390a was also included in the analysis (Figure 3 A).Areas agroinfiltrated with constructs expressing amiRNAs from pri or from chimeric OsDSL and OsDSL-D2 precursors showed intense bleaching and similar reduction in the accumulation of chlorophyll a compared to controls (Figure 3 B and 2 C).In contrast, other agroinfiltrated areas displayed weaker bleaching and lower chlorophyll a content except those from controls and from areas expressing amiR-NbSu from OsDSL-D6 which did not display any sign of silencing (Figure 3 B and C).RNA blot analysis of RNA preparations fr om agr oinfiltrated leaves revealed that amiR-N As accum ulated to high le v els in samples expressing the pri or chimeric OsDSL and OsDSL-D2 precursors, while amiRNA le v els dropped significantly in the rest of samples except for those expressing the 35S:OsDSL-D6-amiR-NbSu construct in which amiR-NbSu was not detected (Figure 3 D).RT-qPCR analysis re v ealed that target mRNA accum ulation significantl y decr eased in all samples expr essing amiRNAs from pri or from chimeric precursors compared to controls, except in samples expressing amiR-NbSu from OsDSL-D4 and OsDSL-D6 (Figure 3 E).Importantly, target mRNA le v els were similar in samples expressing amiR-NAs from pri or from chimeric OsDSL and OsDSL-D2 precursors (Figure 3 E).Taken together, these results indicate that the OsDSL region can also be shortened --in this case down to 14 nt --in AtMIR390a -based precursors without compromising silencing efficacy.

Functional analysis of AtMIR390a -based amiRNA pr ecursors with shortened BS
Next, we analyzed if shortening the BS region of pri-AtMIR390a could affect amiRNA production and silencing efficacy.Se v eral constructs for e xpressing amiR-NbSu and amiR-NbDXS were generated, in which the 448-nt BS region of pri-AtMIR390a was progressi v ely shortened down to 2 nt (Figure 4 A).Similar intense bleaching and reduced chlorophyll a le v el was observed in areas expressing amiR-NbSu or amiR-NbDXS from the pri or BS precursors (Figures 4 B and  E).In contrast, their le v els dropped significantly or were undetectable when expressed from BS-D7 / BS-D17 or BS-D23 / BS-D31 , respecti v ely (Figure 4 D) and induced significantly lo wer tar get mRNA do wnregulation (Figure 4 E).All together these results indicate that a complete BS region is necessary for accurate precursor processing to produce high le v els of amiRN As, w hile further shortening this BS region drastically reduces amiRNA accumulation.

Ne w high-thr oughput v ectors f or expr essing amiRNAs from AtMIR390a -based precursors only including the BS
Previous high-throughput AtMIR390a -based amiRNA 'B / c' vectors included the complete 521 bp sequence of the pri-AtMIR390a pr ecursor ( 16 ).Her e, we de v eloped two additional 'B / c' vectors for the efficient cloning and expression of amiRNAs from shortened precursors only including AtMIR390a BS (Supplementary Figure S3).pENTR-BS-AtMIR390a-B / c Ga teway-compa tible entry vector was generated for direct cloning of amiRNA inserts and subsequent recombination into the pr eferr ed Gate way e xpression vector including the regulatory features of choice.pMDC32B-BS-AtMIR390a-B / c vector was generated for the direct cloning of amiRNA inserts into a binary

T r ansient expr ession of amiRNAs from shortened, MIR390based chimeric precursors triggers effective silencing of N. benthamiana endogenous genes and of pathogenic viral RNAs
Because our previous results showed that the efficiency of amiRNA production and induced silencing of N. benthamiana endogenous genes were maintained in shortened precursors including the BS or OsMIR390-D2 regions, the new  'B / c' vectors were used to systemically test the effect of combining both regions in a single precursor.The resulting sh ortened c himeric MIR390 -based precursor of only 89 nt was named shc-MIR390 precursor (or simply ' shc ') (Figure 5 A).First, we introduced into pMDC32B-BS-AtMIR390a-B / c the sequence corresponding to amiR-NbSu or amiR-NbDXS including the OsMIR390-D2 region to generate the 35S:shc-amiR-NbSu or 35S:shc-amiR-NbDXS constructs, respecti v ely.Both constructs were independently agroinfiltrated into N. benthamiana leaves together with control constructs, and induced similar strong bleaching, amiRNA accumulation and target mRNA downregulation in agroinfiltrated areas than their respecti v e controls e xpressing these same amiRNAs from the full-length pri precursor (Figure 5 B-D).To confirm the correct processing of shc precursors and compare it to that of the pri , sRNA libraries from leaves expressing 35S:pri-amiR-NbSu , 35S:pri-amiR-NbDXS , 35S:shc-amiR-NbSu and 35S:shc-amiR-NbDXS wer e pr epar ed and sequenced.Inter estingly, in samples expressing 35S:shc-amiR-NbSu or 35S:shc-amiR-NbDXS, the majority (95% and 90% respecti v ely) of 19-24-nt (+) reads mapping within ±4 nt of the 5 end of the amiRNA guide corresponded to authentic 21-nt amiR-NbSu or amiR-NbDXS (Figure 5 E).These results indicate a high accuracy in the processing of shc precursors which was similar to that observed for the pri precursors (Figure 5 E, Figure S5).
Next, we aimed to test if amiRNAs pr oduced fr om the shc precursor could also be applied to silence exogenous transcripts, such as viral RNAs.Here, the sequence corresponding to the most efficient anti-tomato spotted wilt virus (TSWV) amiRNA (amiR-TSWV) identified in previous amiRNA screenings ( 41 ) was expressed from the shc precursor in a N. benthamiana leaf inoculated two days later with TSWV (Supplementary Figure S6A).At 7 days postinoculation (dpi), all leaves inoculated with TSWV and expressing amiR-TSWV either from the pri or shc precursors did not display virus-induced symptoms, while leaves inoculated with TSWV and expressing the control amiR-GUS Nb displayed multiple local lesions typical of TSWV infection (Supplementary Figure S6B).AmiRN A accum ulation analyzed by sRNA northern blot showed that amiR-TSWV accumulated to similar le v els in tissues e xpressing 35S:shc-amiR-TSWV or 35S:pri-amiR-TSWV (Supplementary Figure S6C).At 20 dpi, TSWV RNA was not detected in upper non-inoculated leaves from plants inoculated with TSWV and expressing 35S:shc-amiR-TSWV or 35S:pri-amiR-TSWV , while plants expressing 35S:pri-amiR-GUS Nb and inoculated with TSWV accumulated high le v els of TSWV RNA (Supplementary Figure S6C).Thus, these results indicate that the shc precursor can be used to express high levels of amiRNAs to induce antiviral resistance in plants.

Stable expression of amiRNAs from shc-MIR390 precursors leads to effective silencing of Arabidopsis endogenous genes
Next, we aimed to further confirm the functionality of shc precursors in a different plant species and when stab ly e xpressed in transgenic plants.For that purpose we used the amiR-AtFT / AtFT and amiR-AtCH42 / AtCH42 silencing sensor systems in Arabidopsis in which the transgenic expression of amiR-AtFT or amiR-AtCH42 from pri-AtMIR390a leads to a significant delay in flowering time or to an intense bleaching due to the targeting of endo genous FL OWERING L OCUS T ( AtFT ) or CHLO-RINA 42 ( AtCH42 ) endogenous genes, respecti v ely ( 16 ).Here, we introduced amiR-AtFT and amiR-CH42 into pMDC32B-BS-AtMIR390a-B / c to generate the 35S:shc-amiR-AtFT and 35S:shc-amiR-AtCH42 constructs, respecti v ely (Figure 6 A).These constructs were transformed independently into Arabidopsis Col-0 plants together with control constructs 35S:pri-amiR-GUS Ath , 35S:pri-amiR-AtFT and 35S:pri-amiR-AtCH42 expressing an amiRNA against GUS (with no predicted off-targets in Arabidopsis), AtFT and AtCH42 , respecti v ely, from pri-AtMIR390a .To systemically compare the processing and induced silencing of amiR-AtFT or amiR-AtCH42 produced from shc and pri precursors , plant phenotypes , amiRN A and target mRN A accumulation and processing efficiency wer e measur ed in Arabidopsis T1 transgenic lines.
All 35S:pri-amiR-AtFT ( n = 40) and 35S:shc-amiR-AtFT ( n = 34) transformants flowered later than the average flowering time of the 35S:pri-amiR-GUS Ath control lines ( n = 48) (Figure 6 B, Table S2), and their average flowering time (54.7 ± 6 and 54.4 ± 3.2, respecti v ely) was not significantly different between each other (Figure 6 C).RNA-blot and RT-qPCR assays re v ealed that both amiR-AtFT and AtFT mRN A accum ulation was similar in lines expressing amiR-AtFT from each of the two precursors (Figure 6 D), and high-throughput sequencing of sRNAs showed that both precursors were efficiently processed, as in both cases 99% of the r eads corr esponded to authentic amiR-AtFT (Figure 6 E, Figure S7).Similarly, all 35S:pri-amiR-AtCH42 ( n = 54) and 35S:shc-amiR-AtCH42 ( n = 38) transformants displayed bleaching of comparable degrees (Figure 6 B and C; Table S2), accumulated similar le v els of amiR-CH42 and AtCH42 mRNA (Figure 6 D), and displayed effecti v e precursor processing (Figure 6 E, Figure S7).Taken together these results indicate that the shc precursor is functionally equivalent to the pri precursor in inducing highly effecti v e silencing of endogenous genes when stably expressed in Arabidopsis.

Unique functionality of the shc precursor in triggering widespread gene silencing in N. benthamiana when expressed from a viral vector
Because cargo capacity of viral vectors is limited, and large inserts are typically eliminated during viral replication in VIGS vectors ( 46 ), we hypothesized that the short (89-nt long) shc precursor fragment could be a more stable insert when included into a viral vector than the longer 521-nt pri precursor.To test this, pri-amiR-NbSu and shc-amiR-NbSu sequences were inserted into a PVX infectious clone to generate the 35S:PVX-pri-amiR-NbSu and 35S:PVXshc-amiR-NbSu constructs, respecti v ely (Figure 7 A).These constructs were expected to express amiR-NbSu from pri and shc pr ecursors, r especti v ely, during PVX infection and induce the bleaching of PVX-infected tissues.In addition, the 35S:PVX-pri-amiR-GUS Nb negati v e control construct was also generated.These three constructs together with the insert-free 35S:PVX construct (Figure 7 A) were independently agroinoculated in one leaf of three N. benthamiana plants.A 'mock' set of plants was agroinfiltrated with the agroinfiltration solution.
Mild leaf curling typical of PVX-induced symptoms was observed in upper non-agroinoculated leaves between 4 and 5 dpa in all plants agroinoculated with the 35S:PVX and 35S:PVX-shc-amiR-NbSu constructs (Figure 7 B).Howe v er, curling in apical leaves from plants agroinoculated with the 35S:PVX-pri-amiR-GUS Nb and 35S:PVX-pri-amiR-NbSu constructs appeared 3-4 days later, between 7 and 9 dpa (Figure 7 B).Interestingly, bleaching of apical leaves appeared at 8-9 dpa only in plants agroinoculated with the 35S:PVX-shc-amiR-NbSu construct, and at 14 dpa extended to most of the apical tissues of infected plants (Figure 7 C).Additional analyses confirmed the decrease in chlorophyll a le v el in bleached tissues (Figure 7 D).Plants were monitored until 28 dpa, and only those expressing the 35S:PVX-shc-amiR-NbSu construct displayed the char acteristic y ellowing deri v ed from NbSu silencing.RNA-blot and RT-qPCR analyses showed that only tissues from plants expressing 35S:PVX-shc-amiR-NbSu accumulated high le v els of amiR-NbSu (Figure 7 E) and low le v els of NbSu mRNA (Figure 7 F).Interestingl y, sRN A sequencing from apical leaves of plants agroinoculated with 35S:PVX-shc-amiR-NbSu re v ealed that the shc-amiR-NbSu precursor inserted into PVX was accurately processed (Figure 7 G, Figure S8).Additional analyses showed that the percentages of sRNAs of (+) or (-) polarity deri v ed from the PVX sequence are similar (48% and 52%, respecti v ely), suggesting that these sRNAs originate from double-stranded RN A (dsRN A) intermediates produced by RN A-dependent RN A pol ymerases during PVX replication (Supplementary Figure S9).In contrast, the majority (89%) of sRNAs deri v ed from the shc-amiR-NbSu sequence included in PVX are of (+) polarity, similarly to when the same precursor is expressed without PVX (Supplementary Figure S9).These results most likely indicate tha t amiR-NbSu origina tes from the shc precursor rather than from dsRNA intermediates of replication.
Finally, the presence of both pri and shc precursors in apical leaves was analyzed by RT-PCR at 7 and 14 dpa with oligonucleotides flanking the precursor insertion site.At 7 dpa, the shc precursor was clearly amplified in plants expressing 35S:PVX-shc-amiR-NbSu , while only the pri precursor of the 35S:PVX-pri-amiR-NbSu samples was barely detected (Figure 7 H).Indeed, at 14 dpa only the shc precursor was detected (Figure 7 H).Importantly, PVX was detected in all plants expressing PVX-based constructs at the two timepoints analyzed (Figure 7 H) thus excluding the possibility that the absence of pri precursors was due to a lack of infection in these samples.Sequencing analysis of RT-PCR products from each of the three PVXshc-amiR-NbSu infected samples showed that no mutations Figur e 6. Functional anal ysis of constructs expressing the amiR-AtFT and amiR-AtCH42 amiRNAs against A. thaliana FL OWERING L OCUS T ( AtFT ) or CHLORINE 42 ( AtCH42 ) from pri and shc precursors.( A ) Nucleotides corresponding to the guide strand of the amiRNA against AtFT and AtCH42 are in light green and light yellow, respecti v el y, w hile nucleotides of AtFT and AtCH42 target mRNAs are in dark green and dar k yellow, respecti v ely.Other details are as in Figure 2 A  accumulated in the shc-amiR-NbSu insert.These results indica te tha t the pri fragments are elimina ted from the PVX vector during viral infection, while the shorter shc fragments are stable both structurally and at a sequence level.

A PVX-based, DNA-free system for inducing widespread silencing in plants with amiRNAs
Ne xt, we e xplored the possibility of triggering widespread silencing in a DNA-free, non-transgenic manner, by us-ing alternati v e methods to agroinfiltration for inoculating shc -containing PVX sequences.For this purpose, a crude extract from bleached apical leaves of each of the three individual plants agroinoculated with 35S:PVXshc-amiR-NbSu was pr epar ed.Each crude extract was used as inoculum to mechanically inoculate one young N. benthamiana plant (Figure 8 A).As controls, similar crude extracts wer e pr epar ed from mock and empty PVX-agroinoculated plants and analyzed in parallel.Inter estingly, all thr ee plants mechanically inoculated with Finally, we wondered if widespread silencing could also be triggered by a ppl ying PVX-shc-amiR-NbSu-containing r aw extr acts through spr aying (Figure 8 A).Remar kab ly, all three plants spr ay ed with crude extracts derived from 35S:PVX-shc-amiR-NbSu agroinoculated plants displayed b leached apical leav es w hich accum ulated authentic shc-amiR-NbSu precursors with no sequence alterations as confirmed by sequencing of the RT-PCR products (Figure 8 C).Importantly, plants spr ay ed with crude extracts from 35S:PVX -agroinoculated plants or mock-inoculated plants displayed only mild PVX-deri v ed symptoms or no symptoms at all, respecti v ely, and no bleaching (Figure 8 C).These results indicate that crude extracts can be directly spr ay ed to leaves to trigger amiR-VIGS in N. benthamiana in a non-transgenic, DNA-free manner.

DISCUSSION
In this work, we have systematically analyzed the minimal structural and sequence r equir ements for producing effecti v e amiRNAs from the 521-nt long AtMIR390a precursor.Our analyses show that highly effecti v e amiRNAs can be pr oduced fr om the considerably shorter shc precursor of only 89 nt, including the BS of AtMIR390a and the DSL of OsMIR390 with a 2-nt deletion.The unique ability of shc precursors to stably and efficiently express authentic amiR-NAs from an RNA viral vector, such as that derived from PVX, highlights the utility of engineering such class of minimal precursors for transgene-free widespread gene silencing in plants.

Effects on miRNA accumulation of sequence modifications at the BS of the precursor
Systema tic muta tional analyses done in se v eral plant pri-miRN As for anal yzing miRN A production have shown that, for base-to-loop processed precursors, sequence modifica tions a t the BS region have more deleterious effects in both foldback processing efficiency and accuracy than mutations in the DSL region, while deletions of nucleotides of the ssRNA segments seem largely neutral ( 4 , 6 ).In se v eral plant MIRNA precursors, a single change in the BS ∼15 nt below the miRNA is sufficient to completely abolish their processing (47)(48)(49), as observed for instance in AtMIR390a , in which a G-to-A point mutation in the base of the foldback resulted in misprocessing of the miR390 / miR390* duplex and subsequent reduction in TAS3 -deri v ed tasiRNA le v els ( 50 ).All in all, these studies support the idea that base-to-loop precursors require a BS of ∼15 nt for DCL1 recognition and accurate first cleavage ( 9 ).Our results generally agree with this, as shortening AtMIR390a BS drastically affects both amiRNA accumulation and induced target mRNA silencing.Still, it is worth to note that some degree of amiRN A accum ulation and target mRN A silencing was observed when expressing amiRNAs from the BS-D7 precursor, suggesting that DCL1 activity may not be completely abolished when processing AtMIR390a -based amiRNA precursors with a BS shorter than 15 nt.Pertinent to this context, it was reported that an artificially shortened form of AtMIR166b precursor with only six nucleotides at the BS was processed in vivo , releasing miR166b and causing de v elopmental defects ( 51 ).Importantly, it seems that an intact BS of AtMIR390a is necessary but sufficient for high amiRN A accum ulation to le v els comparable to those obtained with the pri-AtMIR390a precursor.Similarly, it was pr eviously r eported that miR171 was efficiently and accurately released from the AtMIR171 foldback in N. benthamiana ( 52 ), thus r einfor cing the idea that ssRNA segments seem dispensable for efficient and accurate miRNA precursor processing.In contrast, in the case of loop-to-base precursors, the BS region is probably dispensable for processing, as shown with the efficient processing of an AtMIR157c -based precursor lacking the BS region ( 53 ).In a more recent genome-wide study, a set of sequence features for the processing of plant pri-miRNA precursors were identified ( 54 ).In particular, specific nucleotide pairs at the precursors' DCL1 cleavage sites were observed, with the identity of the nucleotides located at the mismatched positions of the BS having a big impact on the processing efficiency.Here, we did not analyze the effect of changing the identity of the nucleotides that are mismatched in AtMIR390a BS .Hence , we cannot rule out that some of these changes might, for instance, increase amiRNA accumulation, as observed for miR172 when expressed from pri-AtMIR172a with intentionally paired nucleotides at the first DCL1 cleavage site ( 54 ).

Effects of sequence modifications at the DSL of the precursor on miRNA accumulation
For base-to-loop processed miRNA precursors, a few reports have shown that their nati v e DSL region can be modified without compromising the processing of the precursor.For instance, eliminating bulges at the DSL does not interfere with miR167 biogenesis from AtMIR167a precursors ( 48 ), while an AtMIR166b precursor with a truncated DSL region of only 22-nt accumulated miR166b to le v els similar to those obtained with the wild-type precursor ( 51 ).Here, deletions of 6-13 nt in the stem of AtMIR390a DSL were tolerated, while further shortening of the stem dramatically decreased or abolished amiRN A bio genesis.Interestingl y, the chimeric pri-AtMIR390a-OsDSL precursor including OsMIR390 DSL produced high amounts of amiR-NbSu and amiR-NbCla in N. benthamiana agroinfiltrated leaves, as r eported befor e for other amiRNA sequences ( 43 ).In this work, we further deleted the distal stem of OsMIR390 while maintaining the terminal loop and found that a 2-nt deletion was tolerated, while longer deletions had a negati v e effect of amiRN A accum ulation.This result suggests that a minimal distal stem length is r equir ed for efficient processing, as proposed before for explaining miR172 misprocessing from an AtMIR172 precursor lacking the terminal loop ( 47 ).

A DN A-free, RN A virus-based system to induce widespread gene silencing in plants
AmiRNAs expressed from viral vectors have been used before in plants for silencing endogenous genes.In particular, nuclear-replicating viruses of DNA genome such as cabbage leaf curl virus (CaLCuV), tomato yellow leaf curl China virus (TYLCCNV) or tomato mottle virus (ToMoV) successfully expressed authentic amiRNAs in N. benthamiana (22)(23)(24), while the cytoplasmic-replicating tobacco mosaic virus (TMV) of RNA genome failed to produce effecti v e amiRNAs in this same host ( 23 ).Inter estingly, ther e is only one example in plants of the use of cytoplasmic-replicating RNA virus [barley stripe mosaic virus (BSMV)] to express amiRNAs ( 55 ), although the accumulation of authentic amiRNAs in vivo was not proven.Still, a major drawback of the mentioned nuclear-replicating DNA viruses for use in amiR-VIGS is their limited host-range and, in some cases, tissue tropism, with viral r eplication r estricted to phloem cells.
Here we used PVX, a plus (+) single-stranded RNA virus infecting 62 plant species of 27 families including important crops in the family Solanaceae (e.g.pota to, toma to, pepper or tobacco) ( 56 , 57 ), to efficiently produce authentic amiRNAs in N. benthamiana as shown by northern blot and high-throughput sRNA sequencing.Where the shc precursors are accurately processed from PVX RNAs to release authentic 21-nt amiRNAs and which DCL is involved in this process is unknown.W ha t is known is that PVX replicates in the cytoplasm and that DCL4 is the primary plant DCL functioning in antiviral defense against RNA viruses and producing 21-nt sRNAs from diced viral RNAs ( 58 , 59 ).It is also known that DCL4, besides its ability for processing long dsRNA precursors, can access fle xib ly structured single-str anded RNA substr ates such as pre-miRNA-like RNAs, as shown in the biogenesis of se v eral sRNAs deri v ed from the satellite RNA of cucumber mosaic virus ( 60 ).Despite DCL4 could be the DCL generating amiR-NbSu from PVX RNAs, we cannot rule out that DCL1 might be tr ansiently tr anslocated to the cytoplasm to process the amiRN A precursor, similarl y to what has been described with the redistribution of Drosha during a Sindbis virus infection in an animal system ( 61 , 62 ).Alternati v ely, DCL1 might also act in the cytoplasm right after its synthesis and before being shuttled to the nucleus.We cannot either rule out that PVX RNA fragments that act as amiRNA precursors or the whole PVX genomic RN A enter transientl y into the nucleus allowing DCL1 processing and release of the amiRNA, although this last scenario seems more unlikely.In any case, 35S:PVX-amiR-NbSu -triggered NbSu silencing in DCL4i knockdown plants is significantly reduced compared to wild-type and DCL1i knockdown N. benthamiana plants (Supplementary Figure S10).These results suggest that DCL4 could be the main DCL responsible for the generation of amiR-NbSu from PVX RNAs in the cytoplasm acting directly on the precursor.In any case, the precision in the processing of the shc precursor from viral RNAs is remar kab le (Figure 7 G and Figure S8), and essentially no 21-nt secondary, phased sRNAs deri v ed from amiRNA targets were detected (Data S3, Figure S11) suggesting a lack of transitivity that and r einfor ces the specificity of the PVXbased amiR-VIGS a pproach.Interestingl y, our data presented in Figure S8 suggests that amiR-NbSu is processed from PVX (+) strand RNAs and not from RdRP-deri v ed dsRNA replica tion intermedia tes.Processing of PVX genomic RNA to produce amiR-NbSu seems unlikely, as this may have a deep impact on virus infectivity.Also, it may be difficult that amiR-NbSu precursor folds properly when embedded within the large genomic RNA.In contrast, based on the design of the PVX-based amiR-VIGS construct, the amiR-NbSu precursor is at the 5 terminus of the subgenomic RNA, which may allow its proper folding for accurate amiR-NbSu release.For all these reasons, amiR-NbSu may be deri v ed from the subgenomic (+) RNA generated during the infection rather than from PVX genomic RNA.
The use of viral vectors in VIGS approaches typically depends on the agroinfiltration of plant tissues with DNA construct(s) for triggering the viral infection ( 46 ).Therefore, although not stably tr ansformed, tr ansgenic plant tissues are produced, what certainly limits the biotechnological applications of this type of strategies.Our system uses crude extracts as inoculum, which are obtained from apical (non-agroinfiltrated) leaves infected with PVX variants carrying the amiRNA precursor.Hence, the use of crude extracts for triggering amiR-VIGS circumvents this limitation, while the deli v ery of raw extracts to plants through spraying should allow to scale up this methodology for large-surface applications in fields.A key aspect to highlight is the high stability of the shc precursor, which is retained in PVX genome e v en after one passage without sequence alterations, despite the well-known proneness of PVX to lose inserted foreign sequences through in planta recombination ( 63 ).Noteworthy, in this work, we used a PVX vector that incorporates a deletion of the amino-terminal end of the CP and a heterologous promoter deri v ed from another potexvirus, BaMV, that was previously shown to contribute to insert stability ( 37 ).In any case, the small size and high sequence stability of the shc precursor in PVX, as well as the high specificity intrinsic of the amiRN A a pproach, make this PVX-based amiR-VIGS strategy an attracti v e alternati v e to classic dsRNA-based VIGS in plants.Indeed, the degree of target silencing induced by PVX constructs expressing amiR-NbSu from the shc precursor or a NbSu transcript fragment of the same size (89 nt) is similar (Supplementary Figure S12).These results support that PVX-based amiR-VIGS with shc precursors is highly effecti v e and may be preferred to classic dsRNA-based VIGS because of its higher specificity, at least when similar size inserts are used.AmiR-N As are computationall y designed to e xclusi v ely silence the intended target(s) while target gene fragments included in classic VIGS constructs typically generate large populations of sRNAs of unpredicted size and sequence that can accidentally target cellular RNAs and induce undesired offtarget effects.

Biotechnological interest of minimal amiRNA precursors
Se v eral plant monocistronic miRNA precursors have been used to express amiRNAs in plants [re vie wed in ( 19 )].Such precursors vary in structure and length, with an average length significantly higher than that of their corresponding miRNA foldback (Supplementary Figure S13A), as most of them correspond to full-length pri-miRNA precursors.To our knowledge, the shc-MIR390 precursor is the shortest of all amiRNA precursors tested in plants so far, being considerably shorter than its parental pri-AtMIR390a precursor (Supplementary Figure S13B).Minimal amiRNA precursors may have several advantages.First, they should be more stable and retained in the viral genome for longer periods, as discussed before for PVX.This advantage may be of particular interest in amiR-VIGS experiments involving viruses with limited cargo capacity.Second, using shorter amiRNA precursor sequences should decrease the cost of producing amiRN A constructs, w hich is particularl y convenient in high-throughput applications where large amounts of amiRNA constructs are generated for plant transformation (64)(65)(66).Here, shc precursors are inserted into BS-AtMIR390a-B / c-based vectors after annealing two oligos of 58 bases, while cloning of amiRNA inserts in previous AtMIR390a-B / c -based vectors r equir ed the annealing of 75 bases-long oligonucleotides ( 16 ).Importantly, the reduction of the oligonucleotide length from 75 to 58 bases should allow for a more cost-effecti v e oligo synthesis and at the lowest production scale (25 nmole) offered by the main oligo-manufacturing companies.And third, minimal amiRNA precursors may be more efficiently synthesized in vitro or in bacteria for topical application to plants, among other possible advantages.In any case, the use of minimal amiRNA precursors in a non-transgenic, DNA-free manner should definitely accelerate the de v elopment of ne xtgeneration RNAi treatments of crops in line with international regulations.

SUPPLEMENT ARY DA T A
Supplementary Data are available at NAR Online.

Figur e 1 .
Figur e 1. Methodolo gy for easil y visualizing and quantifying the silencing activity of amiRNAs deri v ed from AtMIR390a -based precursors.( A and B ) Steps for the analysis of fully agroinfiltrated leaves or patches, respecti v ely.

Figure 2 .
Figure 2. Functional analysis of constructs expressing amiR-NbSu or amiR-NbDXS against N. benthamiana SULPHUR ( NbSu ) or DXS ( NbDXS ), respecti v ely, from AtMIR390a -based precursors with shortened distal stem-loop (DSL) region.( A ) Diagram of AtMIR390a -based amiRNA constructs including the base-pairing of amiRNAs and target mRNAs.Nucleotides corresponding to the guide strand of the amiRNA against NbSu and NbDXS are in blue and orange, respecti v el y, w hile nucleotides of target mRNAs are in dark green.The arrows indicate the amiRNA-predicted cleavage site.( B ) Diagrams of the amiRNA precursors with nucleotides corresponding to the amiRNA guide and amiRNA* strands are in blue and gr een, r especti v ely, while those from AtMIR390a are in black.The two nucleotides specific to each amiRNA construct that are modified for preserving authentic pri foldback secondary structur e ar e in r ed.The shapes corr esponding to pri basal stem (BS) or DSL r egions ar e in light blue.The size in nucleotides of the DSL region of each precursor is gi v en. ( C ) Photos at 7 days post-agroinfiltration (dpa) of leaves agroinfiltrated with different constructs.'Yes' and 'No' labels indicate the presence or absence of bleached patches, respectively.( D ) Bar graph showing the relati v e content of chlorophyll a in patches agroinfiltrated with different constructs ( pri-amiR-GUS Nb = 1.0).Bars with the letter 'b' or 'a' are significantly different from that of the corresponding control samples pri-amiR-GUS Nb or pri-amiR-NbSu / pri-amiR-NbDXS , respecti v ely ( P < 0.05 in pairwise Student's t -test comparisons).( E ) Northern blot detection of amiR-NbSu and amiR-NbDXS in RNA preparations fr om agr oinfiltra ted leaves a t 2 dpa.The graph at top shows the mean ( n = 3) + standar d de viation amiRNA relati v e accumulation ( pri-amiR-NbSu = 1.0; pri-amiR-NbDXS = 1.0).Bars with a letter 'a' are significantly different from that of sample pri-amiR-NbSu or pri-amiR-NbDXS .One blot from three biological replicates is sho wn. ( F ) Tar get mRN A accum ula tion in agroinfiltra ted leaves.Mean rela ti v e le v el ( n = 3) + standard error of NbSu or NbDXS mRNAs after normalization to PROTEIN PHOSPHATASE 2A ( PP2A ), as determined by quantitati v e RT-PCR (qPCR) ( pri-amiR-GUS Nb = 1.0 in all comparisons).Other details are as in (D).
3 C).In contrast, weaker bleaching was observed in ar eas expr essing any of the two amiR-NAs from BS-D7 , or amiR-NbSu from BS-D17 , BS-D23 or BS-31 (Figure 4 B), corresponding with slight decreases in chlorophyll a content compared to controls (Figure 4 C).No bleaching or changes in chlorophyll a content were observed in leaves expressing amiR-NbDXS from BS-D17 , BS-D23 or BS-31 (Figure 4 B and C).Both amiR-NbSu and amiR-NbDXS amiRNAs expressed from the pri or BS precursors accumulated to similarly high levels and induced comparable target mRNA silencing (Figure 4 D and

Figure 3 .
Figure 3. Functional analysis of constructs expressing amiR-NbSu or amiR-NbDXS against N. benthamiana SULPHUR ( NbSu ) or DXS ( NbDXS ) from chimeric precursors including intact AtMIR390a basal region and shortened OsMIR390 distal stem-loop (DSL) region.( A ) Diagrams of the amiRNA precursors.Nucleotides from OsMIR390 are in grey.The shapes corresponding to OsMIR390a distal stem-loop regions are in light green.Other details are as in Figure 2 B. ( B ) Photos at 7 days post-agroinfiltration (dpa) of leaves agroinfiltrated with different constructs.( C ) Bar graph showing the relati v e content of chlorophyll a in patches agroinfiltrated with different constructs ( pri-amiR-GUS Nb = 1.0).Other details are as in Figure 2 D. ( D ) Northern blot detection of amiR-NbSu and amiR-NbDXS amiRNAs in RNA preparations from agroinfiltrated leaves at 2 dpa.Other details are as in Figures 2 D and E. ( E ) Target mRNA accumulation in agroinfiltrated leaves.Other details are as in Figures 2 D andF.
Figure 3. Functional analysis of constructs expressing amiR-NbSu or amiR-NbDXS against N. benthamiana SULPHUR ( NbSu ) or DXS ( NbDXS ) from chimeric precursors including intact AtMIR390a basal region and shortened OsMIR390 distal stem-loop (DSL) region.( A ) Diagrams of the amiRNA precursors.Nucleotides from OsMIR390 are in grey.The shapes corresponding to OsMIR390a distal stem-loop regions are in light green.Other details are as in Figure 2 B. ( B ) Photos at 7 days post-agroinfiltration (dpa) of leaves agroinfiltrated with different constructs.( C ) Bar graph showing the relati v e content of chlorophyll a in patches agroinfiltrated with different constructs ( pri-amiR-GUS Nb = 1.0).Other details are as in Figure 2 D. ( D ) Northern blot detection of amiR-NbSu and amiR-NbDXS amiRNAs in RNA preparations from agroinfiltrated leaves at 2 dpa.Other details are as in Figures 2 D and E. ( E ) Target mRNA accumulation in agroinfiltrated leaves.Other details are as in Figures 2 D andF.

Figure 4 .
Figure 4. Functional analysis of constructs expressing the amiR-NbSu amiRNA against N. benthamiana SULPHUR ( NbSu ) or DXS ( NbDXS ) from AtMIR390a precursors with shortened basal stem (BS) region.( A ) Diagrams of the amiRNA precursors.The size in nucleotides of the BS region of each precursor is gi v en.Other details are as in Figure 2 B. ( B ) Photos at 7 days post-agroinfiltration (dpa) of leaves agroinfiltra ted with dif ferent constructs.( C ) Bar graph showing the relati v e content of chlorophyll a in patches agroinfiltrated with different constructs ( pri-amiR-GUS Nb = 1.0).Other details are as in Figure 2 D. ( D ) Northern blot detection of amiR-NbSu and amiR-NbDXS amiRNAs in RNA preparations from agroinfiltrated leaves at 2 dpa.Other details are as in Figures 2 D and E. ( E ) Target mRNA accumulation in agroinfiltrated leaves.Other details are as in Figures 2 D andF.
Figure 4. Functional analysis of constructs expressing the amiR-NbSu amiRNA against N. benthamiana SULPHUR ( NbSu ) or DXS ( NbDXS ) from AtMIR390a precursors with shortened basal stem (BS) region.( A ) Diagrams of the amiRNA precursors.The size in nucleotides of the BS region of each precursor is gi v en.Other details are as in Figure 2 B. ( B ) Photos at 7 days post-agroinfiltration (dpa) of leaves agroinfiltra ted with dif ferent constructs.( C ) Bar graph showing the relati v e content of chlorophyll a in patches agroinfiltrated with different constructs ( pri-amiR-GUS Nb = 1.0).Other details are as in Figure 2 D. ( D ) Northern blot detection of amiR-NbSu and amiR-NbDXS amiRNAs in RNA preparations from agroinfiltrated leaves at 2 dpa.Other details are as in Figures 2 D and E. ( E ) Target mRNA accumulation in agroinfiltrated leaves.Other details are as in Figures 2 D andF.

Figure 5 .
Figure 5. Functional analysis of constructs expressing the amiR-NbSu amiRNA against N. benthamiana SULPHUR ( NbSu ) or DXS ( NbDXS ) from pri and shc precursors.( A ) Diagrams of the pri and shc amiRNA precursors.The total length of each precursor is indicated.Other details are as in Figure 2 B. ( B ) Photos at 7 days post-agroinfiltration (dpa) of leaves agroinfiltrated with the two constructs.( C ) Northern blot detection of amiR-NbSu and amiR-NbDXS amiRNAs in RNA preparations from agroinfiltrated leaves at 2 dpa.Other details are as in Figures 2 D and E. ( D ) Target mRN A accum ulation in agroinfiltrated leaves.Other details are as in Figures 2 D and F. ( E ) amiRNA processing from AtMIR390a -based precursors.Pie charts show percentages of reads corresponding to expected, accurately processed 21-nucleotide mature amiR-NbSu and amiR-NbDXS (blue or orange sections, respecti v ely) or to other 19-24-nucleotide sRNAs (gray sectors).
Figur e 6. Functional anal ysis of constructs expressing the amiR-AtFT and amiR-AtCH42 amiRNAs against A. thaliana FL OWERING L OCUS T ( AtFT ) or CHLORINE 42 ( AtCH42 ) from pri and shc precursors.( A ) Nucleotides corresponding to the guide strand of the amiRNA against AtFT and AtCH42 are in light green and light yellow, respecti v el y, w hile nucleotides of AtFT and AtCH42 target mRNAs are in dark green and dar k yellow, respecti v ely.Other details are as in Figure2 A. ( B ) Representati v e images of Arabidopsis T1 transgenic plants expressing amiRNAs from different precursors.Left, 45-dayold adult plants expressing amiR-GUS At or amiR-AtFT.Right, 10-day-old T1 seedlings expressing amiR-AtCH42 and showing bleaching phenotypes of di v erse degrees.( C ) Phenotyping analysis.Left, box plot r epr esenting the mean flowering time of Arabidopsis T1 transgenic plants expressing amiR-GUS At or amiR-AtFT from different precursors.Pairwise Student's t -test comparisons are represented with the letter 'a' if significantly different ( P < 0.05) and the letter 'b' if not ( P > 0.05).Right, bar graph r epr esenting, for each line, the proportion of seedlings displaying a se v ere (b lack areas), intermediate (dark gray areas), or weak (light gray areas) bleaching phenotype, or with wild-type appearance (white areas).( D ) amiRNA and target RNA accumulation.Left, bar graph showing the relati v e accumulation of amiR-AtFT [mean relati v e le v el ( n = 3) + standard devia tion amiRNA rela tive accumula tion, pri-amiR-AtFT = 1.0] and of AtFT mRNA [mean relati v e le v el ( n = 3) + standard error of AtFT mRNAs after normalization to ACTIN 2 ( AtACT2 ), as determined by quantitati v e RT-qPCR, pri-amiR-GUS At = 1].Right, bar graph showing the relati v e accumulation of amiR-AtCH42 [mean relati v e le v el ( n = 3) + standard devia tion amiRNA rela tive accumula tion, pri-amiR-AtCH42 = 1.0] and of AtCH42 mRNA [mean relati v e le v el (n = 3) + standard error of AtCH42 mRNAs after normalization to A CTIN 2 ( AtA CT2 ), as determined by quantitati v e RT-qPCR, pri-amiR-GUS At = 1].( E ) amiRNA processing from pri or shc precursors.Pie charts show percentages of reads corresponding to expected, accurately processed 21-nucleotide mature amiR-AtFT and amiR-AtCH42 (green or yellow sections, respecti v ely) or to other 19-24-nucleotide sRNAs (gray sectors).

Figure 7 .
Figure 7. Functional analysis of Potato virus X (PVX) constructs expressing amiRNAs from pri and shc precursors.( A ) Diagram of PVX-based constructs.pri-amiR-GUS Nb , pri-amiR-NbSu and shc-amiR-NbSu cassettes are shown in grey, dar k b lue and light blue box es, r espectively.PVX genes RdRp, TGB and CP are represented in white boxes, and CP promoter from Bamboo mosaic virus (BaMV) with a white arrow.( B ) Two-dimensional line graph showing, for each of the three-plant sets listed, the percentage of symptomatic plants per day during 14 days.( C ) Photos at 14 days post-agroinfiltration (dpa) of sets of three plants agroinfiltrated with the different constructs.( D ) Bar graph showing the relati v e content of chlorophyll a in apical leaves from plants agroinfiltrated with different constructs (Mock = 1.0).Bar with the letter 'a' are significantly different from that of the corresponding mock control samples ( P < 0.05 in pairwise Student's t -test comparison).( E ) Northern blot detection of amiR-NbSu in RNA preparations from apical leaves collected at 14 dpa and pooled from three independent plants.( F ) Target NbSu mRN A accum ulation in RN A preparations from apical leaves collected at 14 dpa and anal yzed individuall y.Mean relati v e le v el ( n = 3) + standar d error of NbSu mRN As after normalization to PR OTEIN PHOSPHATASE 2A ( PP2A ), as determined by RT-qPCR (mock = 1.0 in all comparisons).Other details are as in D. ( G ) amiR-NbSu processing from the shc precursor included in PVX ( 35S:PVX-shc-amiR-NbSu construct).Pie chart show percentages of reads corresponding to expected, accurately processed 21-nucleotide mature amiR-NbSu (blue sections) or to other 19-24-nucleotide sRNAs (gray sectors).( H ) RT-PCR detection of PVX, pri and shc precursors in apical leaves at 7 and 14 dpa.RT-PCR products corresponding to the control NbPP2A are also shown (bottom), as well as control amplifications of pri and shc fragments from plasmids.

Figur e 8 .
Figur e 8. Non-transgenic, DN A-fr ee widespr ead gene silencing through amiR-VIGS.( A ) Experimental set up to test gene silencing trigger ed by PVXshc-amiR-NbSu .( B ) Widespread NbSu silencing induced by mechanically inoculated crude extracts.Left, photos at 14 days post-agroinfiltration (dpa) of sets of three plants mechanically inoculated with different crude extracts obtained from agroinoculated plants.Right, RT-PCR detection of PVX and shc precursors in apical leaves at 14 dpa.RT-PCR products corresponding to the control NbPP2A are also shown (bottom).( C ) Widespread NbSu silencing induced by spr ay ed crude extr acts.Left, photos at 14 days post-agroinfiltration (dpa) of sets of three plants spr ay ed with different crude extracts obtained fr om agr oinocula ted plants.Right, RT-PCR detection of PVX and shc precursors in apical leaves a t 14 dpa.Other details are as in (B).