Down-regulation of tomato STEROL GLYCOSYLTRANSFERASE 1 perturbs plant development and facilitates viroid infection

Abstract Potato spindle tuber viroid (PSTVd) is a plant pathogen naturally infecting economically important crops such as tomato (Solanum lycopersicum). Here, we aimed to engineer tomato plants highly resistant to PSTVd and developed several S. lycopersicum lines expressing an artificial microRNA (amiRNA) against PSTVd (amiR-PSTVd). Infectivity assays revealed that amiR-PSTVd-expressing lines were not resistant but instead hypersusceptible to the viroid. A combination of phenotypic, molecular, and metabolic analyses of amiRNA-expressing lines non-inoculated with the viroid revealed that amiR-PSTVd was accidentally silencing the tomato STEROL GLYCOSYLTRANSFERASE 1 (SlSGT1) gene, which caused late developmental and reproductive defects such as leaf epinasty, dwarfism, or reduced fruit size. Importantly, two independent transgenic tomato lines each expressing a different amiRNA specifically designed to target SlSGT1 were also hypersusceptible to PSTVd, thus demonstrating that down-regulation of SlSGT1 was responsible for the viroid-hypersusceptibility phenotype. Our results highlight the role of sterol glycosyltransferases in proper plant development and indicate that the imbalance of sterol glycosylation levels favors viroid infection, most likely by facilitating viroid movement.


Introduction
Viroids are plant pathogens composed of a small (250-400 nucleotides), circular, single-stranded RNA genome without protein-coding capacity (Navarro et al., 2021a). They induce symptoms in infected plants such as stunting, chlorosis, leaf curling, bark alterations, and size reduction of fruits and tubers (Flores et al., 2005). Viroids replicate through an RNA-based rolling-circle mechanism (Branch and Robertson, 1984), move from cell to cell via plasmodesmata and This paper is available online free of all access charges (see https://academic.oup.com/jxb/pages/openaccess for further details) over long distances through the phloem stream (Pallás and Gómez, 2017), and are classified into the Pospiviroidae or Avsunviroidae families, whose members replicate and accumulate in the nucleus and chloroplasts, respectively (Flores et al., 2015;Daròs, 2016). Viroid pathogenesis is a complex process regulated by multiple factors including: (i) alteration of the expression of genes involved in the plant defense response, stress response, cell wall structure, and chloroplast function, among others (Owens et al., 2017); (ii) the direct interaction of viroid genomic RNA with host factors (Adkar-Purushothama and Perreault, 2020); (iii) sequencespecific cleavage by viroid-derived small RNAs (vd-sRNAs) of complementary host mRNAs corresponding to genes involved in development; and (iv) the induction of ribosomal stress (Cottilli et al., 2019). RNA-based resistance to viroids (reviewed in Dalakouras et al., 2015;Flores et al., 2017) has been achieved through the transgenic expression of antisense RNAs (Matoušek et al., 1994), hammerhead ribozymes (Atkins et al., 1995;Yang et al., 1997;Carbonell et al., 2011), dsRNA ribonucleases (Sano et al., 1997), and hairpin RNAs of viroid sequence (Carbonell et al., 2008;Schwind et al., 2009;Adkar-Purushothama et al., 2015), or by exogenous application of large amounts of dsRNA molecules of viroid sequence to leaves (Carbonell et al., 2008). More recently, a large screening of multiple artificial microRNAs (amiRNAs) targeting sites distributed along viroid RNAs identified several amiRNAs that were highly active in agroinfiltrated leaves when co-expressed with an infectious viroid transcript (Carbonell and Daròs, 2017).
Plants contain three major species of sterols, named β-sitosterol, stigmasterol, and campesterol, and some members of the Solanaceae family also include important amounts of cholesterol (Moreau et al., 2002;Benveniste, 2004). Sterols accumulate in different forms, as free sterols (FSs) with a free β-hydroxyl group at C-3 position on the backbone, conjugated esters, sterol glycosides (SGs), and acyl SGs. SGs are produced by UDP-glucose:sterol glycosyltransferases (SGTs), which catalyze the transfer of a glucose residue from UDPglucose to the free hydroxyl group at position C-3 of FSs (Moreau et al., 2002;Benveniste, 2004). Sterols are important for regulating plant growth and development, as changes in cellular sterol composition affect a variety of cellular processes, such as vascular and stomatal patterning, cell division, expansion, and polarity, cell-to-cell connectivity, and hormonal control, among others (Ramirez-Estrada et al., 2017). Sterols are also key structural components of cellular membranes, regulate different membrane functions (e.g. passive or active transport across the membrane, activity of membranes associated with proteins), and changes in their relative proportion are known to alter membrane biophysical properties (Roche et al., 2008;Grosjean et al., 2015). Hence, sterols also play an important role in the plant response to a diverse list of abiotic and biotic stresses, including thermotolerance, drought, metal ions, hydrogen peroxide, and bacterial or fungal pathogens. However, the role of sterols in plant defense against viruses or viroids is unknown (Altabella et al., 2022).
Here, we aimed to engineer tomato plants highly resistant to potato spindle tuber viroid (PSTVd), the type species of the Pospiviroidae family (Gross et al., 1978). PSTVd naturally infects economically relevant crops such as potato (Solanum tuberosum) and tomato (Solanum lycopersicum), and is currently classified as a quarantine pathogen in certain regions of the world (Tsuda and Sano, 2014). A construct expressing a highly active anti-PSTVd amiRNA in Nicotiana benthamiana (Carbonell and Daròs, 2017) was introduced into tomato plants to generate multiple stably transformed lines that were analyzed for antiviroid resistance. Infectivity assays revealed, contrary to our expectations, that all amiRNA lines were hypersusceptible to PSTVd. A combination of phenotypic, molecular, and metabolic analyses of several amiRNA-expressing lines non-inoculated with the viroid revealed that the amiRNA was accidentally silencing the tomato STEROL GLYCOSYLTRANSFERASE 1 (SlSGT1) gene, which caused late developmental and reproductive defects such as dwarfism, leaf epinasty, and reduced fruit size and weight. Moreover, two independent transgenic tomato lines, each expressing a different amiRNA specifically designed to target SlSGT1, were also hypersusceptible to PSTVd, thus confirming that down-regulation of SlSGT1 was responsible for the hypersusceptibility phenotype.

Generation and phenotyping of tomato cv. Moneymaker transgenic plants
Agrobacterium tumefaciens LBA4404 transformed with 35S:amiR-PSTVd was co-cultured with tomato cotyledons. Explant preparation, selection, and regeneration were performed as previously described (Ellul et al., 2003). Transformants were selected in hygromycin-containing medium, and then propagated in soil for seed production and for the infection studies. Non-transgenic controls (NTCs) were in vitro-regenerated tomato plants obtained in parallel with the transgenic plants. Phenotyping analyses were done using four NTCs and four independent amiRNA lines.

Viroid infection assays
A CF11 cellulose-treated RNA extract was obtained from N. benthamiana tissue infected with PSTVd RG1 strain (GenBank accession no. U23058). Tomato plants were inoculated by mixing 5 μl of the infectious extract with 5 μl of a 10% carborundum solution (in 50 mM K 2 HPO 4 ) on one leaf and evenly spreading the inoculum with a glass rod. After inoculation, plants were monitored for viroid symptom appearance for 4 weeks. The two youngest leaves were sampled for RNA preparation and analysis.

RNA preparation
Total RNA from S. lycopersicum leaves was isolated in extraction buffer (1 M guanidinium thiocyanate, 1 M ammonium thiocyanate, 0.1 M sodium acetate, 5% glycerol, 38% water-saturated phenol), followed by chloroform extraction. RNA was precipitated in 0.5× isopropanol for 20 min. Triplicate samples from pools of two leaves were analyzed.

Real-time RT-qPCR
Real-time reverse transcription followed by quantitative PCR (RT-qPCR) was done essentially as described by López-Dolz et al. (2020) in a QuantStudio 3 Real-Time PCR System (Thermo Fisher Scientific). Primers used for RT-qPCR are listed in Supplementary Table S1. Target RNA expression levels were calculated relative to two S. lycopersicum reference genes (SlACT and SlEXP) using the delta delta cycle threshold comparative method of the QuantStudio Design and Analysis Software (version 1.5.1.; Thermo Fisher Scientific).

Small RNA blot assays
Total RNA (20 μg) was separated in 17% polyacrylamide gels containing 0.5× Tris/borate EDTA (TBE) and 7 M urea and transferred to a positively charged nylon membrane. Probe synthesis using [γ-32 P]ATP (PerkinElmer) and T4 polynucleotide kinase (Thermo Fisher Scientific) and northern blot hybridizations were done at 38 °C in PerfectHyb Plus Hybridization Buffer (Sigma-Aldrich) as described previously (Montgomery et al., 2008;Carbonell et al., 2014). A Typhoon IP Imager System (Cytiva) was used to produce digital images from radioactive membranes. Oligonucleotides used as probes for sRNA blots are listed in Supplementary Table S1.

5ʹ-RLM-RACE
RNA ligase-mediated rapid amplification of 5ʹ cDNA ends (5ʹ-RLM-RACE) was done using the GeneRacer kit (Life Technologies) as described by Carbonell et al. (2015), except that the 5ʹ end of cDNA specific to SlSGT1 was directly amplified in a single PCR using the GeneRacer 5ʹ and gene-specific AC-569 oligonucleotides. 5ʹ-RLM-RACE products were gel purified and cloned using the Zero Blunt TOPO PCR Cloning Kit (Life Technologies), introduced into Escherichia coli DH5α, screened for inserts, and sequenced. Control PCR reactions to amplify SlACT were done using oligonucleotides AC-280 and AC-281. The sequences of the oligonucleotides used are listed in Supplementary Table S1.

Sterol analysis
For leaf sterol measurements, a pool of two apical leaves from four NTCs and four amiR-PSTVd independent S. lycopersicum lines was analyzed. Sterols were extracted and quantified by GC-MS in three technical replicates for each genotype (n=3), as previously described (Ramirez-Estrada et al., 2017).

RNA sequencing and data analysis
Total RNA from four wild-type and four amiR-PSTVd independent S. lycopersicum lines was analyzed for quantity, purity, and integrity with a 2100 Bioanalyzer (RNA 6000 Nano kit; Agilent), and submitted to BGI (Hong Kong, China) for strand-specific library preparation and mRNA sequencing (RNA-seq) in the DNBSEQ Platform. After quality analysis with FastQC (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/), and adapter removal and low-quality trimming of raw reads with cutadapt (Martin, 2011), clean read pairs were mapped to the S. lycopersicum genome (version 4.0) using HISAT2 (Kim et al., 2015), and non-uniquely mapped pairs were discarded. Numbers of read counts uniquely mapped to each tomato gene were obtained with htseq-count (Anders et al., 2015) using the iTAG4.0 gene model annotation. Genes not having at least 1 TPM (transcripts per kilobase per million mapped reads) in either the four wild-type replicates or the four amiR-PSTVd lines were filtered out. Differential expression analysis was done with DESeq2 (Love et al., 2014), with a false discovery rate of 1%. Differential gene expression analysis results are shown in Supplementary Dataset S2. An MA plot was generated using ggplot2 (Wickham 2016) in RStudio (https://www.rstudio.com/). Gene Ontology analysis was done using ShinyGo v0.75 (http://bioinformatics.sdstate.edu/go/) (Ge et al., 2020) with a false discovery rate of 5%.

Accession numbers
Solanum lycopersicum genes and corresponding locus identifiers are:

Transgenic tomato plants expressing an amiRNA against PSTVd are hypersusceptible to viroid infection
A construct expressing a highly effective anti-PSTVd amiRNA selected from a previous functional screening in N. benthamiana (Carbonell and Daròs, 2017) was introduced into S. lycopersicum cv. Moneymaker (Fig. 1A). Eight independent transgenic T 1 lines were generated, all of which had a phenotype that was indistinguishable from that of NTC tomato plants at 10 days post-transplanting (Fig. 1B). Northern blot analysis of RNA preparations obtained from apical leaves revealed that all lines accumulated amiR-PSTVd to similar levels (Fig. 1C). Four of the lines (#1, #2, #5, and #16) were selected for further analyses.
To analyze the antiviroid resistance of each independent line, three individuals (propagated from cuttings) of each transgenic line were inoculated with PSTVd. In parallel, three NTC plants were mock-or PSTVd-inoculated. The appearance of typical PSTVd symptoms (leaf epinasty and chlorosis) in distant non-inoculated tissues was recorded during the 4 weeks post inoculation (wpi). Surprisingly, amiRNA plants displayed symptoms earlier than NTC plants ( Fig. 2A). For instance, 37.5% of amiRNA plants showed symptoms as early as 11 days post inoculation (dpi), and 100% showed symptoms at 13 dpi. In contrast, in the NTC plants, the first symptoms were observed in 50% of the plants at 13 dpi, and only at 14 dpi were all the NTC plants symptomatic ( Fig. 2A). At the end of the experiment (4 wpi), PSTVd-infected amiRNA lines were clearly shorter than NTC plants (Fig. 2B).
PSTVd accumulation was analyzed by RT-qPCR in selected amiRNA lines and NTCs at 2 wpi and 4 wpi ( Fig. 2C). At 2 wpi, PSTVd accumulated to significantly lower levels in all amiRNA lines compared with infected NTCs, suggesting that amiR-PSTVd interferes with viroid infection, as predicted. However, this protective effect was transient because at 4 wpi PSTVd accumulation was similar in all infected plants (Fig.  2C). Importantly, the loss of the effect was not due to a decrease in amiR-PSTVd levels at later stages of infection, as amiR-PSTVd accumulated to similar levels at the two time points ( Supplementary Fig. S1). Because viroid symptomatology in tomato positively correlates with the expression of pathogenesis-related (PR) proteins such as pathogenesis-related protein 1 (PR1) and ribosomal stress-inducing factors such as NAC082 (Cottilli et al., 2019;Vázquez Prol et al., 2021), the accumulation of SlPR1 and SlNAC082 mRNAs was analyzed in all plants ( Fig. 2C). As expected, at 2 wpi both SlPR1 and SlNAC082 were induced in infected NTCs compared with mock NTCs. Interestingly, at the same time point, SlPR1 and SlNAC082 accumulated to significantly higher levels in all amiRNA lines. Similar results were obtained at 4 wpi, although the increase of SlPR1 and SlNAC082 levels in amiRNA lines relative to NTCs was more modest. All together, these results indicate that amiR-PSTVd-expressing lines are hypersusceptible to viroid infection. They also reveal that SlPR1 and SlNAC082 are induced in PSTVd-infected plants to levels that correlate with viroid symptomatology.

AmiR-PSTVd-expressing lines display late developmental defects
In parallel, NTCs and amiRNA lines were also grown for seed propagation in greenhouse conditions. Unexpectedly, at around 6 weeks post transplanting (wpt) of the in vitro plantlets,  all amiRNA lines showed slight dwarfing and leaf curling, which became obvious at 8 wpt (Fig. 3A). The mean height of amiRNA plants was significantly reduced (1.9-fold) compared with the NTCs, and the internode length of the amiRNA lines was also significantly reduced compared with that of the NTCs (Fig. 3B). In addition, tomato fruits from the amiRNA lines were significantly smaller and had a lower fresh weight than fruits from NTCs (Fig. 3C, D). In conclusion, amiR-PSTVd expressing lines showed obvious developmental defects at later stages of growth, including dwarfism, leaf epinasty, and reduced fruit size and weight.

Endogenous UDP-glucose:sterol glycosyl transferase 1 is down-regulated in amiR-PSTVd-expressing lines
The unexpected phenotype of the tomato transgenic lines prompted us to hypothesize that amiR-PSTVd might be accidentally targeting (and down-regulating) one or more endogenous transcripts involved in the induction of the observed late developmental defects. We reasoned that, if they exist, putative off-target(s) of amiR-PSTVd would have a high sequence complementarity with the amiRNA and be differentially underexpressed in amiR-PSTVd-expressing lines.
To reveal the identity of the putative off-target(s), first we used the TargetFinder tool (Fahlgren and Carrington, 2010) to generate a genome-wide list of all targets with a relatively high sequence complementarity with amiR-PSTVd (Supplementary Dataset S3). TargetFinder-predicted off-targets are ranked based on a Target Prediction Score (TPS) assigned to each amiRNA-target interaction. The TPS ranges from 1 to 11, that is, from the highest to the lowest degree of complementarity between the amiRNA and the putative target RNA. Indeed, amiR-PSTVd was designed with the P-SAMS amiRNA Designer tool (Fahlgren et al., 2016) because none of the interactions between amiR-PSTVd and cellular RNAs had a TPS <4, which is the arbitrary cut-off that P-SAMS uses to assign high specificity to the designed amiRNA. However, we cannot completely rule out the possibility that amiR-PSTVd-target RNA interactions with a TPS >4 may be productive and lead to the down-regulation of the corresponding target RNA. Table 1 lists amiR-PSTVd most probable off-target transcripts with a TPS ≤6 based on TargetFinder analysis.
Next, transcript libraries from four NTCs and four amiR-PSTVd-expressing lines were generated and analyzed. Differential gene expression analysis was done by comparing the transcript libraries from the NTCs with those from the amiR-PSTVd expressing lines. In total, 223 genes were differentially expressed in plants expressing amiR-PSTVd, with 176 genes being differentially overexpressed and 47 differentially underexpressed (Fig. 4A, Supplementary Dataset S2). Interestingly, Solyc06g007980.4 was the only differentially underexpressed gene included in the list of most probable off-targets (Table  1). Indeed, this gene was the fifth most differentially underexpressed (Supplementary Dataset S2) and the third potential off-target with lowest TPS (Supplementary Dataset S3). This gene encodes the tomato sterol 3-beta glycosyltransferase 1 (SlSGT1), one of the four isoenzymes that catalyzes the glycosylation of the free hydroxyl group at the C-3 position of sterols to produce SGs in tomato (Ramirez-Estrada et al., 2017). RT-qPCR analyses confirmed that SlSGT1 mRNA accumulated to significantly lower levels in amiRNA-expressing lines compared with NTCs (Fig. 4B). Next, we used 5ʹ-RLM-RACE to test for amiR-PSTVd-directed off-target cleavage of SlSGT1. If amiR-PSTVd is cleaving SlSGT1 mRNA, this analysis will detect the 3ʹ cleavage products. Cleavage products of the expected size (439 bp) were detected in samples expressing amiR-PSTVd, but not in NTC samples. Sequencing analysis confirmed that the majority (6 out of 7) of the sequences comprising these products contained the canonical 5ʹ end position predicted for amiR-PSTVd-guided cleavage (Fig. 4C).
Finally, SGs and FSs for the main tomato sterol species (cholesterol, campesterol, stigmasterol, and β-sitosterol) were analyzed and quantified in both NTC and amiRNA transgenic lines (Supplementary Table S2). As expected from SlSGT1 down-regulation, the content of bulk SGs was significantly reduced by ~50% in amiRNA lines compared with NTCs (Fig.  4D). Conversely, the content of major FSs was significantly increased in amiRNA lines (Fig. 4D), thus confirming the reduced glycosylation activity in SlSGT1-silenced amiRNA lines. All together, these results indicate that amiR-PSTVd down-regulates SlSGT1, which affects sterol glycosylation and most likely causes the late developmental defects observed in the tomato transgenic lines.

Is SlSGT1 a direct target of PSTVd during infection?
Notably, the late development defects of the amiRNA lines resembled the symptomatology observed in natural PSTVd infections. Thus, we hypothesized that SlSGT1 might be targeted during PSTVd infection in tomato and that SlSGT1 down-regulation may be, at least in part, responsible for the developmental defects observed in natural PSTVd infections. Because vd-sRNAs are known to associate with ARGONAUTE (AGO) proteins to target host genes (Navarro et al., 2021a), we thought that maybe vd-sRNAs with a sequence similar to amiR-PSTVd may exist in natural PSTVd infections, associate with AGO1, and target SlGST1. Thus, we searched public high-throughput sRNA sequencing data and found three PSTVd-derived sRNAs of (-) polarity with high sequence similarity to amiR-PSTVd (Fig. 5A) that were in the top 10 of most highly enriched vd-sRNAs in AGO1-immunoprecipitated fractions from N. benthamiana tissue infected with PSTVd (Minoia et al., 2014). Still, the lowest TPS from the interac-tions between PSTVd-sRNAs and SlGST1 corresponded to PSTVd-sRNA (-) (261-281) with a relatively high score of 8. 5ʹ-RLM-RACE analysis confirmed that SlSGT1 3ʹ cleavage products could be detected only in infected transgenic lines expressing amiR-PSTVd, and not in infected NTCs (Fig.  5B). In addition, we observed that SlSGT1 mRNA accumulation in NTC plants infected with PSTVd was not significantly different from that of mock-inoculated NTCs (Fig.  5C). Nonetheless, it is possible that SlSGT1 mRNA accumulation remains unaltered because down-regulation by a putative PSTVd-sRNA may be compensated by an induction of SlSGT1 transcription upon viroid infection. Thus, we also performed RT-qPCR with SlSGT1 intron-specific primers and found that SlSGT1 primary transcripts accumulate to similar levels in mock-and PSTVd-infected NTC plants (Supplementary Fig. S2). These results suggest that SlSGT1 expression is not induced in PSTVd-infected plants. Taken together, our results do not support that SlSGT1 may be a direct target in natural PSTVd infections in tomato.

Transgenic tomato plants expressing an amiRNA specifically designed against SlSGT1 are also hypersusceptible to viroid infection
To confirm that down-regulation of SlSGT1 induces hypersusceptibility to PSTVd, two transgenic S. lycopersicum cv. Micro-Tom lines expressing amiR-SGT1-1 or amiR-SGT1-2 amiRNAs specifically designed to target SlSGT1 (Fig. 6A) were challenged with PSTVd. Specifically, three individuals of each transgenic line and three NTCs were inoculated with PSTVd, and three other NTCs were mock inoculated. Plant response to PSTVd infection was analyzed as before, that is, recording the day of symptom appearance for each plant and analyzing the accumulation of PSTVd, and of SlPR1 and SlNAC082 mRNAs, at 2 wpi and 4 wpi. As a control, the accumulation of SlSGT1 was also analyzed at the same time points. As for the amiR-PSTVd lines, both amiR-SGT1-expressing lines displayed symptoms earlier than the NTC plants (Fig. 6B). For example, 75% of amiR-SGT1-1 and 100% of amiR-SGT1-2 plants showed symptoms as early as 10 dpi. In contrast, the first symptoms in NTC plants were observed in 50% of the plants at 12 dpi, and only at 13 dpi were all the NTC plants symptomatic (Fig. 6B). At the end of the experiment (4 wpi), both PSTVd-infected amiR-SGT1 lines were clearly more dwarfed than the NTC plants (Fig. 6C). RT-qPCR analysis showed that PSTVd accumulation was similar in all infected plants, at both 2 wpi and 4 wpi, whereas the accumulation of SlPR1 and SlNAC082 mRNAs was significantly greater in both amiR-SGT1 lines compared with the NTCs at the two time points (Fig. 6D). Finally, SlSGT1 mRNA accumulation was significantly reduced in both amiR-SGT1 lines compared with the NTCs, in both the absence and the presence of PSTVd (Fig. 6D).

Discussion
In this work, several S. lycopersicum cv. Moneymaker transgenic lines expressing an amiRNA against PSTVd (amiR-PSTVd) were generated. Phenotypic and molecular analyses revealed, contrary to our expectations, that amiR-PSTVd-expressing lines were not resistant but instead hypersusceptible to the viroid. Intriguingly, non-infected amiR-PSTVd lines showed late developmental defects including leaf epinasty, dwarfism, and reductions in fruit size and weight, which are also typical of PSTVd infection. A combination of transcriptome profiling  with RT-qPCR, 5ʹ-RLM-RACE, and metabolomic analyses revealed that these phenotypic defects were caused by the accidental down-regulation of endogenous SlGST1 by amiR-PSTVd-expressing lines.

SlSGT1 is required for proper plant and fruit development in tomato
Previous studies indicate that a proper balance between the levels of GSs and FSs is crucial for normal cell function and plant development. For instance, the ugt80A2,B1 Arabidopsis double mutant, which accumulates low GSs levels, displays diverse morphological and biochemical seed phenotypes (DeBolt et al., 2009), alterations in the male gametophyte (Choi et al., 2014), and aberrant root epidermal cell patterning (Pook et al., 2017). In our work, down-regulation of SlGST1 in amiR-PSTVd-expressing tomato plants induced late development phenotypes such as leaf epinasty, dwarfing, and reduced fruit size, similar to the phenotypes described in amiR-SGT1-expressing tomato plants (Chávez et al., 2023). Shortened plant height and leaf area were previously observed in Withania somnifera plants transiently expressing a VIGS construct against endogenous SGTs , whereas over-expression of W. somnifera SGT1 caused enhanced growth and expansion of leaves (Saema et al., 2016), thus reinforcing the idea of the essential role of SGTs in proper plant development. Here, as observed in our metabolic analyses, downregulation of SlSGT1 caused an imbalance of GSs and FSs which might alter membrane biophysical properties and negatively affect the function of membrane-localized proteins such as the receptors of the plant hormones cytokinins, ethylene, and brassinosteroids (Takeuchi et al., 2021). Interference with hormone receptor function might affect the growth hormone signaling pathway, which could hinder the growth of amiR-PSTVd plants. Indeed, the dwarfing and leaf epinasty phenotype observed in the amiR-PSTVd lines is typical of plants affected in brassinosteroid signaling (Zhu et al., 2013;Carbonell et al., 2015). Yet, the significantly higher levels of free campesterol, the main precursor of brassinosteroids, in amiR-PSTVd lines (Supplementary Fig. S3) suggest that brassinosteroid biosynthesis may not be negatively affected at least by the altered campesterol levels.
On the other hand, a Gene Ontology analysis on our RNAseq data revealed that several cellular component categories related to membranes were significantly altered (Supplementary Fig. S4; Supplementary Dataset S4), which is not surprising given the important roles of sterols in cellular membranes. Moreover, we identified several cytochrome P450-encoding genes that were misregulated (Supplementary Dataset S2). As cytochrome P450s are membrane-bound monooxygenases, the alteration of membranes due to SlGST1 down-regulation may also directly affect the stability of these enzymes. Interestingly, several cytochrome P450 enzymes catalyze essential oxidative reactions in the biosynthesis or catabolism of several plant hormones, such as brassinosteroids, abscisic acid, jasmonyl-isoleucine, and strigolactones (Bak et al., 2011). Pertinent to this context, our RNA-seq data indicate that CYP722 cytochrome P450 (Solyc02g084930), which is involved in brassinosteroid biosynthesis (Fujioka and Yokota, 2003;Nelson and Werck-Reichhart, 2011), is significantly down-regulated in amiR-PSTVd lines, while CYP707A1 (Solyc08g005610) and CYP94B3/CYP94C1 (Solyc03g111290/ Solyc06g074420), which participate in abscisic acid catabolism (Okamoto et al., 2006) and jasmonyl-isoleucine inactivation and attenuation of the jasmonic acid signaling cascade (Bruckhoff et al., 2016), respectively, are up-regulated in the same lines (Supplementary Dataset S2). Thus, it cannot be ruled out that altered cytochrome P450 balances may affect hormone biosynthesis or catabolic pathways. In any case, future research is needed to better dissect the effect of cellular membrane destabilization (due to reduced sterol glycosylation) on the stability of membrane components as well as the specific contribution of alterations in diverse hormone pathways in the phenotype of amiR-PSTVd lines.

Viroid-induced symptoms positively correlate with SlPR1 and SlNAC082 expression levels but not with viroid titer
Interestingly, both amiR-PSTVd-and amiR-SlSGT1-expressing plants showing more severe symptoms than NTCs accumulated similar or lower levels of PSTVd at 2 wpi and 4 wpi (Figs 2C, 6D). These data indicate that symptom development is not correlated with PSTVd accumulation. Although this result might seem surprising, this observation was previously reported in experiments comparing the degree of symptoms induced by different PSTVd strains in tomato plants (Gruner et al., 1995), and more recently also in N. benthamiana, Prunus persica (peach), and tomato plants infected with hop stunt viroid (Gomez et al., 2008), peach latent mosaic viroid (Rodio et al., 2006), or citrus exocortis viroid (Vázquez Prol et al., 2020), respectively. In any case, our results suggest that viroidinduced symptoms in distant tissues depend on the presence of the viroid but do not correlate with viroid titer. This could indicate that symptom development in distant tissues may be triggered when a certain amount of viroid accumulates in this tissue above a threshold, but is not enhanced by further viroid accumulation.
Viroids trigger plant defense responses and induce the expression of PR proteins such as PR1 (Granell et al., 1987;Lisón et al., 2013), whose induction in tomato correlates with viroidinduced symptomatology, as has been observed for several viroids (Vázquez Prol et al., 2021). Recently, the induction of the ribosomal stress mediator SlNAC082 was observed upon CEVd infection in tomato and correlated with viroid-induced symptoms (Cottilli et al., 2019;Vázquez Prol et al., 2020), indicating that viroids can induce alterations in the biogenesis of ribosomes. Here, both SlPR1 and SlNAC082 were induced in PSTVd-infected NTCs and significantly over-induced in amiR-PSTVd-or amiR-SlGST1-expressing plants infected with PSTVd (Figs 2C, 6D).

Viroid movement may be facilitated in SlSGT1-silenced plants
During infections, viroids move from inoculated to distal tissues, where symptoms such as leaf epinasty and chlorosis develop. Viroid movement includes intracellular movement (and subcellular compartmentalization for replication), exit of viroid progeny to neighboring cells, and entry to vascular tissue for long-distance trafficking (Pallás and Gómez, 2017). For instance, viroids can move from cell to cell via plasmodesmata, as was shown for PSTVd in tomato and tobacco mesophyll (Ding et al., 1997). Our results show that SlGST1 down-regulation impacts the cellular sterol profile (Supplementary Table  S2), with a decrease in SGs and an increase in FSs (Fig. 4D). Because SGs and FSs are primary structural elements of cell membranes (Roche et al., 2008;Grosjean et al., 2015), it is possible that cytoplasmic membranes are altered in SlSGT1silenced plants. This could affect membrane permeability and fluidity, and somehow facilitate cell-to-cell and long-distance transit of viroids through plasmodesmata, whose correct function depends on a specific membrane composition of lipids, including sterols (Grison et al., 2015). Indeed, cytopathic effects at the plasma membrane in association with CEVd in infected Gynura aurantiaca were described more than four decades ago (Semancik and Vanderwoude, 1976). It is also of note that viroid systemic infections generally take several weeks and are slow compared with their virus counterparts infecting the same hosts, as viruses usually reach distal tissues in just a few days. This difference could be related to viroid steric difficulties in transiting the plasmodesmata in the absence of movement proteins, such as those encoded in plant virus genomes, that facilitate the process. It is tempting to speculate that the sterol imbalance may disturb plasmodesmal control, which could be used by viroids to move faster from cell to cell through plasmodesmata in infected tissues. In this scenario, PSTVd may reach distant tissues more rapidly in SlSGT1-silenced plants and so incite symptoms earlier than in wild-type plants. This could explain why symptom development was accelerated in SlSGT1-silenced plants (Fig. 7). However, our RT-qPCR and 5ʹ-RLM-RACE analyses in PSTVd-infected NTCs could not prove that SlSGT1 is a direct target of PSTVd in natural infections. A close examination of recently published degradome datasets from PSTVd-infected tomato and N. benthamiana failed to identify degradome signatures of SlSGT1 mRNAs by PSTVd-sRNAs (Navarro et al., 2021b). Still, it is possible that both PSTVd-sRNAs and SlSGT1 mRNAs may be accumulating in a non-uniform manner in tomato tissues such that cleaving of SlSGT1 mRNAs might occur in only a subset of tissues. In this scenario, mRNA cleavage would be more difficult to detect than in the amiR-PSTVd plants, where amiRNA expression occurs in most (if not all) cells. In addition, SlSGT1 silencing may need to occur in only certain cell types to induce PSTVd-like phenotypes. Finally, it is also possible that PSTVd-sRNAs could block SlSGT1 translation or epigenetically regulate SlSGT1 expression, or that other SlSGT isoforms could be regulated upon natural viroid infections. For all these reasons we cannot rule out the possibility that SlSGT1 could be a direct target of PSTVd during natural infections.

Roles of SGTs in the stress response
SGTs have been recently linked to enhanced salt stress resistance, as reported for Arabidopsis plants overexpressing the W. somnifera SGT gene (Mishra et al., 2021). The role of SGTs in biotic stress responses has also been reported in several plant species. While silencing of several members of the SGT gene family led to a decrease in W. somnifera basal immunity against Alternaria alternata , overexpression of W. somnifera SGT in Arabidopsis and tobacco plants was associated with increased resistance against Alternaria brassicicola and Spodoptera litura, respectively (Pandey et al., 2014;Mishra et al., 2017). It is not known whether the resistance effects in these reports were due to the modified levels of SGs or induced by the concomitant changes observed for other plant defense compounds. Pertinent to this observation, it was also described that resistance against the bacterial necrotroph Botrytis cinerea is enhanced in the Arabidopsis ugt80A2;B1 double mutant, which is impaired in SG biosynthesis. In this case the resistance phenotype was explained by the transcriptional and metabolic changes induced by an increase in jasmonic acid and camalexin levels observed in this double mutant (Castillo et al., 2019).
Sterols have also been linked to plant virus replication (for a recent review see Altabella et al., 2022), as the formation of viral replication complexes involves the recruitment of high levels of sterols. Moreover, it is known that the inhibition of sterol biosynthesis reduces tombusvirus replication in yeast and plants (Sharma et al., 2010). In addition, it is possible that modifying the sterol profiles of a plant could ultimately increase its antiviral resistance, as suggested previously (Altabella et al., 2022). Therefore, our amiR-PSTVd or amiR-SlSGT1 plants could be a very useful tool to assess in future experiments whether low SlSGT1 levels induce enhanced antiviral resistance in tomato. A better understanding of the impact of sterol profiles on the biophysical properties of cellular membranes, and of the specific role of GSs predominant in Solanaceae species, will facilitate the development of enhanced antiviral resistance for next-generation crops.

Supplementary data
The following supplementary data are available at JXB online.
Dataset S3. amiR-PSTVd predicted off-targets. Dataset S4. Gene Ontology datasets (differentially expressed genes in each biological process, cellular component, and molecular function category). Fig. S1. Analysis of amiR-PSTVd accumulation during PSTVd infection in amiR-PSTVd-expressing tomato lines and NTCs. Fig. S2. Accumulation of SlSGT1 intron-containing primary transcripts in non-transgenic S. lycopersicum control plants (NTCs) during potato spindle tuber viroid infection. Fig. S3. Campesterol analysis. Fig. S4. Gene Ontology analysis. Table S1. Name, sequence, and use of oligonucleotides used in this study. Table S2. Content of glycosylated and free forms of cholesterol, campesterol, stigmasterol, and β-sitosterol in leaves of NTC plants and amiRNA lines.