Exchange of Small Regulatory RNAs between Plants and Their Pests^,

Regulatory small RNAs are found to exchange between plants and their pests.

Abstract

Regulatory small RNAs are well known as antiviral agents, regulators of gene expression, and defenders of genome integrity in plants. Several studies over the last decade have also shown that some small RNAs are exchanged between plants and their pathogens and parasites. Naturally occurring trans-species small RNAs are used by host plants to silence mRNAs in pathogens. These gene-silencing events are thought to be detrimental to the pathogen and beneficial to the host. Conversely, trans-species small RNAs from pathogens and parasites are deployed to silence host mRNAs; these events are thought to be beneficial for the pests. The natural ability of plants to exchange small RNAs with invading eukaryotic organisms can be exploited to provide disease resistance. This review gives an overview of the current state of trans-species small RNA research in plants and discusses several outstanding questions for future research.

SMALL REGULATORY RNA BACKGROUND

Small regulatory RNAs (sRNAs) are numerous in plants. They usually range in size from 21 to 24 nucleotides and serve as key regulators of gene expression. sRNAs are involved in myriad processes, including development, cell type designation, responses to abiotic stress, and silencing of repetitive elements. sRNAs are processed from longer precursor RNAs (either the helical stem regions of self-complementary single-stranded RNAs or double-stranded RNAs [dsRNAs]) by endonucleases in the Dicer-like (DCL) protein family. DCL endonucleases produce an initial short duplex RNA. One of the two short RNA strands forms a complex with a protein in the Argonaute (AGO) family. The AGO-sRNA complex then identifies target RNAs based on complementarity between target and sRNA. sRNAs can be categorized based on differences in their biogenesis and differences in their modes of targeting (Fig. 1).

Figure 1.

Schematic overview of endogenous small RNA biogenesis and molecular functions in plants. The rounded rectangle with the mRNA represents the open reading frame. Figures are not drawn to scale; miRNA and siRNA duplexes have two-nucleotide 3′ overhangs. DRM, Domains rearranged methyltransferase. For brevity, many details are omitted; see Rogers and Chen (2013), Matzke and Mosher (2014), and Borges and Martienssen (2015) for detailed discussions of plant small RNA biogenesis.

MicroRNAs (miRNAs) in plants are processed from RNA polymerase II-transcribed primary RNAs. A region of the primary transcript forms an imperfect hairpin structure that is recognized by the DCL1 endonuclease. DCL1, along with several accessory proteins, liberates a miRNA/miRNA* duplex. The duplex is disassembled, with the mature miRNA becoming bound to an Argonaute (AGO) protein, most frequently AGO1. Once the mature miRNA is bound to an AGO protein, the miRNA* is typically separated from the complex and degraded (for a more detailed review of plant miRNA biogenesis, see Rogers and Chen, 2013). The resulting miRNA/AGO complex directs post-transcriptional regulation of mRNAs and long noncoding RNAs. Target selection is primarily based on complementarity between the miRNA and target RNA (Mallory et al., 2004; Liu et al., 2014). Plant AGO proteins are endonucleases that cut target RNAs. (Tang et al., 2003; Baumberger and Baulcombe, 2005; Qi et al., 2005). This target slicing destabilizes the RNA. The association of AGO/miRNA complexes with mRNAs can also cause translational repression and in certain cases trigger the biogenesis of secondary short interfering RNAs (siRNAs; Fig. 1). Most plant miRNAs are 21 nucleotides long. miRNAs of 22 nucleotides also sometimes occur, but sizes other than 21 or 22 nucleotides are much less common.

Besides miRNAs, many other sRNAs are produced and used by the plant DCL/AGO system. These are collectively termed siRNAs. Plant siRNAs are typically generated from dsRNA and can be processed by multiple DCLs. They are distinguished from miRNAs by the absence of a precisely processed stem-loop precursor. Also, unlike miRNAs, which target RNAs distinct from their own precursors, plant siRNAs typically target transcripts from the same loci where they originate. Two major classes of plant siRNAs have been widely recognized: secondary siRNAs and RNA polymerase IV-dependent siRNAs (p4-siRNAs; Fig. 1).

Secondary siRNA biogenesis depends on an initial AGO-miRNA or AGO-siRNA interaction with a target RNA. This interaction stimulates the activity of a specific RNA-dependent RNA polymerase (RDR6) on the target, creating dsRNA. Typically, the dsRNA is processed into siRNA duplexes by both DCL4 (which makes 21-nucleotide-long duplexes) and DCL2 (which makes 22-nucleotide-long duplexes). The resulting population of 21- to 22-nucleotide secondary siRNAs can be bound to AGO proteins and target additional copies of the original transcript as well as other transcripts with sufficient complementarity. The result is a positive feedback loop where an initial miRNA or siRNA trigger can amplify its effects and cause potent gene silencing. Not all AGO/miRNA or AGO/siRNA targets spawn secondary siRNAs. For reasons that remain murky, targeting by a 22-nucleotide miRNA or siRNA (Chen et al., 2010; Cuperus et al., 2010) and/or multiple target sites on the same RNA (Axtell et al., 2006) promote secondary siRNA biogenesis; in contrast, targeting at single sites by typical 21-nucleotide miRNAs does not promote secondary siRNA accumulation. Secondary siRNA biogenesis does not strictly require AGO-catalyzed slicing of the precursor (Axtell et al., 2006; Arribas-Hernández et al., 2016), but slicing often occurs. When it does occur at a single predominant site, the resulting dsRNA production will all begin at the same position. Because the relevant DCLs liberate secondary siRNA duplexes sequentially from the dsRNA terminus, the resulting siRNA population has 5′ and 3′ ends at regularly defined 21- to 22-nucleotide intervals. This property is known as phasing, and the resulting secondary siRNAs are thus known as phased siRNAs (Fei et al., 2013). In some cases, secondary siRNAs can also target mRNAs that are distinct from their precursor RNA; these siRNAs have been called trans-acting siRNAs (tasiRNAs; Vazquez et al., 2004). We regard tasiRNAs and phased siRNAs as subsets within the more general class of secondary siRNAs.

The third major type of plant sRNAs are the p4-siRNAs. Both the biogenesis and function of p4-siRNAs are distinct from miRNAs and secondary siRNAs. RDR2 is attached to RNA polymerase IV and generates a short (∼40 nucleotides) dsRNA. This dsRNA is in turn processed into a 24-nucleotide-long siRNA duplex by DCL3. A mature 24-nucleotide p4-siRNA is then loaded onto a specialized AGO in the AGO4 clade. A major distinctive feature of p4-siRNAs is that they are not known to target mRNAs or to function outside of the nucleus. Instead, they function in the nucleus to target non-protein-coding nascent RNAs. Successful p4-siRNA/AGO targeting is thought to lead to the recruitment of de novo DNA methyltransferases to the local chromatin, causing de novo DNA methylation. The outcome of p4-siRNA function is therefore DNA modification of sequence-similar loci. Secondary siRNAs can sometimes be captured by the p4-siRNA-specific AGOs and cause DNA methylation of homologous loci (Wu et al., 2012; McCue et al., 2015). Thus, all three major types of plant sRNAs are connected to each other: miRNAs can stimulate secondary siRNAs, and secondary siRNAs can act like p4-siRNAs. Although p4-siRNAs are very abundant in plants, they are not currently known to be involved in trans-species small RNA interactions and thus will not be discussed further here. Interested readers are directed to the excellent reviews by Borges and Martienssen (2015) and Matzke and Mosher (2014) for details on p4-siRNAs.

PATHOGEN GENE SILENCING INDUCED BY HOST sRNAS

Since the discovery of RNA interference (RNAi) about two decades ago, evidence has accumulated to indicate a role for sRNAs in plant defense. Small RNA-mediated immunity is best understood in antiviral defense (Baulcombe, 2004; Ding, 2010). Plants infected with viruses acquire immunity by producing DCL-dependent and virus-derived siRNAs, which guide AGO proteins to viral RNAs and thus help to arrest the infection (Guo et al., 2019). However, it was not until recently that a role of sRNAs was established in plant defense during infections by cellular pathogens, especially eukaryotic pathogens including fungi and oomycetes.

An early example of pathogen gene silencing induced by host sRNAs was from the observation that native miRNAs produced from human erythrocytes translocate into the malaria-causing parasite Plasmodium falciparum and inhibit pathogen gene expression (LaMonte et al., 2012). Whether endogenous sRNAs in plants could mediatetrans-species RNAi remained unknown until the report from Zhang et al. (2016b), which described two cotton (Gossypium hirsutum) miRNAs, miR159 and miR166, that conferred resistance to the fungal pathogen Verticillium dahliae (Fig. 2). miR159 and miR166 are induced upon the fungal infection. Importantly, they were detected in fungal hyphae isolated from the infected cotton tissues and predicted to target specific transcripts encoding virulence-related proteins in the fungus. As a result, these miRNAs promoted resistance to V. dahliae. This defense mechanism seems to be conserved in Arabidopsis (Arabidopsis thaliana), in which miR159 and miR166 were also induced by V. dahliae infection (Zhang et al., 2016b). Furthermore, knockdown mutants of miR166 showed compromised resistance to the fungal pathogen, indicating that miR166 contributed to plant defense, possibly through trans-species gene silencing.

Figure 2.

Examples of host plant-to-pathogen (top) and pathogen/parasite-to-host plant (bottom) movement and function of trans-species small RNAs. The rounded rectangles of mRNAs represent open reading frames. Figures are not drawn to scale; miRNA and siRNA duplexes have two-nucleotide 3′ overhangs. All experimentally confirmed target RNAs or target RNA families are shown and are listed as they appear from top to bottom and from left to right in each part of the figure. MYB, Myelobastosis (Arabidopsis loci: AT5G06100 and AT3G11440); HD-ZIP III, Homeodomain Leu-zipper class III (Arabidopsis loci: AT2G34710, AT1G30490, AT5G60690, AT4G32880, and AT1G52150); TAS, Trans-acting siRNA (Arabidopsis loci: AT2G27400, AT1G50055, AT2G39675, and AT2G39681); PPR, Pentatricopeptide repeat (Arabidopsis loci: AT1G62590, AT1G62910, AT1G62914, AT1G62930, AT1G63080, AT1G63130, AT1G63150, and AT1G63400); PRXIIF, Peroxiredoxin2-F (Arabidopsis locus: AT3G06050); MPK, Mitogen-activated kinase (Arabidopsis loci: AT1G10210 and AT1G59580); MPKKK, Mitogen-activated kinase kinase kinase; WAK, Wall-associated kinase (Arabidopsis locus: AT5G50290); BIK1, Botrytis-induced kinase1 (Arabidopsis locus: AT2G39660); TIR1, Transport inhibitor related1 (Arabidopsis locus: AT3G62980); AFB, Auxin-related F-box (Arabidopsis loci: AT3G26810 and AT1G12810); SEOR1, Sieve element occlusion related1 (Arabidopsis locus: AT3G01680); SCZ, Schizorhiza (Arabidopsis locus: AT1G46264); HiC-51, Isotrichodermin C-15 hydroxylase (V. dahliae locus: VDAG_09950); Clp-1, Calpain1 (V. dahliae locus: VDAG_09736); Bc-VPS51, B. cinerea vacuolar protein sorting51 (B. cinerea locus: BC1G_10728); Bc-DCTN, B. cinerea dynactin complex large subunit (B. cinerea locus: BC1G_10508); Bc-SAC1, B. cinerea suppressor of actin-like (B. cinerea locus: BC1G_08464); LTR, Long-terminal repeat retrotransposon.

Following the discovery that plant miRNAs could enter invading V. dahliae cells and induce gene silencing, studies in Arabidopsis showed that siRNAs were also used to silence target genes in the necrotrophic fungal pathogen Botrytis cinerea (Cai et al., 2018) and the hemibiotrophic oomycete pathogen Phytophthora capsici (Hou et al., 2019). Two siRNAs derived from the noncoding, secondary siRNA-spawning loci TAS1 and TAS2 were found to target genes involved in vesicle trafficking in B. cinerea; as a result, overexpression of these siRNAs in Arabidopsis led to reduced virulence of the fungal pathogen (Cai et al., 2018). During the infection of P. capsici, a pool of siRNAs generated from a few transcripts of pentatricopeptide repeat (PPR) encoding genes was induced as a defense response (Hou et al., 2019). These PPR-derived siRNAs represent a large diversity of sequences and presumably silence pathogen genes using a shotgun approach. Multiple PPR-derived siRNAs were predicted to target several genes in P. capsici, some of which contribute to pathogen development and colonization. The PPR-derived siRNAs also have predicted targets in V. dahliae, indicating that they may contribute to broad-spectrum resistance to diverse pathogens. Both the TAS-derived and PPR-derived siRNAs are secondary siRNAs, whose biosynthesis depends on the activity of RDR6. rdr6 mutants of Arabidopsis exhibited hypersusceptibility to B. cinerea, P. capsici, and V. dahliae (Ellendorff et al., 2009; Cai et al., 2018; Hou et al., 2019). The contribution of RDR6 to plant immunity is likely attributed to its essential role in the production of secondary siRNAs that silence pathogen mRNAs.

External application of both dsRNAs and sRNAs has been reported to induce gene silencing in pathogens (Koch et al., 2016; Wang et al., 2016; Rosa et al., 2018), raising the question whether dsRNAs instead of mature siRNAs are being transferred from plants to pathogens. In Arabidopsis, PPR-derived siRNAs were produced by DCL4 and its cofactor Double-Stranded-RNA-Binding Protein4(DRB4). In addition to the rdr6 mutant, the drb4 mutant of Arabidopsis also showed hypersusceptibility to P. capsici (Hou et al., 2019) and V. dahliae (Ellendorff et al., 2009). Because the drb4 mutant still accumulates the precursor dsRNA, these findings indicate that mature siRNAs or siRNA duplexes, rather than dsRNAs, are likely the antimicrobial agents. Taken together, the current data suggest that trans-species RNAi between plant hosts and fungal/oomycete pathogens is an integral component of plant immunity. In particular, since the endogenous functions of many plant secondary siRNAs are unclear, their main activity may be trans-species gene silencing.

APPLICATION OF HOST-INDUCED GENE SILENCING IN PLANT DISEASE CONTROL

Although the natural role of endogenous sRNAs in trans-species gene silencing is a recent discovery in plants, substantial efforts have been invested to engineer plant sRNAs designed to inhibit the infection by cellular pathogens and parasites. This strategy, termed host-induced gene silencing (HIGS), involves transgenic plants developed to produce dsRNA precursors with homology to target mRNAs in invading pests. It is presumed that the dsRNA precursors produce artificial sRNAs, which are then taken up by the invaders and induce gene silencing using the pathogen’s RNA silencing machinery. Some of these plants were indeed found to exhibit enhanced resistance to the targeted organism.

The first successful example of HIGS for nonviral pests was reported in Arabidopsis, where the expression of specific hairpin RNAs induced the silencing of a gene encoding a secretory peptide in root-knot nematodes (Meloidogyne spp.) and resulted in disease resistance (Huang et al., 2006). dsRNAs designed to target a cytochrome P450 gene in cotton bollworm (Helicoverpa armigera) and a vacuolar ATPase gene in coleopteran insect pests were also found to significantly reduce the infestation of these insects in cotton (Baum et al., 2007) and maize (Zea mays; Mao et al., 2007). Following these early studies, HIGS has been demonstrated to be a promising tool in controlling parasitic weeds. Different from nematodes and insects, parasitic plants exploit host plants by connecting to the host’s vasculature and taking up water and nutrients. Through the connected vasculature, hosts and pathogens exchange RNA molecules including mRNAs and sRNAs (Kim and Westwood, 2015). Transgenic lettuce (Lactuca sativa) expressing hairpin RNAs that target the GUS reporter gene was found to be able to silence GUS expression in the parasitic plant Triphysaria versicolor (Tomilov et al., 2008). A similar approach was successfully used to induce gene silencing in Orobanche and Cuscuta species, thus limiting parasitic growth (Aly et al., 2009; Alakonya et al., 2012).

HIGS has also been successfully employed to confer resistance to specific fungi and oomycete pathogens. Barley (Hordeum vulgare) and wheat (Triticum spp.) expressing artificial siRNAs were found to be more resistant to the biotrophic fungal pathogen Blumeria graminis (Nowara et al., 2010) and the hemibiotrophic fungal pathogen Fusarium graminearum (Koch et al., 2013). A similar approach was used in tomato (Solanum lycopersicum) and cotton to enhance resistance to wilting disease caused by V. dahliae (Zhang et al., 2016a; Song and Thomma, 2018). Furthermore, successful silencing of targeted genes, and hence increased resistance, was reported in the hemibiotrophic oomycete pathogen Phytophthora infestans using potato (Solanum tuberosum) plants expressing siRNA-producing hairpin RNAs (Jahan et al., 2015). These numerous examples of HIGS support engineering sRNAs of plants as a practical tool in agricultural biotechnology. The success of HIGS in such diverse systems also indicates that sRNAs may frequently move from host to pathogen/parasite. Thus, we expect the number of known natural examples of host-to-pathogen/parasite sRNA movement to continue to grow.

Nearly all of the reports of successful HIGS have used a strategy of producing siRNAs from long hairpin RNA precursors. Long hairpin-induced RNAi primarily engages the DCL4/DCL2-dependent secondary siRNA pathway, and to a lesser extent the DCL3-dependent p4-siRNA pathway, to produce a potent mixture of 21-, 22-, and 24-nucleotide-long siRNAs (Fusaro et al., 2006). However, empirical evidence showing that the plant-produced 21-, 22-, and 24-nucleotide siRNAs are directly responsible for HIGS is lacking. Further experiments using dcl and drb mutants are required to demonstrate whether intact dsRNAs can enter the pathogen/parasite and be subsequently processed by the endogenous machinery of the recipient organism for gene silencing.

PATHOGEN/PARASITE-TO-HOST sRNA MOVEMENT

Small RNA movement has also been shown to occur from pathogens and parasites (Fig. 2), which employ sRNAs as one of their numerous molecular strategies to subvert host defenses and remodel host physiology to their advantage. In at least two known cases, these strategies include the delivery of sRNAs into host tissues to silence host mRNAs. The pathogenic fungus B. cinerea accumulates several sRNAs during plant infection (Weiberg et al., 2013). Some of these sRNAs have complementarity to host immune signaling mRNAs that are down-regulated during B. cinerea infection. The incoming B. cinerea sRNAs interact with the host’s AGO1 protein, and Arabidopsis ago1 hypomorphic mutants are resistant to B. cinerea infection (Weiberg et al., 2013). A dcl1/dcl2 double mutant of B. cinerea lost accumulation of the relevant trans-species sRNAs and became avirulent (note that B. cinerea DCL1 and DCL2 are not functionally homologous to plant DCL1 and DCL2). Altogether, these data show that B. cinerea hijacks host sRNA machinery with its own sRNAs. An analysis of mRNA 5′ ends has also indicated that the oomycete Plasmopara viticola has a great number of sRNAs that may target host (grapevine, Vitis spp.) mRNAs for cleavage during infection (Brilli et al., 2018).

Another example of parasite-to-host movement of sRNAs comes from the parasitic plant Cuscuta campestris. Cuscuta spp. is a widespread genus of obligate parasitic plants that attach to host plant stems using a specialized organ called haustorium. Cuscuta haustoriaare permissive and known to permit bidirectional movement of mRNAs, proteins, and secondary metabolites (Kim and Westwood, 2015). Cuscuta campestris produces many miRNAs that specifically accumulate in haustorial tissues (Shahid et al., 2018). Some of these miRNAs specifically target host mRNAs involved in pathogen defense, vascular system function, and hormone signaling. Examples include Sieve Element Occlusion Related1 (SEOR1), which functions to reduce sap loss from wounded phloem (Ernst et al., 2012), and Auxin F-Box Related3 (AFB3), which couples auxin sensing to transcriptional responses (Parry and Estelle, 2006). C. campestris growth is increased on Arabidopsis seor1 or afb3 mutants (Shahid et al., 2018). These data suggest that trans-species miRNAs from C. campestris function to silence host genes in order to increase parasite growth and fitness. Interestingly, many of the C. campestris trans-species miRNAs are 22 nucleotides long and induce secondary siRNA accumulation from their targets. Secondary siRNA accumulation requires host DCL4 and RDR6, which provides evidence that the C. campestris miRNAs are acting inside host cells. An attractive hypothesis is that secondary siRNA induction serves to strengthen the silencing caused by the initial miRNA from the parasitic plant. However, reduction or removal of the secondary siRNAs (by mutation in host dcl4 or rdr6) fails to noticeably affect C. campestris growth (Shahid et al., 2018). Thus, the importance of the secondary siRNAs in this interaction remains unknown.

MECHANISM(S) OF sRNA MOVEMENT BETWEEN PATHOGENS AND PLANT CELLS

Small RNAs are known to be highly mobile molecules that traffic intercellularly and systemically in plants (Liu and Chen, 2018). In plants, intercellular movement involves sRNA transfer through plasmodesmata, and systemic movement is believed to occur in the phloem. The phenomenon of HIGS as well as the natural examples of trans-species sRNA movement raise the question: How are sRNAs transferred between different cellular organisms? Recent studies suggest a potential role of extracellular vesicles (EVs) in the translocation of sRNAs from Arabidopsis to P. capsici (Hou et al., 2019) and B. cinerea (Cai et al., 2018).

EVs are membrane-bound particles containing transmembrane proteins and soluble cargoes. They are released by donor cells into their surrounding environment. EVs have been extensively characterized in animals and humans, where they are classified based on their size and origins: exosomes, which are 30 to 100 nm in diameter; shedding microvesicles, which are 100 nm to 1 μm in diameter; and apoptotic bodies, which are 50 nm to 5 μm in diameter (Théry et al., 2009; Ressel et al., 2019). Microvesicles are shed from the plasma membrane; apoptotic bodies contain parts of dying cells formed during programmed cell death; and exosomes are formed in the cytosol by inward budding of the limiting membrane of endocytic compartments, leading to vesicle-containing endosomes called multivesicular bodies (MVBs). The vesicles are released into the environment when MVBs fuse with the plasma membrane (Raposo and Stoorvogel, 2013; Ressel et al., 2019). EVs have been shown to be key players in intercellular communication in animal systems. In particular, EVs of animal cells participate in the transport of proteins, lipids, mRNAs, miRNAs, and other noncoding RNAs (Valadi et al., 2007; Yáñez-Mó et al., 2015). These discoveries led to the interesting possibility that EVs may also shuttle sRNAs between plant hosts and their invading pathogens or parasites.

Plant EVs were first successfully isolated by Rutter and Innes (2017) from apoplastic wash fluid recovered from Arabidopsis leaves. Proteomic analysis of the purified EVs revealed enrichment of defense-related proteins. Examples of these proteins include the syntaxin AtSYP121/PENETRATION1 (PEN1), which is related to papillae formation during fungal infection (Assaad et al., 2004), the ATP-binding cassette transporter PEN3, which accumulates around haustoria of powdery mildew (Underwood and Somerville, 2013), and RPM1-INTERACTING PROTEIN4 (RIN4), which participates in immune responses triggered by bacterial infection (Mackey et al., 2003). These results suggest that plant EVs are associated with biotic stress responses.

In addition to defense-related proteins, EVs in Arabidopsis also carry sRNAs (Cai et al., 2018; Baldrich et al., 2019; Hou et al., 2019). A comparison of sRNA profiles in apoplast and EVs revealed that sRNAs are differentially secreted and enriched in EVs (Baldrich et al., 2019). Interestingly, the majority of sRNA cargos in EVs are only 10 to 17 nucleotides in length (Baldrich et al., 2019). These so-called tiny RNAs (tyRNAs) appear to be processing by-products of miRNA precursors and mRNAs. Whether they regulate gene expression is currently unknown. sRNA profiling of AGO-associated sRNAs has not found any evidence that AGO proteins bind 10- to 17-nucleotide RNAs in vivo (Mi et al., 2008; Wang et al., 2011), so it seems likely that tyRNAs function differently than miRNAs and siRNAs. Although miRNAs and siRNAs were also detected in EVs, their abundance was relatively low compared with tyRNAs (Baldrich et al., 2019). Furthermore, specific miRNAs and siRNAs were found to be enriched in EVs, indicating a possible sorting mechanism.

Two PPR-derived siRNAs that silence target genes in P. capsici were found to be EV cargos in Arabidopsis (Hou et al., 2019). Their abundance increased in EVs isolated from P. capsici-infected leaves, consistent with the observation that PPR-derived siRNAs were induced during infection. The tasiRNAs that target mRNAs in the fungal pathogen B. cinerea were also present in EVs (Cai et al., 2018). However, secondary siRNAs, including multiple PPR-derived siRNAs and tasiRNAs, were highly enriched in apoplast but not in EVs (Baldrich et al., 2019), indicating that these siRNAs were actively secreted by plant cells but possibly through an EV-independent pathway. On the other hand, EV cargos may change in infected tissues and EV-dependent sRNA secretion could still be involved in defense responses.

Following the release from plant cells, sRNAs need to enter pathogen cells, or vice versa in cases where pathogen sRNAs enter plant hosts. Biotrophic/hemibiotrophic filamentous pathogens and parasitic plants form specialized infection structures called haustoria (Catanzariti et al., 2006; Yoshida et al., 2016), which extend into host cells and may act as the gateway for sRNA delivery. Fungal/oomycete haustoria are enveloped by the extrahaustorial membrane (EHM), a modified plant plasma membrane, and are separated from plant cells by the extrahaustorial matrix (EHMx), in which active material exchange is believed to occur (Koh et al., 2005). For example, haustoria provide portals for nutrient uptake from the host to the pathogen and the delivery of virulence effectors from the pathogen to the host (Catanzariti et al., 2006). On the other hand, antimicrobial agents produced by the host are also targeted to accumulate in the EHMx in order to arrest infection. Systematic studies of fluorescence-tagged cell components during oomycete infection, including Phytophthora sojae, Peronospora parasitica, and Phytophthora parasitica, showed the accumulation of Golgi stacks, endoplasmic reticulum, secretory vesicles, and MVBs in the infected plant cells adjacent to the EHM (Takemoto et al., 2003; Lu et al., 2012). Similarly, the accumulation of secretory vesicles and MVBs in the proximity of the EHM was also observed in Arabidopsis during the infection of Golovinomyces orontii (Micali et al., 2011). These findings indicate that plant sRNAs, together with other antimicrobial proteins and metabolites, could be targeted to the EHM and released into the EHMx. It is noteworthy that fungal/oomycete haustoria are encased by the pathogen cell wall; it remains to be determined how the pathogen cell wall may function as a barrier for the endocytosis of host sRNAs, especially those shuffled as cargos in EVs.

Although they share the same name, the haustoria of parasitic plants are very distinct from those of pathogenic fungi and oomycetes. One fundamental difference is that parasitic plant haustoria function to bridge the vascular systems of the host and parasite, thus enabling large-scale flow of water, minerals, carbohydrates, and other compounds into the parasitic plant (Yoshida et al., 2016). Another fundamental difference is that parasitic plant haustoria are multicellular, macroscopic organs that contain several different specialized cell types. The trans-species miRNAs that are delivered from C. campestris to hosts thus move in the opposite direction relative to the major flow of material across the haustorium; therefore, their delivery is not by simple bulk flow. The mature Cuscuta spp. haustorium contains several distinct tissues in close contact with host cells (Dawson et al., 1994). The main endophyte body is a large papillar organ that penetrates the host epidermis and is lodged within host cortical tissue. Long filamentous searching hyphae cells are projected from this main body, which thread through the middle lamellae of host tissues and can penetrate into host cells. Searching hyphae that contact host xylem differentiate into xylem, while searching hyphae that contact host phloem branch form finger-like projections that surround the host phloem (Dawson et al., 1994). Each of these distinct Cuscuta spp. tissues is in direct contact with host cells; which ones actually deliver trans-species miRNAs is not clear. The tips of searching hyphae that penetrate host cells have numerous plasmodesmatal connections between parasite and host (Dawson et al., 1994); thus, one possibility is that miRNA delivery in this system occurs through plasmodesmata. Another possibility is delivery by EVs. Further studies are required to understand the cell and molecular biology of trans-species sRNA movement in fungal, oomycete, and parasitic plant systems.

EVOLUTION OF TRANS-SPECIES sRNAS

Few studies have explicitly studied the evolution of trans-species small RNAs. Endogenous host secondary siRNAs induced upon infection of P. capsici have the capability to alter pathogen gene expression mediated by the induction of miR161 triggering siRNA accumulation from PPR transcripts (Hou et al., 2019). PPR proteins are primarily known as sequence-specific RNA-binding proteins that are critical for mRNA maturation in plastids (Miranda et al., 2018). However, some PPR proteins also play roles in abiotic and biotic defense (Kobayashi et al., 2007; Tang et al., 2010). The pool of secondary siRNAs made from PPR transcripts is extremely diverse, which might increase the chances of successfully targeting of pathogen mRNAs (Hou et al., 2019). The PPR gene family is large, but only members of a small clade spawn secondary siRNAs. This clade evolves at a faster rate compared with other clades of PPR genes (Dahan and Mireau, 2013), further suggesting that PPR-derived secondary siRNA accumulation could be under selective pressure to maximize sequence diversity rather than to maintain pairing between any particular siRNA and pathogen mRNA. Indeed, PPR-derived secondary siRNAs have complementarity to mRNAs from diverse pathogens. The induction of miR161 requires an intact pattern recognition receptor complex on the plant cell surface (Hou et al., 2019). Furthermore, the miR161 level was increased in Arabidopsis treated with the bacterial elicitor flg22 (Li et al., 2010). These findings suggest that miR161 induction, and hence elevated accumulation of PPR-derived siRNAs, is a basal immune response that confers broad-spectrum resistance (Hou et al., 2019). PPR-based secondary siRNA production is commonly seen from diverse plants (Xia et al., 2013), which suggests that this could be an ancient evolutionary response that continues to confer effective resistance. Interestingly, although the existence of PPR-derived secondary siRNAs is quite conserved, the trigger miRNA, miR161, is restricted to the Brassicaceae (Chávez Montes et al., 2014; Kozomara and Griffiths-Jones, 2014). The conservation of PPR-derived secondary siRNAs implies that some other miRNA(s) serves as the trigger in non-Brassicaceae plants. One possibility is that the descendants of a single ancestral miR161 have diversified substantially in sequence in different plant lineages, confounding the normal rules for miRNA family assignment but retaining the ability to target PPR mRNAs. The identification of expressed small RNAs from some basal lineages of plants (ferns and lycophytes) that have some sequence homology to miR161 (You et al., 2017) is consistent with this hypothesis.

Secondary siRNAs from Arabidopsis TAS1 and TAS2 noncoding RNAs can target mRNAs from B. cinerea (Cai et al., 2018). The trigger for TAS1/TAS2 secondary siRNA accumulation is miR173. Some TAS1/TAS2-derived secondary siRNAs are also able to target PPR transcripts (Allen et al., 2005; Howell et al., 2007); thus, the two pools of secondary siRNAs are linked (Fig. 2). Similar to PPR-derived secondary siRNAs, there is no evidence that the sequence of any particular TAS1/TAS2-derived secondary siRNA is under strong purifying selection. Both miR173 and the TAS1 and TAS2 loci are only found in Arabidopsis. However, miR173 is a member of a larger superfamily of miRNAs, members of which are found in many other plants, and most of which are triggers for PPR-derived secondary siRNAs (Xia et al., 2013). This is further evidence that the Arabidopsis-specific miR173/TAS1/TAS2 cascade is really a component or offshoot of the more widespread PPR secondary siRNA system. It is also important to consider that because secondary siRNA accumulation seems to be a general defense response, there may be subsets of secondary siRNAs that have been specifically selected to target mRNAs from certain individual pathogens. This would result in host production of a massive number of secondary siRNAs with only a few being actually effective for any particular infection. An interesting hypothesis on how a host responds to pathogen infection is that the accumulation of these antimicrobial sRNAs may not be under any strong selective pressure but that the host responds with a shotgun approach and by chance is able to target many different kinds of transcripts. Alternatively, this may also suggest that because pathogenic transcripts are under such high selective pressure to alter their target sites, the hosts have evolved a way to produce a wide variety of sequence variations in an attempt to always remain one step ahead of their pathogens.

A counterpoint to the shotgun approach of using secondary siRNAs is the use of miRNAs in trans-species small RNA silencing. MicroRNAs by definition are loci where just one or two functional small RNAs are made from a single precursor (Axtell and Meyers, 2018). This contrasts with the large number of siRNAs spawned from a single secondary siRNA precursor. The case described by Zhang et al. (2016b), where cotton miR159 and miR166 are active against V. dahliae mRNAs, is especially curious from an evolutionary point of view. Both miR159 and miR166 are ancient miRNAs with complete sequence conservation between ferns, gymnosperms, and angiosperms (Cuperus et al., 2011; Chávez Montes et al., 2014). miR159 targets endogenous Myeloblastosis transcription factor mRNAs, while miR166 targets endogenous Homeodomain leucine-zipper class III mRNAs; these target relationships are also universally conserved among land plants (Floyd and Bowman, 2004; Jones-Rhoades and Bartel, 2004). These conserved regulatory interactions are critical for plant development (Mallory et al., 2004; Allen et al., 2007). It seems clear that these two miRNAs have been selected for their roles in endogenous gene regulation. Moreover, any pathogen that has attacked any land plant in the last ∼250 million years would have been exposed to both miR159 and miR166 as well as a handful of other ultraconserved plant miRNAs. It is possible that miR159 and miR166 function in both endogenous gene regulation and pathogen defense. However, the possibility that miR159 and miR166 targeting of V. dahliae mRNAs is accidental should not be dismissed.

The parasitic plant C. campestris also uses miRNAs as trans-species effectors. Unlike miR159 and miR166, the miRNAs that C. campestris uses as effectors are not present in other plants (Shahid et al., 2018); instead, they seem to be specialized for use in trans-species silencing. But how could single parasite-derived miRNAs persist over evolutionary time? Presumably, the targeting event is not beneficial for the host, so one might expect that host target sites would have been selected over time to diminish complementarity to the incoming parasite miRNAs. Another possibility is that the trans-species C. campestris miRNAs target regions of host mRNAs that code for highly conserved or critical amino acids. The fact that some C. campestris miRNAs silence homologous target mRNAs in both Arabidopsis and Nicotiana benthamiana (Shahid et al., 2018) supports the hypothesis of conserved target sites. If that were the case, non-synonymous changes to the target site might affect host fitness due to disruption of the relevant proteins. Thus, purifying selection would act to maintain the non-synonymous nucleotides within target sites, and sequence changes in target sites might be constrained to only synonymous site variation.

Alternatively, it may also be possible that C. campestris does not produce these trans-species miRNAs in order to affect host physiology but instead to protect itself from the translation of certain incoming host mRNAs. Many host-derived mRNAs are present within Cuscuta spp., and there does not seem to be selectivity with respect to which specific mRNAs are imported (Kim et al., 2014). Whether these incoming host mRNAs can be translated once they enter Cuscuta spp. is not known. If translation of host mRNAs did occur, the appearance of foreign proteins could be detrimental. In that case, it would be advantageous for the parasite to have mechanisms in place to rapidly degrade particularly dangerous host mRNAs. In this scenario, the trans-species miRNAs would defend the parasite rather than attack the host. Testing this alternative hypothesis will be an interesting goal for the future.

THE PREVALENCE OF TRANS-SPECIES RNAi IN PLANTS

Current research, although mainly centered on the model plant Arabidopsis and several examples of eukaryotic invaders (including fungi, oomycetes, and parasitic weeds), has established a role of trans-species RNAi during the host-pathogen arms race. However, whether trans-species RNAi also contributes to plant defense against prokaryotic pathogens remains unclear. The intriguing observation that dsRNAs expressed in Escherichia coli could induce gene silencing in nematode larvae that feed on them raised the possibility that dsRNAs may be transferred from prokaryotic pathogens to their hosts (Timmons and Fire, 1998). Plant mutants deficient in sRNA biogenesis, including dcls and hen1, were also hypersusceptible to infection by the bacterial pathogen Pseudomonas syringae (Navarro et al., 2008). Interestingly, three effector proteins produced by P. syringae were shown to possess RNAi suppression activity in Arabidopsis (Navarro et al., 2008). Effectors are essential virulence proteins produced by microbial pathogens to defeat host immunity (Dodds and Rathjen, 2010; Dou and Zhou, 2012). Effectors with RNAi suppression activities have been reported in numerous viruses (Burgyán and Havelda, 2011), several species of Phytophthora (Qiao et al., 2013; Xiong et al., 2014; Zhang et al., 2019), and the fungus Puccinia graminis (Yin et al., 2019). The identification of these viral, Phytophthora, or fungal suppressors of RNA silencing supports plant sRNAs as antimicrobial agents in the corresponding pathosystems. The fact that P. syringae produces effectors that can target the sRNA pathway in Arabidopsis indicates that sRNAs may also contribute to plant defense against P. syringae. It is important to note that although bacteria do not have eukaryote-like RNA-silencing machineries, some of them encode AGO or AGO-like enzymes (Swarts et al., 2014) that may guide gene silencing triggered by plant sRNAs, should they be transferred into bacterial cells during infection. Indeed, bacterial gene silencing triggered by animal sRNAs has been reported in gut microbiota. Extracellular miRNAs of mouse and human feces were found to be functional to regulate bacterial gene expression in the gut (Liu et al., 2016). In addition, miRNAs enclosed in plant-derived exosome-like nanoparticles could enter specific bacterial species in the gut microbiota for gene silencing, thus altering the composition of the microbial community (Teng et al., 2018). It would be interesting to determine whether plant bacterial pathogens have AGO-like proteins and how plant sRNAs may affect bacterial gene expression during infection.

Although accumulating evidence supports a role of trans-species RNAi in host-pathogen interactions, it may not be involved in all pathosystems. For example, dcl and ago mutants of the fungal pathogen Zymoseptoria tritici were still able to cause disease on wheat with uncompromised virulence activities (Kettles et al., 2019). Moreover, dsRNAs externally applied in vitro or generated from transgenic wheat plants did not induce target gene silencing in the fungal pathogen (Kettles et al., 2019). The lack of trans-species RNAi between some fungal pathogens and their host is not surprising because many fungal species have lost the canonical RNA-silencing machinery. In a study that analyzed 54 fungal genomes (Nunes et al., 2011), 13 were found to have lost all the RNA-silencing genes such as the DCL, AGO, and RDR families. Among them is the fungal pathogen Ustilago maydis that infects maize. It is intriguing to hypothesize that the lack of an RNAi mechanism could be a consequence of a co-evolutionary arms race with the hosts. In any case, these findings indicate that trans-species RNAi or HIGS may not be effective against some specific plant pathogens.

CONCLUDING REMARKS

Multiple plant pathosystems have been shown to involve the exchange of small RNAs between cellular organisms. Exploiting this small RNA transfer may prove to be a useful strategy to engineer crops with improved disease resistance. This is a new and fast-moving area of research, and many questions await answers (see Outstanding Questions). We are excited to see the progress in this area in the coming years.

There are several critical limitations in our current understanding of these processes. One is that most of the existing studies demonstrating trans-species exchange of sRNAs between plants and their pests have used Arabidopsis. Going forward, it will be important to assess whether the initial examples from this model system are applicable to a wider array of plants, especially crops. Second, only a limited number of plant pests have to date been scrutinized for trans-species small RNA activity during infections/infestations. It will be important to keep an open mind for paradigms of trans-species small RNA functions that have not yet been described in the few studies to date.