The convergence of autophagy, small RNA and the stress response – implications for transgenerational epigenetic inheritance in plants

Introduction to autophagy

Autophagy is process that involves proteins and other cytoplasmic components in the cell being delivered to a lysosome (in animals) or vacuole (in yeast and plants) for degradation. It is a survival mechanism that plays a crucial role in nutrient recycling, development, cell homeostasis, defence against pathogens and toxins (1). Autophagy consists of three stages: the phagophore, autophagosome, and autolysosome (animal) or autophagic body (plant) (Figure 1) (2). Each stage is regulated by autophagy-related (ATG) proteins. Much is known about the mechanism of autophagy in animal systems, but less is known about it in plants.

Figure 1

Model for P0-mediated autophagy for AGO1 degradation.

The general miRNA pathway is a miRNA produced from miRNA precursor (pre-miRNA) to produce a miRNA duplex, and then assembled into an RNA-induced silencing complex (RISC) with AGO1 and other components. The miRNA* is erased during the process. The remaining miRNA forms a miRISC with AGO1 to determine the target mRNA. The P0 F-box-like domain binds unloaded or loaded AGO1 to the P0-SCF complex for ubiquitination. The ubiquitinated AGO1 might be recognised by the autophagy receptor and embed into the phagophore, and subsequently be transferred by the autophagosome. The outer membrane of the autophagosome merges with the membrane of the vacuole to form an autophagic body. AGO1 is finally degraded inside of the autophagic body.

However, autophagy is a highly consistent mechanism in eukaryotic cells, and recent evidence has shown similar autophagic mechanisms in plants. For example, the Arabidopsis NBR1 protein, which is an ortholog of the mammalian autophagic cargo receptor, interacts with Atg8 ubiquitin-like protein through AIM (Atg8-interacting motif) for triggering selective autophagy (3). Furthermore, several atg mutants of Arabidopsis have shown accelerated leaf senescence and hypersensitivity to nutrition starvation, which was considered to trigger autophagy (1). Finally, Atg genes of Arabidopsis also have been shown to play important roles in biotic and abiotic stress responses and plant development (4–7), suggesting that autophagy crosstalks with many pathways.

Introduction to small RNAs in plants

MicroRNAs (miRNAs) and short-interfering RNAs (siRNAs) are small RNAs in animals and plants that have critical roles in development, epigenetic regulation, and pathogen defence (8). They are both 20–24 nucleotides (nt) long and function in either post-transcriptional gene silencing (PTGS) or RNA-directed DNA methylation (RdDM). They differ in the type of RNA from which they are derived and in some of the proteins that produce them and that execute their functions. In general, both miRNAs and siRNAs perform PTGS through cleavage or translational inhibition of RNAs, while siRNAs can also facilitate transcriptional gene silencing (TGS) by directing epigenetic modifications to genomic or viral DNA (9, 10).

In mammalian cells, the Drosha-DGCR8 complex binds primary miRNA (pri-miRNA) and processes pri-miRNA to the precursor miRNA (pre-miRNA). Pre-miRNA binds with Exportin-5-RanGTP complexes, which transfers the pre-miRNA from the nucleus to the cytoplasm (8). After Drosha-DGCR8 processing, DICER1 cleaves the pre-miRNA to generate a double-stranded form of miRNA duplex. In plants, Dicer-like 1 (DCL1) works with a double-stranded RNA-binding domain (dsRDB) – HYL1 – to generate a pre-miRNA and miRNA duplex in the nucleus (8). In animals and plants, the miRNA duplex is loaded into ARGONAUTE (AGO) and the complementary strand of miRNA (miRNA*) is erased, forming a RNA-induced-silencing complex (RISC) with a mature miRNA (miRISC) (8). Finally, the mature RISC goes on to cleave or translationally-inhibit mRNAs that have homology to the miRNA.

The siRNA pathway starts with a double-stranded RNA precursor that is processed by various DICERs in animals and plants (11). Endogenous (i.e., originating from the genome) siRNA, such as trans-acting siRNA (tasiRNA), plays an identical role to miRNA in downregulating endogenous mRNAs through PTGS (12, 13), while production of viral siRNAs are responsible for cleavage of viral genomic RNA in the plant as part of the virus defence response (14, 15). There are four Dicer-like proteins in Arabidopsis thaliana compared to the one in animals, and DCL2 and DCL4 have been reported to facilitate the viral defence response in a process called RNA-mediated resistance (16–18). However, siRNAs can also direct epigenetic change of the genome or viral DNAs. In plants, the siRNAs, which are generated by RNA-DEPENDENT RNA POLYMERASE 2 (RDR2), RNA polymerase IV (Pol IV), and Dicer-like 3 (DCL3), target homologous DNA sequences for epigenetic remodelling, which can result in RdDM (19, 20). Unlike PTGS, which uses ARGONAUTE1 (AGO1), the siRNAs are incorporated into ARGONAUTE 4 or 6 (AGO4, or AGO6) in the RdDM pathway. AGO4 and the other components of the RdDM complex can reprogram DNA methylation, histone modifications and higher order chromatin state, resulting in TGS (21–23).

Inherited stress responses in plants

As plants are unable to move to new locations they have evolved rapid adaptive biological systems to deal with stressful changes in their local environment. These environmental changes can be grouped into biotic (e.g., pathogen infection, herbivory) and abiotic (e.g., temperature, nutrient availability, toxin levels, drought). Each stress elicits a response that can alter the gene expression program at a contained location or throughout the whole plant and ultimately leads to phenotypic change. The phenotypic changes aim to physiologically counter the stress or can even lead to a developmental change that usually involves an earlier progression to flowering and seed dispersal than unstressed plants (24). A large amount of studies have characterised the molecular changes in various stress responses in plants, including hundreds of alterations to small RNAs (miRNAs and siRNAs) (25). Changes to both global and locus-specific epigenetic states have also been described in various stress responses (26).

However, several studies have indicated that the progenies from stressed parental plants have inherited the molecular responses to that stress (26). This is presumed to be an adaptive mechanism that allows the next generation to be prepared for the environment in which they will live. Examples of this transgenerational memory have been shown with plants that are exposed to a wide range of stresses, both abiotic and biotic (26, 27). The precise mechanisms of the inherited responses are not understood but it is thought that the transfer of stress-induced RNAs, metabolites, hormones or epigenetic modifications in the gametes is involved. It is important to note that not all stress responses are inherited so there must also be some mechanisms through which they are reset between generations. Furthermore, some prominent researchers have put forward stringent criteria that they suggest are required to ensure that there is unambiguous evidence for transgenerationally-inherited stress responses (28).

In general, the offspring of stressed plants have often been found to have increased global DNA methylation as well as localised hypomethylation (26). DNA methylation at CG dinucleotides is highly stable in plants and reduction in global CG methylation caused by mutations in the enzymes responsible for its maintenance, MET1 and DDM1 can persist for more than five generations (29, 30) (Box 1). In spite of this making the CG methylation systems apparently ideal for facilitating transgenerational epigenetic inheritance in plants, it is the more dynamic non-CG (CHG and CHH sites, where H is A, T, or C) DNA methylation systems that have been connected with inherited stress responses.

In plants, the RdDM pathway plays an important role in setting non-CG methylation on genomic DNA. 24-nt siRNA biogenesis depends on Pol IV, RDR2, and DCL2-4. Downstream of the siRNA-production, other RdDM components, including the DNA methyltransferase DRM2, AGO4, and Pol V, are associated with the siRNAs to target the genomic loci, which can cause TGS (21). Deletion of components of the RdDM pathway have been shown to reduce or prevent inherited stress responses to herbivory, drought, heat, cold and ultra violet C radiation (31, 32). However, it is currently not clear whether it is the small RNAs themselves, the siRNA-directed epigenetic modifications –or both – that create the transgenerational memory. Clues perhaps lie in the recent description of RdDM that occurs in pollen, egg cells and early embryos in Arabidopsis that serves to establish repeat-element silencing as well as methylation at alleles that display transgenerationally-stable epigenetic states (epialleles) (33, 34).

Finally it should be noted that there is overlap in the CG, and non-CG methylation mechanisms as small RNAs were shown to guide CG remethylation of repeat elements in MET1 null mutant plants (35). Therefore, future research may reveal a role for CG methylation in the inherited stress responses.

Mechanism of selective autophagic degradation of core components of PTGS

The autophagy receptors (p62, and NDP52) bind with substrates and ATG8 family proteins, and integrate the complexes into the membrane of the autophagosome for selective degradation (39, 40). Gibbings et al. (2012) (36) showed that the autophagy receptor NDP52 was responsible for targeting unloaded AGO2 for degradation. Moreover, DICER significantly co-localised with autophagy receptor NDP52 in HeLa cells. The co-localisation foci increased 3.2-fold in cells treated with RAP, suggesting that NDP52-dependent autophagy targets DICER. Moreover, bioimage evidence showed that DICER and AGO2 were enriched in autophagosomes and autolysosomes of HeLa cells when treated with CQ inhibitor, indicating that DICER and AGO2 are subjected to NDP52-dependent autophagic degradation (36).

MiRNA loading in RISC determines AGO stability

The interactions between small RNAs and their associated proteins are not only required for small RNA biosynthesis and function. Recent work has shown that the interactions also coordinate the turn-over of the proteins themselves. DGCR8 null mouse ESCs lacked mature miRNA accumulation and displayed significant reduction of AGO2 protein levels compared with wild-type ESCs (37). However, the AGO2 protein levels increased when DGCR8 null ESCs were complemented with Flag-DGCR8 (37). In addition, introduction of miRNA precursor or siRNA duplexes in the ESCs resulted in an increase of AGO2 protein levels, indicating that the AGO2 stability depended on the loaded miRNAs (37). Finally, this study also implicates autophagy in the reduction of AGO2 as its degradation was blocked by inhibition of the lysosome, but not of the proteosome. Therefore, while the protein machinery of small RNA biogenesis will determine the amount of small RNAs in a cell, the opposite can also occur, i.e. the level of small RNAs can determine the level of proteins. This self-regulating system is thought to achieve a homeostasis of the system but also there is evidence that DICER-ARGONAUTE complexes that do not carry small RNAs can actually interfere with the normal function of those that do (36, 41, 42).

Viral suppressor triggers autophagic AGO1 degradation in plants

In plants, PTGS is commonly used as a virus defence system. To counter this, viruses have evolved methods to suppress the RNA-mediated resistance, resulting in gene silencing suppression of the miRNA and siRNA pathways (43, 44). Most viral suppressors have the ability to bind to miRNA and siRNA duplexes, preventing the small RNA duplexes from loading into the AGO proteins, such as p19, 2b, p21, and HC-Pro (45, 46). P0 is a viral suppressor of Polerovirus that contains an F-box-like motif, and hijacks the host S-phase kinase-associated protein 1 (SKP1) and cullin 1 (CUL1) to form the SCF complex, which is involved in ubiquitination (Figure 1) (47). In addition, P0 has been reported to interact with AGO1 of plants (the major ribonuclease III for miRNA- and siRNA-mediated cleavage) and trigger AGO1 degradation (38, 48). Pazhouhandeh et al. showed that mutated P0 F-box motif inhibits gene silencing suppression (47). However, P0-mediated AGO1 degradation was not affected by treatment with ubiquitination inhibitor MG132 in an experiment that was similar to the mouse cell-line work of Martinez et al. (37), but suppressed by MLN-4924, an inhibitor of neddylation, which is an analogous process to ubiquitination (38, 49). These results imply that inhibition of CUL1 neddylation resulted in impaired AGO1 degradation (38). Furthermore, Derrien et al. (2012) demonstrated that cysteine protease inhibitor E-64d, blocks the activity of lysosomal hydrolases, treatment stabilises AGO1 protein levels in the presence of P0 in the cell, indicating that P0 triggers AGO1 degradation by the autophagic pathway. The destabilisation of AGO1 involves neddylation and autophagy but not ubiquitination-proteosome degradation.

The virally encoded suppressor protein P0 contains an F-box-like domain to hijack SKP1 and CUL1 to form an SCF complex and interact with AGO1 (Figure 1) (48, 49). The P0-SKP1 interaction is required for virus pathogenicity and depletion of SKP1 resulted in host-resistance to Polerovirus (47). Mutations in the F-box-like domain of P0 caused less progeny of viral RNA, with milder symptoms in the host plant than in the wild-type virus, and also inhibited suppression of PTGS (47). P0 requires the ND and the PAZ protein domains of AGO1 for AGO1 interaction and destabilisation (48). However, P0 does not affect Dicer-like 1 (DCL1; DICER homolog) stabilisation, although the PAZ domain also exists in DCL1, suggesting that the ND domain plays an adjacent role with the PAZ domain in P0 recognition and regulation (48).

Bioimaging has provided more information on the AGO1 autophagy mechanism. This technique showed that AGO1 co-localises with ATG genes in vesicles (38). ATG8a covalently attaches to the lipid phosphatidylethanolamine (PE) to bind to autophagosome membranes by means of its lipid moiety. Co-expression of GFP-AGO1 and red fluorescence protein (RFP)-ATG8a fusion proteins showed that AGO1 and ATG8a co-localised in autophagic vesicles. The vesicles become larger when treated with E-64d (38). In addition, ATG8a co-immunoprecipitated with AGO1 in E-64d-treated Arabidopsis plants, indicating that AGO1 interacts with ATG8a (38).