The HIGH EXPRESSION OF OSMOTICALLY RESPONSIVE GENE15–HISTONE DEACETYLASE9 complex associates with HYPONASTIC LEAVES 1 to modulate microRNA expression in response to abscisic acid signaling

After binding to nascent primary microRNA (miRNA) transcripts (pri-miRNAs), the miRNA biogenesis protein HYPONASTIC LEAVES 1 (HYL1) acts as a scaffold that recruits the HIGH EXPRESSION OF OSMOTICALLY RESPONSIVE GENE15 (HOS15)–HISTONE DEACETYLASE9 (HDA9) complex to regulate the expression and processing of miRNAs in response to abscisic acid (ABA).

Abstract

The regulation of microRNA (miRNA) biogenesis is crucial for maintaining plant homeostasis under biotic and abiotic stress. The crosstalk between the RNA polymerase II (Pol-II) complex and the miRNA processing machinery has emerged as a central hub modulating transcription and cotranscriptional processing of primary miRNA transcripts (pri-miRNAs). However, it remains unclear how miRNA-specific transcriptional regulators recognize MIRNA loci. Here, we show that the Arabidopsis (Arabidopsis thaliana) HIGH EXPRESSION OF OSMOTICALLY RESPONSIVE GENE15 (HOS15)–HISTONE DEACETYLASE9 (HDA9) complex is a conditional suppressor of miRNA biogenesis, particularly in response to abscisic acid (ABA). When treated with ABA, hos15/hda9 mutants show enhanced transcription of pri-miRNAs that is accompanied by increased processing, leading to overaccumulation of a set of mature miRNAs. Moreover, upon recognition of the nascent pri-miRNAs, the ABA–induced recruitment of the HOS15–HDA9 complex to MIRNA loci is guided by HYPONASTIC LEAVES 1 (HYL1). The HYL1–dependent recruitment of the HOS15–HDA9 complex to MIRNA loci suppresses expression of MIRNAs and processing of pri-miRNA. Most importantly, our findings indicate that nascent pri-miRNAs serve as scaffolds for recruiting transcriptional regulators, specifically to MIRNA loci. This indicates that RNA molecules can act as regulators of their own expression by causing a negative feedback loop that turns off their transcription, providing a self-buffering system.

IN A NUTSHELL.

Background: MicroRNAs (miRNAs) are small noncoding RNAs with a length of 21–24 nucleotides that play critical roles in mRNA silencing and translational suppression. In plants, miRNAs are transcribed by RNA polymerase II (Pol-II) and processed by the microprocessor, which includes HYPONASTIC LEAVES 1 (HYL1) as a core factor. To maintain homeostasis in the plants and ensure proper developmental programming, the microprocessor is tightly regulated at the transcriptional and posttranscriptional levels. This regulation is particularly important when the plants face environmental stresses, and miRNAs function as key regulatory elements.

Question: A forward genetic screen identified the HIGH EXPRESSION OF OSMOTICALLY RESPONSIVE GENE15 (HOS15)–HISTONE DEACETYLASE9 (HDA9) complex, a well-defined stress-responsive complex, as a component of miRNA biogenesis that interacts with HYL1. We therefore wondered how the HOS15–HDA9 complex acts in the miRNA pathway, how HYL1 is involved in this process, and whether this regulation is important during signaling of the stress-related hormone abscisic acid (ABA).

Findings: We showed that the HOS15–HDA9 complex is a conditional suppressor of miRNA biogenesis under ABA treatment in Arabidopsis thaliana. HYL1 recognizes the nascent primary miRNAs (pri-miRNAs) and guides the HOS15–HDA9 complex to miRNA gene loci. This recruitment suppresses miRNA gene expression and pri-miRNA processing. hos15 hda9 mutants show an increase in pri-miRNA transcription and processing that leads to the overaccumulation of a set of mature miRNAs. Thus, our findings indicate that the nascent pri-miRNAs could serve as scaffolds for recruiting transcriptional regulators specifically to miRNA gene loci.

Next steps: Defining why only a subset of miRNAs appears to be affected by the HOS15–HDA9 complex, even under ABA treatment, is a compelling open question. Additionally, understanding the precise molecular mechanism by which the HOS15–HDA9 complex suppresses pri-miRNA processing requires further investigation.

Introduction

MicroRNAs (miRNAs) are master regulators of cell homeostasis that coordinate plant development and growth via the silencing of numerous transcripts (Bologna and Voinnet 2014; Li and Zhang 2016). In plants, one of the essential functions of miRNAs is to modulate responses to environmental biotic and abiotic stressors (Khraiwesh et al. 2012; Manavella et al. 2019). RNA polymerase II (Pol II) transcribes MIRNA genes, producing primary miRNA transcripts (pri-miRNAs), which are characterized by a unique stem–loop structure containing partially double-stranded RNA (dsRNA). This structure is recognized by the miRNA processing machinery, which, through coordinated cleavage steps, produces mature biologically active miRNAs (Rogers and Chen 2013; Achkar et al. 2016; Wang et al. 2019; Mencia et al. 2022). The activity of the miRNA processing machinery depends on 3 main components, the Dicer-like RNase III endonuclease DICER-LIKE1 (DCL1), the nuclear dsRNA–binding protein HYL1, and the single-zinc finger containing protein SERRATE (SE). Several proteins of the miRNA pathway associate with MIRNA loci and promote recruitment of DCL1 to nascent pri-miRNAs (Fang et al. 2015; Bhat et al. 2020; Cambiagno et al. 2021; Gonzalo et al. 2022; Mencia et al. 2022).

Recently, PRE-MRNA PROCESSING PROTEIN40 (PRP40), the U1 snRNP auxiliary protein, was found to promote the recruitment of SE and DCL1 to MIRNA loci, affecting the Pol II-mediated transcription of miRNA genes (Stepien et al. 2022). The association of these processing proteins with the chromatin at MIRNA loci results in the cotranscriptional processing of nascent pri-miRNAs, although frequently such processing also occurs posttranscriptionally (Gonzalo et al. 2022). The processing of nascent pri-miRNAs, which requires the early recruitment of miRNA processing factors to the MIRNA encoding loci, is promoted by the presence of a 3-stranded nucleic acid structure known as R-loops near the transcription start sites of miRNA genes and is assisted by PRP40 (Gonzalo et al. 2022; Stepien et al. 2022). In addition to regulation by general transcription factors (TFs), MIRNA transcription is regulated specifically by proteins such as CELL DIVISION CYCLE 5 (CDC5), the transcription factor CYCLING DOF FACTOR 2 (CDF2), NEGATIVE ON TATA LESS2A (NOT2A), and GENERAL CONTROL NON-REPRESSED PROTEIN 5 (GCN5), among others (Achkar et al. 2016; Wang et al. 2019). However, it is largely unknown how these TFs can recognize specific MIRNA promoters, which collectively lack signature cis-regulatory elements that differentiate them from other genes. One possibility is that the core components of the miRNA processing machinery help in recruiting regulatory factors to give the system specificity.

HYL1 is one of the miRNA biogenesis proteins commonly found associated with chromatin, likely indirectly through its interaction with the nascent pri-miRNA (Bhat et al. 2020). Whether the recruitment of HYL1 to the nascent pri-miRNAs depends on the formation of the processing complex or is the consequence of a direct recognition of the naked stem–loop transcript is unknown. In either case, the HYL1–pri-miRNA interaction causes HYL1 to interact with the transcriptional machinery, potentially regulating gene expression at the level of chromatin (Bielewicz et al. 2023). In turn, the ability of HYL1 to interact with the pri-miRNA dsRNA stem and mature miRNA duplex is a dynamic light-dependent turnover mechanism that constitutes a central hub to fine-tune miRNA activity in plants (Manavella et al. 2012; Cho et al. 2014; Yang et al. 2014; Achkar et al. 2018; Choi et al. 2020; Tomassi et al. 2020). HYL1 function is not limited to miRNA biogenesis, as it also plays a critical role in miRNA-induced inhibition of translation by promoting enrichment of ARGONAUTE 1 in the polysome, skotomorphogenic growth, and apical hood formation (Sacnun et al. 2020; Vacs et al. 2021; Yang et al. 2021). Moreover, HYL1 interacts with a large set of additional RNA–binding proteins, including RS40 and RS41, 2 serine/arginine-rich splicing factor proteins, and ENHANCED STRESS RESPONSE 1 (HOS5) proteins, indicative of the broad range of HYL1 functions (Chen et al. 2015).

Chromatin remodeling is a vital process controlling chromatin architecture and epigenetic state, regulating gene expression by impacting the accessibility of the transcriptional machinery to genomic DNA. Histone acetyltransferases (HATs), histone deacetylases (HDACs), and histone methyltransferases (HMTs) play primary roles in the chromatin remodeling process (Hsieh and Fischer 2005). Chromatin modifications such as deacetylation and methylation regulate chromatin accessibility and recruitment of both TFs and Pol II and ultimately also regulate MIRNA transcription (Luo et al. 2013). HIGH EXPRESSION OF OSMOTICALLY RESPONSIVE GENES 15 (HOS15) is a multifunctional Arabidopsis (Arabidopsis thaliana) putative ortholog of the human TRANSDUCIN-BETA-LIKE1 (TBL1), a component of the NUCLEAR RECEPTOR CO-REPRESSOR 1/SILENCING MEDIATOR OF RETINOIC ACID AND THYROID HORMONE RECEPTOR (NCoR1/SMRT) complex, which is a repressive coregulatory module for multiple transcription pathways (Zhu et al. 2008). In Arabidopsis, HOS15 facilitates the recruitment of histone deacetylases to the promoters of target genes involved in a variety of biological pathways, including cell proliferation, immunity, flowering, and leaf senescence (Park et al. 2019; Lim et al. 2020; Shen et al. 2020). In addition, HOS15 serves as an E3 ubiquitin ligase substrate receptor in response to stresses such as drought and cold by mediating the degradation of OPEN STOMATA 1 (OST1) and HISTONE DEACETYLASE2C (HD2C), respectively (Park et al. 2018b; Ali et al. 2019). This dual activity of HOS15 is evident in the plant response to cold stress, in which HOS15 promotes a change in chromatin from a repressive to a permissive state. In normal conditions, the POWERDRESS (PWR)/HOS15/HD2C repressor complex negatively regulates the expression of multiple COLD-REGULATED (COR) genes by promoting histone deacetylation and chromatin compaction. However, under cold stress, HOS15 triggers the proteasomal degradation of HD2C. This permits the positioning of HATs at COR–encoding loci to increase histone acetylation and the consequent transcriptional activation of COR genes (Park et al. 2018b; Lim et al. 2020).

Like HYL1, HOS15 interacts with the splicing factor RS40, which bridges chromatin and the spliceosome (Park et al. 2018a). Moreover, immunoprecipitation tandem mass spectrometry (IP-MS) experiments analyzing HOS15 and its chromatin-acting partners HISTONE DEACETYLASE 9 (HDA9) and POWERDRESS (PWR) reveal many shared coprecipitated RNA–binding proteins, including RS40, RS41, Pol II, and HYL1 (Mayer et al. 2019). These interactions suggest a crosstalk between the chromatin-remodeling machinery and pri-miRNA processing, occurring at the transcriptional level. Consistent with this view, transcriptomic analyses of hos15-2 and hda9-1 loss of function mutants demonstrate a significant alteration in the transcripts involved in gene silencing by RNA and splicing pathways (Mayer et al. 2019).

Here, we show that HOS15 and HDA9 play important roles in the miRNA pathway by selectively regulating the expression of MIRNA genes. Our data show that hos15 and hda9 mutants showed moderate, but consistent, alterations in miRNA expression that were exacerbated under abscisic acid (ABA) treatment. This amplification of the effects on miRNA biogenesis caused by ABA–induced HOS15/HDA9 deficiency was consistent with the critical roles of these proteins during ABA signaling (Ali and Yun 2020; Baek et al. 2020; Khan et al. 2020). Surprisingly, we found that the recruitment of the HOS15–HDA9 complex to the MIRNA loci, and the concomitant transcriptional change, depended on the recognition of nascent pri-miRNAs by HYL1. We observed that, during this recognition, both HOS15 and HDA9 interacted with HYL1 to suppress the ABA–responsive processing of pri-miRNAs. These findings establish HYL1 as a scaffold that specifically recruits chromatin modifiers to MIRNA loci, modulating cotranscriptional pri-miRNA processing. They also indicate that the nascent pri-miRNAs are critical for recruiting transcriptional regulators to MIRNA loci. In the case of the HOS15–HDA9 complex, its RNA–dependent recruitment under ABA treatment represses selected MIRNA loci, causing a negative feedback loop. Still, our findings may go beyond the ABA–dependent recruitment of HOS15–HDA9 and point to a more general mechanism in which cotranscriptional recognition of the nascent pri-miRNAs by the processing complex allows the recruitment of specific transcriptional regulators to MIRNA loci.

Results

A forward genetic screen identified HOS15 as a component of the miRNA pathway in Arabidopsis

Previously, we performed a forward genetic screening using plants carrying a reporter system in which an artificially designed miRNA silenced the firefly luciferase gene, helping us isolate plants with impaired miRNA activity (Manavella et al. 2012). Among the isolated mutants, we identified numerous genes encoding miRNA-related proteins (Manavella et al. 2012; Francisco-Mangilet et al. 2015; Karlsson et al. 2015; Re et al. 2020; Tomassi et al. 2020). Here, using mapping by sequencing, we localized the causal mutation of one of the isolated mutants to a small region of chromosome 5 (Supplemental Figure S1A). Within this region, we found a typical EMS-induced C > T transition that results in a nonsense mutation of arginine 43 (R43X) in the gene encoding HOS15 (AT5G67320) (Fig. 1A). This mutant, named hos15-3 hereafter, showed an increase in the luciferase activity of the reporter, although this increase was moderate when compared to previously isolated mutants (Fig. 1, B and C). Morphologically, hos15-3 presented reduced stature, delayed flowering, and shorter stems with packed siliques (Fig. 1D, Supplemental Fig. S1, B and D). Most of these phenotypes, likely pleiotropic, are moderate or hardly observable in the hos15-2 alleles, which likely represent a hypomorphic allele (Fig. 1B, Supplemental Fig. S1, B, C and D). Similarly, we did not detect any morphological alterations in plants overexpressing HOS15 (Supplemental Fig. S1, B and C).

Figure 1.

Characterization of the hos15 mutant. A) Gene structure of genomic locus encoding HOS15 (gHOS15) showing a T-DNA insertion site in hos15-2, a nonsense mutation at amino acid residue 43; arginine (R) in hos15-3, and the target region for the CRISPR system (hos15^CRISPR). Gray boxes, exons; black lines, introns; flanking blue boxes, 5′ and 3′ UTR regions. B, C) Bioluminescence activity as measured with a CCD camera in 14-d-old B) and 25-d-old C)hos15-3 mutants and control plants (reporter and 35S_pro:Luciferase). The color ranging from low (blue) to high (white) represents the luminescence intensity. D) Phenotypic characterization of control lines (reporter and Col-0), and hos15 mutants (hos15-2, hos15-3, and hos15^CRISPR) having fully expanded siliques. Bar = 5 cm. To facilitate comparison and observation, each plant was photographed separately, digitally extracted, and mounted on a single panel with a black background.

To further show that the phenotype observed in hos15-3 was caused by the mutation we detected, we complemented the mutant plant with a wild-type (WT) copy of the gene, and this reverted the phenotype (Supplemental Fig. S1D). In addition, we created an additional null allele in the third exon of the gene using the CRISPR/Cas9 system targeting guide RNAs (Fig. 1A). The mutants presented a phenotype similar to that of hos15-3 (Fig. 1B and Supplemental Fig. S1B). Unexpectedly, HOS15 transcript levels were drastically elevated in the CRISPR lines, showing the presence of feedback regulation triggered by the lack of active protein (Supplemental Fig. S1E).

HOS15 and the histone deacetylases HDA9 and HDA19 interact with the miRNA biogenesis protein HYL1

To further explore the role of HOS15 in the miRNA pathway, we analyzed its cellular localization using confocal microscopy. As previously reported for a repressor complex protein (Zhu et al. 2008), we detected HOS15 mainly in the nucleus (Fig. 2A). Within the nucleus, we observed HOS15 colocalized with the miRNA biogenesis factors HYL1, SE, and DCL1 (Fig. 2A). Still, such colocalization did not occur in the so-called dicing bodies, the proposed place of pri-miRNA processing (Fig. 2B). Because colocalization does not necessarily imply an interaction between HOS15 and miRNA biogenetic machinery, we tested whether HOS15 interacts directly with some miRNA biogenesis pathway–related proteins. We first performed a yeast 2-hybrid (Y2H) assay with HOS15 as bait against several miRNA biogenesis factors as prey. While we did not detect the interaction of HOS15 with most proteins tested, we observed that HOS15 could interact with HYL1 and the histone deacetylase HD2C, used as the positive control (Fig. 2, C and D). We further demonstrated the interaction of HOS15 with HYL1 with coimmunoprecipitation (Co-IP) assays (Fig. 2, E to H). These results resemble previous results showing that HYL1 interacts with TOPLESS (TPL) and TOPLESS-RELATED (TPR1 and TPR2), Lish-WD40 proteins with functions in chromatin remodeling (Bielewicz et al. 2023). This suggests that HYL1 is prone to interact with regulatory factors, likely at the chromatin level.

Figure 2.

HYL1 associates HOS15 in the nucleus but HOS15 does not interact with SE, DCL1, or DDL. A and B) Confocal microscopy images showing the nuclear localization of eGFP- and mCherry-tagged versions of HOS15 that were transiently transformed into N. benthamiana leaves. Images show the localization of A) HOS15 with DCL1, HYL1, and SE and B) HOS15 with SE. Bars = 5 μm. C) Interaction between HOS15 and HYL1 in the yeast 2-hybrid assay. The assay was performed with HYL1 as prey and HOS15 as bait. An EV construct was used as a negative control, with HOS15 as bait and HD2C as prey for the positive control. Yeast cells were grown in the presence of LT, LTH + 0.1 mm 3AT deficiency media for 5 d, after which images were taken. D) Yeast 2-hybrid assays. Assays were performed with SE, DDL1, DCL1, or HEN1 as prey and HOS15 as bait for monitoring their interactions. An EV construct was used as a negative control, HOS15 as prey, and HD2C as bait for the positive control. Yeast was grown in the presence of LT, LTH + 0.1 mm 3AT deficiency media for 5 d, after which images were taken. E) HOS15 interacts with HYL1 in vivo. N. benthamiana leaves were infiltrated with Agrobacterium harboring 35S_pro:HOS15-HA and 35S_pro:HYL1-FLAG for transient expression. Protein extracts (input) were IP with an anti-HA antibody and resolved by SDS–PAGE. The antibodies used to detect the proteins are noted on the right of the blots. F to H) In vivo Co-IP of HOS15 with SE, HEN1, and DCL1. F) HOS15-HA and SE-FLAG, G) HOS15-HA and DDL-FLAG, and H) HOS15-HA, HEN1-GFP, and DCL1-GFP, proteins were transiently coexpressed in N. benthamiana. Protein extracts (input) were IP with an anti-HA antibody and resolved by SDS–PAGE. The antibodies used to detect the proteins are noted on the right of the blots.

HOS15 was reported to have dual functions as an E3-ligase substrate receptor and transcriptional corepressor protein (Park et al. 2018b). This suggested the possibility that HOS15 regulates the miRNA pathway either by affecting the protein stability of a microprocessor component, likely HYL1, or by directly affecting expression at the transcriptional level. The former scenario is unlikely, as HYL1 degradation appeared to be independent of ubiquitination (Cho et al. 2014; Jung et al. 2022). We found that the level of HYL1 was slightly decreased in hos15-2 mutant, suggesting that the E3 ligase activity of HOS15 is irrelevant to HYL1 degradation (Supplemental Fig. S2).

Therefore, we checked for HYL1 interaction with histone deacetylases, well-known HOS15 partners in transcriptional regulation (Mayer et al. 2019; Park et al. 2019; Lim et al. 2020). Y2H and co-IP experiments revealed that HYL1 interacts not only with HOS15 but also with the RPD3 histone deacetylases HDA9 and HDA19, known partners of HOS15 (Fig. 3, A and B). HOS15–HDA9 is a well-defined stress-responsive complex for drought stress and the ABA signaling pathway (Zhu et al. 2008; Park et al. 2018b; Ali et al. 2019; Baek et al. 2020; Khan et al. 2020; Lim et al. 2020; Shen et al. 2020). Thus, we retested the HOS15–HYL1 interaction with Co-IP experiments under ABA treatment. We found that ABA triggers a stronger interaction between HOS15 and HYL1 (Fig. 3C). These results indicated that complex formation among HYL1, HOS15, and RPD3-type HDACs is promoted under ABA treatment. To check whether HYL1 influences H3 acetylation, we quantified histone acetylation under ABA treatment in hyl1-2, hos15-2, and hda9 mutants by immunoblot analysis. As expected, both hos15-2 and hda9, which were used as positive controls, showed highly acetylated histone levels as compared to that of WT (Fig. 3D). We observed a similar increase in histone acetylation in hyl1-2, especially after ABA treatment (Fig. 3D). In the Co-IP experiments, HOS15 only interacted with HYL1 among all tested microprocessor proteins, implying that it is possible that a fraction of HYL1 not associated with the microprocessor interacts with the HOS15–HDA9 complex.

Figure 3.

HYL1 forms a complex with the RPD3 histone deacetylase family. A) Yeast 2-hybrid assays. Assays were performed with HDA9, HDA19, or HD2C as prey and HYL1 as bait for monitoring their interactions. An EV was used as the negative control, with HYL1 as bait and SE as prey for the positive control. Yeast cells were grown in the presence of LT, LTH + 0.1 mm 3AT deficiency media for 5 d, after which images were taken. B) In vivo Co-IP of HYL1 with HDA9 and HDA19. HYL1-FLAG, HDA9-HA, and HDA19-HA proteins were transiently coexpressed in N. benthamiana. Protein extracts (input) were IP with anti-FLAG antibodies and resolved with SDS–PAGE. The shown immunoblots were developed with target antibodies. C) The protein extracts indicated (input) were IP with an anti-MYC antibody and resolved with SDS–PAGE. The shown immunoblots were developed with target antibodies. D) Immunoblotting showing histone acetylation levels using an antibody specific for AcH3 in 1-wk-old seedlings of Col-0, hyl1-2, hos15-2, and hda9. Arabidopsis lines were treated with 50 μm ABA for 3 h. An anti-H3 antibody was used as a loading control.

HYL1 binding to nascent pri-miRNAs allows the recruitment of HOS15–HDA9 to MIRNA loci

Due to the role of HOS15 and its accessory histone deacetylase in transcriptional regulation, it appeared likely that the activity of HOS15 and HDA9 in the miRNA pathway occurs at transcriptional levels. In agreement, a HDA9 chromatin Immunoprecipitation (ChIP)-seq analysis revealed the associations of HDA9 with many MIRNA loci (Supplemental Figs. S3, A to D). Similarly, ChIP experiments showed that HOS15 can also associate with MIRNA loci (Supplemental Fig. S4A and S4B). However, there were no common features such as a specific sequence or positioning distance that can explain the apparently selective binding of this protein to MIRNA loci (Supplemental Fig. S3B). Thus, how the HOS15–HDA9 complex recognizes MIRNA loci, which do not have any distinctive regulatory feature with respect to other genes, is an interesting question. This implies that an additional feature is necessary to allow the HOS15–HDA9 complex to recognize the MIRNA genes and be specifically recruited to these loci.

Recently, the crosstalk between transcription and pri-miRNA processing was demonstrated to play a critical role in miRNA biogenesis (Gonzalo et al. 2022; Stepien et al. 2022). HYL1, which interacted with HOS15–HDA9 and MIRNA loci (Figs. 2, 3 and Supplemental Fig. S4, A and B), is known to associate with MIRNA loci during cotranscriptional pri-miRNA processing (Bhat et al. 2020) (Gonzalo et al. 2022). Thus, we predicted that HYL1 may serve as a scaffold directing the HOS15–HDA9 complex to the MIRNA loci through direct interaction. To test this hypothesis, we performed ChIP-qPCR assays to quantify the recruitment of HOS15 and HYL1 to MIRNA loci in hyl1-2 and hos15-2 mutants. We found that HOS15 associates with MIRNA loci such as MIR156, MIR167, and MIR171, yet these interactions are lost in the absence of HYL1 (Fig. 4, A and E). Conversely, HYL1 association with the selected MIRNA loci was not impaired in the hos15-2 and hda9-1 mutants, as expected from a general miRNA biogenesis factor (Fig. 4, A, B, and C). In agreement with increased interaction of HOS15–HDA9 with HYL1 in the presence of ABA (Fig. 3C), ChIP experiments revealed that ABA enhances the association of HYL1 and HOS15, although not HDA9, with MIR156 and MIR159 loci (Fig. 4, B, D, and E). The presence of HYL1 remains an essential requirement for HOS15 association to MIRNA loci even under ABA treatments (Fig. 4E). Interestingly, while the association of HYL1 to MIRNA loci was not impaired by the absence of HOS15, the mutation of this gene caused a reduction of the effect of ABA on HYL1 association with MIRNA loci (Fig. 4B). Altogether, our results indicate that HYL1 is responsible for the HOS15–HDA9 to the MIRNA loci, a recruitment that is facilitated by ABA.

Figure 4.

HYL1 binding to nascent pri-miRNAs allows the recruitment of HOS15–HDA9 to MIRNA loci. A) ChIP assay was performed to analyze HOS15 and HYL1 association with 3 MIRNA loci. IP was performed using anti-GFP (right panel) and anti-HYL1 antibodies (left panel) in hyl1-2 and hos15, respectively, each compared with the control plant Col-0. IgG was used as the control. B) ChIP assay using an anti MYC antibody to detect the association of HYL1-6XMYC with MIRNA loci in Col-0 and hos15-2 plants incubated for different periods of time with ABA or a Mock solution (time point 0). C) ChIP assay of MIRNA156a and MIRNA159b loci using anti-HYL1 antibody in hda9-2 mutant compared to the Col-0 control plant. D) ChIP assay of MIRNA156a and MIRNA159b loci using an anti-GFP antibody in HDA9–GFP expressing plants after 3 h of 50 μm ABA treatment (+) compared to untreated (−) plants. E) ChIP assay of MIRNA156a and MIRNA159b loci using an anti-HOS15 antibody in hyl1-2 mutant plants and Col-0 control plants at different time points during ABA treatment. F) ChIP assay for HYL1-6MYC lines pretreated with RNAse and α-amanitin and IP with an anti-MYC antibody compared to nonpretreated plants, with or without ABA treatment. In all cases, the amount of DNA in the IP complex was determined by qPCR and is presented as the fold enrichment after normalization. Values of anti-GFP and anti-HYL1 IP samples are normalized to the input and expressed as relative to the IP signal using anti-rabbit IgG antibody (IgG line). The data represent means (±Sd) from 3 biological replicates with 3 technical repeats. Significant differences were determined using an unpaired, 2-tailed, Student's t-test (*P< 0.05, **P < 0.01). ns, not significant.

Perhaps the most surprising result came from testing the interactions of HYL1 with the MIRNA loci in plants treated with RNAse and α-amanitin to eliminate the nascent pri-miRNA from the loci. Our results showed that the treatment with RNAse and α-amanitin disrupted the association of HYL1 with MIRNA loci, indicating that the association of HYL1 with these loci is indirect and is dependent on the nascent pri-miRNAs (Fig. 4F). Thus, given the ability of HOS15 and HDA9 to interact with HYL1, we reasoned that the HOS15–HDA9 complex recognizes MIRNA loci, guided by a pri-miRNA/HYL1 complex rather than by recognizing specific DNA motifs. These results position HYL1 as a scaffold protein that allows the specific recruitment of a chromatin modifier complex to MIRNA loci. Our data also provide a potential mechanism by which MIRNA genes can be specifically distinguished from other coding genes by the nascent pri-miRNAs. Since HYL1 is a dsRNA binding protein, it is likely that the typical hairpin structure of the nascent pri-miRNAs confers the specificity. Interestingly, many miRNAs follow a processing pathway that appears to be independent of HYL1, especially under stress (Re et al. 2019). Thus, these loci may be poorly targeted by the HOS15–HDA9 complex.

HOS15 and HDA9 suppress ABA–responsive miRNA accumulation

So far, our results indicate that HYL1 association with nascent pri-miRNAs triggers the recruitment of HOS15–HDA9 to MIRNA loci. This recruitment, which appears to be promoted under ABA treatment when the complex is most active, could ultimately trigger histone deacetylation and a transcriptional reprogramming of the MIRNA loci. To test the effect of HOS15–HDA9 in the miRNA pathway, we first quantified the levels of miRNA and miRNA targets in hos15 and hda9 mutants grown for 12 d under long-day photoperiod (Supplemental Data Set 1). Our RT-qPCR and RNA sequencing analyses showed that only a small subset of miRNA-targeted mRNAs were moderately overaccumulated in the mutants (Fig. 5A and Supplemental Data Set 2, A and B). Similarly, under control conditions, the mutation of HOS15 did not appear to have a strong effect on miRNA accumulation, contrary to the effect using hyl1-2 mutants as a control (Fig. 5B). A further analysis of additional mutant alleles (hos15-3 and the CRISPR edited lines) by small RNA blots (Fig. 5C and Supplemental Fig. S5A) and small RNA sequencing (Fig. 5, D and E, and Supplemental Fig. S5, B and C, and Data Set 3) confirmed the minor effect of the mutations on miRNA accumulation in control conditions. These observations paralleled the subtle increase in the luciferase reporter activity measured in hos15-3 mutants (Fig. 1, B and C), suggesting that the effect of HOS15–HDA9 in the miRNA pathway is negligible in control conditions.

Figure 5.

HOS15 and HDA9 suppress miRNA expression upon ABA treatment. A) Expression levels of miRNA target genes in control (reporter and Col-0) and mutant (hos15-2 and hos15-3) plants grown in control conditions, measured using RT-qPCR. Error bars represented with means (±Sd) 3 biological replicates with 3 technical repeats and asterisks (*) designate the significance by a 2-tailed, unpaired t-test followed by the Holm–Sidak correction (*P < 0.05, **P < 0.01). B) Scatter plot of miRNA levels and their target gene expression levels in hos15-2 and hyl1-2 mutant plants. Data points show the fold change (log₂) relative to WT plants. C) Small RNA blots for detecting endogenous miRNA levels measured with RT-qPCR in mutants (hos15-3, hos15^CRISPRL1, and hos15^CRISPRL2) and control lines (reporter and Col-0). The miRNAs detected are indicated on the right, and the relative abundance of each miRNA is indicated above each band as calculated by measuring the band intensity using ImageJ and normalized to the corresponding control plant. U6 was used as a loading control. D) Small RNA sequencing analysis. Mean expression levels of individual miRNAs showing the up- and downregulated miRNA in hos15-3 and hos15-2 plants relative to Col-0 plants. Black lines indicate the median of the expression levels. Each open circle corresponds to a single miRNA or collapsed miRNA family. E) Venn diagram of differentially expressed genes in the mutant lines (hos15-3 and hos15-2) representing the overlapping miRNAs among the samples; Up: upregulated genes; Down: downregulated genes. F) Small RNA sequencing analysis of plants treated with ABA for different periods of time and compared to Mock-treated plants. Black lines indicate median expression levels. Each open circle corresponds to a single miRNA or collapsed miRNA family. G) Small RNA sequencing analysis showing the median expression levels of individual miRNAs in hda9-1, hos15-2, and hyl1-2 as compared to WT plants with or without ABA. Left (SET 1) and right (SET 2) panels show the results from 2 independent sRNA-seq experiments. Black lines indicate median expression levels. H) Heatmap of differentially expressed miRNAs from 2 independent sRNA-seq experiments (SET1 and SET2) in hda9-1, hos15-2, and hyl1-2 plants as compared to WT plants treated with and without ABA. The color bar represents miRNA expression levels (red, upregulation; blue, downregulation; white, no change). I) Small RNA blots for detecting endogenous miRNA levels in mutant (hyl1-2, hos15-2, and hda9-1) and control (Col-0) plants treated with or without ABA. The miRNAs detected are indicated on the left, and the relative abundance of each miRNA is indicated above each band as calculated by measuring the band intensity using ImageJ and normalized to the corresponding control plant. U6 was used as a loading control.

This subtle effect under control conditions is not surprising given that HOS15–HDA9 is a well-defined stress-responsive complex for drought stress and the ABA signaling pathway (Zhu et al. 2008; Park et al. 2018b; Ali et al. 2019; Baek et al. 2020; Khan et al. 2020; Lim et al. 2020; Shen et al. 2020). Furthermore, we found the HOS15–HDA9 interaction with HYL1 and MIRNA loci was promoted under ABA treatment (Figs. 3C and 4, B, E, and F). HYL1 also appeared to be regulated by ABA and hyl1-2 plants which showed hypersensitivity to ABA treatment (Lu and Fedoroff 2000). Thus, it is reasonable to hypothesize that the HOS15–HDA9–HYL1 complex acts in the miRNA pathway by regulating miRNA expression in response to ABA. To test this hypothesis, we performed small RNA sequencing analyses using hyl1-2, hos15-2, and hda9-1 mutants treated with ABA. First, we analyzed a publicly available miRNA-sequencing kinetic data set from plants treated with ABA (GSE145208) to define the effect of ABA treatment on miRNA accumulation over time. In wild-type plants, there was a global change in miRNA levels after ABA treatment, with many up- and downregulated miRNAs (Fig. 5F and Supplemental Data Set 4). Still, a distinctive upregulation of miRNAs such as miR159, miR162, miR166, miR169, and miR395 was observed 3 h after treatment, with a gradual reduction from 6 to 24 h (Fig. 5F). This observation, plus the fact that the peak of ABA signaling before desensitization also occurs 3 h after treatment (Ali et al. 2019; Ali et al. 2020), encouraged us to select this time point for our small RNA sequencing experiments. The hyl1-2 plants showed a severe reduction in miRNA production that was independent of the hormone treatment, because of such a strong miRNA biogenesis–deficient mutant (Fig. 5, G, H, and I). Conversely, hos15-2 and hda9-1 displayed a moderate increase in the abundance of many miRNAs (median expression change of ∼0.6 and ∼0.7, respectively) under ABA treatment (Fig. 5, G and H, and Supplemental Data Set 5). ABA–responsive miRNAs such as miR159, miR160, miR162, miR166, miR172, miR395, miR397, miR399, mir403, and miR869 were among the most highly accumulated miRNAs in both mutants under ABA treatment (Fig. 5H). The small RNA blot analysis of selected ABA–responsive miRNAs confirmed the increased miRNA levels in hos15-2 and hda9-1 upon ABA treatment (Fig. 5I). Furthermore, we observed that the ABA–responsive miRNAs were not upregulated in plants overexpressing HDA9 (Supplemental Fig. S5D). These results suggested that the HOS15–HDA9 complex acts specifically to repress the accumulation of miRNAs under ABA treatment. Furthermore, our data indicate that the effect of hos15 and hda9 mutations on the miRNA biogenesis pathway is not general and may only act conditionally.

HOS15 and HDA9 deficiency causes a negligible alteration of the levels of pri-miRNAs

Thus far, our data show that the interaction of HOS15–HDA9 with HYL1 affects the miRNA pathway. Under control conditions, this complex appears to have only a negligible impact on the miRNA pathway. This is likely a consequence of HOS15–HDA9 acting as a stress-specific complex that may preferentially interact with HYL1 under specific conditions, as suggested by our data (Fig. 3C). Consistent with this idea, under ABA treatment, where the interaction of HOS15–HDA9 with HYL1 was promoted (Fig. 3C), the complex appeared to have a stronger effect in repressing miRNA accumulation.

Given that HOS15 and HDA9 are known transcriptional corepressors, we speculated that their association with HYL1 can suppress the transcription of specific MIRNA genes. Thus, the elevated levels of miRNAs detected in the mutants after ABA treatment may be a consequence of enhanced transcription of miRNAs targeted by HOS15–HDA9 in the mutant background. To test whether the HOS15–HDA9–HYL1 complex regulates miRNA biogenesis at the transcriptional level, we performed a transcriptomic analysis of hyl1-2, hos15-2, and hda9-1 mutants with or without ABA treatment (Supplemental Data Set 6). As previously reported for miRNA-processing–deficient mutants, hyl1-2 plants overaccumulated unprocessed pri-miRNAs. Conversely, hos15-2 and hda9-1 mutants showed, with some exceptions, a profile similar to the wild-type plants under both the control and ABA-treated conditions (Fig. 6A and Supplemental Fig. S6A). A comparison of the transcriptome and sRNA sequencing experiments revealed that besides a few exceptional MIRNAs, such as MIR3932b, MIR825, and MIR5026, the expression level of MIRNA genes did not parallel the levels of the mature miRNAs in the mutants (Supplemental Fig. S6, B and C, and Supplemental Data Set 7). In contrast, we confirmed the inverse correlation between the levels of pri-miRNAs and mature miRNAs in the hyl1-2 mutant (Supplemental Fig. S6B).

Figure 6.

pri-miRNA transcription is promoted in hos15 and hda9 plants although pri-miRNA steady levels remain stable in the mutants. A) Scatter plot showing the steady levels of pri-miRNA transcripts in hyl1-2, hos15-2, and hda9-1 mutants with 50 μm or without (Mock) ABA as measured by RNA sequencing analysis. The y-axis shows the Log₂ ratio between mutant and Col-0. Each data point represents an individual pri-miRNA. B, C) ChIP assay of Pol-II occupancy at MIR159a, MIR164a, and MIR166a loci in Col-0, hyl1-2, hos15-2C), and hos15-3B) plants with or without 100 μm ABA (Mock) using an anti-Pol-II antibody. Anti-IgG IP samples and reporter plants were used as the control. The results are expressed as fold enrichments compared with the input. Error bars represented with means ± Sd of biological triplicates, and asterisks (*) designate the significance determined by a 2-tailed, unpaired t-test (*P < 0.05, **P < 0.01).

However, this apparent lack of transcriptional response in the mutants may be inconclusive. Increased MIRNA transcription is difficult to detect by measuring the steady levels of pri-miRNAs because the miRNA processing complex efficiently processes them. In this sense, the accumulation of a specific miRNA rarely correlates with high levels of its pri-miRNA. Accumulation of pri-miRNAs is commonly observed if the processing complex itself is impaired, such as in DCL1, HYL1, or SE mutants, but not often in plants with impaired miRNA biogenesis regulatory factors. To investigate this possibility and determine whether HOS15–HDA9 is implicated in the transcriptional regulation of MIRNA genes, we performed a Pol-II occupancy assay in mutant plants to quantify the amount of transcriptional complex on MIRNA loci; this measures transcriptional activity rather than steady levels of pri-miRNAs. Supporting the role of HOS15–HDA9 repressing MIRNA transcription, we observed a significant overaccumulation of Pol-II at 3 selected MIRNA loci in hos15-2 and hos15-3 mutants both in control conditions and under ABA treatment (Fig. 6, B and C). We selected these 3 loci for 3 reasons: the processing complex is associated with these loci (Cambiagno et al. 2021), these miRNAs are cotranscriptionally processed (Gonzalo et al. 2022), and these miRNAs are increased in the mutants under ABA treatment (this work). These features make them good candidates to be targeted by HOS15, as HYL1 will be associated with them. The transcription of MIRNA genes was enhanced in hda9-1 and hos15-2, even though the steady levels of pri-miRNAs in the mutants were comparable to those of wild-type plants.

Pri-miRNA processing is promoted in hos15 and hda9 mutants

The apparent unchanged levels of pri-miRNAs in hda9-1 and hos15-2 (Fig. 6A) appear to be inconsistent with enhanced MIRNA transcription (Fig. 6B) and the overaccumulation of miRNAs in hos15 and hda9 under ABA treatment (Fig. 5, G to I). We hypothesized that the inconsistency among enriched Pol-II occupancy, steady pri-miRNA levels, and mature miRNA accumulation under ABA treatment was due to an enhanced rate of pri-miRNA processing. To test this hypothesis, we investigated whether miRNA processing activity was affected in hos15 and hda9 mutants. First, we isolated the nuclear fractions from Col-0, hos15-2, and hda9-1 plants (Supplemental Fig. S7A) and incubated them with UTP-radioactive–labeled pri-miR172c or pri-miR159b to monitor the processing activity in the absence of HOS15 and HDA9. We found that, compared to WT, the processing of pri-miR172c and pri-miR159b was enhanced in hda9 mutants, while HOS15 deficiency resulted in a slightly, but consistent, increased processing activity (Fig. 7A). To ensure that the processed fragments corresponded to the products of microprocessor activity, we purified recombinant full-length DCL1 (Supplemental Fig. S7B) and performed in vitro processing assays with pri-miR172c (Supplemental Fig. S7C). The cleavage pattern of pri-miR172c was almost identical to that of Figure 7A, confirming the efficiency of the assay. Then, to test the effect of ABA treatment under processing, we isolated nuclear fractions from Col-0, hos15-2, and hda9-1 plants treated with or without ABA (Supplemental Fig. S7D) and repeated the pri-miRNA processing assay with UTP-radioactive–labeled pri-miR172c, pri-miR159b, and pri-miR398a using the nuclear fractions. We found that the ABA treatment accelerated the processing activity in both hos15 and hda9 mutants (Fig. 7B). These results implied that pri-miRNA processing of the pri-miRNAs we tested was enhanced in hos15 and hda9 mutants under ABA treatment.

Figure 7.

ABA treatment causes enhanced pri-miRNA processing in hos15 and hda9 mutants. A) In vitro pri-miRNA processing assay performed with increased relative concentrations (1×, 2×, and 4×) of crude nuclear protein extracts from Col-0, hos15-2, and hda9-1 plants. Numbers on the left show molecular sizes in bases. Black arrows indicate the unprocessed pri-miRNA, and black asterisks mark mature miR172c and miR159b. B) In vitro pri-miRNA processing assay performed with crude nuclear protein extracts from Col-0, hos15-2, and hda9-1 plants treated with 50 μm or without (Mock) ABA. Numbers on the left show molecular sizes in bases. Red arrows indicate the unprocessed pri-miRNA while red asterisks mark mature miRNAs. Labels underneath the panels show the radiolabeled pri-miRNA used in each experiment. Bottom panels show overexposed images of the ∼21 nt region. C) Precision of miRNA processing is represented as the Log₂-transformed ratio of mutant imprecision/wild-type imprecision at each pri-miRNA hairpin for the following mutant/wild-type combinations; hyl1-2, hos15-2, and hda9-1 in the presence or absence of ABA. Each dot represents individual miRNAs.

As an enhancement in processing could lead to an increase of misprocessed miRNAs, we analyzed our small RNA sequencing data looking for miRNA processing imprecision. As previously reported (Liu et al. 2012), hyl1-2 mutants, used as a positive control, displayed a general and clear profile of imprecisely processed miRNAs (Fig. 7C). In contrast, hos15-2 and hda9-1 did not reveal processing imprecision in the general pool of miRNAs, indicating that the enhanced processing activity detected was not aberrant. These results support the scenario in which the absence of HOS15 and HDA9 leads to an increment in pri-miRNA processing to compensate for the enhanced MIRNA transcription caused by the mutation. This mechanism would maintain the steady levels of pri-miRNAs while producing an increased level of mature miRNAs. Taken together, our data suggest that the HOS15–HDA9 complex is a conditional suppressor of ABA–responsive pri-miRNAs processing. Whether this effect on processing is an indirect feedback loop caused by enhanced pri-miRNA transcription or a direct effect of the complex over the processing machinery by affecting HYL1 remains to be addressed.

Discussion

HYL1 was first identified as a double-stranded RNA–binding protein regulating miRNA transcription and stability (Han et al. 2004). Later, further research using HYL1 loss of function mutants revealed many biological aspects of this protein, positioning it as one of the major miRNA biogenesis components in plants. Recently, a few studies also hint at the possibility that HYL1 acts outside the miRNA biogenesis regulating miRNA-mediated translation inhibition and MIRNA transcription (Yang et al. 2021; Bielewicz et al. 2023). Interaction assays revealed that HYL1 associates with TPL, TPR1, TPR2, and TPR4, which are Lish-WD40 proteins known to repress transcription (Bielewicz et al. 2023), indicating a tight link between the miRNA processing machinery and transcription. Still, whether HYL1 has a direct effect on transcription or promotes the recruitment of transcriptional regulators remains unclear. With the increasing evidence of an early assembly of the microprocessor at MIRNA loci and the cotranscriptional processing of nascent pri-miRNAs, the transcriptional machinery may promote the association of HYL1 to active MIRNA loci, or vice versa.

Here, we showed that HOS15, another Lish-WD40 protein, interacts with HYL1 to regulate a subset of miRNAs. HYL1 also interacts with histone deacetylases HDA9 and HDA19, which are also known to interact with HOS15 and TPL–TPR complexes. The interactions of HYL1 and HDA9/19 with HOS15 and TPL–TPRs suggest that these alternative chromatin modifier complexes could regulate specific MIRNA loci. We found that HYL1 association with MIRNA loci relies on nascent pri-miRNAs and that such associations allow the recruitment of the HOS15–HDA9 chromatin modifying complex to specific MIRNA loci, especially under ABA treatment.

Interestingly, our results suggest that many of the transcriptional regulatory functions attributed to HYL1 may reflect its ability to promote the recruitment of transcription regulators to MIRNA loci by its capacity to interact both with them and with the dsRNA region of the nascent pri-miRNA hairpins. As HYL1 associates with the nascent pri-miRNA at the chromatin level, only those pri-miRNAs processed cotranscriptionally may be prone to the action of the HOS15–HDA9 complex. Given that the ratio of cotranscriptional vs. posttranscriptional processing varies between pri-miRNAs and environmental conditions (Gonzalo et al. 2022), the HYL1–dependent recruitment of the HOS15–HDA9 complex to miRNA loci may also fluctuate. Furthermore, the stem–loop secondary structure of the pri-miRNAs likely provides the signal that is recognized by HYL1 at MIRNA loci (Fig. 8). However, it is possible that other proteins of the microprocessor known to be recruited by elongator/mediator complexes such as DCL1 and HST (Fang et al. 2015; Cambiagno et al. 2021) act as a scaffold to recruit HYL1 to the loci. Taken together, our results suggest not only that the transcriptional machinery can promote miRNA biogenesis by recruiting the microprocessor to the chromatin but also that the microprocessor itself can also stabilize transcriptional regulators establishing unexpected 2-way crosstalk between both processes, as suggested for PRP40 (Stepien et al. 2022).

Figure 8.

HYL1 recruits the HOS15–HDA9 chromatin modifier complex to the MIRNA genes to regulate the loci expression. A model showing how transcriptionally inactive MIRNA loci A) can be transcribed by Pol II B) to produce a pri-miRNA transcript that folds to form a hairpin structure C). The double-stranded stem of the pri-miRNAs is recognized by the dsRNA–binding protein HYL1 and the DCL1-containing processing complex C). The association of the processing complex to the nascent pri-miRNA triggers processing and the subsequent production of mature miRNAs that are loaded into AGO1 D). In some loci, and specifically with ABA treatment, HYL1 acts as a scaffold to recruit HOS15 and HDA9 to MIRNA loci E). This interaction allows the recognition of MIRNA loci by the HOS15–HDA9 chromatin remodeling complex, which in turn will change the histone acetylation profile E). This process leads to the repression of MIRNA gene transcription, taking them to a repressive state A). Whether the HOS15–HDA9 complex could also regulate the proteostasis of HYL1 remains to be explored E).

The enhanced association of HYL1 and HOS15 with the MIRNA loci upon ABA treatment suggests the possibility that a hormonal signal is necessary for the transition from the active transcriptional state to the repressed state at some MIRNA loci. Considering that HDA9 interacts with ABA–responsive transcription factors such as ABI4 at the chromatin during ABA treatment (Baek et al. 2020; Khan et al. 2020), such transcription factors may bind ABA–responsive MIRNA loci and prime the crosstalk between the HOS15–HDA9 complex and nascent pri-miRNA-associated HYL1 under ABA signaling. Interestingly, even under ABA treatment, only a subset of miRNA appeared to be affected by HOS15–HDA9. Defining why only a subset is affected is not a trivial task.

The evidence suggests that a specificity mechanism may not even exist, but the set of miRNAs regulated by HOS15–HDA9 could change depending on the tissue or the condition. It could be expected that HOS15–HDA9 only regulate a subset of miRNAs that are in the intersection of several essential conditions: as HOS15–HDA9 requires HYL1 to recognize miRNA loci, only those tissues or conditions expressing the 3 proteins are suitable for such regulation, for example, ABA treatment. HOS15–HDA9 requires HYL1 to recognize MIRNA loci; thus, only those miRNAs that depend on HYL1 for their processing (Re et al. 2019) can be targeted by HOS15–HDA9. Similarly, as the recruitment of HYL1, and therefore HOS15–HDA9, relies on the nascent pri-miRNA, only those transcriptionally active MIRNA loci can be targeted. In the same sense, as HYL1 must associate with a nascent pri-miRNA to recruit HOS15 to the chromatin, all those MIRNA loci processed exclusively at the posttranscriptional level (Gonzalo et al. 2022) will escape HOS15–HDA9 regulation. Since whether or not a miRNA is processed cotranscriptionally depends on the environmental conditions (Gonzalo et al. 2022), a locus that could potentially recruit HYL1–HOS15–HDA9 under ABA treatment may not do so under control conditions. Finally, the HOS15–HDA9 complex acts by changing histone acetylation. Thus, not all MIRNA loci may have a suitable histone landscape for HOS15–HDA9 regulation even if recruited. Similar rules may also apply to the recruitment of other regulatory factors by HYL1 defining a specific set of targeted miRNAs depending on the tissue or environmental signaling. Given how narrow the set of MIRNA loci satisfying these requirements could be, it is not surprising that many miRNA regulatory factors regulate only a subset of miRNAs and that we do not detect a general effect in control conditions. Still, the detailed mechanism of the HOS15–HDA9–HYL1 complex that suppresses the pri-miRNA processing under ABA signaling should be further investigated in connection with the proteostasis of the microprocessor.

Interestingly, we found that hos15 and hda9 mutants accumulate more mature miRNAs under ABA treatment, even though the levels of pri-miRNAs were mostly unaltered (Figs. 5 and 6). Measuring steady levels of pri-miRNAs does not accurately reflect the transcriptional activity of pri-miRNAs because any fluctuation in activity can be easily masked by the high efficiency of the processing complex. This removes any excess of pri-miRNAs, maintaining negligible levels of full-length pri-miRNAs in most plants, except in mutants of core miRNA processing proteins. Our data showed that a deficiency of the HOS15–HDA9 complex increased MIRNA transcription, yet the enhanced processing activity maintained pri-miRNAs at near homeostatic levels by producing more miRNAs. This process likely differs from the previously reported function of the SWI2/SNF2 ATPase chromatin remodeler, BRAHMA, that unwinds the secondary structure of pri-miRNAs and subsequently suppresses miRNA biogenesis (Wang et al. 2018) and the IMITATION SWITCH (ISWI) component, FORKHEAD-ASSOCIATED DOMAIN 2 (FHA2), that directly interacts with DCL1 and HYL1 to suppress pri-miRNA processing (Park et al. 2021).

Altogether, our data indicate that HOS15–HDA9 can regulate a subset of MIRNA genes at both the transcriptional and processing levels upon interaction with HYL1 at the chromatin. However, the physiological relevance of HYL1 interaction with HOS15 and HDA9 and their subsequent effect on miRNA accumulation remains unclear. Many miRNAs regulated in a given experimental condition may not deeply impact the homeostasis of the plant. However, the mechanism of recruitment/regulation we identified is potentially mostly active in transcriptionally active genes. Thus, under ABA treatment, it is likely that all the miRNAs strongly responding to this hormone by increased transcription will be actively targeted by HYL1. Such targeting, which is critical for miRNA processing, will ultimately lead to the specific recruitment of HOS15–HDA9 leading to the repression of the genes and returning the miRNAs to homeostatic levels poststimulus. It would be interesting to define whether the repressor complex recruitment is homogeneous or whether a certain transcription threshold has to be reached to trigger its recruitment. In this sense, a high transcription rate coupled with cotranscriptional processing of a pri-miRNA may elevate the amount of HYL1 at a given MIRNA locus, increasing the chances that HOS15–HDA9 recognize and decrease its expression via a buffering mechanism. The fact that both the speed of transcription and the frequency of cotranscriptional processing change under stress may cause a more favorable landscape for the association of HYL1 to the MIRNA loci, explaining our results under the hormone treatment.

Materials and methods

Plant materials and growth conditions

The hos15-2, hda9-1, and hyl1-2 T-DNA insertion mutants and the transgenic 35S_pro:HYL1-6MYC line used in this study are previously described (Park et al. 2019; Jung et al. 2022). The hos15-3 mutant is an EMS-induced mutant in the luciferase reporter system transformed into Col-0 plants (Manavella et al. 2012). hos15^crispr was generated using the CRISPR-Cas9 system (Wu et al. 2018). All plants used for the analysis were 12-d-old seedlings grown on ½× Murashige and Skoog (MS) media containing 1% (w/v) sucrose, pH 5.7, under a long-day photoperiod (16 h light/8 h dark) at 22 °C and 100 to 120 μmol m⁻² s⁻¹ light intensity. For the miRNA and mRNA sequencing experiments, 12-d-old seedlings were sprayed with 50 µm ABA for 3 h. All treatments were in liquid ½× MS media for co-IP, small RNA blot, RT-qPCR, and ChIP experiments. The concentration and time of the treatment were specified for each experiment. For the amanitin treatment, 14-d-old seedlings were rooted up and soaked into ½× Murashige and Skoog medium (Duchefa Biochemie) or ½× MS medium containing 10 μm α-amanitin (Sigma-Aldrich). Seedlings were harvested after 3 h of incubation. Total RNA was extracted using the RNeasy Plant Mini Kit (Qiagen, Hilden, Germany) including the DNase I treatment step and was used for RT-qPCR analyses.

Yeast 2-hybrid assay

For the yeast 2-hybrid experiments, all genes in this study were initially cloned in pDONR/Zeo (Invitrogen) using gateway BP reaction (Invitrogen) and later transformed into the yeast 2-hybrid destination vectors pDEST22 and pDEST32 (Invitrogen) as activation and DNA–binding domain (AD and BD, respectively) using the LR reaction following the manufacturer's protocol. Later, the AD and BD plasmids were transformed into Saccharomyces cerevisiae (YRG2) strain. Protein–protein interactions were determined by monitoring the growth of yeast colonies on SD/-Trp-Leu (Sc-TL) or SD/-Trp-Leu-His (Sc-TLH; Takara Bio, Kusatsu, Japan) agar media containing 3-amino-1,2,4-triazole (3-AT; 25 mm). An empty vector (EV) was used as a negative control.

Coimmunoprecipitation assays

For the coimmunoprecipitation assays (co-IP), we used 3-w-old Nicotiana benthamiana leaves which were coinfiltrated with Agrobacterium tumefaciens (GV3101 strain) expressing the plasmid of interest using combinations of 35S_pro:HOS15-3xHA, 35S_pro:HYL1-FLAG, 35S_pro:SE-FLAG, 35S_pro:HEN1-GFP, 35S_pro:DCL1-GFP, 35S_pro:HDA9-3xHA, and 35S_pro:HDA19-HA. Total protein was extracted using IP buffer (100 mm Tris-Cl, pH 7.5, 150 mm NaCl, 1 mm EDTA, 0.5% NP-40, 3 mm DTT, and protease inhibitor cocktail supplemented with 1 mm PMSF) and used for immunoprecipitation using the required antibody depending on the experiment after aliquoting the input protein. The immunoprecipitated (IP) protein and input were mixed with sample buffer and loaded to 10% SDS–PAGE, and the immunoblots were treated with the antibody of concern and developed using ECL solution.

Chromatin immunoprecipitation assay

The ChIP analyses were performed as described in Komar et al. 2016 with a slight modification. Nuclei from 1.5 g of 12-d-old seedlings were extracted accordingly after crosslinking with a 1% formaldehyde solution using 2 vacuum cycles of 10 min each followed by glycine treatment for 5 min under vacuum to stop the crosslinking reaction. Immunoprecipitations were performed using the magnetic beads (BIORAD Cat. # 161-4011) incubated with the antibody indicated in each experiment. We used normal IgG from different species according to the antibody host species used as a negative control to calculate the fold enrichment compared to the IgG. The IP DNA was purified and quantified using RT-qPCR.

For the ChIP-seq analysis, we used publicly available bigwig files (GSE145208). Further, we generated bed files for upregulated, downregulated miRNA in the miRNA-seq analysis along with total miRNA genes, and generated a matrix and heatmaps with the bigwig files using bedtools (Quinlan and Hall 2010).

RNA and miRNA analysis

Total RNA was extracted from plants (harvest timing is described in each experiment) with the RNeasy Plant Mini Kit (Qiagen) and the Plant total RNA extraction kit (Xenohelix, Incheon, Korea) and then treated with DNase (Sigma, St. Louis, MO, USA). Briefly, 2 µg RNAs were used for the synthesis of first-strand cDNA using the Thermoscript RT-PCR System (Invitrogen, Carlsbad, CA, USA). Quantitative PCR was performed using the SYBR Green PCR Master Mix kit (SYBR Green Supermix; Bio-Rad, Hercules, CA, USA) according to the instructions and the CFX96 real-time PCR detection system (Bio-Rad). The expression of TUBULIN8 (AT5G23860) or UBQ10 (AT4G05320) was used as the endogenous control. The RT-qPCR experiments were performed using 3 independent replicates.

For the sRNA-seq analysis, raw sRNA-seq reads were quality checked by FastQC and trimmed to remove 3′ adapter sequences using skewer v.0.2.2; reads of length 18 to 30 nt were kept. Trimmed reads were aligned to the Arabidopsis TAIR10 genome using ShortStack v.3.8.5. (Johnson et al. 2016). Overlaps were counted using the feature Count function of the subread package. For plotting, miRNA counts were normalized by the number of mapped reads in each library using the TPM method. Differential expression analysis was carried out with DESeq2 using unnormalized read counts, and the Wald test was used for statistical testing. P-values were adjusted with the Benjamini–Hochberg method.

For the RNA-seq analysis, RNA-seq libraries were prepared with TruSeq Stranded mRNA Sample Preparation Kit and paired-end sequenced at 75 bp read length with an Illumina NextSeq platform. Reads were quality checked with FastQC, and adapters were removed with skewer v.0.2.2. Reads were aligned using STAR v.2.6 to the Arabidopsis TAIR10 genome. Quantification was performed with the Cufflinks package v.2.2.1. RNA and miRNA sequencing information is found in Supplemental Data Sets 2, 3, 5 to 7.

Small RNA blot analysis

Small RNA blot analyses were performed as described previously (Choi et al. 2020; Park et al. 2021). Briefly, total RNA was extracted from 14-d-old seedlings of Col-0, hyl1-2, hos15-2, and hda9-1 using the TRIzol reagent (Invitrogen). The aqueous phase of the extract was precipitated with isopropanol for 20 min at −20 °C. Precipitates were washed twice with 80% ethanol and dissolved in 50% formamide. The concentration of the purified RNA was measured using a Nanodrop (NanoDrop One Microvolume UV-Vis Spectrophotometer; Thermo Scientific, Waltham, MA, USA). Then, 15 μg of total RNA was separated in a 13% denaturing urea–PAGE gel and transferred to a nylon membrane (Amersham Hybond-XL Membranes; Cytiva, Marlborough, MA, USA). The membranes were hybridized overnight using 5′-end ³²P-labeled antisense DNA oligonucleotides as probes. T4 polynucleotide kinase (New England Biolabs, Ipswich, MA, USA) was used for the 5′-end labeling of DNA oligonucleotides with [γ⁻³²P]ATP (PerkinElmer, Waltham, MA, USA). ULTRAhyb Ultrasensitive Hybridization Buffer (Ambion, Austin, TX, USA) was used for the hybridization of the membrane. Blots were washed twice with the washing buffer (2× sodium chloride–sodium phosphate–EDTA [pH 7.4] and 0.1% SDS) for 20 min each. ³²P-signals were detected using a phosphor-imager Typhoon scanner (Amersham, IP Biomolecular Imager, Cytiva). Alternatively, miRNAs were detected by nonradioactive blots as previously described (Tomassi et al. 2017).

In vitro pri-miRNA processing assay

pri-miRNA sequences were obtained from miREX2.0 (Zielezinski et al. 2015). 5′-end T7 promoter sequence containing pri-miRNA sequences were PCR amplified and cloned into pTop V2 (Enzynomics, Daejeon, Korea). The plasmids were digested with the appropriate restriction enzymes and agarose gel purified for in vitro transcription, performed using the MEGAscript T7 Transcription Kit (Ambion) following the manufacturer's instructions. Transcripts were internally labeled with [α-³²P]UTP (PerkinElmer) during the in vitro transcription. In vitro-transcribed RNA was purified using a Xenopure Nucleotide Removal Kit (Xenohelix) and then used for the in vitro pri-miRNA processing assay, performed as described previously (Zhu et al. 2013) with a few modifications. In vitro transcribed pri-miRNA was folded by heating for 2 min at 95 °C and cooling for 1 h at room temperature. In vitro pri-miRNA processing assays were carried out in a total volume of 20 μL with in vitro pri-miRNA processing buffer (20 mm Tris-HCl pH 7.4, 53 mm KCl, 4 mm MgCl₂, 1 mm DTT, 7.5 mm ATP, and 1 mm GTP), including the [α-³²P]UTP–labeled pri-miRNA transcripts (20,000 cpm). The nuclear fraction was mixed with the reaction mixture and incubated at 37 °C for 2 h. The processed RNA products were purified using the Xenopure Nucleotide Removal Kit (Xenohelix) and separated in a 13% denaturing urea-PAGE gel. The gel was dried using a gel dryer (GD 2000 Vacuum Gel-Drying System; Hoefer, Holiston, MA, USA) at 60 °C for 1.5 h. ³²P-signals were detected using a phosphor-imager scanner (Amersham, Cytiva).

Subcellular fractionation

Sucrose gradient subcellular fractionation was performed as described previously (Xu et al. 2012) with a few modifications. Briefly, frozen ground samples of 14-d-old Col-0, hos15 to 2, and hda9-1 were homogenized in 3 mL of lysis buffer (20 mm Tris-HCl pH 7.4, 20 mm KCl, 2.5 mm MgCl₂, 2 mm EDTA, 250 mm sucrose, and 25% [v/v] glycerol) supplemented with 1 mm PMSF, 5 mm 1,4-dithiothreitol (DTT; Roche, Basel, Switzerland), and EDTA-free protease inhibitor cocktail (Roche). The lysate was sequentially filtered through 2 layers of mesh nylon filter (SHA NY2030, OHKI) and Miracloth (Millipore, Burlington, MA, USA) and then centrifuged at 1,000 g for 10 min at 4 °C. The supernatant was transferred into new tubes and further centrifuged at 20,000 g for 30 min at 4 °C to obtain the cytoplasmic fraction (Supernatant). The pellet after the first centrifugation was gently washed 4 times with 5 mL of nuclei resuspension buffer (NRBT, 20 mm Tris-Hcl pH 7.4, 2.5 mm MgCl₂, 250 mm sucrose, 25% [v/v] glycerol, and 0.2% [v/v] Triton X-100). Washed pellets were resuspended in 500 μL of NRB2 buffer (20 mm Tris-HCl pH 7.4, 10 mm MgCl₂, 250 mm sucrose, and 0.5% Triton X-100) and supplemented with protease inhibitor cocktail and 5 mm 2-mercaptoethanol (2-ME, Sigma-Aldrich). The suspension was layered on the top of 500 μL of NRB3 buffer (20 mm Tris-HCl pH 7.4, 10 mm MgCl₂, 1.7 m sucrose, and 0.5% [v/v] Triton X-100), supplemented with protease inhibitor cocktail and 5 mm 2-ME, and centrifuged at 16,000 g for 45 min at 4 °C. The supernatant was removed, and the pellet was lysed with nuclei lysis buffer (NLB, 50 mm Tris-HCl pH 7.4, 10 mm EDTA, 0.05% [v/v] SDS, 0.1% Triton X-100) supplemented with protease inhibitor cocktail to obtain the nuclear fraction. The subcellular fractionation was confirmed using immunoblot using the nuclear marker α-histone H3 (1:20,000, AS10 710; Agrisera, Vännäs, Sweden). The nuclear fraction was used for the in vitro pri-miRNA processing assay.

Confocal microscopy

Subcellular localization and colocalization of HOS15 and miRNA biogenesis proteins were observed in transgenic A. thaliana and transiently transformed N. benthamiana plants, respectively. GFP and mCherry were excited at 480 nm or 552 nm, and emission was collected at 503 to 531 nm or 602 to 630 nm, respectively. All assays were performed on a TCS SP8 confocal microscope (Leica, Solms, Germany).

Accession numbers

HOS15, AT5G67320; HDA9, AT3G44680; and HYL1, AT1G09700. Sequence data from this article can be found in the GenBank/EMBL data libraries under accession numbers GSE145208 and GSE145208. The RNA-seq and sRNA-seq data from this article can be found in the SRA database under accession number PRJNA956989.

Author contributions

A.J.G, J.P, D.-J.Y, and P.A.M conceived and designed the study. A.J.G, Z.E.B, M.K.H, C.M, J.P, and Z.-Y.X performed all the experiments and validations; Z.E.B, C.M, and A.L.A analyzed the sequencing data; P.A.M, S.W.Y, and D.-J.Y supervised the work and secured project funding; Z.E.B, P.A.M, S.W.Y, and D.-J.Y wrote the manuscript with input from all authors.

Supplemental data

The following materials are available in the online version of this article.

Supplemental Figure S1. Characterization of the hos15 mutant.

Supplemental Figure S2. HYL1 accumulation in HOS15 and HDA9 mutants.

Supplemental Figure S3. HDA9 association with MIRNA loci.

Supplemental Figure S4. HYL1 and HOS15 coassociation with MIRNA loci.

Supplemental Figure S5. miRNA expression under control conditions and ABA treatment.

Supplemental Figure S6. pri-miRNA transcript levels are not drastically changed in hos15 and hda9 mutants upon ABA treatment.

Supplemental Figure S7. Enhanced pri-miRNA processing in hos15 and rhda9 mutants.

Supplemental Data Set 1. Col-0, hos15, hda9, and hyl1 miRNA_target gene_expression, both sets.

Supplemental Data Set 2A. hos15, hda9, and hyl1 mRNA-seq_differential gene expression, set 1.

Supplemental Data Set 2B. hos15, hda9, and hyl1 mRNA-seq differential gene expression set 2.

Supplemental Data Set 3. HOS15 miRNAome.

Supplemental Data Set 4. Col-0 miRNA expression following ABA treatment.

Supplemental Data Set 5. Col-0, hos15, hda9, and hyl1 small RNA sequencing analysis.

Supplemental Data Set 6. hos15, hda9, and hyl1_pri-miRNA_tptm_bothsets.

Supplemental Data Set 7. hos15, hda9, hyl1 small RNA sequencing, and pri-miRNA correlations.

Supplemental Data Set 8. Statistical analysis results.

Supplemental Data Set 9. Primer list.

Funding

This work was supported by grants from the National Research Foundation of Korea (NRF) funded by the Korean Government (2022R1A2C3004098 to D.-J.Y. and NRF-2020R1A2B5B01002592 and NRF-2018R1A6A1A03025607 to S.W.Y). This work was also supported by a grant from ANPCyT (PICT2020-SERIEA-00757, Agencia Nacional de Promoción Científica y Tecnológica, Argentina). P.A.M. and A.L.A. are members of CONICET; A.J.G is a fellow of the same institution. This research was supported in part by the Brain Korea 21 (BK21) FOUR (Fostering Outstanding Universities for Research) program from the Ministry of Education (MOE, Korea). M.K.H. and C.M. are recipients of a fellowship awarded by the BK21 FOUR program.