In all experiments, Arabidopsis ecotype Columbia was used as the wild-type (WT) control. All plant seeds were germinated on MS medium supplemented with agar (1%) and sucrose (3%) at pH 5.8. All plants were grown in soil at 22°C with a 16 h:8 h, light:dark photoperiod.

For the rrp42 null mutant, one sgRNA target (C1: CCAACAGCTGAACCGACATTTGG) in RRP42 gene was selected and cloned into the pHSN401 (Xing et al., 2014). For the largest possibility of getting the non-mosaic mutants, we cloned target C1 into the pHEE401 vector as described by Wang et al. (2015) later. In addition, we also selected another gRNA target (C2: AGTTCACTTCAACCCGATAAAGG) in RRP42 gene, and generated another pHEE401 vector with two gRNA expression cassettes targeting the two adjacent sites (C1 and C2) of RRP42 gene (Wang et al., 2015). The construct was transformed into WT plants by the floral dip method (Clough and Bent, 1998). The putative transformants were screened on MS plates contained with 25 μg ml^-1 hygromycin B. To detect mutation on targeted sequence, the genomic DNA was isolated from rosette leaves of about 20-day-old T1 transformants. For the sequence analysis of target C1, a 526 bp genomic DNA region containing the target site was amplified by PCR using the primers 42LP (5′-GGCTCTAGGCTAATGGTTCAG-3′) and 42RP (5′-CTGCTCCACTTTTGCCACCCA-3′). We used restriction endonuclease PvuII to digest PCR products for primary screens and obtained a few candidate lines for sequence analysis. For the target C2, we sequenced it directly.

For the rrp42 knock-down mutants, we designed three different amiRNAs (amiRNA1: TTTCGTTTGGTTAACCCGCAT; amiRNA2: TTTCGTTTGGTTAACCGACAT; amiRNA3: TATATAGATACAGCTGCGCTC) to knockdown the expression levels of RRP42 in Arabidopsis using WMD (web microRNA designer)¹. Then, according the sequence of amiRNAs designed, we acquired the corresponding primers (I–IV) of amiRNA1, amiRNA2, and amiRNA3 (Supplementary Table S1). The amiRNA foldback fragments were generated using the miR319a vector as a template and the following primers for amplification: A:5′AATTATCTAGAACACACGCTCGGACGCAT-3′. B:5′-AATTAATCCCATGGCGATGCCTT-3′. Each amiRNA corres-ponding primers was designed using the Web MicroRNA Designer 3 oligo design algorithm, and then ligated into pMDC99-32A vectors which harbor a dual 35S promoter. Detailed information regarding the use of overlapping PCR and each primer set is available on the Web MicroRNA Designer 3 web-site. Transformation was performed as described above.

For the observation of ovules, inflorescences were harvested and fixed in 4% glutaraldehyde (in 12.5 mM cacodylate, pH 6.9), and a vacuum was applied for the initial 1 h, after which they were in fixative overnight at room temperature. Then, the inflorescences was dehydrated through a common ethanol series with 30 min per step. After the dehydration, the dehydrated inflorescences was cleared in 1:2 (v/v) benzyl alcohol:benzyl benzoate overnight at room temperature. The dissected pistils were mounted with immersion oil, and observed using a Zeiss LSM710 META confocal laser scanning microscope with a 488 nm argon laser.

For DAPI staining of pollen grains, pollen grains from dehisced anthers were dissected and stained in DAPI solution (PIPES 50 mM, DAPI 5 mg/ml, DMSO 10%, EGTA 5 mM, and NP-40 0.1%) and incubated for 15 min. And then, the stained pollen grains were observed on a Leica HQ stereomicroscope equipped with a 40× optic.

The leaf serration was quantified according to Hasson et al. (2011). We took the leaves of the plants and took pictures of them, and then we used the Digimizer software to measure the leaf area.

Chlorophyll was isolated from the leaves and measured according to a previously described method (Grbic and Bleecker, 1995). Extracts were obtained from 100 mg of fresh tissue from the first and second cauline leaves from 42-day-old plants and immerged in 2 mL of 80% (v/v) acetone overnight at room temperature. Chlorophyll was measured by Elisa (Power Wave XS2). Chlorophyll content analysis were repeated at least in three independent experiments.

Total RNA was extracted according to the protocol of Oñate-Sánchez and Vicente-Carbajosa (2008). One microgram of RNA was used as the template to produce cDNA with the TaKaRa oligo(dT) primer and M-MLV reverse transcriptase. We used the SYBR Premix Ex Taq (TaKaRa, Japan) and the ABI 7500 Real Time PCR to perform quantitative real-time PCR (qRT-PCR). AT4G34270 was used as an internal control. The qRT-PCR analysis used two technical replicates and three biological replicates. The primers used in qRT-PCR are shown in Supplementary Table S2.

The density of mesophyll cells and chlorophyll fluorescence analysis were observed using a Leica HQ and Zeiss 510 META confocal laser scanning microscope, respectively. Trypan blue staining was performed as described in Koch and Slusarenko (1990). For leaf anatomy light microscopy, the first cauline leaf of 6-week-old WT, a42-1 and a42-3 plants were fixed in formaldehyde/glutaraldehyde fixative (1% [v/v] glutaraldehyde and 4% [w/v] paraformaldehyde in 0.05 mol L^-1 phosphate buffer, pH 7.2). Then, the tissues were dehydrated in ethanol and embedded in Spurr’s resin (SPI-CHEM). Thin sections were cut on a microtome (Leica EM UC7), and observed using light microscopy (ZEISS Scope A1). Materials were also sliced to perform the experiment of ultrathin sections (LKB-8800). Before being examined with a transmission electron microscope (JEM-123O), they were stained with alkaline lead citrate and uranyl acetate. For scanning electron microscopy, plant material was prepared according to a previously described method (Serrano-Cartagena et al., 2000). We used a scanning electron microscope (Hitachi S-3400N) to take micrographs.

RNA decay assays in WT, a42-1 and a42-3 mutants were measured in rosette leaves of 6-week-old plants. This protocol was adapted from a previous report from Johnson et al. (2000). Leaves were incubated in buffer (1 mM KCl, 15 mM sucrose, 1 mM sodium citrate, 1 mM Pipes, pH 6.5) for 30 min before addition of cordycepin (150 mg ml^-1, Sigma). About 30 s vacuum was applied to the all samples. Four leaves were frozen in liquid nitrogen every 15 min for 45 min, and then stored at -80°C for the use of RNA isolation and qRT-PCR.

To express the RRP42-GFP fusion protein under the control of the 35S promoter, the full-length coding sequence except the stop codon was cloned from the cDNA of WT plants using the primers: GFP-LP (5′-TCTAGAATGGGGCTTTCTCTTGGGGA-3′) and GFP-RP (5′-GGTACCAGATTCGTCTTCGCAGGCCT-3′). The PCR products were fused to the pSuper 1300-GFP vector. The GFP is tagged at its C terminus. The protoplast extraction and plasmid transformation procedures were performed according to a previously described method (Kim and Somers, 2010). We used the floral dip method to obtain the stable Arabidopsis transformants, and putative transgenic plants were screened on MS plates containing 25 mg L^-1 hygromycin. We used a Zeiss 510 META confocal laser scanning microscope to observe the GFP fluorescence of the transgenic plants and the RFP and GFP fluorescence of the transgenic protoplasts. To confirm that the localization of RRP42 was unaffected by GFP fusion, the 35S:RRP42 vector was transformed into the WT plants. The 35S:RRP42 stable Arabidopsis transformants displayed no obvious phenotype with the 35S:RRP42:GFP stable Arabidopsis transformants.

For the GUS staining, the RRP42 promoter:GUS gene was constructed using PCR amplification of a fragment 1533 bp upstream from the initial codon of RRP42. We used the following primers for amplification: PRO-LP (5′-CCCAAGCTTATGTGTTTTCATCAGTCCTTACCG-3′) and PRO-RP (5′-GCGTCGACCACTAATACTACACAGAGAACG-3′); the fragment was then cloned into a pCAMBIA 1391 vector. Transformation was performed as described above. Ovules for microscopy were performed as described by Wu et al. (2006). The materials were observed using a microscope (Olympus SZX16-DP72), and a Canon digital camera (PowerShot G12) recorded the digital images.

To investigate the role of exosome component RRP42 (AT3G07750) in Arabidopsis development, we used an egg cell-specific promoter-controlled (EPC) CRISPR/Cas9 system to acquire rrp42 mutant (Wang et al., 2015). The pHEE401-42 construct with one gRNA was transformed into WT plants, only three non-mosaic mutants were obtained of nearly 300 T1 lines. We chose one line (named as pHEE42-1) which was identified as heterozygotes for further study. In the heterozygote rrp42/RRP42, a G was inserted into the Cas9 editing targets, which were located at 255 and 256 bp in the coding regions of RRP42, leading to reading frame shifts (Figures 1A,B). To eliminate the influence of the Cas9 gene, we isolated rrp42/RRP42 on MS plates containing hygromycin (Figure 1C). The non-resistant plants were used for further study. RRP42 was essential for development of the female gametophyte. The selfed heterozygote rrp42/RRP42 produced seeds and aborted ovules in a 1:1 (252:265) ratio (Figure 1D; three non-mosaic mutants showed identical phenotypes), and in the progeny of selfed heterozygote rrp42/RRP42, the proportion of heterozygous and WT plants was almost 1:1 (50:47). We also found that the rrp42 mutant allele was normally transmitted through the male parent, but couldn’t transmitted through the female (Figure 1E). Furthermore, according to the previous description of Smyth et al. (1990), most of embryo sacs were at the four-celled stage at flower developmental stage 14 in WT pistils. While the rrp42/RRP42 female gametophytes arrested (n = 134) after the first mitosis (two-nucleate stage, 46.3%; Figure 1F) and less frequently at one-nucleate (2.2%), four-nucleate (8.9%). Additionally, there was no obvious morphological difference in rrp42/RRP42 and WT pollen grains (Supplementary Figure S1A). These observations suggest that the synchrony of female gametophyte development was impaired in rrp42 pistils.

We also transformed the construct pHSN42 with one gRNA and pHEE-42 with two gRNA into WT plants, respectively. The heterozygote rrp42/RRP42 from pHSN42 construct transformation was named as pHSN42-1 while another heterozygote from the transformation of pHEE-42 construct with two gRNA was named as pHEE-2g-42-1. They showed the identical phenotype as that in pHEE42-1 heterozygote. In these three heterozygotes rrp42/RRP42: pHEE-42-1, pHSN42-1, and pHEE-2g-42-1, the mutational pattern were different with each other (Figure 1B and Supplementary Figures S1B,C). These data indicated that the null mutation of rrp42 is lethal and RRP42 has an essential role for development of the female gametophyte in Arabidopsis.

To address the functions of RRP42 during vegetative growth, we designed three different amiRNAs targeting RRP42 using web microRNA designer² (Figure 2A). We obtained at least 50 T1 lines with similar phenotypes. Next, we chose one line from the lines targeted by amiRNA1, amiRNA2 and amiRNA3, respectively, and referred to as a42-1, a42-2, and a42-3 for the next step of analysis. Knock-down of RRP42 expression induced various morphological abnormalities, which were specifically evident on both the cauline and rosette leaves (Figures 2D–F). The abnormal phenotypes of rrp42 knock-down mutants could be grouped into three classes according to the severity of the abnormalities. The levels of RRP42 in a42-1 with a severe defect decreased to about 10% of the WT plant, while levels in a42-3 with a mild defect were about 70% of WT (Figure 2B). Thus, the expression levels of RRP42 were consistent with the severity of rrp42 knock-down mutant defects. These results indicated that the defects of rrp42 knock-down mutants were induced by the repressive activity of RRP42.

We also tested the expression levels of MRP and snoRNA31, which were the known nucleus target RNAs of exosome and accumulated in rrp4^iRNAi, rrp41^iRNAi, and rrp44a mutant (Chekanova et al., 2007; Kumakura et al., 2013). The levels of MRP and snoRNA31 were significantly increased in different degrees in three mutants (Figure 2C). The most accumulation of both RNAs was in a42-1, while the least accumulation was in a42-3. These data further show that the phenotype is the result of RRP42 downregulation.

Leaf morphologies of rrp42 knock-down mutant are shown in Figure 2. The cotyledons of rrp42 knock-down seedlings displayed wrinkled surface and a reduced leaf size (Figures 2D,G), and the root growth was no obviously affected in mutant seedlings (Supplementary Figure S2A). The most observable defects in rrp42 knock-down mutant were the variegated and serrated cauline leaves of older plants (Figures 2F,J). These defects were also visible in rosette leaves (Figures 2E,I). The a42-1 mutant with a severe defect was serrated and distorted beginning at the seventh or eighth rosette leaf, leaf size in a42-1 mutant was also reduced (Figures 2I,K). The a42-2 mutant with a moderate defect was beginning at the eleventh or twelfth rosette leaf (Figure 2I), while the a42-3 with a mild defect was almost normal except for one or two weakly variegated cauline leaves (Figure 2F). The abnormal phenotypes of rrp42 knock-down mutant become more severe at later growth stages. After bolting, the stem of a42-1 was a little twisted compared with WT, and the final height of a42-1 was below that of the WT plant (Figure 2H). These results suggested that knock-down of RRP42 seriously affected the plant growth and leaf development.

Scanning electron microscopy showed that the surface of the mutant leaves was wrinkled, extremely so in the case of the a42-1, whose lamina was completely crumpled (Figure 3A). Nevertheless, no obvious differences compared to the WT were observed for the size and morphology of the a42-1 adaxial and abaxial epidermal cells (Figures 3B,C). We also analyzed internal leaf anatomy by means of cross sections, a42-1 and a42-2 leaves displayed disruptions in cell arrangement and morphology, as well as the overall leaf shape. There were larger air spaces between cells from mutants compared to the WT and the area of palisade cell wall surface was larger than that in WT leaves because of the contraction in the middle position of mutant cells (Figure 3D). According to the microscopic and chlorophyll fluorescence analysis of mesophyll cells of the first and second cauline leaves, we found a reduced density of mesophyll cells in a42-2 (Figures 4A–C). However, we found no significant reduction of chlorophyll content in the first and second cauline leaves from 6-week-old a42-1 or a42-2 mutant plants as compared with WT plants (Figure 4D). We did not observe increased cell death in rrp42 knock-down mutants by trypan blue staining (Supplementary Figure S3), therefore, we concluded that the increased air spaces and irregular cell morphologies were responsible for the variegation and the distorted surface observed in rrp42 knock-down mutant leaves.

We also examined leaf chloroplast ultrastructure in the first variegated cauline leaf of a42-1 and WT by transmission electron microscopy and found that chloroplasts in a42-1 exhibited reduced starch grains, but they were similar in numbers, size, and morphology to those of WT (Supplementary Figures S4A–D). Only a few chloroplasts in the a42-1 mutant displayed enlarged thylakoid lamellas (Supplementary Figure S4E), a trait never observed in WT.

We also generated a construct harboring AT3G07750 the full-length coding sequence under the control of 35S promoter and the GFP was tagged at its C terminus. At least ten homozygous transgenic over-expression (OE) lines were obtained, and RRP42 transcript levels were measured using qRT-PCR (Supplementary Figure S2B). Under normal conditions, the morphologies of OE1 and OE2 displayed no obvious difference with WT plants.

To investigate the spatial and temporal expression patterns of RRP42 in plant tissue, a vector in which a 1533-bp promoter fragment of RRP42 was fused with the GUS gene was constructed. Seven independent transgenic lines were obtained, and at least four lines were detected. GUS activity was observed in all of the organs. We found that RRP42 was strongly expressed in the leaves (Figures 5B–F). It also had a high expression in the ovules (Figure 5A). The qPCR analysis also showed that RRP42 were expressed at higher levels in the cauline leaf (Figure 5F). A similar expression pattern was observed in four independent lines. These data were consistent with the results that the RRP42 may play an important role in female gametophytes and leaf development in Arabidopsis.

To investigate the intracellular localization of RRP42, we constructed an RRP42-GFP fusion protein expressed under the control of the cauliflower mosaic virus 35S promoter. RRP42-GFP and red fluorescent protein (RFP)-AHL22, an AT -rich DNA sequence (AT)-hook motif nucleus-localized protein (Xiao et al., 2009), were transiently coexpressed in living Arabidopsis protoplasts. In contrast to the nuclear distribution of RFP-AHL22, we found that the RRP42-GFP fusion protein was localized to both the cytoplasm and nucleus (Figure 5G). Consistent with this, the localization of RRP42-GFP was also in the cytoplasm and nucleus in the 9-day-old stable Arabidopsis transformants root cells (Figure 5H).

The phenotypic defects of rrp42 knock-down mutants prompted us to use qRT-PCR to measure the level of transcripts that encode proteins participating in photosynthesis, starch synthesis, cell wall assembly, and leaf morphogenesis. The 42 days rosette leaves of WT, a42-1 and a42-3 were used to perform the analysis. We found that expression of the XTHs and EXPAs, both involved in cell wall assembly, were at least two-fold higher in a42-1. The expression of these mRNAs were almost unchanged or a little higher in a42-3 with a mild phenotype (Figure 6A). Most noticeably of all, the expression of XTH19, EXPA10, and EXPA11 were up to about 20-fold in a42-1 compared with WT (Figure 6A). Additionally, the expression of the genes related to photosynthesis and starch synthesis were almost normal in the two mutants as compared with WT (Table 1). While the expression of CUC1, a gene involved in leaf blade margin, was increased about six-fold in a42-1 (Table 1). The expression levels of CUC2 and CUC3 were nearly unchanged among these mutants.

To investigate the putative function of RRP42 in the degredation of XTH and EXPA mRNAs, three independent experiments were performed in which excised rosette leaves from a42-1, a42-3, and WT plants were incubated with cordycepin, a compound that strongly blocked the process of transcription (Zhang et al., 2010). We tested XTH19, EXPA10 and EXPA11 which were high accumulated in mutant plants. The decay of XTH19, EXPA10, and EXPA11 mRNAs in a42-1 leaves was clearly slower than in WT leaves. As expected, the differences in decay kinetics are evident at the 15 and 30 min point (Figures 6B–D), suggesting that RRP42 has a role in the decay of these mRNAs. We also tested the mRNA decay of EXPL1 (expansion-like 1) which has a comparable half life as control (Xu et al., 2006; Xu and Chua, 2009). There were no obviously differences at each point in mutants and WT plants (Supplementary Figure S5). The decay of XTH19, EXPA10, and EXPA11 mRNAs was not completely arrested in the two mutants; this could be the consequence of the residual function of RRP42 in the rrp42 knock-down mutant. Alternatively, RRP42 may not be the only pathway for the decay of these genes at the developmental state tested. In a42-3 mutant, the rate of these mRNA decay was nearly same or only a little slower compared with WT plant (Figures 6B–D), this is consistent with that morphology of a42-3 rosette leaves was almost normal.

Additionally, to investigate whether the knock-down of RRP42 could affect the expression levels of mRNAs up-regulated in other exosome subunit mutants, we randomly chose a number of mRNAs up-regulated in the rrp41 and rrp4 mutants, and examined their transcript levels in a42-1 and a42-3 by qRT-PCR. The results showed that these mRNA levels were almost equal to WT plants. We also examined the expression levels of mRNAs up-regulated in rrp41l. As expected, only slight increases or decreases occurred in these mRNA levels compared with WT plants (Supplementary Table S3). Conversely, we also tested the expression of XTH7, XTH18, XTH19, EXPA3, EXPA10, and EXPA11 in rrp41l. We found that their expression level were basically the same as that of the WT (Supplementary Table S4).