T-DNA insertion lines (atlsg1-2: Salk_114083 and atnmd3: WiscDsLox257G09) were obtained from Arabidopsis Biological Research Center. Wild-type (Accession Columbia-0) and mutant seeds were surface-sterilized in 50% bleach solution for 5 min and rinsed with water five times. The sterilized seeds were sown on agar-solidified half-strength Murashige and Skoog medium and incubated at 4°C for 3 days before being transferred to a growth chamber at 22°C with a 16-h-light/8-h-dark photoperiod. For soil-grown plants, 12-day-old seedlings growing on a petri dish were transferred to soil and grown in a growth room at 22°C with a 16-h-light/8-h-dark photoperiod.

For T-DNA line WiscDsLox257G09, the mutant allele was verified by PCR genotyping and by sequencing the PCR products. T-DNA insertion was detected using the left border primer (LB) combined with gene- specific primers LP and RP. Primer sequences are given in Supplementary Table S2.

Leaf size, cell number, and size of abaxial epidermis were measured in the fifth leaf of 4-week-old plants growing in soil. A small droplet of superglue was applied on the glass slide. The leaf abaxial side was tightly imprinted on the slide coated with superglue for 30 s and the leaf was then quickly removed. The imprinted slides were observed and photographed with a microscope (Axio Imager Z2). Data from 5 to 8 leaves were used for statistical analysis.

To generate promoter deletion fusion constructs, different promoter deletions were cloned into the pENTRTM/D-TOPO vector using pENTRTM directional TOPO^® cloning kit (Invitrogen) and then subcloned into the binary vector pMDC162 by the LR recombination reaction. The plasmids were transferred into Agrobacterium tumefaciens and Arabidopsis plants were transformed by the floral dipping method (Clough and Bent, 1998).

Leaf kinematic analysis was performed essentially as described by De veylder et al. (2001) with some modifications. Leaves were submerged in 10 μM FM4-64 for 10 min and destained in water. The epidermal cells of stained leaf abaxial side were observed and photographed with a microscope (Axio Imager Z2). The experiment was repeated three times with similar results obtained. Results from only one set of the experiment were presented in this study.

Whole seedlings were incubated in ice-cold 90% acetone for 2 h, washed in 100 mM Na₃PO₄ (pH 7.0) and subsequently immersed in 5-bromo-4-chloro-3-indolyl-β-D-glucuronide (X-Gluc) buffer (100 mM Na₃PO₄ buffer, pH 7.0, 10 mM Tris, pH 8.0, 1 mM EDTA, 0.05% Triton X-100, 1 mg/ml X-Gluc) at 37°C overnight. Chlorophyll was cleared in 70% ethanol. The cleared samples were photographed with a Nikon SMZ25 stereomicroscope.

Preparation of plant materials for flow cytometric assays was performed as described in Dolezel et al. (2007). In brief, leaves were quickly chopped with a razor blade in ice-cold Galbraith’s buffer [45 mM MgCl₂, 20 mM MOPS, 30 mM sodium citrate, and 0.1% (vol/vol) Triton X-100, pH 7.0]. The homogenate was filtered through 70 μM nylon mesh. Fifty microgram per milliliter propidium iodide (PI) combined with 50 μg/ml of RNase A was added into the filtered homogenate and mixed for flow cytometry analysis. The sample was analyzed in a BD LSRFortessa flow cytometer equipped with a 50 mW 561 nm laser.

Total RNA was extracted from the first pair of 11-day-old wild-type plant and 13-day-old mutant leaves using an RNasy Plant Mini Kit (Qiagen). Residual DNA was removed with DNase I (NEB) and 2 μg of cleaned RNA were used for reverse transcription. RNA sequencing was performed according to Chen et al. (2013). For real-time PCR, reverse transcription was conducted using the SuperScript III first-strand synthesis SuperMix (Invitrogen) and PCR was conducted using Power SYBRgreen PCR Master Mix (Applied Biosystems) in a 7900 HT Fast Real-Time PCR System (Applied Biosystems). ACTIN2 was used as the internal control. Three biological replicates were performed for real-time PCR analysis. For checking the transcript level of AtNMD3 in the T-DNA insertion line WiscDsLox257G09, RNA was extracted from the wild-type plant and the mutant seedlings and real-time PCR was performed as described above.

In a genetic screen for mutants defective in lateral root response to drought and abscisic acid, we isolated a mutant, dig6 (drought inhibited growth of lateral roots), which showed reduced lateral root numbers. Map-based cloning identified that the mutation occurred in the AtLSG1-2 gene (Zhao et al., 2015). The dig6 mutant and a T-DNA insertion mutant allele atlsg1-2 (Salk_114083) had nearly identical phenotypes (Zhao et al., 2015). Here, we focused on the T-DNA insertion mutant atlsg1-2. Young leaves of the mutant exhibited significantly retarded growth and mature leaves remained small (Figures 1A,B). By checking the fifth leaves of the wild-type and atlsg1-2 mutants in detail, we found that the leaf area of the mutant was half that of the wild-type, as shown in Figure 1C. To determine which factor contributed to leaf size reduction, we evaluated cell size and cell number and found that in the mutant, cell size and cell number decreased to 73.5 and 70.9% that of the wild-type plant, respectively.

Cell size and cell number are closely associated with cell expansion and cell proliferation activity, respectively. We conducted a leaf kinematic analysis to investigate how loss of function of AtLSG1-2 affects cell division and expansion. Leaf development is divided into three stages: cell division, expansion, and maturation (Beemster et al., 2005). We evaluated leaf development in the first pair of atlsg1-2 leaves. As shown in Figure 2, compared to the wild-type, leaf emergence was delayed by 2 days in the atlsg1-2 mutant. Furthermore, leaf growth was relatively slower in the mutant and cell size and cell number also increased more slowly compared with the wild-type, suggesting that cell proliferation and expansion activity were simultaneously suppressed in the atlsg1-2 mutant (Figure 2). Furthermore, the cell expansion phase was noticeably delayed in the atlsg1-2 mutant (Figure 2 and Supplementary Figure S1). In wild-type plants, cell size was approximately 100 μm² 5–8 days after stratification (DAS), but after 9 DAS, cell size rapidly increased, suggesting that cell differentiation has begun. In the atlsg1-2 mutant, cell size from 7 to 11 DAS was similar to that of the wild-type plant from 5 to 8 DAS, and although cell size increased after 11 DAS, the rate was apparently slower (Figure 2). As shown in Supplementary Figure S1, whereas most of the cells in 9-DAS-old wild-type leaves had already underwent expansion, those in the mutant leaves began to expand only at 13-DAS. This suggests that cell expansion activity was apparently impaired in the atlsg1-2 mutant.

To substantiate the inhibitory effects of AtLSG1-2 deficiency on the normal progression of cell division to differentiation, we examined the CYCB1;1::GUS activity in the wild-type and atlsg1-2 mutants; CYCB1;1 is specifically expressed in dividing cells (Ferreira et al., 1994). We found that GUS signals greatly diminished in 12-day-old wild-type leaves but were still strongly expressed in the 14-day-old mutant leaves (Figure 3). As shown in Supplementary Figure S1, many cells in 13-DAS-old mutants still underwent division whereas the cells in wild-type plant were expanded. This was consistent with CYCB1;1 staining results. Therefore, loss of function of AtLSG1-2 hindered the transition from cell division to expansion.

The leaf kinematic analysis showed that cell expansion processes were suppressed in the atlsg1-2 mutant. Because cell expansion is closely related with the endoreduplication process, we examined the ploidy distribution by flow cytometry in the first pair of wild-type and atlsg1-2 mutant leaves. Leaf samples were collected during cell division, cell expansion, and maturation phases. As shown in Figure 4, the 2C population decreased more slowly in atlsg1-2 mutants than in wild-type leaves, while the 4C population slowly increased in the mutant. This suggests that cell expansion phase was delayed in the mutant. Although both 8C and 16C populations represent the onset of endoreduplication, the 16C population contributes minimally and thus will not be discussed here. The atlsg1-2 mutant plants also had a more slowly increasing 8C population such that the final 8C frequency was around 30% in wild-type leaves but only about 20% in the atlsg1-2 mutant (Figure 4), suggesting that the endoreduplication process was affected in the mutants. The ploidy distribution measured by flow cytometry can also be expressed as an endoreduplication index (EI), which represents the average number of endocycles undergone by a given nucleus. The EI, developmentally regulated during leaf growth, was lower in the atlsg1-2 mutant throughout development, indicating that the average number of endocycles is reduced in the atlsg1-2 mutant.

To verify these results, we also checked the ratio of trichomes with different numbers of branches because trichome branch number is positively correlated with ploidy level (Bramsiepe et al., 2010). In the first leaf, the ratio of two-branched trichome was significantly higher in atlsg1-2 mutant leaves than in wild-type plants (Figure 5). In atlsg1-2 mutants, most trichomes are two or three-branched, while wild-type plant leaves contain a relatively high ratio of four-branched trichomes and small number of five-branched trichomes (Figure 5). These results combined with those from flow cytometry suggest that an AtLSG1-2 deficiency suppresses endoreduplication.

Both leaf kinematic analysis and pCYCB1;1::GUS expression showed that the progression from cell division to differentiation was delayed in the atlsg1-2 mutant, suggesting that the expression of those genes associated with this process might be differentially regulated. Because leaf emergence in the mutant was delayed 2 days compared with wild-type plants, to eliminate the time difference in leaf emergence, we chose the first pair of 11-day-old wild-type and 13-day-old atlsg1-2 mutant leaves for RNAseq and qRT-PCR analysis. In 11-day-old wild-type plant leaves, small amount of cells in the tip of wild plant leaves expand and undergo differentiation. RNAseq results showed that cell-cycle process was obviously perturbed in the atlsg1-2 mutant (Supplementary Table S1). We further checked the expression of those genes specifically expressed in proliferation stage (Beemster et al., 2005), which includes CYCA2;3, CYCA3;2, CDKB2;1, CYCB1;5, CYCB2;1, and CYCB2;4, etc. Most of these genes were highly expressed in atlsg1-2 mutant (Supplementary Table S1). While the cause and effect relation of these events is unclear, one possibility is that AtLSG1 deficiency probably affects ribosome biogenesis and compromises the capacity of protein synthesis. The insufficient ability for protein translation may cause slower proliferation and delayed cell-cycle exit, which resulted in higher expression of cell-cycle genes later in proliferation stage in the mutant. Since RNA-seq analysis was only performed once, the expression patterns of some genes specifically expressed in proliferation stage were further confirmed by using real-time PCR (Figure 6). It could be concluded that the prolonged cell division phase in the atlsg1 mutant correlated with high levels of these genes specifically expressed in proliferation stage.

Leaf kinematic analysis showed that the onset of endoreduplication was affected in the atlsg1-2 mutant, suggesting that the expression of some genes associated with this process might be altered. Our RNA-seq analysis (Supplementary Table S1) showed that several key regulators of endoreduplication including KRP1, CYCA2;3, CDKB1;1, and LGO were differentially regulated in atlsg1-2 mutants. Among these genes, KRP1 and LGO are positive regulators of endoreduplication. KRP1 encodes a cyclin-dependent kinase inhibitor protein that negatively regulates cell division and promotes endoreduplication (Weinl et al., 2005). LGO is a member of a plant-specific cell-cycle-inhibitor-family SIAMESE and is essential for endoreduplication of sepal giant cells. Its loss has caused the absence of giant cells in leaves and sepals (Roeder et al., 2010). RNA-seq data showed that the transcript levels of KRP1 and LGO were reduced considerably in atlsg1-2 mutants. On the other hand, the expression of CDKB1;1 and CYCA2;3, which encode negative regulators of endoreduplication, significantly increased in the mutant (Supplementary Table S1). CYCA2;3, encoding A-type cyclin, could form a functional complex with CDKB1;1 and suppress the onset of endoreduplication (Boudolf et al., 2009). Other genes associated with endoreduplication were also differentially expressed in the mutant (Supplementary Table S1); these genes were associated with signaling pathways or transcription factors. For example, CALMODULIN LIKE 42 (CML42), a calmodulin-related calcium sensor: a cml42 knockout mutant had abnormal trichomes with increased branching (Dobney et al., 2009), suggesting that it is a negative regulator of trichome branching. Its expression was significantly upregulated in the atlsg1-2 mutant. HOMEODOMAIN GLABROUS 11 and 12, belonging to the HD-ZIP IV family, also negative regulators of trichome branching (Nakamura et al., 2006), had upregulated expression levels in the atlsg1-2 mutant. These results from RNA-seq analysis suggest that the expression of these genes might be affected in the mutant. We thus conducted real-time PCR to validate and quantify the expression level of these genes. Our real-time PCR results are consistent with those of the RNAseq analysis (Figure 6). Therefore, the differential expression of these endoreduplication regulatory genes may lead to disturbed endoreduplication as observed in the mutant.

LSG1 is known to mediate the export of the ribosome export factor NMD3 from the nucleus in yeast, but how plant NMD3 homologs interplay with LSG1 remains unknown. To understand the genetic interaction between AtLSG1-2 and AtNMD3 in Arabidopsis, we obtained an atnmd3 knockdown mutant (WiscDsLox257G09) because a atnmd3 knockout mutant is lethal (Chen et al., 2012). In the atnmd3 knockdown mutant, a T-DNA fragment was inserted in its 3′-untranslated region. The expression level of the AtNMD3 transcripts in the homozygous line was examined by RT-PCR using gene-specific primers (Supplementary Figure S2 and Table S2) and was found significantly reduced (Supplementary Figure S2). Nonetheless, this mutant showed only subtle phenotype changes at a very early stage such as retarded growth, delayed leaf emergence, and stunted roots (Figure 7). We generated an atlsg1 atnmd3 double mutant by genetic crossing of the two single mutants and examined the phenotype of the resulting double mutant. The growth of the primary root of the double mutant was very slow and leaf emergence was also significantly delayed in the early seedling stage (Figures 7A,B). The stature of the adult double mutant was much shorter than either single mutant (Figure 7C). The dwarf and bushy phenotypes along with short siliques and low fertility (Figures 7C,D) were not seen in either single mutant. We thus conclude that the attenuation of AtNMD3 function enhances the phenotypes caused by AtLSG1-2 loss. We examined the phenotypes of the first pair of leaves in more details. As shown in Figure 7E and Supplementary Figure S3, atlsg1-2 leaf area was reduced, whereas that of the atnmd3 mutant was similar to that of the wild-type. The leaf area of the double mutant was considerably less than either single mutant. Although cell size of the double mutant was similar to that of the atlsg1-2 mutant, cell number was significantly reduced compared with either single mutant (Figure 7E). These results suggest that AtLSG1-2 and AtNMD3 synergistically control cell proliferation activity during leaf development.