A hierarchical transcriptional network activates specific CDK inhibitors that regulate G2 to control cell size and number in Arabidopsis

How cell size and number are determined during organ development remains a fundamental question in cell biology. Here, we identified a GRAS family transcription factor, called SCARECROW-LIKE28 (SCL28), with a critical role in determining cell size in Arabidopsis. SCL28 is part of a transcriptional regulatory network downstream of the central MYB3Rs that regulate G2 to M phase cell cycle transition. We show that SCL28 forms a dimer with the AP2-type transcription factor, AtSMOS1, which defines the specificity for promoter binding and directly activates transcription of a specific set of SIAMESE-RELATED (SMR) family genes, encoding plant-specific inhibitors of cyclin-dependent kinases and thus inhibiting cell cycle progression at G2 and promoting the onset of endoreplication. Through this dose-dependent regulation of SMR transcription, SCL28 quantitatively sets the balance between cell size and number without dramatically changing final organ size. We propose that this hierarchical transcriptional network constitutes a cell cycle regulatory mechanism that allows to adjust cell size and number to attain robust organ growth. Cell division and expansion are carefully coordinated during organ growth. Here, the authors show that the SCL28 and SMOS1 transcription factors interact to regulate cyclin-dependent kinase inhibitors and promote cell size increase over division in Arabidopsis.

D uring organ growth and development, cell proliferation is intricately controlled in space and time. Governed by developmental programs [1][2][3] and influenced by active plant responses to environmental conditions 4,5 , the number of cells produced during organ growth is set by the cell cycle speed during proliferation and the point when cells exit to cellular differentiation. Controlled cell proliferation requires coordinately regulated gene expression within and upon exit from the cell cycle. Generally, there are two main groups of genes showing waves of transcription during the cell cycle; G1/S-specific and G2/ M-specific genes 6 . The G1/S-specific genes facilitate initiation and progression of DNA replication and are typically regulated by the activity of E2F transcription factors 7,8 . Generally, E2F dimerizes with Dimerization Partner (DP) proteins to activate or repress their target genes, depending on their association with the Retinoblastoma-Related (RBR) repressor protein 9 . On the other hand, most G2/M-specific genes are positively or negatively regulated by MYB3R family of transcription factors in plants 10,11 . Some members of MYB3Rs are specifically expressed during G2/ M and act as transcriptional activators, while others act as transcriptional repressors of G2/M-specific genes [12][13][14] . Recent studies identified the two main groups of transcription factors, E2Fs and MYB3Rs, which had been studied separately, as part of the same multi-protein complex in Arabidopsis 11 . This E2F-MYB3R complex is evolutionarily related to the DREAM (DP, Retinoblastoma-like, E2F, and MuvB) complex reported in Drosophila and human cells 11,15,16 . The metazoan DREAM complex plays a predominant role in repressing both G1/S-and G2/M-specific genes, thus promoting cell cycle exit and maintaining cellular quiescence 17,18 . The DREAM complex in Arabidopsis shows significant differences from metazoan complexes, which include involvement of plant-specific subunits and existence of diversified complexes with different subunit compositions 9,11,15 .
Transcriptional regulation during cell cycle generally constitutes multi-layered hierarchical networks, in which master regulators regulate other transcription factors, which further regulate each other or downstream genes [19][20][21] . Notably, studies in yeasts showed that cell cycle transcriptional activators that function during one stage of the cell cycle regulate transcriptional activators that function during the next stage, forming a connected regulatory network that is itself a cycle 22 . However, in plants, such a hierarchical network composed of cell cycle transcription factors, E2F and MYB3R, has not yet been uncovered. Exploring the transcription network during cell cycle may uncover missing important factors and hidden mechanisms governing plant-specific cell cycle regulation.
In this work, we identified a mitosis specific GRAS family transcription factor, called SCARECROW-LIKE28 (SCL28). In accordance with our work, a recent report identified the same genes being directly regulated by MYB3Rs 23 . Here, we demonstrate that SCL28 acts in association with the AP2-type transcription factor, AtSMOS1, to directly activate transcription of a specific set of SMR family genes, encoding plant-specific cyclindependent kinase (CDK) inhibitors 24 . This regulatory network inhibits the G2 to M phase transition during the cell cycle and promotes the onset of endoreplication, an atypical cell cycle consisting of repeated DNA replication without mitosis 4 . Our study identified a G2/M regulatory pathway that controls cell cycle length, and is likely to be important in optimizing cellular functions by setting cell size and number during organ growth.

Results
Identification of a mitotic GRAS-type transcription factor. By analyzing transcripts specific in mitotic cells, we have identified an Arabidopsis GRAS family transcription factor that we designated E1M 25 . A recent report looking for mitotic genes in the root meristem uncovered the same gene called SCL28 23 , the name hereafter also adopted in this study. In a synchronized culture of Arabidopsis MM2d cells, we showed that this gene exhibited G2/M-specific transcript accumulation, which closely resembles that of the mitotic cyclin CYCB1;2 (Fig. 1a). As for most G2/M-specific genes, the so-called mitosis-specific activator (MSA) element that serves as a binding motif for MYB3Rs 10,26 were repeatedly present within the proximal promoter regions of SCL28 (Fig. 1b). Binding of MYB3R to the SCL28 promoter is supported by our published data from chromatin immunoprecipitation (ChIP) with MYB3R3 followed by high-throughput sequencing (ChIP-Seq; Fig. 1b) 11 and DNA affinity purification sequencing (DAP-Seq) data reported for MYB3R1, MYB3R4, and MYB3R5 ( Supplementary Fig. 1a) 27 . In addition, loss of MYB3R activators (myb3r1/4 double mutant) or MYB3R repressors (myb3r1/3/5 triple mutant) resulted in significant down-or upregulation of SCL28, respectively (Fig. 1c). GUS reporter activity driven by SCL28 promoter decreased significantly in the myb3r1/ 4 double mutant and became essentially undetectable when the MSA elements in the SCL28 promoter had been deleted (Fig. 1d). Collectively, these observations support the idea that SCL28 is a direct target of both activator and repressor type MYB3Rs in Arabidopsis. Similar to the G2/M-specific CYCB1;1-GFP accumulation 28 , we observed patchy pattern of SCL28-GFP signal from a construct driven by native SCL28 promoter in root meristem, suggesting cell cycle-regulated protein accumulation (Fig. 1e). Taken together, as has been recently shown 23 , SCL28 is a mitotic gene directly regulated by MYB3Rs.
To evaluate whether MYB3Rs act as part of DREAM complex on SCL28 transcription, we searched lists of target genes bound by potential DREAM components, RBR and TESMIN/TSO1-LIKE CXC 5 (TCX5), as defined by ChIP-Seq experiments reported previously 15,29 . This examination revealed that TCX5, but not RBR, showed a significant association to SCL28 locus in vivo. However, SCL28 showed no significant change in expression in neither rbr, tcx5 tcx6 double, nor e2fa e2fb e2fc triple mutants, indicating that SCL28 transcription depends exclusively on MYB3Rs ( Supplementary Fig. 1b).
SCL28 strongly affects cell size. To analyze the biological function of SCL28, we generated transgenic plants overexpressing SCL28 under the strong RPS5A promoter (proRPS5A::SCL28). These plants, herein designated SCL28 OE , showed general growth retardation both in seedling and adult stages (Fig. 2a). Cell size in these plants was significantly enlarged in all examined organs and tissues, such as root tip, leaf mesophyll, embryo, and inflorescence stem ( Fig. 2b and Supplementary Fig. 2). An increased cell size was apparent in both post-mitotic differentiated cells and proliferating cell populations in root and shoot apical meristems ( Fig. 2c and Supplementary Fig. 2). As described later, the increased cell size in SCL28 OE leaves was associated with elevated levels of cellular ploidy induced by enhanced endoreduplication. The loss-of-function scl28 mutant ( Supplementary Fig. 3) showed an opposite cellular phenotype to SCL28 OE , having cells with significantly reduced size (Fig. 2b, c and Supplementary Fig. 2). However, unlike SCL28 OE , the overall stature of scl28 mutant plants was largely normal and indistinguishable from that of wild type (WT) plants (Fig. 2a).
To explore the developmental origin of cell size differences, we performed kinematic analysis on growing first leaf pairs by monitoring size and number of palisade cells (Fig. 2d-g). This analysis showed that cells in scl28 leaves were smaller than WT already during initial stages of organ development when most cells are actively proliferating (Fig. 2d). During this early proliferative stage, increasing cell number was also more rapid in scl28 than WT, consistent with a higher cell division rate in scl28 leaves compared with WT (Fig. 2e, f). Conversely, duration of cell proliferation remained largely unchanged between scl28 and WT (Fig. 2e, f). Therefore, the increased cell number in scl28 leaves is due to accelerated cell division, rather than increased duration of cell proliferation. When leaf area was compared, however, scl28 and WT showed no clear difference throughout the course of leaf development (Fig. 2g). In scl28 leaves, accelerated cell division was balanced by reduced cell size, thus maintaining total organ size unchanged during leaf development. The cellular effect of SCL28 OE was generally opposite to the scl28 mutant. The number of palisade cells per leaf was significantly reduced in SCL28 OE plants (Fig. 2e) due to severely inhibited cell proliferation (Fig. 2f). Although dramatic cell enlargement partially counteracted the reduced cell number, leaf area was still reduced in SCL28 OE compared with WT (Fig. 2d, g). We also analyzed cell size and number of epidermal pavement cells during leaf development in scl28 and SCL28 OE lines and found the response of this cell type to altered SCL28 activity was comparable to what we have shown for palisade cells (Supplementary Fig. 4).
In the root tip, cell size difference was similarly apparent among SCL28 OE , scl28, and WT (Fig. 2c, h, i). This cell size difference was maintained along the apical-basal axis of the root from cells adjacent to the quiescent center to those at the transition zone, as well as in differentiated cells at distal positions (Fig. 2h). Therefore, the behavior of cells in the root meristems of SCL28 OE and scl28 was generally consistent with those observed in developing leaves.
Effects of SCL28 on mitotic cell cycle and endocycle. To directly explore the role of SCL28 in cell cycle progression, we performed live-cell imaging of root meristem cells after introgression of the PCNA-GFP cell cycle marker into SCL28 OE and scl28 mutant. The PCNA-GFP line allows quantification of cell cycle stages Arabidopsis MM2d cells were synchronized by aphidicolin treatment and used for qRT-PCR analysis to quantify SCL28 mRNA levels (lower). Synchronous progression of cell cycle was monitored by measuring mitotic index (upper). As a representative of G2/M-specific genes, CYCB1;2 was also analyzed in the same way for comparison. Expression data are shown as averages from three technical replicates (±SD). b MYB3R3 binds to the upstream region of SCL28. The ChIP-Seq profile of MYB3R3 indicates its direct binding to the proximal promoter region of SCL28. The nucleotide sequence around the ChIP-Seq peak contains MSA elements repeated four times, of which core motifs (AACGG) are shown by orange (forward orientation) and green (reverse orientation) letters. c Levels of SCL28 transcripts are significantly affected by mutations in myb3r genes. Transcript levels of SCL28 and KNOLLE were analyzed by qRT-PCR in plants 12 days after sowing (DAS) that carry mutations of myb3r genes in the indicated combinations. Expression data were normalized to ACT2 and shown as averages from three biological replicates (±SD). Statistical significance was determined using two-sided Student's t test. *P < 0.05, **P < 0.01. d Promoter activity of SCL28 requires the presence of MYB3R activators and MSA elements. GUS reporter expression was analyzed in wild type (WT) and myb3r1/4 plants carrying proSCL28::GUS and WT plants carrying proSCL28ΔMSA::GUS, in which all MSA elements in the SCL28 promoter were mutated. GUS staining around shoot apex and root tip regions is shown at higher magnification in the lower panels. Multiple independent lines for each construct showed similar difference in GUS staining. Scale bar indicates 100 µm. e Patchy pattern expression of SCL28-GFP protein. Root meristems of proSCL28::SCL28-GFP and proCYCB1;1::CYCB1;1-GFP plants were analyzed by confocal laser scanning microscopy (CLSM) after counterstaining of the cell wall with propidium iodide (PI). Green and red signals indicate fluorescence of GFP and PI, respectively. Similar expression patterns of SCL28-GFP were confirmed in multiple independent lines. Scale bar indicates 100 µm.
NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-29316-2 ARTICLE based on its intracellular fluorescent patterns 30 . During G1, PCNA-GFP appears as uniform fluorescence within the whole nuclei, which alters into dotty and speckled nuclear signals during early and late S phase, respectively. During G2, fluorescence again becomes uniform within the nuclei, then disappears upon onset of mitosis and remains undetectable until the exit from mitosis. By our definition, the G1 period begins with re-appearance of nuclear fluorescence after mitosis and ends with emergence of dotty GFP signals, whereas the G2 period begins with conversion of dotty into uniform nuclear PCNA-GFP signals and ends with disappearance of this nuclear fluorescence. Based on these definitions, we analyzed data from live-cell imaging and calculated the average duration of each cell cycle phase in WT root meristem cells to be 5.1 h in G1, 2.5 h in S, and 11.3 h in G2 (Fig. 3a, b). G2/ M duration was oppositely affected in the scl28 and SCL28 OE lines, shortened 19% in the former and lengthened 33% in the later. In contrast, G1 length showed only limited alteration, but became longer in the scl28 mutant (Fig. 3a, b), which is consistent with the existence of an additional cell size checkpoint at G1/S that compensates for shortened G2 31 . This suggests that SCL28 actively inhibits progression through G2 and prevents entry into mitosis. In agreement with the role of SCL28 in G2 duration of the mitotic cell cycle, our ploidy analysis of developing leaves showed that SCL28 also affects endoreplication (Fig. 3c, d, and Supplementary Fig. 5). During leaf development, SCL28 OE plants initiated earlier endoreplication, indicated by 8 C cell emergence as early as 8 days after sowing (DAS) and thereafter consistently showing higher cellular ploidy levels compared with WT. Elevated ploidy levels were associated with dramatically increased cell size in SCL28 OE leaves as shown earlier (see Fig. 2b, d). In the scl28 mutant, ploidy levels were not affected during early stages of leaf development and only showed a modest decrease at 20 DAS. During the earliest stage (8 DAS) before onset of endoreplication, we observed a smaller proportion of 4 C cells in scl28 compared with WT (Fig. 3c), which supports the cell cycle analysis, showing reduced G2 duration (Fig. 3a). In summary, our data suggest that SCL28 has a role to inhibit the G2 to M phase transition.
SCL28 acts together with AtSMOS1 as a heterodimer. Phylogenetic analysis of GRAS family proteins indicated that SCL28 constitutes a unique clade together with SMOS2 in rice (Supplementary Fig. 6a). Rice smos2 was initially identified as a mutant showing reduced organ size with component cells smaller than those in WT 32 . Another rice mutant, smos1, with a phenotype similar to that in smos2, has loss-of-function mutation in a unique gene encoding an AP2-type transcription factor 33 . Because a physical interaction between SMOS1 and SOMS2 has been reported 32 , we postulated that SCL28 may act through interaction with an Arabidopsis protein orthologous to SMOS1. When AP2-type transcription factors were phylogenetically analyzed, we found At2g41710 to be the likely Arabidopsis ortholog b Cell size analysis of leaf palisade tissue from WT, scl28, and SCL28 OE plants. First leaf pairs from plants at 10 or 18 DAS were cleared and analyzed by differential interference contrast (DIC) microscopy. Scale bar indicates 20 µm. c Cell size analysis of root meristem from WT, scl28, and SCL28 OE plants at 5 DAS. Primary roots were stained with PI, and analyzed by CLSM. Regions surrounded by yellow rectangles are shown at a higher magnification in the lower panels. Scale bars indicate 100 µm (upper) and 10 µm (lower). d-g Kinematic analysis of leaf growth in WT, scl28, and SCL28 OE plants. Time-course changes of palisade cell area (d), number of palisade cells per leaf (e), cell division rate (f), and leaf area (g) were analyzed in developing first leaf pairs. Values are averages from the data of leaves from ten different plants (±SD), in each of which more than 20 cells were analyzed. Inset shows the same data with an expanded y-axis scale. h Spatial distribution of cell size across root meristems in WT, scl28, and SCL28 OE plants at 5 DAS. Mean cell length was calculated at each position along the cortical cell file, which is defined by counts of cortical cells from the quiescent center. Values are averages from the data analyzed in 21 roots from different plants (±SD), in each of which more than 30 cells were analyzed. i Quantitative analysis of cortical cell length within root meristems. Boxplot was generated using data collected from roots of 21 different plants, in each of which more than 20 cells were analyzed. In boxplot, 'boxes' represent the interquartile range (IQR), midline indicates the median, and whiskers extend to the largest and smallest observations within 1.5 × IQR. Statistical significance compared with WT was determined using two-sided Student's t test. **P < 0.01. of SMOS1 ( Supplementary Fig. 6b). To test the physical interaction between SCL28 and At2g41710, we performed yeast twohybrid and bimolecular fluorescence complementation (BiFC) assays to obtain clear results showing that they indeed interact in yeast and in plant cells, respectively (Fig. 4a, b). We concluded that the observed protein-protein interaction is evolutionarily related to the SMOS1-SMOS2 interaction in rice and named At2g41710 as AtSMOS1.
To examine AtSMOS1 function, we analyzed a T-DNA insertion mutant for this gene ( Supplementary Fig. 7) and found a small cell size phenotype that closely resembled that of scl28 ( Fig. 4c-f). However, the overall stature of atsmos1 was largely indistinguishable from WT plants, as is the case for scl28 ( Supplementary Fig. 8a). To analyze epistasis between scl28 and atsmos1, we performed genetic analysis by quantitatively comparing phenotypes of scl28, atsmos1 and scl28 atsmos1 double mutants. These single and double mutants showed essentially equivalent phenotypes in terms of palisade cell size (Fig. 4c, d) and cell size in the root meristem (Fig. 4e, f). This indicates that, as with SCL28, AtSMOS1 also impacts cell size. The lack of additive effect between atsmos1 and scl28 suggests that these proteins may act in the same pathway, which is consistent with the interaction between these proteins. We then tested whether the enlarged cells in SCL28 OE relies on the presence of AtSMOS1, and indeed in the atsmos1 mutant background, the strong effects of SCL28 OE on cell size of leaf palisades (Fig. 4g, h) and root meristem cells (Fig. 4i, j), as well as whole plant growth, completely diminished ( Supplementary  Fig. 8b). Collectively, these observations strongly suggest that SCL28 and AtSMOS1 function cooperatively to regulate cell size in different plant organs.
To analyze AtSMOS1 expression, we generated plants expressing AtSMOS1-GFP driven by its own promoter. These transgenic plants showed nuclear localization of AtSMOS1-GFP in various tissues and organs such as root meristem, developing leaves, and cotyledons (Fig. 4k). However, unlike SCL28, AtSMOS1-GFP was uniformly expressed in meristematic cells, suggesting a cell cycle-independent AtSMOS1 expression. Nonetheless, cells expressing AtSMOS1 and SCL28 were overlapping, supporting the idea that they interact to form a heterodimer in vivo. We also noted an E2F binding element in the AtSMOS1 promoter, which binds RBR in published ChIP-Seq data 29 . We found that E2FB binds to AtSMOS1 promoter in our ChIP-qPCR experiment and that AtSMOS1 is significantly upregulated in the e2fa e2fb e2fc triple mutant ( Supplementary Fig. 9), suggesting that AtSMOS1 is indeed regulated by E2Fs. One explanation for  Numbers of biological replicates for calculating length of G1, early S, late S, and G2/M, respectively, are 14, 12, 12 and 11 for WT, and 12, 10, 10, and 12 for SCL28 OE . Statistical significance was determined and shown as in (a). c Ploidy analysis of scl28 and SCL28 OE plants. First leaf pairs of WT, scl28, and SCL28 OE plants were subjected to flow cytometric analysis to determine ploidy distribution during leaf development. Data are shown as averages from ten biological replicates (±SD). Statistical significance compared with WT was determined using two-sided Student's t test. *P < 0.05, **P < 0.01. d Timecourse change in cellular ploidy levels during leaf development. Data presented in (c) were used for calculating endoreplication index, which represents mean number of endoreplication cycles per leaf cells. Data are shown as averages from ten biological replicates (±SD). Statistical significance compared with WT was determined and shown as in (c). insignificant change of AtSMOS1 mRNA in rbr and tcx5 tcx6 might be that RBR-and DREAM-dependent regulation on E2Fs is developmental specific, and was not apparent at the seedling stage.
Downstream targets of SCL28. Studies of smos1 and smos2 rice mutants suggest a role in post-mitotic cell expansion through the regulation of the Oryza sativa PHOSPAHATE INDUCED1 (OsPHI-1) gene 32 . Because we observed SCL28 expression specific to meristematic cells, we postulated that SCL28 should largely influence actively-proliferating cells before the onset of cell expansion. To identify the downstream targets of SCL28, we analyzed genome-wide gene expression changes in scl28, SCL28 OE , and atsmos1 by conducting RNA-sequencing (RNA-Seq) and microarray experiments (Supplementary Data 1).
Considering the epistatic phenotypes among scl28, SCL28 OE , and atsmos1 mutants, the critical downstream genes of SCL28 should be affected in scl28 and atsmos1 in a similar manner, as well as in SCL28 OE in an opposing manner. To reveal the genes satisfying these expression criteria, we conducted an overlapping analysis of differentially expressed genes in scl28, atsmos1, and SCL28 OE , and identified 21 genes that are significantly (adjusted P value < 0.05) downregulated in both scl28 and atsoms1, and upregulated in k Root meristem    (Fig. 5a and Supplementary Fig. 10a), as well as 26 genes upregulated in both scl28 and atsoms1, and downregulated in SCL28 OE (Fig. 5b and Supplementary Fig. 10b). Among these genes, we focused on SIAMESE-RELATED2 (SMR2), encoding a member of SMR family proteins, as proteins of this family represent plant-specific CDK inhibitors and some members, such as SIM and SMR1, positively affect cell size and negatively influence cell division, reflecting the observed effect of SCL28 1,24 . According to RNA-Seq and microarray data, in addition to SMR2, we found other SMR family members as candidates of downstream genes, with decreased expression levels in both scl28 and atsmos1 and increased levels in SCL28 OE . To further analyze SMR family genes, we conducted quantitative reversetranscription PCR (qRT-PCR) analysis to examine expression of all members of this family in scl28, atsmos1, and SCL28 OE , and identified at least seven members (SMR1, SMR2, SMR6, SMR8, SMR9, SMR13, and SMR14) out of all 17 as potential target genes of SCL28 and AtSMOS1 (Fig. 5c-e). Consistent with this conclusion, we also observed reduced expression of proSMR2::SMR2-GFP and proSMR13::SMR13-GFP in root meristems in scl28 or atsoms1 mutant backgrounds compared with WT (Supplementary Fig. 11). KIP-RELATED PROTEINs (KRPs) constitute another family of CDK inhibitors in addition to SMRs in plants. The general view is that KRP proteins primarily connect with CDKA to negatively regulate G1/S 34 , while SMR proteins inhibit G2/M by associating with both CDKA and CDKB1 24,35 . There are seven KRP members encoded in the Arabidopsis genome, and some we did find deregulated in the scl28, atsmos1, and SCL28 OE plants. However, in contrast to the SMRs, the change in expression of KRPs were generally moderate and inconsistent with the activating function of SCL28-AtSMOS1 heterodimer, and thus likely to be an indirect pleiotropic effect ( Supplementary Fig. 12). Therefore, the SCL28 and AtSMOS1 may directly affect expression of SMR members, but not those of the KRP family, consistent with our results that SCL28 preferentially affects G2 duration in meristematic cells.
Genome-wide mapping of SCL28 and AtSMOS1 binding sites. In order to define direct targets for SCL28 and AtSMOS1, combined with transcriptome analysis, we performed ChIP-Seq assays using proRPS5A::SCL28-GFP and proAtSMOS1::AtSMOS1-GFP lines to identify their binding genomic loci. Consistent with our emphasis on downregulation of SMR genes, which suggests that SCL28 and AtSMOS1 act as transcriptional activators, the majority of binding sites are located in promoter regions proximal to transcription start sites (TSS) (Supplementary Fig. 13) and located within nucleosome-free and highly accessible chromatin regions ( Supplementary Fig. 14a-d). Their enrichment levels at TSS correlated well with mRNA levels of corresponding genes, suggesting a positive correlation between their binding and transcriptional activity (Supplementary Fig. 14e, f).
The ChIP-Seq data analysis uncovered 463 and 4,287 genes as targets of SCL28 and AtSMOS1, respectively (Supplementary Data 2). Comparing these target genes revealed the presence of a significant overlap, a set of 308 common targets that accounts for 66% of SCL28 targets (Fig. 6a). On one hand, this result confirms our genetic analysis, suggesting that SCL28 function largely relies on the interaction with AtSMOS1. On the other hand, AtSMOS1 may show a wider range of functions both dependent on and independent of SCL28. This interpretation is also supported by our RNA-Seq data showing larger number of downregulated genes in atsmos1 compared to scl28 and these downregulated genes have a relatively small overlap (see Fig. 5a). We also found only a limited overlap between common targets identified by ChIP-Seq and those that show regulated expression in transcriptome analysis of scl28, atsmos1 and SCL28 OE lines (Fig. 6b). Gene ontology (GO) enrichment analysis of the 308 common targets revealed overrepresented GO terms related to cell cycle, such as "regulation of mitotic nuclear division" and "regulation of DNA endoreduplication" (Supplementary Fig. 15). This overrepresentation of cell cycle-related GO terms largely relies on the presence of five common genes that belong to such GO categories. All these genes were found to be members of SMR family genes, SMR2, SMR4, SMR6, SMR8, and SMR9, all of which, except SMR4, showed significant downregulation in both scl28 and atsmos1 and upregulation in SCL28 OE in our qRT-PCR experiments described above (see Fig. 5). Though three SMR genes-SMR1, SMR13, and SMR14-with significant expression changes in all lines failed to fulfill the criteria of ChIP-Seq data analysis, visual inspection revealed recognizable peaks for both SCL28 and AtSMOS1 in ChIP-Seq profiles, suggesting these SMRs are also direct common targets ( Fig. 6c and Supplementary  Fig. 16). The ChIP-Seq peaks were observed at either 5′ (SMR1, SMR2, and SMR4), 3′ (SMR6, SMR9, and SMR14) or both 5′ and 3′ (SMR8 and SMR13) regions of the target SMR loci. In most cases, we found that the positions of ChIP-Seq peaks adjacent to SMR loci coincided for SCL28 and AtSMOS1, suggesting that they associate with the same sites. Our ChIP-Seq data also showed that none of the KRP genes were found as common targets of SCL28 and AtSMOS1, further confirming that SCL28-AtSMOS1 directly regulates some SMRs but not KRPs.
To find common binding sites and sequence motifs for SCL28 and AtSMOS1, we compared the precise positions of ChIP-Seq peaks along the gene structure. Dot-plot analysis of the common targets showed frequent clustering of SCL28 and AtSMOS1 peaks at the same positions relative to TSS and transcription end site (TES; Fig. 6d). Similar analysis at a genome-wide scale confirmed that SCL28 frequently targets the same sites as AtSMOS1 around the TSSs (Supplementary Fig. 17a). Within the common peaks for SCL28 and AtSMOS1, a DNA motif of C(a/t)T(a/t)GGATNC(c/t) (a/t) could be identified as an overrepresented cis element (Fig. 6e). This motif is indeed present among more than 40% of the common targets, but is much less frequent among targets of either SCL28 or AtSMOS1, which contain quite different cis elements based on sequence overrepresentation analysis (Supplementary Fig. 17b, c). This suggests that the formation of SCL28-AtSMOS1 heterodimer generates a specific sequence preference for DNA binding to regulate transcription of a specific set of targets. This or closely related motifs are present in all seven SMR target loci at positions frequently coinciding with the peak summits of SCL28 and AtSMOS1 (Fig. 6c and Supplementary  Fig. 16), suggesting that SCL28 and AtSMOS1 recognize this motif and bind to their targets as a heterodimer.
We then tested whether chromatin association of SCL28 depended on AtSMOS1 and vice-versa by conducting ChIP-qPCR experiments in scl28 and atsmos1 backgrounds (Fig. 6f). As expected, both SCL28 binding on SMR2 promoter in the atsmos1 and AtSMOS1 binding in the scl28 mutant backgrounds were abolished. Collectively, our data supports SCL28 and AtSMOS1 functional heterodimer binding to specific sets of SMR target genes to activate transcription and negatively regulate cell cycle progression at the G2 to M cell cycle transition.
SMRs are critical downstream effectors for SCL28 and AtS-MOS1. If SMR genes directly bound by SCL28 and AtSMOS1 are crucial for mediating their effects on cellular phenotype, transcript levels of those SMR genes should be altered accordingly with cell size phenotype observed in our epistasis analysis (see Fig. 4c-j). Confirming this idea, downregulation of target SMR ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-29316-2 genes-SMR2, SMR9 and SMR13-were quantitatively equivalent in scl28, atsmos1, and double scl28 atsmos1 mutants (Fig. 7a). In addition, strong activating effects of SCL28 OE on target transcription was totally abolished under atsmos1 mutant background, providing the molecular basis for AtSMOS1-dependent action of SCL28 on cell size (Fig. 7b). To further elucidate that AtSMOS1 acts together with SCL28 on SMR genes, we also analyzed the proSMR2::LUC reporter co-transfected in protoplasts with pro35S::SCL28 and/or pro35S::AtSMOS1 constructs. The most prominent and significant activation of luciferase was only observed when both plasmids expressing SCL28 and AtS-MOS1 were simultaneously transfected, supporting the conclusion that SCL28 and AtSMOS1 fulfill their function as a transcriptional activator by forming a heterodimer (Fig. 7c).
To directly address whether these SMR genes are critical for SCL28 function, we performed genetic analysis focusing on SMR1, SMR2 and SMR13, which are significantly downregulated in scl28 and upregulated in SCL28 OE . For each smr mutant, we could not detect any abnormalities in cell size ( Supplementary  Fig. 18) Fig. 6 Binding of active SCL28-AtSMOS1 heterodimer to target SMR genes. a Venn diagram showing significant overlap between genes bound by AtSMOS1 and SCL28. Significance of the overlap was evaluated using one-sided hypergeometric tests (no adjustment). b Venn diagram showing limited overlap between common target genes bound by SCL28 and AtSMOS1 and those downregulated in both scl28 and atsmos1 and upregulated in SCL28 OE . SMR genes found in overlapping and non-overlapping regions are also shown. c ChIP-Seq profiles for SCL28 and AtSMOS1 around SMR2, SMR8, and SMR9 loci. Orange arrowheads indicate DNA motifs perfectly matching C(a/t)T(a/t)GGATNC(c/t)(a/t), identified as an enriched motif in the SCL28/AtSMOS1 common targets, whereas blue arrowheads indicate motifs matching the enriched motif with one base mismatch (see Fig. 6e). d Density plot showing overlap of SCL28 and AtSMOS1 using hexagonal binning routine. Each dot represents the distance from the peak midpoint to the nearest gene. The y-axis shows location of the AtSMOS1 peak midpoint compared with gene position, while the x-axis indicates location of SCL28 peak midpoint compared with nearest gene. The large number of dots occurs along the positive correlation line, showing the co-occurrence pattern of SCL28 and AtSMOS1. e DNA motif overrepresented in common targets for SCL28 and AtSMOS1. HOMER motif search identified a specific common SCL28/AtSMOS1 binding motif with a P value of 1e−184 (one-sided hypergeometric tests, no adjustment), which was found in 40.13% of common targets. f ChIP-qPCR analysis was performed on SMR2 loci using plants carrying proRPS5A::SCL28-GFP under WT and atsmos1 backgrounds, and those carrying proAtSMOS1::AtSMOS1-GFP under scl28 and atsmos1 backgrounds. Amplified regions (−150, +150, and cds) in ChIP-qPCR are shown by double-headed arrows. UBQ locus was analyzed in the same way as a negative control without binding. Data are shown as averages from three technical replicates (±SD). Statistical significance was determined using two-sided Student's t test. *P < 0.05. n.s., not significant. NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-29316-2 ARTICLE NATURE COMMUNICATIONS | (2022) 13:1660 | https://doi.org/10.1038/s41467-022-29316-2 | www.nature.com/naturecommunications smr1/2/13 triple mutant, there was a significant reduction in cell size in multiple tissues, including leaf palisades (Fig. 7d, e) and root meristem (Fig. 7f, g), suggesting functional redundancy among SMR genes. To address the link between SCL28 and SMRs, we studied the cell size phenotype when these mutations were combined. In leaf palisade cells, the triple smr1/2/13 mutant showed stronger reduction in cell size than the scl28 mutant, which did not further reduce when all mutations combined in the smr1/2/13 scl28 line, supporting the idea that cell size regulation by SCL28 is mediated by SMR1/2/13 (Fig. 7d, e). Conversely, scl28 mutation still significantly decreased cell size in the root meristem when combined with smr1/2/13 (Fig. 7f, g). Therefore, SMR1/2/13 contribution downstream of SCL28 appears to be larger in palisade tissue than root meristem, where additional SCL28-regulated SMRs may play a role in cell size control. These tissue-specific differences indicate the contribution of developmental regulation that positions different sets of SMRs downstream of SCL28.
Dose-dependent control of cell size and cell number by SCL28 to maintain organ size homeostasis. In the scl28 mutant, leaf growth is essentially normal, but the constituent cells became small and more numerous (see Fig. 2d, e, g), suggesting that SCL28 is pivotal to maintain organ size homeostasis by regulating the balance between cell size and cell number. For a regulator designed to tune the balance between cellular parameters, one assumes that it acts dose dependently. To explore whether or not SCL28 levels quantitatively affect cell size, we analyzed plants heterozygous for scl28 (scl28/+ plants). Heterozygous plants showed reduced SCL28 transcript levels that is approximately half of that in WT (Fig. 8a). This SCL28 downregulation resulted in a clear reduction of cell size in root meristem and leaf palisade tissue, with an intermediate cell size in heterozygous plants between those in WT and scl28 homozygous plants (Fig. 8b, c and Supplementary Fig. 19a, b).
Overexpression of SCL28 causes a phenotype opposite to scl28, dramatically reducing the number of cells that became Reporter plasmid Expression plasmids Fig. 7 Cell size control by SCL28-AtSMOS1 is mediated by SMRs. a Double scl28 atsmos1 mutation downregulates SMR genes, equivalently to each single mutation. qRT-PCR analysis of SMR2, SMR9 and SMR13 transcripts was performed for comparing expression levels in WT, scl28, atsmos1, and scl28 atsmos1 plants. Data are shown as averages from three biological replicates (±SD). Statistical analysis was performed for each SMR gene based on one-way ANOVA and Tukey's test, and significant differences (P < 0.05) were shown in different letters above the bars. b Upregulation of SMR genes in SCL28 OE requires presence of AtSMOS1. qRT-PCR analysis was performed for comparing expression levels between atsmos1, SCL28 OE plants, and those carrying atsmos1 and SCL28 OE in combination. Two independent lines were used for analyzing SCL28 OE / atsmos1. Data are shown as averages from three biological replicates (±SD). Statistical analysis was performed as in (a). c Co-expression of SCL28 and AtSMOS1 activates the SMR2 promoter in the protoplast transient assay. proSMR2::LUC reporter plasmid was transfected into protoplasts prepared from T87 cells together with expression plasmids of SCL28 and AtSMOS1 as indicated. Schematic presentation of reporter and expression plasmids as shown on the left. Data are shown as averages from four biological replicates (± SD). Statistical analysis was performed as in (a). d DIC observations of leaf palisade cells in WT, scl28, smr1/2/13, and smr1/2/13 scl28 plants at 22 DAS. Scale bar indicates 20 µm. e Quantification of palisade cell area in the first leaf pairs from plants with indicated genotypes. Boxplot was generated using data collected from leaves of ten different plants, in each of which more than 60 cells were analyzed (midline = median, box = IQR, whiskers = 1.5 × IQR). Different letters above the boxplots indicate significant differences (P < 0.05) based on one-way ANOVA and Tukey's test. f CLSM observation of cortical cell files in root meristems from WT, scl28, smr1/2/3, and smr1/2/13 scl28 plants at 7 DAS. Scale bar indicates 20 µm. g Quantification of cortical cell length in root meristems from plants with indicated genotypes. Boxplot was generated using data collected from roots of multiple individual plants (midline = median, box = IQR, whiskers = 1.5 × IQR). Numbers of plants analyzed were 10, 12, 12, and 11 for WT, scl28, smr1/2/3, and smr1/2/13 scl28, respectively. In each plant, more than 40 cells were analyzed for calculating average, which were further used for statistical analysis. Different letters above boxplots indicate significant differences as in (e). enlarged. To examine the outcome when SCL28 expression is only moderately increased, we utilized plants from different T2 lines carrying proSCL28::SCL28-GFP in scl28 mutant background and showing varying levels of SCL28-GFP expression. We quantitatively compared levels of SCL28 transcript and cellular parameters of palisade tissues in each individual plant (Fig. 8d, e and Supplementary Fig. 19c). This showed that SCL28 expression positively correlates with cell size and negatively with leaf cell number. Both positive and negative correlations were statistically significant. Due to compensatory changes in cell number and size, overall leaf size was not dramatically affected by moderate alteration of SCL28 expression (Fig. 8f). Therefore, we concluded that a finely-graded expression of SCL28 sets the balance between cell size and number without altering organ growth as a whole. This mechanism may enable plants to achieve an optimal balance between cell size and number, possibly depending on developmental status and environmental conditions.

Discussion
Cell size at division depends on coordination between cell growth and cell cycle progression, which is actively maintained in a cell autonomous manner in both uni-and multicellular organisms [36][37][38] . Here, we identified a molecular mechanism for cell size regulation in Arabidopsis that relies on a hierarchical transcriptional activation of CDK inhibitors. In this pathway, SCL28 expression is specifically confined to mitosis through the MSA element in its promoter controlled by MYB3Rs. In turn, SCL28 associates with AtSMOS1, a transcription factor with cell cycle-independent expression, and this dimer defines the binding site present in a set of SMR genes encoding CDK inhibitors to activate their expression and thereby act as a brake against the cell cycle engine predominantly driven by CDK activity (Fig. 9). The SCL28-AtSMOS1-SMR axis uniquely affects only cell cycle progression and cell doubling time, but not the exit from cell proliferation. Therefore, it acts to set the balance between cell number and size during organ development without significantly qRT-PCR analysis was performed using WT plants, and those heterozygous (scl28/+) and homozygous (scl28) for scl28 at 10 DAS. Data are shown as averages from three biological replicates (±SD). Statistical analysis was performed based on one-way ANOVA and Tukey's test and significant differences (P < 0.05) are shown in different letters above bars. b Quantification of palisade cell area in scl28, scl28/+ and WT plants. Boxplot was generated using cell size data from 1st leaf pairs of multiple individual plants at 10 DAS (midline = median, box = IQR, whiskers = 1.5 × IQR). Numbers of plants analyzed were 15, 18, and 22 for scl28, scl28/+ and WT, respectively. In each plant, more than 70 cells were analyzed for calculating average, which were further used for statistical analysis. Different letters above boxplots indicate significant differences (P < 0.05) based on one-way ANOVA and Tukey's test. c Quantification of cortical cell length in root meristems of scl28, scl28/+ and WT plants. Boxplot was generated using cell size data from roots of multiple individual plants at 7 DAS (midline = median, box = IQR, whiskers = 1.5 × IQR). Numbers of plants analyzed were 16, 17, and 12 for scl28, scl28/+ and WT, respectively. In each plant, more than 30 cells were analyzed for calculating average, which was further used for statistical analysis. Different letters above the boxplots indicate significant differences as in (b). d Correlation between SCL28 mRNA levels and palisade cell size analyzed by scatterplots, where each dot corresponds to each individual plant from different transgenic lines carrying proSCL28::SCL28-GFP under scl28 background. In addition to proSCL28::SCL28-GFP plants (shown in gray dots), WT (blue dot), scl28 (red dot), and SCL28 OE plants (green dot) were also analyzed. e Correlation between SCL28 mRNA levels and number of palisade cells per leaf was analyzed by scatterplot as in (d). Symbols are same as in (d). f Correlation between SCL28 mRNA levels and leaf area was analyzed by scatterplot as in (d). Symbols are same as in (d). Values in (d-f) represent Pearson correlation coefficients (r s ) and its two-sided P value without adjustment (p). impacting final organ size. Whether SCL28 is part of a cell size sensor and how it is coupled to the measurement of cell growth remain to be determined.
We provide multiple lines of evidence supporting the conclusion that SCL28 acts as a brake on the G2/M transition to set cell size: (1) measurements of cell cycle phases by following PCNA-GFP using live-cell imaging showed that scl28 mutant and SCL28 OE shortened or lengthened the G2 phase, respectively; (2) kinematic analysis of cellular parameters during leaf development of scl28 and SCL28 OE showed a clear alteration in cell doubling time without any effect on cell cycle exit; (3) the molecular mechanisms identified here are fully consistent with the cell cycle inhibitory function, SCL28, forming a dimer with AtSMOS1 and directly activating the transcription of SMR genes whose products inhibit CDK activity; (4) in support of this, the scl28 mutant phenocopied and interacted in an epistatic fashion with the triple smr1/2/13 mutant; and 5) SCL28 regulates the number of endoreplication cycles, which is typically enhanced by G2 inhibition. Previous reports suggest that SCL28 plays a positive role in root growth and promotes the G2/M cell cycle progression 23,39 . These conclusions are inconsistent with our data and the small cell size phenotype observed in both studies, which may originate from using different methods relying on fixed samples.
Another example of a transcription factor inhibiting cell cycle at G2/M is a NAC-type transcription factor, SUPPRESSOR OF GAMMA RESPONSE 1 (SOG1), acting as a central regulator of DNA damage response in Arabidopsis. Similar to SCL28, SOG1 transcriptionally activates SMR family genes to inhibit cell cycle progression upon DNA damage 40 , but its targets, SMR5 and SMR7, are not regulated by SCL28, which instead directly regulates seven other SMR genes. Notably, in contrast with SOG1, which is activated specifically upon DNA damage, SCL28 negatively regulates cell cycle, thereby affecting cell size during normal plant development. The target of rapamycin (TOR) signaling pathway is a primary regulator of cell growth and proliferation and is involved in cell size regulation in yeast 41 . In Arabidopsis, TOR, acting through YAK1 kinase, regulates the transcription of SMR genes and thereby G2/M transition 42 . However, the set of YAK1-targeted SMRs overlap with those regulated by SOG1 rather than those regulated by SCL28. Therefore, our study places SCL28 as part of a novel pathway for regulating G2/M progression that determines cell size and cell number during plant organ development.
Variability in cell size at birth or stochastic changes in key regulators necessitate mechanisms that achieve cell size homeostasis by linking cell cycle progression to cell growth 38 . The way cell size variations are corrected has long been debated. The principal possibilities are using (i) a cell cycle timer, (ii) cell size measurements or (iii) correcting cell size by adding a constant increase to daughters regardless of initial sizes 36 . Comparing predicted outcomes of these theoretical models with time-lapse imaging data on cell size in plant meristematic cells identifies possible co-existence of a mix of these mechanisms that may act at multiple control points of the cell cycle 31,43 . The molecular device for cell size control may rely on cell size-dependent accumulation of an activator or dilution of a cell cycle inhibitor molecule. The 'inhibitor dilution model' postulates that cell growth dilutes such inhibitor protein until its concentration meets a threshold, thereby permitting cell division when cells reach an appropriate size 44 . This hypothesis has been verified in yeast and mammalian systems 45,46 . In Arabidopsis, recent studies show segregation of set amounts of chromatin-bound KRP4 CDK inhibitors act as a cell size sensor that is diluted by cell growth and is responsible for cell size-dependent cell cycle regulation at G1/S 47 .
Live-cell imaging of cell cycle markers in Arabidopsis shoot apical meristem demonstrated that both G1/S and G2/M transitions are regulated in a cell size-dependent manner. Accumulation of two waves of CDK activities underpinned by D-type cyclin and mitotic CDKB1, their interactors or downstream effectors were proposed as molecular components for cell size-dependent G1/S and G2/M regulation, respectively 31 . SCL28 may participate in regulating G2/M transition in this model, in which dilution of an inhibitor rather than accumulation of an activator is the principle behind. How SCL28 is connected to monitoring of cell size remains to be determined. Stochastic variability in the accumulation of key regulators can also set cell size and cell fate differences in a developing organ. An example is ATML1, a transcription factor that is necessary for the patterning of giant cells within the sepal epidermis by overcoming a threshold in G2 cells and activating the downstream SMR1, also called LOSS OF GIANT CELLS FROM ORGANS (LGO) 48 . Similar to ATML1, SCL28 shows a finely-tuned dose-dependent effect on cell size by accurately distinguishing gene copy numbers of SCL28 and reflecting its graded overexpression. It is possible that the molecular ruler in an inhibitor dilution model potentially involving SCL28 is set by the synthesis rate limited by gene copy number, as shown for Whi5 in budding yeast 45 .
The human functional orthologue of Whi5 is pRB, which also controls cell size via its synthesis mostly after G1 and subsequent dilution during growth 46 . In Arabidopsis, both RBR and E2FB were shown to strongly affect cell size in proliferating cells 49,50 , but as opposed to SCL28 that solely acts on cell proliferation rate, the RBR-E2F pathway also regulates cell cycle exit, establishment of quiescence, and cellular differentiation 29,51,52 . The repressortype MYB3Rs-MYB3R1, MYB3R3 and MYB3R5-when combined with RBR and E2FC form the DREAM complex that represses mitotic genes during or after developmental cessation of cell proliferation 11 . Whether the RBR/E2F/DREAM pathway directly regulates the cell size in proliferating cells or its cell size effect is through the regulation of cell cycle exit, remains to be established. It will be also important to know how the RBR/E2F/ DREAM and the SCL28/AtSMOS1 pathways are coordinated. In addition to cell size during proliferation, SCL28 also strongly affects the size of terminally differentiated cells. In the scl28 mutant, we observed a reduced final cell size in various organs, whereas moderately higher SCL28 expression led to its increase. Reduced cell size caused by increased cell division, as observed in scl28, has been repeatedly reported in Arabidopsis organ development. Examples include plants with inhibited RBR expression, those overexpressing CYCD3;1, or cooverexpressing E2FA and DPA [50][51][52][53] . In many of these cases, the cellular phenotype is due to prolonged cell proliferation and delayed cell differentiation, leading to additional cell division that is not balanced by cell size increase due to developmental decrease of cell expansion. On the contrary, we showed with kinematic analysis during leaf development that the increased number of smaller cells observed in scl28 leaves is caused by accelerated cell proliferation, but not by prolonged cell proliferation. Therefore, it is puzzling how loss of SCL28 affects final cell size that is largely determined by post-mitotic cell expansion occurring in the absence of SCL28.
Our observations are reminiscent of the phenomena reported in Drosophila wings, where the alteration of cell number is compensated by changes in cell size such that final organ size is essentially unchanged. Based on this and similar observations, it has been proposed that animal organ size may be controlled by 'total mass checkpoint' mechanisms, which operate at the level of whole organs 54 , although the underlying molecular details remain unknown. Instead, the well-established mechanism of 'compensation' in plants shows post-mitotic cell expansion that is enhanced by reduced cell proliferation 55 . A small number of larger cells observed in SCL28 OE leaves may be a prime example of compensation. One possible explanation for compensatory alteration of cell expansion is that, during cell proliferation, cells establish memory that is maintained during the elongation phase. An example of this can be seen in plants overexpressing KRP2, where final cell size may be influenced by the size of mitotic cells that are already altered during proliferation 56 . We observed larger and smaller mitotic cells both in root meristem and young leaves of SCL28 OE and slc28, respectively, which may be maintained until terminal differentiation, thus affecting final cell size. A possible molecular mechanism to create such memory is cell sizedependent alteration in gene expression. It was shown in yeast that cell size changes due to ploidy levels are associated with altered gene expression related to functions of cell wall and extracellular matrix 57 . Similarly, in Arabidopsis root, ploidy levels were correlated with expression of genes related to chromatin and cell expansion, such as ion transport and cell wall modifications 58 . Therefore, it is plausible that SCL28 action in meristematic cells has long-lasting effects through cell expansion on final cell size in differentiated tissues. Another possible mechanism for cell size memory involves a monitoring system for cellular dimensions that shares components in proliferating and expanding cells. Supporting this idea, it has been suggested that cells expand to a target size by a mechanism requiring a cell size measurement device, from an approach combining quantitative analysis and mathematical modeling of variability in final cell size of Arabidopsis root 59 . Further careful studies are required to clarify how final cell size is affected by expression levels of SLC28 that is confined to proliferating cells.
In summary, we identified a novel hierarchical MYB3R-SCL28/ AtSMOS1-SMR transcriptional regulatory pathway leading to cell cycle inhibition at G2/M that regulates cell size and number during organ development without dramatically altering organ size. This mechanism may help to adjust cell size to optimize cellular and tissue performance such as metabolic functions or mechanical properties.
qRT-PCR analysis. Total RNA was extracted using TRIzol reagent (Thermo Fisher) and used for synthesis of first-strand cDNA with ReverTra Ace qPCR RT Master Mix (Toyobo), according to the manufacture's instruction. A real-time qPCR was conducted using Thunderbird SYBR qPCR Mix (Toyobo) on StepO-nePlus Real-Time PCR Systems (Applied Biosystems). Three biological replicates were run for each measurement, and results were normalized to the expression levels of UBQ5 mRNA unless otherwise mentioned. Primers used for qRT-PCR are listed in Supplementary Table 1.
BiFC assay. For construction of plasmids used for BiFC assays, the entire coding regions of SCL28 and AtSMOS1 were amplified by PCR, and cloned into donor vectors pDONR201 and pDONR207 using BP Clonase II (Thermo Fisher), respectively. The cloned fragments were then transferred into the destination vectors pGWnY and pGWcY 61 using LR clonase II (Thermo Fisher) to generate C-terminal fusions of YFP fragments. Transient gene expression using Arabidopsis leaf mesophyll protoplasts was performed as described previously 62 . The transfected protoplasts were incubated overnight in the dark at 22°C. YFP fluorescent was observed with fluorescent optics on a BX54 microscope (Olympus).
LUC reporter assay. Preparation of protoplasts from Arabidopsis T87 suspension cultured cells and polyethylene glycol-mediated gene transfer were performed as described previously 63 . An empty vector containing 35 S promoter (pJIT-60) and 35 S:hRLUC plasmid expressing the humanized Renilla LUC (hRLUC) were used as a negative control and an internal control, respectively. Protoplasts (1.5 × 10 5 for each transfection) were co-transfected with 15 µg each of the LUC reporter and expression plasmids, then incubated at 22°C for 20 h before measuring LUC activities, which was performed using the Dual-Luciferase Reporter Assay system (Promega) and Luminoskan Ascent luminometer (Thermo Fisher). LUC activity was normalized according to hRLUC activity in each assay, and the relative ratio was determined.
Plasmid construction. To construct GUS fusion reporters, the upstream region of SCL28 (2.2 kb) was amplified by PCR and cloned into the pENTR/D-TOPO vector (Thermo Fisher), then transferred to pBGGUS 10 through Gateway LR reaction to create proSCL28::GUS. The PCR-based site-directed mutagenesis was performed to change all four MSA core motifs, AACGG, in the SCL28 promoter to ATTGG, resulting in generation of proSCL28ΔMSA::GUS.
To prepare the construct expressing SCL28 fused to GFP at its C-terminus (SCL28-GFP) under the control of the native promoter, the entire SCL28 genomic region (5.3 kb) containing 2.2-kb promoter was amplified by PCR using genomic DNA from Arabidopsis (Col), and cloned into pENTR/D-TOPO. The resulting plasmid was then used for In-Fusion reaction (Takara Clontech) to insert PCR- amplified GFP fragments at the C-terminus of the SCL28 coding sequence, to create entry plasmid containing the proSCL28::SCL28-GFP fusion construct, which was then transferred to the binary vector pGWB501 64 .
To prepare the AtSMOS1-GFP fusion construct driven by its own promoter, the entire AtSMOS1 genomic region (4.4 kb) containing 1.2 kb promoter was amplified by PCR using genomic DNA from Arabidopsis (Col) and cloned into pDONR201 through Gateway BP reaction (Thermo Fisher). The resulting plasmid was then used for In-Fusion reaction (Takara Clontech) for inserting PCR-amplified GFP fragment at the C-terminus of AtSMOS1, to create entry plasmid containing proAtSMOS1::AtSMOS1-GFP fusion construct, which was then transferred to the binary vector pPZP211-GW 65 .
For construction of a LUC reporter plasmid, promoter region of SMR2 (2.0 kb) was amplified by PCR and cloned into HindIII-BamHI interval of pUC-LUC 12 to obtain proSMR2::LUC. For construction of expression plasmids of AtSMOS1 and SCL28 (pro35S::AtSMOS1 and pro35S::SCL28), the entire coding regions were amplified by PCR using cDNA prepared from Arabidopsis T87 cells as a template. The resulting PCR fragments were cloned into pENTR/D-TOPO and then transferred to pJIT60 12 at the site downstream of 35S promoter through the Gateway LR reaction.
To create proRPS5A::SCL28-GFP, the entire coding sequence of SCL28-GFP fusion was amplified by PCR using proSCL28::SCL28-GFP plasmid as a template and cloned into pENTR/D-TOPO. The resulting plasmid was then used for the Gateway LR reaction to transfer the SCL28-GFP fragment downstream of RPS5A promoter in pPZP221 13  Transcriptome analysis. Microarray analyses were performed using an ATH1 GeneChip (Affymetrix). Total RNA was extracted from whole seedlings by RNeasy Plant Mini Kit (Qiagen), and reverse-transcribed, labeled with an Affymetrix 3′ IVT Express Kit (Affymetrix), and used for hybridization to the chip according to the supplier's protocol. Data analysis was performed using Microarray Suite ver. 5 (Affymetrix) and GeneSpring 7.1 (Agilent Technologies). For transcriptome profiling in WT and SCL28 OE plants, whole seedlings at 9 DAS were analyzed with three biological replicates. Genes with FDR < 0.05 were defined as differentially expressed genes in SCL28 OE compared with WT plants.
For RNA-Seq analyses, total RNA was extracted as above and used to construct cDNA libraries with the TruSeq RNA Library Preparation Kit v2 (Illumina, United States) according to the manufacturer's protocol. For transcriptome profiling in WT, scl28, and atsmos1 plants, whole seedlings at 9 DAS were analyzed with three biological replicates. The libraries were sequenced using the NextSeq500 sequencer (Illumina, United States). Raw reads containing adapter sequences were trimmed using bcl2fastq (Illumina, United States), and nucleotides with low-quality (QV < 25) were masked by N using the original script. Reads shorter than 50 bp were discarded, and the remaining reads were mapped to the cDNA reference using Bowtie with the following parameters: "-all-best-strata" 66 . The reads were counted by transcript models. Differentially expressed genes were selected based on the adjusted P-value calculated using edgeR (version 3.20.9) with default settings 67 .
Computational analysis of ChIP-Seq data. Single-end sequencing of ChIP samples was performed using Illumina NextSeq 500 with a read length of 76 bp. Reads were quality controlled using FASTQC (http://www.bioinformatics.babraham.ac.uk/projects/ fastqc/). Trimmomatic was used for quality trimming. Parameters for read quality filtering were set as follows: minimum length of 36 bp; mean Phred quality score greater than 30; and leading and trailing bases removal with base quality below 5. Reads were mapped onto the TAIR10 assembly using Bowtie 68 with mismatch permission of 1 bp. To identify significantly enriched regions, we used MACS2 69 . Parameters for peak detection were set as follows: number of duplicate reads at a location: 1; mfold of 5: 50; q value cutoff: 0.05; extsize 200; and sharp peak. Visualization and analysis of genomewide enrichment profiles were performed with Integrated Genome Browser. Peak annotations such as proximity to genes and overlap of genomic features, including transposons and genes were performed using BEDTOOLS INTERSECT. NGSplot was used to profile enrichment at TSSs and along the gene 70 . Spatial binding of the AtS-MOS1 and SCL28 peaks were performed by position-wise comparison using a binning approach and plotted in hexplot. De novo motif analysis of both SCL28 and AtSMOS1 binding regions were screened using HOMER 71 .
Kinematic analysis of leaf growth. After plants were fixed in a 9:1 of ethanol and acetic acid solution and cleared with Hoyer's solution (a mixture of 100 g chloral hydrate, 10 g glycerol, 15 g gum arabic, and 25 mL water), we performed microscopic observations using 1st or 2nd leaves as described previously 13 . After whole leaf images were captured, palisade cells at positions one-fourth and three-fourth from the tip of each leaf were observed with differential interference contrast (DIC) microscope (BX51, Olympus). The captured images were analyzed using ImageJ (ver.2.1.0; rsb.info.nih.gov/ij) and the average size of palisade cell, the number of cells in the uppermost layer of palisade tissue per leaf, and cell division rate were calculated according to methods described previously 72 .
Ploidy analysis. For ploidy analysis, whole leaves from plants at 8-20 DAS were used. Nuclei were isolated by chopping whole leaves with razor blade in Nuclei extraction buffer of CyStain UV precise P kit (Sysmex), filtered through 30 μm mesh, and stained with DAPI by adding Staining buffer of CyStain UV precise P kit. After incubation for 10 min, samples were analyzed with CyFlow Ploidy Analyzer (Sysmex) according to the manufacturer's instructions. For estimating population of nuclei in each peak, baselines of ploidy distribution profiles were calculated using the polynomial trendline function of Microsoft Excel. The value of baseline at each position was subtracted from the corresponding value of raw data for calculating the total count in each peak.
Analysis of meristem cell number and cell size in roots. To visualize cell outlines, roots from plants at five DAS were stained with 0.05 mg mL −1 propidium iodide and observed by confocal laser scanning microscopy (CLSM), using an inverted fluorescence microscope (Eclipse Ti2, Nikon) equipped with a confocal scanning unit (A1, Nikon). The resulting images were processed using ImageJ software to measure cell length in the root meristem. Root meristem size was measured by counting the number of cortical cells between the quiescent center and the first elongated cell.
Cell cycle analysis by time-lapse imaging. Seedlings at six DAS carrying PCNA-GFP were transferred onto an MS medium in a glass-bottom dish, and root meristem cells were observed by CLSM as described above. Time-lapse images were acquired every 30 min for 15 h. To avoid long-time imaging that possibly damages the samples, the duration of the G1/S and G2/M phases was measured separately. Based on PCNA-GFP fluorescence patterns, duration of each cell cycle phase was analyzed for epidermal cells in hair cell lineage.