NGATHA-LIKEs Control Leaf Margin Development by Repressing CUP-SHAPED COTYLEDON2 Transcription1

The leaf margin is a fascinating feature of leaf morphology, contributing to the incredible diversity of leaf shapes and forms. As a central regulator of plant organ separation and margin development, CUP-SHAPED COTYLEDON2 (CUC2), a NAM, ATAF1, 2, CUC2 (NAC)-family transcription factor, governs the extent of serrations along the leaf margin. CUC2 activity is tightly regulated at transcriptional and posttranscriptional levels. However, the molecular mechanism that controls CUC2 transcription during leaf development has not been fully elucidated. Here we report that Arabidopsis (Arabidopsis thaliana) NGATHA-LIKE1 (NGAL1) to NGAL3, which are three related B3 family transcription factors, act as negative regulators of leaf margin serration formation. Over-expression of NGALs led to “cup-shaped” cotyledons and smooth leaf margins, whereas the triple loss-offunction mutant ngaltri exhibited more serrated leaves than the wild type. RNA-sequencing analyses revealed that the expression levels of a number of transcription factor genes involved in leaf development are regulated by NGALs, including CUC2. Comparative transcriptome analyses further uncovered a significant overlap between NGALand CUC2-regulated genes. Moreover, genetic analyses using various combinations of gainand loss-of-function mutants of NGALs and CUC2 confirmed that CUC2 acts downstream of NGALs in promoting the formation of leaf-margin serrations. Finally, we demonstrate that NGAL1 directly binds to the CUC2 promoter causing repressed CUC2 expression. In summary, direct CUC2 transcriptional repression by NGAL1 characterizes a further regulatory module controlling leaf margin development.

Leaves are essential organs that play central roles in plant growth and development (Byrne, 2012). In nature, a wide array of leaf traits, including sizes and shapes, enable the adaptation of higher plants to vastly different natural environments and habitats (Chitwood and Sinha, 2016;Nikolov et al., 2019). Leaf margin, meaning the edge of the leaf blade, is a major leaf trait and can adopt a fascinatingly diverse range of shapes from smooth, serrated, and lobed to compound with leaflets (Nikolov et al., 2019). The regulation of leaf margin formation has long been a subject of interest to plant biologists, and numerous factors including phytohormones, transcription factors (TFs), and microRNAs (miRNAs) have been shown to govern the elaboration of leaf margins (Bar and Ori, 2014;Fouracre and Poethig, 2016;Nikolov et al., 2019). Studies in the model plant Arabidopsis (Arabidopsis thaliana) and its relative Cardamine hirsuta suggest that the patterning of leaf margin is likely governed by conserved genetic factors (Blein et al., 2008;Vlad et al., 2014;Nikolov et al., 2019). In Arabidopsis, the NAM, ATAF1, ATAF2, CUC2 (NAC)-family TFs CUP-SHA-PED COTYLEDON1 (CUC1), CUC2, and CUC3 function as key regulators in organ boundary formation (Aida et al., 1997;Vroemen et al., 2003;Hibara et al., 2006;Nikovics et al., 2006). CUC1 and CUC2 genes were originally isolated through the intriguing "cup-shaped cotyledon" phenotype of the cuc1 cuc2 double mutants (Aida et al., 1997;Hake, 2019). Loss-of-function mutations of CUC2 or CUC3 give rise to smooth leaf margins, indicating that they promote leaf margin outgrowth and serration formation, whereas the loss of CUC1 alone does not impact leaf margin development (Nikovics et al., 2006;Kawamura et al., 2010;Hasson et al., 2011). Although CUC2 and CUC3 have partially redundant functions in determining leaf margin, the timing of their actions appears to be distinct Maugarny-Calès et al., 2019). CUC2 is critical for leaf margin patterning and the initiation of tooth outgrowth at the early stages of leaf development, whereas CUC3 participates in the maintenance of tooth outgrowth at later stages Maugarny-Calès et al., 2019).
The phytohormone auxin intimately regulates leaf margin development (Hay et al., 2006;Bilsborough et al., 2011). During early leaf development, the establishment of auxin maxima along the leaf margin, which depends on the activity of auxin efflux carrier PIN FORMED1 (PIN1), is crucial for the formation of serrations (Hay et al., 2006). Disruption of auxin distribution by pharmacological treatments with auxintransport inhibitors produces leaves with smooth margins (Hay et al., 2006). Without a functional PIN1, leaf-margin serrations cannot be formed even when CUC2 expression is up-regulated (Hay et al., 2006;Bilsborough et al., 2011). Along the margin of emerging leaves, a negative feedback loop in which CUC2 promotes the formation of auxin maxima via PIN1, and in turn auxin antagonizes CUC2 transcript accumulation, eventually leads to an interspersed pattern of auxin maxima and CUC2 expression levels, corresponding to leaf margin outgrowths and sinuses (Bilsborough et al., 2011). In addition, mutations in auxininflux transporters or modulation of auxin responses also affect leaf margin serrations (Kasprzewska et al., 2015;Tameshige et al., 2016). miRNA also participates in leaf margin regulation (Palatnik et al., 2003;Laufs et al., 2004;Mallory et al., 2004;Nikovics et al., 2006). CUC1 and CUC2 transcripts contain recognition sites for miR164s, including miR164A, miR164B, and miR164C, and are posttranscriptionally targeted by miR164s, whereas CUC3 is not a target of miR164 in Arabidopsis (Laufs et al., 2004;Mallory et al., 2004). The absence of miR164A activity leads to the stabilization and increased accumulation of CUC2 transcript, giving rise to more serrated leaf margins, whereas the over-expression of miR164s cause smooth leaf margins, indicating that miR164s negatively regulate leaf margin formation through posttranscriptional repression of CUC2 (Laufs et al., 2004;Nikovics et al., 2006). Mutants harboring miRNA164-resistant CUC2 variants, such as the cuc2-1D mutant, display deeply serrated leaf margins (Nikovics et al., 2006;Larue et al., 2009). In addition to miR164s, CUCs are negatively regulated by CINCINNATA-like (CIN-like) TEOSINTE BRANCHED1, CYCLOIDEA, and PCF (TCP) TFs at multiple levels (Koyama et al., 2007(Koyama et al., , 2010Rubio-Somoza et al., 2014). First, CIN-like TCPs negatively affect the expression levels of CUCs through direct activation of miR164A (Koyama et al., 2007(Koyama et al., , 2010. Moreover, at the posttranslational level, TCP4 directly interacts with CUC2 and CUC3, interfering with the formation of functional CUC dimers (Rubio-Somoza et al., 2014). Interestingly, TCPs are also targets of miR319s (Koyama et al., 2017). The complex interaction between the miR319-TCPs module and the miR164-CUCs module highlights the multilayered regulation of leaf margin development. Despite the tremendous progress in our understanding of the regulation of CUC2 activities, much remains poorly understood regarding the regulatory mechanism of CUC2 gene expression.
The Arabidopsis NGATHA1 (NGA1) to NGA4 and NGATHA-LIKE1 (NGAL1) to NGAL3 TFs are members of the RAV subfamily of the plant-specific B3 family TFs (Swaminathan et al., 2008;Alvarez et al., 2009;Trigueros et al., 2009). NGA TFs are involved in the development of gynoecium (Alvarez et al., 2009;Trigueros et al., 2009). Loss of the NGA genes also results in enhanced leaf margin serration, whereas the ectopic expression of NGA genes induce cotyledon fusions (Alvarez et al., 2009;Trigueros et al., 2009). NGAs act redundantly with CIN-like TCPs in limiting the marginal growth of leaves and other lateral organs (Alvarez et al., 2016). Interestingly, over-expression studies have also pointed to NGALs as negative regulators of leaf margin serration. The over-expression of NGAL3, also known as DEVELOPMENT-RELATED PcG IN THE APEX 4 (DPA4), leads to smooth leaf margins, whereas a loss-of-function allele gives a slightly more serrated leaf-margin phenotype (Engelhorn et al., 2012). The identification of an activation tagging gain-of-function allele of NGAL1/ABNORMAL SHOOT2 (ABS2) also showed that the elevated expression of NGAL1/ABS2 prevents the formation of leaf margin serration (Shao et al., 2012). However, how the three NGALs work together in regulating leaf margin development remains unclear.
In this work, we further investigated the molecular mechanism underlying the regulation of leaf margin development by the three NGAL TFs. We discovered that strong over-expression of each of the three NGAL genes can lead to a striking cup-shaped cotyledon phenotype, reminiscent of the cuc1 cuc2 double mutants (Aida et al., 1997;Hake, 2019). In addition to NGAL1/ABS2, over-expression of NGAL2 and NGAL3 also led to a smooth leaf margin, whereas the triple loss-of-function mutant ngaltri, defective in all three NGALs, showed more pronounced leaf margin serration, indicating that NGALs act redundantly as negative regulators of leaf margin serration formation. Transcriptome profiling revealed strong negative correlations between the expression of NGALs and CUC genes. Genetically, a loss-of-function cuc2 mutation suppressed the leaf margin phenotype, as well as the abnormal transcriptional response in the NGALs triple-mutant ngaltri. The introduction of a gain-of-function CUC2 allele, cuc2-1D, into ngaltri caused dramatic enhancement of leaf margin outgrowths and lobed leaves in the ngaltri cuc2-1D quadruple mutant. Moreover, ngaltri triple mutation exacerbated the gynoecium development defects of cuc2-1D. Finally, we demonstrate that NGAL1 directly binds to the promoter of CUC2, resulting in repressed CUC2 expression. Together, our results establish a regulatory scheme whereby NGALs negatively regulate leaf margin serration formation in a redundant fashion by directly repressing the expression of CUC2.

Over-expression of NGALs Leads to the Formation of Cup-Shaped Cotyledons
Previously, we have shown that NGAL1/ABS2 regulates petal development in Arabidopsis (Shao et al., 2012). NGAL1/ABS2 is one of three closely related B3 TFs, alongside NGAL2 and NGAL3 ( Fig. 1A; Swaminathan et al., 2008). To further dissect the functions of these TFs, we first took a gain-of-function approach, constructed vectors of each individual NGAL gene under the control of the constitutive Cauliflower Mosaic Virus 35S promoter (p35S:NGAL1, p35S:NGAL2, and p35S:NGAL3), and analyzed the consequences of NGAL1/ABS2, NGAL2, and NGAL3 overexpression (OE), respectively. Unexpectedly, when screening for T1 transgenic lines, we observed that NGAL1 OE seedlings showed varied degrees of cotyledon fusion, with the most severe type displaying a complete cotyledon fusion (i.e. cup-shaped cotyledons; Fig. 1B). The cotyledon fusion phenotypes and cup-shaped cotyledons were also observed with NGAL2 and NGAL3 OE lines (Fig. 1, D and F). We categorized NGAL OE lines into three phenotypic types. Type I refers to transformants displaying fully fused cup-shaped cotyledons, whereas type II refers to those with partially fused cotyledons or a single cotyledon. NGAL OE lines with two separated cotyledons were considered as weak OE lines. In OE lines of all three NGALs, the severity of cotyledon fusion correlated with the extent of NGALs over-expression ( Fig. 1, C, E, and G). Notably, the penetrance of cotyledon fusion varied among the OE transformants of the three NGALs (Table 1). NGAL1 OE lines showed the highest frequency (23.46%, 57 of 243) of type-I phenotype, whereas NGAL3 OE lines showed the lowest frequency (1.25%, 3 of 240) of type-I phenotype. The cupshaped cotyledon phenotype was originally described in the cuc1 cuc2 double mutant (Aida et al., 1997). Similar to the cuc1 cuc2 mutant, transgenic plants with highly elevated expression of the three NGAL genes and cup-shaped cotyledons did not survive the cotyledon stage. Taken together, these findings suggest that NGALs play a role in specifying cotyledon organ separation and boundary determination.

NGALs Regulate Leaf Margin Development
Whereas strong NGAL OE lines exhibited cotyledon fusion, weaker OE lines of NGALs displayed smoother leaf margins compared with the wild type ( Fig. 2A; Supplemental Fig. S1A). This is consistent with our previous report that the elevated expression of NGAL1/ABS2 in activation tagging line abs2-1D caused smooth leaf margins (Shao et al., 2012). The weaker OE lines of NGALs showed reduced seed set but were nonetheless viable.
To establish the functions of NGAL TFs in a loss-offunction context, we identified transfer DNA (T-DNA) insertion mutants for the three NGAL genes and designated them ngal1-2, ngal2-1, ngal3-1, and ngal3-2, respectively (Supplemental Fig. S1, B-D). Reverse transcription-quantitative PCR (RT-qPCR) analysis confirmed that NGAL transcript levels were greatly reduced in the corresponding T-DNA mutants (Supplemental Fig.  S1, E-G). Phenotypically, the ngal1-2 and ngal2-1 single mutants were indistinguishable from wild type under our growth conditions (Supplemental Fig. S1, H and I). Consistent with a previous finding, both ngal3-1 and ngal3-2 displayed slightly more evident leaf serrations compared with those of the wild type (Supplemental Fig. S1, H and I;Engelhorn et al., 2012). To investigate the genetic redundancy of NGALs, we generated a ngal1-2 ngal3-1 double mutant and a ngal1-2 ngal2-1 ngal3-1 triple mutant (referred to as ngaltri hereafter). Two quantifiable parameters of leaf margin serration, the number of leaf margin serrations and the ratio of tooth area over total leaf area, were used to compare the leaf margin morphology between the wild type and single, double, and triple ngal mutants (Zheng et al., 2016). The extent of leaf serration increased with the increased loss of NGAL genes (Fig. 2, B-D; Supplemental Fig. S1J). Moreover, expressing a NGAL1-GFP fusion gene driven by the NGAL1 native promoter (pNGAL1:NGAL1-GFP) was sufficient to complement the leaf serration phenotype of ngaltri (Supplemental Fig. S2). These findings indicate that NGALs share redundant roles in suppressing leaf margin serration.
Finally, to determine whether NGALs affect tooth outgrowth during early stages of leaf development, we examined emerging young leaves (the fifth, sixth, seventh, and eighth rosette leaves of 19-d-old seedlings) in wild type, abs2-1D/1 heterozygote, and ngaltri. We did not use the abs2-1D homozygote due to its sterility and extremely stunted growth (Shao et al., 2012). In all the leaves examined, abs2-1D/1 showed smooth leaf margins and ngaltri showed more pronounced leaf serrations than those of the wild type (Supplemental Fig.  S3). The difference in leaf margin morphology between the three genotypes was most pronounced in the youngest eighth rosette leaf (Supplemental Fig. S3). Together, results from our gain-of-function and loss-of-function mutant analyses indicate that the three NGALs work together in the control leaf margin development and NGALs are negative regulators of leaf margin outgrowth.

NGALs Modulate Transcript Levels of CUC Genes
NGALs belong to the B3 family of TFs and are potential transcription repressors (Swaminathan et al., 2008). To elucidate the role of NGALs in transcriptional regulation, we performed RNA-seq analyses using tissues including the shoot apical meristem and emerging young leaves in wild type, abs2-1D/1 heterozygote, and ngaltri. Based on the RNA-seq data and RT-qPCR analysis, among the three NGALs, only NGAL1 is overexpressed in abs2-1D/1 (Supplemental Fig. S4). In pairwise comparisons, genes whose expression levels are increased or decreased more than twofold and have a false discovery rate (FDR) , 0.05 are considered as differentially expressed genes (DEGs). When compared with wild type, 1,009 and 681 DEGs were identified in abs2-1D/1 and ngaltri, respectively (Supplemental Table S1). The expression changes of most DEGs in abs2-1D/1 were reversed in ngaltri (Fig. 3A). A similar pattern was also observed when comparing the expression changes of DEGs identified in ngaltri in abs2-1D/1 (Fig. 3B). Hormone metabolism, development, and cell wall were among the most highly enriched functional categories in DEGs identified in abs2-1D/1 and ngaltri ( Fig. 3C; Supplemental Table S2). In addition, a number of TF genes that are known regulators of leaf development were found in the NGALs-regulated genes (Fig. 3D). Among these, the leaf margin key regulators, CUC2 and CUC3, were repressed in abs2-1D/1 but up-regulated in ngaltri ( Fig. 3D; Nikovics et al., 2006;Hasson et al., 2011). The alterations in CUCs expression levels were further verified with RT-qPCR. We found significantly lowered CUC2 and CUC3 transcript levels in abs2-1D/ 1 compared with those in the wild type, whereas expression levels of CUC2 and CUC3, as well as CUC1, were increased in ngaltri (Fig. 3E). Next, we analyzed the effect of NGAL1 over-expression on CUC2 tissue expression pattern. A reporter line expressing the b-glucuronidase (GUS) gene driven by the CUC2 promoter (pCUC2:-GUS) was crossed with abs2-1D/1, and the expression pattern of pCUC2:GUS in the F1 segregating population was analyzed by GUS staining. As previously reported, pCUC2:GUS signal was detected in the basal part of emerging young leaves, and the highest expression of pCUC2:GUS was observed at the leaf sinus in wild-type background ( Fig. 3F; Nikovics et al., 2006). By contrast, we only detected traces of pCUC2:GUS signal at the base of emerging young leaves in the abs2-1D/1 background (Fig. 3F). In addition, using the complementation line ngaltri pNGAL1:NGAL1-GFP, we analyzed the expression pattern of NGAL1. Intriguingly, the highest expression of NGAL1-GFP was detected at the tips of teeth along the leaf margin, coinciding with the low expression of CUC2 at the tips of teeth (Fig. 3G). These results suggest that NGALs may control leaf margin development through modulating the expression of leaf development-related genes, particularly the CUC genes.

The Regulation of Leaf Margin Development by NGALs Is CUC2-dependent
Because CUC2 plays a more prominent role in tooth formation, especially in the patterning of leaf margin at early stages of leaf development, we next sought to investigate the functional relationship between NGALs and CUC2 Maugarny-Cales et al., 2019). To this end, we first took a genetic approach by constructing quadruple mutant of ngaltri and cuc2-101, a T-DNA knockout allele of cuc2, and examined the genetic interaction between NGALs and CUC2 (Zheng et al., 2016). ngaltri mutants have more serrated leaves, whereas loss of CUC2 leads to smooth leaf margins in cuc2-101 (Fig. 4, A and B). Notably, in ngaltri cuc2-101, both the number and the size of leaf teeth were markedly reduced compared with those in the ngaltri, suggesting the activity of CUC2 is necessary for the manifestation of leaf margin phenotype caused by ngaltri (Fig. 4, A-C).
Next, we examined the functional relationship between NGALs and CUC2 at the transcriptome level. Pairwise comparison of the RNA-seq data of cuc2-101 and wild type revealed that 477 genes were differentially expressed in cuc2-101 (Supplemental Table S3). Among the 477 CUC2-regulated genes, 238 (nine expected randomly) were also regulated by NGALs (Fig. 4D). This significant overlap between NGALs-and CUC2-regulated genes (P 5 6.286116e2294, hypergeometric test) suggests that they may control a common transcriptional program. Moreover, heatmap and scatter plot analyses comparing the expression levels of the 681 NGALs-regulated genes in ngaltri cuc2-101 versus cuc2-101 revealed that the overall effects of ngaltri on the gene expression was diminished in the cuc2-101 background, when compared with the wild-type background (Fig. 4, E and F; Supplemental Table S4). To quantify the effects of cuc2-101 on the NGALs regulate leaf margin development. A, Rosette leaves of wild-type (WT) and weak NGAL OE lines. B, Rosette leaves of wild type and ngaltri. In A and B, starting with the third leaf, leaves were arranged in the order of initiation. C, Silhouettes of the eighth, ninth, and tenth rosette leaves (R8, R9, and R10) in wild type and ngaltri. Bars 5 1 cm (A and B) and 0.5 cm (C). D, Quantification of leaf margin serrations of R8, R9, and R10 in wild type and ngaltri with two parameters, the ratio of tooth area over total leaf area, and the number of leaf margin serrations. Data are means 6 SD (n $ 8). Asterisks indicate significance by Student's unpaired t test (**P , 0.01 and ***P , 0.001).
NGALs-regulated genes, we define those genes that are DEGs in ngaltri versus wild type, but not DEGs in ngaltri cuc2-101 versus cuc2-101 as CUC2-dependent genes. Astonishingly, 96.5% of down-regulated genes and 82.2% of up-regulated genes, which together account for 93.7% of total DEGs in ngaltri, are CUC2 depenent (Fig. 4G). These data indicate that NGALs regulate gene expression in a largely CUC2-dependent manner. Together, genetic and molecular evidence support the hypothesis that CUC2 acts downstream of NGALs in the pathway that controls leaf margin development.
NGALs Down-regulate CUC2 Transcript Level Independently of miR164s A well-established regulatory module for CUC2 is that the accumulation of CUC2 transcript is negatively regulated by miR164s (Nikovics et al., 2006). To test whether NGALs negatively regulate CUC2 transcript accumulation through the miR164 pathway, we took advantage of a gain-of-function allele of CUC2, cuc2-1D, which carries a point mutation at the miR164 target site (Larue et al., 2009). cuc2-1D accumulates more CUC2 transcript and exhibits more serrated leaves, in MapMan functional categories in DEGs found in abs2-1D/1/wild type and ngaltri/wild type. D, Leaf development genes that are regulated by abs2-1D/1 and ngaltri in opposite directions. E, RT-qPCR analysis of CUC1, CUC2, and CUC3 expression levels in wild type, abs2-1D/1, and ngaltri. Gene expression levels are shown as fold changes with respect to the expression levels in wild type. Data are means 6 SD (n 5 3). Asterisks indicate significance by Student's unpaired t test (**P , 0.01 and ***P , 0.001). F, pCUC2:GUS expression in wild-type and abs2-1D/1 backgrounds. G, Expression pattern of NGAL1-GFP in two independent ngaltri pNGAL1:NGAL1-GFP complementation lines. Wild type served as a negative control for detecting GFP fluorescence. Bars 5 500 mm.
contrast with the smooth leaf margin phenotype of the loss-of-function cuc2-101 allele (Larue et al., 2009). We generated ngaltri cuc2-1D quadruple mutant, and ngaltri cuc2-1D displayed a more dramatic leaf-margin phenotype compared with either ngaltri or cuc2-1D (Fig. 5, A and B). Leaves of ngaltri cuc2-1D were lobed with deep indentations along the leaf margins (Fig. 5, A  and B). Quantification of leaf margin characteristics showed that the number of serrations was increased in cuc2-1D, ngaltri, and ngaltri cuc2-1D to a similar extent compared with that in the wild type (Fig. 5C). However, the tooth area/leaf area ratio was increased in ngaltri cuc2-1D compared with cuc2-1D or ngaltri, consistent with the exaggerated indentations along leaf margins (Fig. 5C). Next, we analyzed CUC2 transcript level in wild type, cuc2-1D, ngaltri, and ngaltri cuc2-1D through RT-qPCR and detected increased CUC2 transcript level in cuc2-1D and ngaltri as expected (Fig. 5D). In ngaltri cuc2-1D, CUC2 transcript level was further elevated compared with that in cuc2-1D and ngaltri (Fig. 5D). The synergistic interaction between ngaltri and cuc2-1D suggests that NGALs likely modulate CUC2 transcript level in a miR164s-independent pathway.

NGALs Act Together with CUC2 in the Regulation of Gynoecium Development
The genetic interaction between NGALs and CUC2 were further confirmed by their involvement in gynoecium development. Arabidopsis wild-type gynoecium is composed of two carpels, fused along the replum and capped at the top with clearly defined stigma and style (Fig. 7). When examining the floral organs of ngaltri cuc2-1D quadruple mutants, we uncovered gynoecium development defects that are far more severe compared with those of ngaltri or cuc2-1D (Fig. 7). When compared with the wild type, ngaltri gynoecia are radially swelled, but remain fully fused (Fig. 7). In cuc2-1D, occasional abnormal outgrowths can be observed on the surface of the carpel, particularly along the replum, as reported ( Fig. 7; Larue et al., 2009). However, in ngaltri cuc2-1D, carpels are often not fused with ovules exposed (Fig. 7). These data indicate that the lack of NGALs enhances the gynoecium defects of cuc2-1D, and NGALs and CUC2 act together in the regulation of gynoecium development.

NGAL1 Directly Binds to the CUC2 Promoter and Represses CUC2 Expression
The strong negative correlations between the expression levels of NGALs and CUCs in NGALs gainand loss-of-function mutants prompted us to test whether CUC2 is a direct target of NGAL1. To this end, we carried out chromatin immunoprecipitation followed by quantitative PCR (ChIP-qPCR) analysis. p35S:NGAL1-HA (HA tag fused at C terminus of NGAL1) and pCUC2:GFP were coexpressed in wildtype mesophyll protoplasts, and ChIP was carried out with an anti-HA antibody. Immunoprecipitated DNA was analyzed by qPCR using six pairs of primers amplifying overlapping fragments (P1-P6) covering the CUC2 promoter and a pair of GFP-specific primers serving as a control (Fig. 8A). We found that two fragments, namely P2 and P4, were significantly enriched in Figure 5. ngaltri enhances leaf margin indentations in cuc2-1D. A, Rosette leaves of wild type (WT), cuc2-1D, ngaltri, and ngaltri cuc2-1D. Starting with the fifth leaf, leaves were arranged in the order of initiation. Bars 5 1 cm. B, Silhouette of R8, R9, and R10 in wild type, cuc2-1D, ngaltri, and ngaltri cuc2-1D. Bar 5 0.5 cm. C, Quantification of leaf margin serrations of R8, R9, and R10 in genotypes shown in B. Leaf margin serrations were quantified as in Figure 2D. Data are means 6 SD (n $ 8). D, RT-qPCR analysis of CUC2 expression in wild type, cuc2-1D, ngaltri, and ngaltri cuc2-1D. CUC2 expression levels are shown as fold changes with respect to the expression level in wild type. Data are means 6 SD (n 5 3). Asterisks indicate significance by Student's unpaired t test (**P , 0.01 and ***P , 0.001).
the immunoprecipitated DNA compared with other fragments of the CUC2 promoter or GFP (Fig. 8B), suggesting a direct association between NGAL1 and these two fragments within the CUC2 promoter.
Next, we tested whether the binding of NGAL1 to the CUC2 promoter could repress CUC2 promoter activity with a GFP-based reporter assay in wild-type mesophyll protoplasts. In this assay, the effector vector p35S:NGAL1-GR expresses a fusion protein with the glucocorticoid receptor (GR) fused to the C terminus of NGAL1 driven by the 35S promoter (Fig. 8C). The GR fusion allows us to control the nuclear localization of NGAL1, providing an inducible system to assess the activity of NGAL1. The reporter vectors express GFP driven by the CUC2 promoter (pCUC2) or a shortened version of CUC2 promoter (pCUC2 D597 ) in which a 597bp region encompassing P2 to P4 was deleted (Fig. 8C). An expression cassette containing UBQ10 promoterdriven mCherry was also included in the reporter vectors as a transfection control (Fig. 8C). When p35S:NGAL1-GR was cotransfected with pCUC2:GFP-pUBQ10:mCherry, signals of GFP fluorescence were markedly reduced in dexamethasone (DEX)-treated protoplasts compared with mock treatments (Fig. 8, D and E), suggesting that NGAL1 can repress CUC2 promoter activity. The partial deletion within the CUC2 promoter did not alter its ability to drive GFP expression (Fig. 8D). Upon DEX treatment, the repression of GFP by NGAL1 was greatly alleviated when GFP was driven by pCUC2 D597 , indicating that the deleted region is necessary for the repression by NGAL1 and that NGAL1 binding sites in the CUC2 promoter are situated in the deleted region, consistent with the ChIP-qPCR data (Fig. 8, D and E). Together, our results demonstrate that NGAL1 can bind to the promoter of CUC2 and repress CUC2 expression. We propose a model in which balanced activities of NGALs and CUC2 controls the outgrowth of leaf margin (Fig. 8F). High levels of NGAL1 cause CUC2 repression and, in turn, smooth leaf margins, whereas low levels or absence of NGAL1 relieve the repression of CUC2, resulting in more serrated leaf margins (Fig. 8F).
system for leaf margin investigations as wild type Arabidopsis plants typically display moderate serrations, or outgrowths, along the leaf margin, which has enabled the identifications of both positive and negative regulators of serration formation through molecular genetic analyses (Nikovics et al., 2006;Kawamura et al., 2010;Bilsborough et al., 2011;Engelhorn et al., 2012;Kasprzewska et al., 2015;Tameshige et al., 2016). Positive regulators such as the NAC-family TFs CUC1, CUC2, and CUC3 promote the formation of leaf margin serrations, and loss-of-function mutants of these genes show smoother leaf margins (Vroemen et al., 2003;Bilsborough et al., 2011). Conversely, when the activities of negative regulators such as miR164s are absent, mutants show more serrated leaf margins (Nikovics et al., 2006). Despite the identification of numerous factors and processes that regulate the leaf margin serration formation, much remains unknown regarding this fundamental process.
In this work, through molecular genetics characterizations, we establish that Arabidopsis NGAL1/ABS2, NGAL2, and NGAL3/DPA4 (together referred to as NGALs), three members of the RAV subfamily of the B3 TF gene family, act redundantly in the regulation of leaf margin development. Gain-of-function NGALs mutants or NGALs over-expression lines display smooth leaf margin, whereas ngaltri, the triple loss-of-function mutants of NGALs, show more serrated leaf margin, indicating that NGALs are negative regulators of leaf margin serration formation. It is interesting to note that the NGA clade of B3 TFs, closely related to the NGAL clade, is also involved in regulation of leaf growth and marginal patterning, suggesting that the RAV subfamily of plant B3 TFs may play a broad role in regulating leaf margin development (Alvarez et al., 2009(Alvarez et al., , 2016Ballester et al., 2015).
During the course of our study, we observed the intriguing cup-shaped cotyledon phenotype in strong NGALs OE lines (Fig. 1), reminiscent of the classic cupshaped cotyledon mutant cuc1 cuc2 (Aida et al., 1997;Nikovics et al., 2006;Hake, 2019). In Arabidopsis, the three CUC genes (CUC1, CUC2, and CUC3) are established as the key regulators of organ boundary determination, including leaf margin (Aida et al., 1997;Nikovics et al., 2006;Hake, 2019). We provide several lines of evidence that strongly suggest that NGALs exert their action through the repression of CUC activities, particularly CUC2. First, we observed that CUC1 to CUC3 transcript accumulated to higher levels in NGALs triple loss-of-function mutants (Fig. 3E). Previously, it was also shown that the loss of NGAL3/ DPA4 leads to slightly increased CUC2 transcript level, whereas NGAL3/DPA4 over-expression caused reduced CUC2 transcript level (Engelhorn et al., 2012). Second, we demonstrated genetically that the exaggerated leaf margin serration formation in NGALs triple loss-of-function mutants is dependent on CUC2 (Fig. 4, A-C). Furthermore, we showed dramatically enhanced leaf margin serration formation and lobe-shaped leaves in ngaltri cuc2-1D quadruple mutants and the suppression of cuc2-1D leaf margin serration by the gain-offunction allele of NGAL1, abs2-1D (Figs. 5 and 6). Because cuc2-1D expresses a miR164-resistant CUC2 variant, these findings suggest a miR164-independent regulation of CUC2 by NGALs (Larue et al., 2009). Third, through RNA-seq, we were able to determine the transcriptional profiles of gain-of-function and loss-offunction mutants of NGALs at the transcriptome level and show that the majority of the transcriptional alterations in ngaltri mutants were dependent on the presence of the CUC2 gene (Fig. 4). Finally, we showed in protoplasts that GFP expression driven by the CUC2 promoter was repressed by NGAL1 (Fig. 8). Moreover, we showed direct binding of NGAL1 to two positions on the CUC2 promoter (Fig. 8). Based on our findings, we proposed a model in which NGALs negatively regulate leaf margin outgrowth through direct repression of CUC2 transcription (Fig. 8F). Given that CUC2 expression is regulated by many factors including plant hormone auxin and miR164, it is possible that NGALs are one of the many pathways that converge on CUC2 and regulate its expression through different mechanisms. We noticed that in abs2-1D/1 cuc2-1D double mutant, the accumulation of CUC2 transcript, albeit much lower than in cuc2-1D, is still considerably higher than in wild type. However, abs2-1D/1 cuc2-1D showed completely smooth leaf margin (Fig. 6). One possible scenario is that the over-expression of NGAL1 in abs2-1D/1 leads to the repression of other tooth promoting factors that function independently of Figure 7. ngaltri enhances gynoecium defects in cuc2-1D. Gynoecium morphology in mature flowers of wild type (WT), cuc2-1D, ngaltri, and ngaltri cuc2-1D was examined with a stereo microscope (A) or a scanning electron microscope (B). Electron micrographs were pseudocolored (blue, stigma; purple, ovule; orange, filamentous structure found in cuc2-1D) to highlight the gynoecium defects in mutants. Bars 5 500 mm. Figure 8. NGAL1 directly binds to the CUC2 promoter and represses CUC2 expression. A, Diagram of the pCUC2:GFP vector used in protoplast ChIP-qPCR assay. Black lines (P1 to P7) indicated the fragments amplified in ChIP-qPCR assay. B, ChIP-qPCR analysis of NGAL1 DNA binding activity to P1 to P7 fragments in the pCUC2:GFP vector. Wild-type (WT) mesophyll protoplasts coexpressing p35S:NGAL1-HA and pCUC2:GFP were used in ChIP-qPCR. Data are means 6 SD (n 5 3). Lowercase letters on each bar indicated statistical differences by one-way ANOVA followed by Tukey's multiple comparisons test (P , 0.001). C, Diagrams of the reporter and effector vectors used in the protoplast reporter assays. Shaded parts indicated the 597-bp region deleted in pCUC2 D597 . D, Effector vector p35S:NGAL-GR was coexpressed with reporter vector pCUC2:GFP-pUBQ10:mCherry or pCUC2 D597 :GFP-pUBQ10:mCherry in wild-type mesophyll protoplasts. DEX-treated or mock-treated protoplasts were examined with florescence microscopy. Experiments were independently conducted three times with similar results. E, Quantification of GFP relative fluorescence intensity of protoplasts treated as in (D). Data are means 6 SD (n . 100). Asterisks indicate significance by Student's unpaired t test (***P , 0.001). F, model of transcriptional repression of CUC2 by NGAL1 leading to inhibited formation of leaf margin serrations.
CUC2. Alternatively, in addition to transcriptional repression, NGAL1 may also modulate CUC2 activity in other fashions.
Unlike CUC2, several TF genes showed reduced expression levels in abs2-1D/1 but their expression levels were unchanged in ngaltri compared with those in wild type (Fig. 3D). It is possible that the steady expression levels of these genes are not subject to repression by NGAL1 and the repression of these genes by NGAL1 is only activated under certain developmental contexts, or in our case through constitutive activation of NGAL1. Although less likely, it may also stem from the ectopic activation of NGAL1 outside of its native expression domain. Future work is necessary to clarify the functional relationship between NGAL1 and these genes.
The proper elaboration of reproductive organs is essential for successful completion of the plant life cycle. Organ initiation and boundary determination are critical for the development of plant reproductive organs (Žádníková and Simon, 2014). In the RAV subfamily of B3 TFs, the NGA clade (NGA1-NGA4) was shown to regulate gynoecium development, particularly style and stigma specification (Alvarez et al., 2009;Trigueros et al., 2009). CUC2 has also been implicated in the regulation of gynoecium development, as a gain-offunction allele of CUC2, cuc2-1D, was associated with moderate defects in gynoecium development (Larue et al., 2009). Although we saw only slight gynoecium defects in NGALs triple loss-of-function mutants, dramatic abnormalities in gynoecium formation were observed in ngaltri cuc2-1D quadruple mutants (Fig. 7). Frequently, the two valves of the carpel failed to properly fuse and form closed ovary structures in ngaltri cuc2-1D quadruple mutants, indicating a strong defect in organ boundary formation.

CONCLUSION
In this work, we present evidence that Arabidopsis NGAL-subfamily B3 transcription factors NGAL1, NGAL2, and NGAL3 work together in the regulation of leaf margin development. Moreover, we identified a regulatory module in which NGAL1 inhibits leaf margin serration formation through direct transcriptional repression of CUC2.
Plants used in protoplast assays and leaf margin analysis were grown on Jiffy-7 Peat Pellets (Jiffy Group) under a 12-h/12-h day/night cycle at 22°C/20°C with a light intensity of ;100 mmol m 22 s 21 in a growth chamber. For all other purposes, plants were grown on a commercial peat moss mix (Pindstrup) under continuous illumination (;100 mmol m 22 s 21 ) in a growth room maintained at 22°C.

Generation of Transgenic Lines
To construct OE lines of NGAL2 and NGAL3, the coding sequence of NGAL2 and NGAL3 were amplified using primers NGAL2F and NGAL2R, and NGAL3F and NGAL3R, respectively. The amplified fragments were cloned into the pBluescript KS1 (pBS) vector, sequenced, and subcloned into a binary vector pBI111L (Yu et al., 2004). The resulting vectors were named pBI111L-p35S:NGAL2 and pBI111L-p35S:NGAL3 and used to transform wild type. To generate ngaltri pNGAL1:NGAL1-GFP complementation lines, a genomic fragment encompassing the promoter (1,240 bp upstream of the start codon) and the opening reading frame of NGAL1 was amplified with primers pNGAL1F and gNGAL1R. This fragment was cloned into a modified pCambia1300 binary vector containing the GFP coding sequence to generate a NGAL1-GFP fusion gene. The resulting vector was named pCambia1300-pNGAL1:NGAL1-GFP and used to transform ngaltri. Arabidopsis transformation was carried out using the Agrobacterium (GV3101)-mediated floral-dip method (Clough and Bent, 1998). For selection of transgenic lines, surface-sterilized T1 seeds were plated on 1/2 Murashige and Skoog medium containing 1% Suc (w/v), 0.8% agar (w/v), and appropriate antibiotics (50 mg L 21 kanamycin for pBI111L-derived or 25 mg L 2 1 hygromycin for pCambia1300-derived vectors).

Vectors Used in Protoplast Assays
To generate pTF486-pCUC2:GFP, the promoter region of CUC2 (from 21,659 bp) was amplified using primers CUC2-PF and CUC2-PR and cloned into pTF486 to replace the 35S promoter (Yu et al., 2008). To generate HBT95-p35S:NGAL1-HA, the coding sequence of NGAL1 was cloned into HBT95-HA via restriction digestion and ligation to obtain the 23HA tag fused to the C terminus of NGAL1 (Yoo et al., 2007). To generate pTF486-pCUC2:GFP-pUBQ10:mCherry, linearized pTF486-pCUC2:GFP, a fragment containing the UBQ10 promoter sequence, and a fragment containing the mCherry coding sequence and NOS terminator were assembled together using the NEBuilder HiFi DNA Assembly Master Mix (New England Biolabs). To generate pTF486-pCUC2 D597 :GFP-pUBQ10:mCherry, first, pCUC2 was removed from pTF486-pCUC2:GFP-pUBQ10:mCherry by restriction digestion. The resulting backbone was then assembled with two CUC2 promoter fragments flanking the deleted region using the NEBuilder HiFi DNA Assembly Master Mix (New England Biolabs). To generate pTF486-35S:NGAL1-GR, first, the coding sequence of GR was amplified from p35S:ZFP5:GR and cloned into pTF486 to replace GFP (An et al., 2012). Next, the NGAL1 coding sequence was cloned into pTF486-35S:GR via restriction digestion and ligation to have GR fused to the C terminus of NGAL1. Primers used in vector construction are listed in Supplemental  Table S5.

RNA Extraction and RT-qPCR
Total cellular RNA was extracted from indicated plant tissues using the TRIzol RNA reagent (Thermo Fisher Scientific). Complementary DNA was synthesized from 1 mg total RNA using the Maxima H Minus complementary DNA Synthesis Master Mix (Thermo Fisher Scientific). qPCRs were performed using the FastStart Essential DNA Green Master (Roche) on a QuantStudio 6 Flex Real-Time PCR System (Thermo Fisher Scientific). Primers for qPCR are listed in Supplemental Table S5. Quantifications of RT-qPCR data were based on three biological replicates using ACT2 as the reference gene.

RNA-seq Analysis
For each genotype, three biological replicates were included in the RNA-seq analysis. Center regions of the rosette (including shoot apical meristem, leaf primordia, and emerging young leaves) of 12-d-old seedlings were harvested and frozen in liquid nitrogen for RNA extraction. Total cellular RNA was extracted using the TRIzol reagent (Thermo Fisher Scientific). RNA-seq library preparation, sequencing, and initial data analysis were performed at SAGENE Biotech in Guangzhou, China. In brief, mRNA purification and RNA-seq library preparation were performed following standard procedures recommended by Illumina. Qualities of sequencing libraries were assessed by the Bioanalyzer 2100 (Agilent). Libraries were sequenced on an Illumina Hiseq4000 instrument and paired-end 150-bp reads were generated. Filtered high-quality clean reads were aligned to the Arabidopsis genome (Arapor-t11.Release.201606) using TopHat2 v2.1.1. The expression levels of individual genes were estimated by Cufflinks v2.2.1. Differentially expressed genes between different genotypes were detected by EdgeR v3.14.0. Genes with jlog2FCj . 1 and FDR , 0.05 are considered as differentially expressed. Functional categories of DEGs were analyzed using the MapMan classification source option on the Classification SuperViewer Tool (http://bar. utoronto.ca/ntools/cgi-bin/ntools_classification_superviewer.cgi). Information of the enriched functional categories is listed in Supplemental Table  S2. The enrichment of CUC2-regulated genes in ngaltri mutant was evaluated by the hypergeometric test performed with the phyper function in R. The original RNA-seq data were submitted to Sequence Read Archive (https:// www.ncbi.nlm.nih.gov/sra) under the accession number PRJNA612928.

Leaf Margin Analysis
The eighth, ninth, and tenth rosette leaves were placed between two transparent sheets and scanned by a CanoScan 9000F scanner (Canon) to generate silhouettes. Leaf areas and tooth areas were measured manually based on leaf silhouettes in ImageJ (National Institutes of Health). Quantifications of leaf margin serrations were performed as previously described (Zheng et al., 2016).

Protoplast Effector/Reporter Assays
The effector (pTF486-35S:NGAL1-GR) and reporter (pTF486-pCUC2 D597 :GFP-pUBQ10:mCherry and pTF486-pCUC2:GFP-pUBQ10:mCherry) plasmids were purified using the ZymoPURE II Plasmid Maxiprep kit (Zymo Research). Arabidopsis mesophyll protoplasts were prepared and transfected as previously described (Yoo et al., 2007). In brief, 200 mL protoplasts (2 3 10 5 mL 21 ) were cotransfected with effector and report plasmids (20 mg each). After transfection, protoplasts were incubated in W5 solution containing 5 mM DEX or equal volume of ethanol for 12 h. GFP and mCherry signals were detected by a fluorescence microscope (DMi8; Leica) equipped with a DFC365 FX charge-coupled device (Leica) using a 203 objective lens (HC PL APO N.A. 0.80). The mCherry fluorescence serves as the transfection control. For each treatment, more than 100 transfected protoplasts with mCherry fluorescence were measured for GFP fluorescence signal intensity using the ImageJ software. Assays were repeated independently three times with similar results.

Gynoecium Morphology Observation
Dissected flowers were examined with a stereoscope (SMZ25, Nikon) equipped with a charge-coupled device camera (DS-Ri2, Nikon) or fixed in formalin acetic alcohol (50% [v/v] ethanol, 10% [v/v] acetic acid, and 3.7% [v/ v] formaldehyde) at 4°C. Fixed floral organs were then dehydrated through a series of ethanol solutions in ascending concentration up to 100%. After critical point drying and sputter coating with gold, samples were examined with a scanning electron microscope (Hitachi S-3400 II).

Supplemental Data
The following supplemental materials are available.
Supplemental Figure S1. Characterization of T-DNA insertional lines of NGALs.
Supplemental Figure S3. Effect of NGALs on tooth formation during early stages of leaf development.
Supplemental Table S5. List of primers used in this study.