Introduction

Plant leaves are initiated from the peripheral region of the shoot apical meristem (SAM), and display great variations in shape and size. They are traditionally classified into two major morphogenetic classes: undivided simple leaves, and compound leaves. Simple leaves have a single lamina that can sometimes be elaborated with less-pronounced incisions such as serrations or lobes on the laminal margin, for example, unlobed leaves in Arabidopsis and broad or lobed leaves in cotton (Fig. 1a–c). Compound leaves or dissected leaves have numerous individual leaflets on a rachis that arises at a node, for example, ternate compound leaves of soybean and pinnately compound leaves of tomato (Fig. 1d, e). The same factors are involved in the formation of serrations and leaflets, and include auxin activity maxima and CUP-SHAPED COTYLEDON (CUC) genes (Blein and Laufs 2008; Kougioumoutzi 2008; Bilsborough et al. 2011; Kasprzewska et al. 2015). Many compound leaf mutants in Cardamine hirsuta and Medicago truncatula have defective separation between the adjacent leaflets leading to the conversion of dissected leaves into lobed simple leaves (Peng et al. 2011; Vlad et al. 2014). Lobed simple leaves are essentially an intermediate shape between serrations and dissected leaves. However, whether gene mutations can cause lobed leaves to become compound leaves requires further investigation.

Fig. 1
figure 1

A variety of leaf shapes in dicotyledons. aA. thaliana, Col-0, bG. hirsutum acc. TM-1, cG. hirsutum acc. Okra cotton, dGlycine max, Williams 82, and eLycopersicon esculentum, Micro-Tom. Scale bars = 1 cm. Okra cotton was used in VIGS assays for investigating the effect of silencing GhAS1, GhAS2, KNOX1 and GhOKRA genes. TM-1 cotton was used in GhAS1 and GhAS2 silencing VIGS assays. Transgenic A. thaliana (Col-0) constitutively expressed GhOKRA resulted in the production of a lobed leaf

Full size image

Cotton plants have a variety of leaf shapes, including broad leaves, as in TM-1, and lobed leaves, as in Okra (Fig. 1b, c). We have found that an HD-ZIP I transcription factor (GhOKRA) controls the formation of deep lobes in cotton (Gossypium) (Chang et al. 2016), which has been independently confirmed by two other groups (Andres et al. 2014, 2016; Zhu et al. 2016). The mutant of GhOKRA in TM-1 results in the production of broad leaves. Compound leaves are non-existent in cotton. The ε clade of HD-ZIP I transcription factors were conserved to regulate leaf shape in many plants (Hofer et al. 2009; Vlad et al. 2014). The GhOKRA homolog, LATE MERISTEM IDENTITY1 (LMI1), was first reported in Arabidopsis thaliana as a floral regulator and was found to influence leaf morphogenesis (Saddic et al. 2006). Other GhOKRA homologs, such as REDUCED COMPLEXITY (RCO) in C. hirsuta and Tendril-less (TL) in pea, which are also LMI1-like genes, are emerging as regulators of lateral organ genesis in compound leaves by affecting lateral organ formation (Hofer et al. 2009; Vlad et al. 2014). The function of these LMI1-like genes appears to be similar, such as blade growth-repression in compound leaves and lobed simple leaves (Vlad et al. 2014; Andres et al. 2016).

Class I KNOTTED1-like homeobox (KNOX1) transcription factors are involved in the maintenance of indeterminate cell fate in the SAMs and developing primordia of complex leaves (Bharathan et al. 2002; Hake et al. 2004; Uchida et al. 2010). The KNOX1 genes are expressed in SAM to maintain the indeterminate nature of meristem cells, and are down-regulated at the position where leaf primordia initiate (Lincoln et al. 1994; Hay and Tsiantis 2010; Sluis and Hake 2015). At the site of leaf initiation, KNOX1 genes are downregulated, and auxin signaling occurs. Auxin regulates KNOX expression (Hay et al. 2006), and KNOX modulates many genes in auxin signaling (Bolduc et al. 2012); thus, they have regulatory interactions. The expression of KNOX1 genes is essential in the developing primordia for leaflet formation in compound leaf plants. However, in simple leaf plants, the KNOX1 genes are absent from leaf primordia (Parnis 1996; Hay and Tsiantis 2006; Kougioumoutzi 2008; Efroni and Lifschitz 2010). Compound leaf development requires organogenic activity during primary morphogenesis and leaflet formation. The KNOX1 genes can be classified into three subclasses; SHOOT MERISTEMLESS (STM)-like, Knotted in A. thaliana (KNAT)2/6-like and KNAT1/BREVIPEDICELLUS(BP)-like. In A. thaliana, ASYMMETRIC LEAVES1 (AS1) and AS2 repress the activity of KNAT1 and KNAT2 in leaves (Byrne et al. 2000, 2002). Both as1 and as2 mutants have abnormal lobed leaves with ectopic expression of BP and KNAT2 (Byrne et al. 2000, 2002). The MYB transcription factors, which are encoded by AS1, interact with the LATERAL ORGAN BOUNDARIES (LOB) domain protein, AS2, and work together as the AS1–AS2 complex. This complex binds KNOX loci, resulting in the recruitment of chromatin remodeling factors, such as HDA6, a histone deacetylase (Luo et al. 2012; Lodha et al. 2013). Both BTB ankyrin genes, BLADE ON PETIOLE1 (BOP1) and BOP2, can activate AS2 directly, but also can repress KNOX independently (Khan et al. 2014). Some simple leaves develop from complex primordia through secondary morphogenesis, as in Lepidium oleraceum, which has KNOX1 expression in leaf primordia to produce marginal outgrowths (Bharathan et al. 2002). In addition, ectopic expression of KNAT1 in A. thaliana transforms simple leaves into lobed leaves, resulting in an increase in leaf complexity (Chuck et al. 1996). In Cardamine hirsute, which has dissected leaves comprising leaflets, a KNOX1 paralogous gene, ChBP, is concurrently regulated by the microRNA164A (MIR164A)/ChCUP-SHAPED COTYLEDON (ChCUC) module and ChASYMMETRIC LEAVES1 (ChAS1). This gene does not occur in A. thaliana, a relative of Cardamine hirsute, which has simple leaves (Rastsomssich et al. 2015). However, it is still unclear whether KNOX1 genes regulate lobe development in simple leaves. Organogenic activity and leaf marginal structure development are known to dictate the final leaf shape; however, there is little information on the role of KNOX1 genes in combination with LMI1-like genes in the formation of different leaf shapes.

To determine whether the LMI1-like gene regulates leaf shape in collaboration with KNOX1 genes, we silenced the KNOX1 genes and the homologous genes of AS1 and AS2 in wild-type (WT) okra leaf cotton, and found that different expression models of the LMI1-like gene and KNOX1 genes control the formation of broad, lobed and compound leaves. We also generated dissected-like leaves in A. thaliana, which further confirms that formation of these leaf shapes is controlled by LMI1-like gene and KNOX1 genes. The present research provides new insights into the formation of different leaf marginal structures, including unlobed leaves, lobed leaves, and leaflets.

Materials and methods

Plant materials

TM-1 with broad leaves is a standard genetic line of Upland cotton (Kohel et al. 1970). Okra cotton with okra leaves, provided by the Institute of Cotton Research of CAAS named as Super Okra, is a Gossypium hirsutum accession. The A. thaliana mutants used in this study, as1-101 and as2-101, were produced with a Landsberg erecta (Ler) background. as1-101 and as2-101 seeds were kindly provided by Lin Xu and Hai Huang (Institute of Plant Physiology & Ecology, SIBS, CAS) (Yue et al. 2000; Sun et al. 2002; Xu et al. 2002). Transgenic A. thaliana constitutively expressing GhOKRA plants were generated previously in a Col-0 background (Chang et al. 2016). The T1 progeny of transgenic A. thaliana with lobed leaves was crossed with as1-101 and as2-101 mutant plants. The leaves of the F1 progeny had more lobes than that observed in constitutively expressing GhOKRA A. thaliana plants. There were many different leaf phenotypes in the F2 populations of these two crosses, such as wild-type, as1-101 or as2-101, and constitutively expressed GhOKRA A. thaliana leaves, as well as some novel leaf types. All materials were grown in the green houses of Nanjing Agriculture University following normal practices.

Phylogenetic analyses of KNOX genes in cotton

The genes in Gossypium raimondii are highly homologous with those in G. hirsutum. Since G. hirsutum has two sub-genomes, and some homologous have not been annotated in the G. hirsutum acc. TM-1 genome, so we analyzed KNOX genes in diploid G. raimondii for next work. G. raimondii genome sequences were downloaded from the Phytozome database (http://www.phytozome.net) (Paterson et al. 2012). Arabidopsis thaliana KNOX protein sequences were downloaded from the Arabidopsis Information Resource website (TAIR) (http://www.arabidopsis.org). All cotton proteins were screened for potential KNOX genes using HMMER software version 3.0 and the Pfam database (Zhang and Wood 2003; Finn et al. 2011, 2016). KNOX protein sequences from G. raimondii were aligned with the homologous proteins from A. thaliana. We used the Maximum likelihood (ML) method to construct a phylogenetic tree in MEGA 6.06 (http://www.megasoftware.net). The bootstrap test of phylogeny was performed with 1000 replications (Tamura et al. 2013).

Cloning of GhAS1, GhAS2 and KNOX1 genes

Homologs of AS1 and AS2 in A. thaliana were identified using HMMER software version 3.0 and the Pfam database to screen all TM-1 proteins (Zhang and Wood 2003; Finn et al. 2011, 2016). We cloned the coding sequences (CDSs) of GhAS1, GhAS2 and some KNOX1 genes from shoot tips of okra cotton using the genome sequence of TM-1 (Zhang et al. 2015). The primers used are listed in Table S1. We used ExTaq DNA Polymerase (TaKaRa, Japan) for PCR. Amplification products were cloned into the pMD19-T vector (TaKaRa, Japan) for sequencing by the Nanjing Jinsite Biotech. Co. Ltd. All of the genes mentioned above were cloned from okra cotton, with separation of the A and the D sub-genomes. Based on the genome sequences of G. raimondii and G. arboreum and the sub-genome sequences in G. hirsutum, we differentiated between the A and the D sub-genome homologs (Table S2). Both the A and the D sub-genome homologs in G. hirsutum are highly homologous with those in G. raimondii. Only a few SNPs exist between them. A silencing construct developed from the A or the D sub-genome would silence two sub-genome homeologs. We therefore chose only one sub-genome for further analysis due to the high sequence similarity of the A and the D sub-genomes.

Virus-induced gene silencing (VIGS) assay

We amplified fragments around 300-bp long from the 3′ ends of GhAS1, GhAS2 and seven KNOX1 genes (GhKNL2-GhKNL8) (Table S2). A ClonExpress II One Step Cloning Kit (C112-01, Vazyme Biotech Co., Ltd) was used to recombine the fragments into EcoRI-XbaI-digested pTRV2 for VIGS assay. All primers used for VIGS vector construction are listed in Table S1. The VIGS assay was carried out according to methods described previously (Liu et al. 2002). Cotton seedlings were grown in a growth chamber (21–25 °C) under long days light cycle (16 h:8 h, light:dark). The Agrobacterium mixture was infiltrated into the abaxial side of cotyledons of 8-day-old cotton seedings by needleless syringes. More than 15 Okra cotton individuals with okra leaves were infiltrated with pTRV2-GhAS1 or pTRV2-GhAS2. Around 15 more TM-1 individuals with broad leaves were infiltrated with pTRV2-GhAS1 or pTRV2-GhAS2. More than 15 Okra cotton individuals with okra leaves were infiltrated with mix of pTRV2-GhKNLs (GhKNL2-GhKNL8). We mixed different KNOX1 genes together in VIGS assays to inhibit functional complementation between paralogous genes as follows; GhKNL2 and GhKNL3 with GhKNL4, GhKNL5 with GhKNL6, and GhKNL7 with GhKNL8. These KNOX1 gene combinations were chosen based on relationships represented by phylogenetic tree (Fig. 2). KNOX1 genes can be classified into three subclasses. We chose these genes belong to the same subclass as a combination. pTRV2 was used as a negative control. Three independent tests were carried out for all VIGS assays.

Fig. 2
figure 2

Phylogenetic tree of KNOX genes of Arabidopsis thaliana and Gossypium raimondii. A maximum-likelihood tree was generated by aligning KNOX protein sequences of G. raimondii and A. thaliana. G. raimondii KNOX proteins were classified into three clades; class I KNOX, class II KNOX and KNATM. Seven KNOX1 genes were selected and named GhKNL2-GhKNL8 in this study. Phylogenetic analysis was carried out using the ML method with 1000 resampling replicates

Full size image

Quantitative real time PCR (qPCR) analysis

To detect the expression of silenced genes (GhOKRA and GhKNL2/3/4), the shoot tips (approximately 6 mm) of Okra cotton individuals infiltrated with either pTRV2-GhAS1 or pTRV2-GhAS2 at the same growth stage were harvested by wiping off the same number of leaves when the leaf shape had changed. We used the Biospin Plant Total RNA Extraction Kit (BioFlux, cat: BSC65S1) to extract total RNA from the shoot tips. First-strand cDNA was generated according to the manufacturer’s instructions using TransScript® One-Step gDNA Removal and cDNA Synthesis SuperMix (TransGen Biotec Co., Ltd., cat: AT311) kits. The cotton Histone 3 gene (His3, GenBank accession number: AF024716) was used as a reference gene. The qPCR primer sequences are listed in Table S1. They are universal for both the A and the D sub-genome homologs. So, we can detect all the two sub-genome homeologs. qPCR products were quantified according to the manufacturer’s instructions using the ABI 7500 Real Time System (Applied Biosystems, USA) and the light cycler fast start DNA Master SYBR Green I kit (Roche, Basel, Switzerland).

Results

Plants silenced in either GhAS1 or GhAS2 produce compound leaves

To explore the relationship between an HD-ZIP I transcription factor (GhOKRA) and other genes in modulating leaf development, we cloned the corresponding cotton homologs of ASYMMETRIC LEAVES1 (AS1), AS2 and seven KNOX1 genes that regulate leaf development in A. thaliana (Table S2). Based on known KNOX gene domains, 22 KNOX genes were identified in the G. raimondii genome (Fig. 2). Of them, seven cotton class I KNOX genes, named GhKNL2 (KNOTTED1-LIKE) to GhKNL8, that have close relationships with the corresponding homologs in A. thaliana were cloned from the wild-type (WT) okra leaf cotton (Fig. 2) (Gong et al. 2014).

There was only one corresponding homolog of either A. thaliana AS1 or AS2 in each sub-genome in cotton. To our knowledge, there are no reports on how these two genes regulate leaf development in cotton. In the GhAS1 and GhAS2 virus induced gene silencing (VIGS) assays, the leaves of all okra leaf plants infiltrated with either pTRV2-GhAS1 or pTRV2-GhAS2 became split from the petiolar sinus and dissected 3 weeks after infiltration (Fig. 3a). Leaf blades at the base of the leaf vein disappeared. Three individual petiolulate leaflets arose at a new distinct node. Each leaflet had a long petiole (Fig. 3a, e). The deeply lobed leaves in the okra leaf cotton were replaced by ternate compound leaves; a phenomenon that has not been observed before in cotton. qPCR analysis showed that the expression of GhAS1 and GhAS2 was lower in pTRV2-GhAS1 and pTRV2-GhAS2-silenced lines compared to controls (Fig. 3b). It is clear that the lobed simple leaves of okra cotton were transformed into compound leaves when either GhAS1 or GhAS2 expression was suppressed. These results demonstrate that GhAS1 and GhAS2 regulate the depth of leaf lobes in cotton. GhAS1 may be involved in leaf adaxial/abaxial polarity in cotton, since individual plants infiltrated with pTRV2-GhAS1 exhibited rumpled and curled leaves (Fig. 3a).

Fig. 3
figure 3

Leaf phenotypes of VIGS GhAS1 or GhAS2- silenced okra cotton plants. a Plant phenotypes from left to right: pTRV2, pTRV2-GhAS1 and pTRV2-GhAS2. Leaf shape of negative control plants (left), okra cotton plants silenced by pTRV2-GhAS1 (middle), and okra cotton plants silenced by pTRV2-GhAS2 (right). The leaf lobes deepened and the blade at the base of the leaf vein disappeared in either GhAS1 or GhAS2-silenced plants. bGhAS1 transcript levels in GhAS1-silenced and negative control plants. GhAS2 transcript levels in GhAS2-silenced and negative control plants. cGhKNL2, GhKNL3, and GhKNL4 transcript levels in GhAS1 or GhAS2-silenced and negative control plants. The expression of GhKNL2, GhKNL3, and GhKNL4 was higher in silenced cotton than in controls. dGhOKRA transcript levels in three individuals silenced either GhAS1 or GhAS2 and negative control. There is no significant difference in comparision with the negative control. Data represent the means ± SDs of three replicated experiments. Asterisks indicate a significant difference at P < 0.05; Double asterisks indicate a significant difference at P < 0.01. e We measured the length of the formed left (L), middle (M) and right (R) petiolules in the fourth leaf of the VIGS plants in one of our VIGS assays by 37 days post infiltration (dpi). There is no petiolules in the control. Data represent the means ± SDs of 15 individuals

Full size image

We used VIGS to explore how GhOKRA modulates leaf development. One month after infiltration with pTRV2 fused with KNOX1 genes (GhKNL2-GhKNL8), the WT okra leaf cotton retained its okra leaf shape (Fig. 4i–l); however, KNOX1 gene-silenced plants grew slowly (Fig. 4a–d), leaves in their shoot apex became compact and the distance between nodes was shorter than in controls (Fig. 4e–h, m). From these experiments, we could not reveal the roles of KNOX1 in lobe development in cotton.

Fig. 4
figure 4

The phenotype of okra cotton silenced by several paralogous genes from the KNOX1 family. Plant phenotypes from left to right: pTRV2, pTRV2-GhKNL2/3/4, pTRV2-GhKNL5/6, and pTRV2-GhKNL7/8. a–d Whole plant scaning; e–h feature of the distance between the nodes; i–l Feature of the first leaf. m We measured the length of the nodes between fifth and sixth nodes of the VIGS plants in one of our VIGS assays by 61 dpi. Data represent the means ± SDs of 15 individuals. Double asterisks indicate a significant difference at P < 0.01. Plants with pTRV2-GhKNLs grew slowly and their leaves became compact in shoot apex of cotton, the distance between the nodes were decreased in comparision with the negative control, while the leaf shape showed no change. Scale bars = 2 cm

Full size image

GhOKRA is necessary for compound leaf formation

GhOKRA controlled okra leaf development in the WT cotton. Ectopic expression of the wild cotton GhOKRA gene under the control of the 35S promoter (Pro35S::GhOKRA) led to a lobed leaf type rather than compound leaves in A. thaliana (Chang et al. 2016). To explore whether GhAS1 and GhAS2 regulate the depth of leaf lobes in collaboration with the GhOKRA gene, we silenced either GhAS1 or GhAS2 in both WT okra leaf cotton (Okra cotton) and mutant broad leaf cotton (TM-1). All GhAS1 or GhAS2-silenced WT okra cotton plants had larger lobe depths than controls, suggesting that the functions of GhAS1 and GhAS2 are the same in leaf lobe development. AS1 and AS2 promote leaf adaxial-abaxial polarity specification and repress KNOX gene expression by forming AS1–AS2 protein complexes in A. thaliana (Fu et al. 2007; Guo et al. 2008). Therefore, we suppose that the AS1–AS2 complex might influence leaf lobe development as an upstream regulator of GhOKRA. All GhAS1 or GhAS2-silencing in broad leaf TM-1 plants caused no change in the leaf lobes, although the second and third leaves of GhAS1-silenced plants were seriously curled (Fig. S1). The broad leaf TM-1 had a mutated non-functional Ghokra genotype which led to a different result compared with Okra cotton after silencing of GhAS1 or GhAS2. These results suggest that GhOKRA is necessary for compound leaf formation.

GhKNL2/3/4 expression was elevated in either GhAS1 or GhAS2-silenced plants

To demonstrate whether the AS1–AS2 complex negatively regulates GhOKRA expression, the transcript levels of GhOKRA in either GhAS1 or GhAS2-silenced individuals were detected. There was no difference in the expression levels of GhOKRA in either GhAS1 or GhAS2-silenced plants compared to the negative control (Fig. 3d), confirming that the AS1–AS2 complex does not regulate GhOKRA. The AS1–AS2 protein complex can repress Knotted in A. thaliana 1 (KNAT1) and KNAT2 activity in leaves (Guo et al. 2008). When KNAT1 was overexpressed in A. thaliana, the leaves showed ectopic stipules (Chuck et al. 1996). The transcript levels of GhKNL2/3/4, which are homologs of KNAT1, were thus analyzed in the shoot tips of either GhAS1 or GhAS2-silenced plants. Compared with the negative control, the expression levels of all three homologs were elevated in the either GhAS1 or GhAS2-silencing plants (Fig. 3c), with a greater increase observed in the GhAS2-silenced plants. These results suggest that GhAS1 and GhAS2 repress the expression of KNOX1 in leaf primordia of cotton and this repression might be responsible for the change from lobed simple leaves to compound leaves in the GhAS1 or GhAS2-silenced plants.

Co-expression of GhLMI1-like and KNOX1 genes produces compound-like leaves in A. thaliana

In A. thaliana, the GhOKRA homolog was secondarily lost through duplication, leading to the evolution of a simple leaf (Vlad et al. 2014). Previously, we generated constitutively expressing GhOKRA A. thaliana plants with a lobed leaf (Chang et al. 2016). To determine whether coordination of the LATE MERISTEM IDENTITY 1 (LMI1)-like and KNOX1 genes influences the formation of different leaf shapes, we crossed heterozygous constitutively expressing GhOKRA A. thaliana plants with homozygous as1-101 and as2-101 mutant plants in a Landsberg erecta (Ler) background (Yue et al. 2000; Sun et al. 2002; Xu et al. 2002). As compared to the WT type A. thaliana (Fig. 5a), the leaves of as1-101 plants were heart-shaped and the leaf edges were slightly curled down with no lobes (Fig. 5b), while the mutant as2-101 rosette leaves were broad (Fig. 5c). F1 plants with a greater number of lobes than constitutively expressing GhOKRA A. thaliana plants were self-pollinated to generate F2 populations. The F2 segregated out several novel leaf types (Table S3) that were different from the WT, as1-101 and as2-101 mutant parents, and constitutively expressing GhOKRA A. thaliana plants. In Fig. 5, we show some typical individuals with novel leaf types from the two F2 populations. Their genotypes were identified by cloning and sequencing (Figs. S2, S3). The rosette leaves of three F2 individuals genotyped as GhOKRA/as1-101 were divided into two or three parts along the proximal–distal axis (Fig. 5e–g). Their leaf lobes extended to leaf petioles. Some of their cauline leaves became compound-like leaves (Fig. 5i–k) and leaf-like structures were also observed at the petioles of cauline leaves (Fig. 5j). One individual from the cross between the as2-101 and constitutively expressing GhOKRA A. thaliana plants showed extremely deeply lobed leaves (Fig. 5h), and their secondary lobes appeared on lobed leaves, thus increasing the leaf complexity (Fig. 5l).

Fig. 5
figure 5

Phenotypes and leaf vein patterns of parents and their F2 individuals. a–d Morphological observations of wild-type Col-0 (a), mutant as1-101 (b), mutant as2-101 (c) and constitutively expressing GhOKRA A. thaliana (d) plants. e–g The F2 individuals of as1-101 and constitutively expressing GhOKRA A. thaliana plants. h The F2 individuals of as2-101 and constitutively expressing GhOKRA A. thaliana plants. i–k The rosette leaf (left) and cauline leaf (right) of F2 individuals of as1-101 and constitutively expressing GhOKRA A. thaliana plants. l The rosette leaf (left) and cauline leaf (right) of F2 individuals of as2-101 and constitutively expressing GhOKRA A. thaliana plants. Rosette leaves were divided from leaf petioles, and cauline leaves became compound leaves in the F2 individuals of as1-101 and constitutively expressing GhOKRA A. thaliana plants. The F2 individual of as2-101 and constitutively expressing GhOKRA A. thaliana plants showed extremely deeply lobed leaves with secondary lobes. m–p Leaf vascular structure of wild-type Col-0 (m), mutant as1-101 (n), mutant as2-101 (o) and constitutively expressing GhOKRA A. thaliana plants (p). q–t Leaf vascular structure of F2 individuals of as1-101 and constitutively expressing GhOKRA A. thaliana plants. Two different leaf vein patterns of F2 individuals indicated different methods of leaflet-like rosette and cauline leaf formation. u Leaf vascular structure of F2 individuals of as2-101 and constitutively expressing GhOKRA A. thaliana plants. Scale bars = 1 cm

Full size image

To further confirm whether these novel leaves were formed from one divided leaf or the fusion of several leaves, we analyzed the leaf vein and found a difference between rosette leaves and cauline leaves in the F2 generated by crossing as1-101 and constitutively expressing GhOKRA A. thaliana. The aberrant rosette leaves in F2 plants had a reduced vascular structure. Lobes were divided between main veins, which indicates that the leaf was formed from a single complete leaf (Fig. 5q, r). The cauline leaves from F2 plants had a complex vascular structure. Each leaflet vein was similar to those of wide-type leaves, which suggests that these cauline leaves became compound-like leaves through the fusion of simple leaves (Fig. 5s, t).

The vascular structure of rosette leaves of F2 plants from the cross between as2-101 and constitutively expressing GhOKRA A. thaliana was more complex. Secondary lobes increased the complexity of the vascular structure (Fig. 5u).

In the A. thaliana F2 populations generated in this study, the phenotypic segregation did not follow a Mendelian pattern. The constitutively expressing GhOKRA A. thaliana plants showed a deeply lobed leaf in the T1 progeny (Chang et al. 2016), but most had disappeared in the T2 progeny. It is due to the limitations of the heterologous overexpression system which we used. Therefore, the number of progeny with dissected leaves was lower than expected (Table S3).

Discussion

Temporal regulation of GhOKRA and KNOX1 genes during compound leaf development

We previously reported that GhOKRA-silenced plants display palmate leaves changed from deeply lobed okra leaves (Chang et al. 2016). Based on the present study, it seems like that elevated expression of KNAT1 homologous genes in lobed leaf primordia results in the formation of dissected leaves. This is the first report of the generation of compound leaves in cotton (Fig. 3a). As we know, transient gene silencing is the limitation of the VIGS method. While the VIGS method seems like have advantages over transgenic plant methods in research into gene temporal regulation for gene function identification. We can explore the temporal expression differences between two genes by observing their phenotypic variation via VIGS. It has been reported that leaf marginal structures, such as leaflets, lobes and serrations, are formed during the phase of primary morphogenesis when leaves are primordia (Bar and Ori 2014). However, in the present study, we observed the phenotypes of leaf marginal structures after the leaves were fully expanded. In our VIGS assay, Okra cotton seedlings were infected with pTRV2-GhAS1, pTRV2-GhAS2 and pTRV2-GhOKRA at the same time. We found that the phenotypes of either GhAS1 or GhAS2 gene-silenced plants were changed from the lobed to compound leaf at the fifth fully expanded leaf (L5). However, in the GhOKRA-silenced plants, the phenotype variations from the lobed to broad leaf were firstly observed at the seventh fully expanded leaf (L7) rather than at L5. These results suggest that silencing of GhAS1 or GhAS2 was effective in the primordium of L5 and those of younger leaves including L6 and L7, whereas silencing of GhOKRA affected leaf morphology only when silencing took place from a very early developmental stage (e.g., primordium of L7). The primordium of L7 is a younger initiating leaf than the primordium of L5. Therefore, it is likely that GhOKRA is involved in leaf primordia development at a very early stage, before KNOX1 genes start to express in the GhAS1 or GhAS2-silensed plants. We speculate that the expression of the LMI1-like gene allows lobe formation or limited lamina growth between leaflets first, and then KNOX1 expresses to maintain the indeterminate nature of meristem cells in the developing primordia for leaflet formation (Fig. 6).

Fig. 6
figure 6

Hypothesis of regulatory model for the LMI1-like and KNOX1 genes. Different expression patterns of the LMI1-like and KNOX1 genes regulate the formation of different leaf marginal structures. The primordia of lobed leaf is only developed when the LMI1-like gene express (Andres et al. 2016), but not KNOX1. The inactivity of the LMI1-like gene from lobed primordia will lead to the development of an unlobed simple leaf type (Chang et al. 2016). The AS1–AS2 protein complex repress the activity of KNOX1 in leaf primordia, but their mutations can remove this repression. This repression might be responsible for the change from lobed simple leaves to compound leaves in the GhAS1 or GhAS2-silenced plants. Orange mark LMI1-like gene; green mark KNOX1 genes; asterisks mark the SAM; dashed panes mark primordia

Full size image

LMI1-like and KNOX1 genes may regulate the development of different leaf shapes

The function of LMI1-like genes has been widely reported in different plants (Hofer et al. 2009; Vlad et al. 2014; Ni et al. 2017). Constitutively expressing homologous genes in A. thaliana produced lobed leaves (Chang et al. 2016; Ni et al. 2017). In the compound leaf plant, Cardamine hirsuta, the homologous gene, RCO, repressed lamina growth at the leaf margin (Vlad et al. 2014). However, the molecular context of the LMI1-like gene is not clear. KNOX1 transcription factors maintain an indeterminate cell fate in SAMs and complex leaf primordia in most plants (Lincoln et al. 1994; Hay and Tsiantis 2010). Based on the present results, a hypothetical model was proposed to illustrate how the LMI1-like and KNOX1 genes regulate the formation of different leaf shapes, including unlobed leaves, lobed leaves and compound leaves (Fig. 6). If primordia only expressed the LMI1-like gene without KNOX1, the leaf shape was lobed (Andres et al. 2016), as in the WT okra cotton (Fig. 1c), and as shown by constitutively expressing LMI1-like genes in A. thaliana (Fig. 5d). Removing the activity of the LMI1-like gene from lobed primordia led to an unlobed simple leaf type (Fig. 6), as seen in A. thaliana and TM-1 (Vlad et al. 2014; Chang et al. 2016). The GhAS1 or GhAS2-silenced plants with lobed leaves produce compound leaves (Fig. 3a). As the GhAS1 and GhAS2 repress the expression of KNOX1 (Fig. 3c). So, the KNOX1 expression in lobed primordia might result in compound leaves (Fig. 6). When primordia expressed KNOX1 without the LMI1-like gene, the leaf shape was either unlobed or lobed, possibly depending on the expression levels of KNOX1 genes. Neither GhAS1 nor GhAS2-silenced TM-1 plants had altered leaf shapes (Fig. S1), while ectopic expression of KNAT1 under the control of the 35S promoter in A. thaliana produced lobed leaves (Chuck et al. 1996). Therefore, co-expression of LMI1-like and KNOX1 genes in primordia could result in compound leaves: we generated leaflet-like A. thaliana leaves by crossing as1-101 and as2-101 mutants with constitutively expressing GhOKRA plants (Fig. 5). The deeply lobed leaves of C. hirsuta rco mutant caused by the fusion of leaflets also supports this hypothetical model (Vlad et al. 2014).

Intricate and complicated factors regulate leaf development

Many factors repress the activity of KNOX1 in leaf primordia, such as polar auxin transport and TEOSINTEBRANCHED1/CYCLOIDEA/PROLIFERATING CELL FACTOR (TCP), as well as the AS1–AS2 protein complex (Scanlon 2003; Guo et al. 2008; Li et al. 2012). POLYCOMB REPRESSIVE COMPLEX (PRC)2 interacts with the AS1–AS2 protein complex to stably silence KNAT1 and KNAT2 in leaf primordia (Lodha et al. 2013). In addition, in pea, the FLORICAULA/LEAFY ortholog, UNIFOLIATA, rather than KNOX1, was found to control the development of compound leaves (Hofer et al. 1997). It has also been reported that CUC genes define boundaries between leaflets in compound leaf plants (Blein and Laufs 2008; Berger et al. 2009). Leaf development is changeable and complicated and is regulated by many factors. Ectopic -expressing GhOKRA A. thaliana plants develop many lobes in leaves in the same direction along the medial–lateral axis (Chang et al. 2016). The A. thaliana F2 progeny showed lobes in two directions; along the medial–lateral and proximal–distal axes (Fig. 5i–l). KNOX1 might affect the position or pattern of GhOKRA expression. The dissected leaves of GhAS1 and GhAS2-silenced okra cotton plants indicate that compound leaves are formed from deeply lobed simple leaves (Fig. 3a). There were two types of leaf vein distribution in A. thaliana F2 progeny (Fig. 5q–u), and leaf vein development also affected leaflet formation (Runions et al. 2017). Whether compound leaves are formed from collections of simple leaves deserves further research (Champagne and Sinha 2004). In this study, we propose that the LMI1-like and KNOX1 genes regulate the formation of several margin structures (Fig. 6).