CLE Peptides in Plants: Proteolytic Processing, Structure-Activity Relationship, and Ligand-Receptor Interaction

Authors

  • Xiaoming Gao,

    1. Tobacco Research Institute, Chinese Academy of Agricultural Sciences/Key Laboratory of Tobacco Biology and Processing, Ministry of Agriculture, Qingdao 266101, China
    Search for more papers by this author
  • Yongfeng Guo

    Corresponding author
    1. Tobacco Research Institute, Chinese Academy of Agricultural Sciences/Key Laboratory of Tobacco Biology and Processing, Ministry of Agriculture, Qingdao 266101, China
      Tel: +86 532 6671 5256; Fax: +86 532 8870 2056; E-mail: yongfeng.guo@gmail.com
    Search for more papers by this author

Tel: +86 532 6671 5256; Fax: +86 532 8870 2056; E-mail: yongfeng.guo@gmail.com

Abstract

Ligand-receptor signaling initiated by the CLAVATA3/ ENDOSPERM SURROUNDING REGION (CLE) family peptides is critical in regulating cell division and differentiation in meristematic tissues in plants. Biologically active CLE peptides are released from precursor proteins via proteolytic processing. The mature form of CLE ligands consists of 12–13 amino acids with several post-translational modifications. This review summarizes recent progress toward understanding the proteolytic activities that cleave precursor proteins to release CLE peptides, the molecular structure and function of mature CLE ligands, and interactions between CLE ligands and corresponding leucine-rich repeat (LRR) receptor-like kinases (RLKs).

  • image

[ Yongfeng Guo (Corresponding author)]

Introduction

In addition to traditional phytohormones such as auxin, ethylene, cytokinins, abscisic acid, gibberellins, brassinosteroids, and jasmonic acid, peptides are important in mediating intercellular communications in plants. Since the isolation of systemin from tomato in 1991 (Pearce et al. 1991), a number of peptides have been identified in plants to initiate ligand-receptor signaling, and control plant development and responses to environmental stimuli (reviewed in Matsubayashi and Sakagami 2006; Bahyrycz and Konopinska 2007). Many of the CLAVATA3/ ENDOSPERM SURROUNDING REGION (CLE) family peptides are involved in cell division/differentiation control in meristematic tissues (Fiers et al. 2007; Jun et al. 2008; Stahl et al. 2009; Hirakawa et al. 2010a; Katsir et al. 2011).

As the founding member of the CLE protein family, CLAVATA3 (CLV3) is a key regulator of stem cell homoeostasis at the shoot apical meristem (SAM) (Clark et al. 1995; Fletcher et al. 1999). The CLV3 signal is perceived and relayed through the CLV pathway to suppress expression of the homeodomain transcription factor gene WUSCHEL (WUS), which functions to promote stem cell identity in SAMs (Mayer et al. 1998; Schoof et al. 2000). Loss-of-function mutations in CLV3 and several of the CLV pathway genes result in massive over-accumulation of stem cells at the SAM (Clark et al. 1993; Clark et al. 1995; Kayes and Clark 1998; Muller et al. 2008).

The CLE genes are broadly distributed in all plant species, from the basal lycophyte Selaginella to both monocots and dicots. Including CLV3, the CLE family in Arabidopsis consists of 32 members (Cock and McCormick 2001; Strabala et al. 2006a; Oelkers et al. 2008). 44 CLE proteins are encoded by the rice (Oryza sativa) genome (Oelkers et al. 2008). CLE proteins have been demonstrated to have extensive functional conservation, and are expressed in diverse tissues where they presumably activate receptors in different pathways (Sharma et al. 2003; Jun et al. 2010; Meng and Feldman 2011).

One of the Arabidopsis CLEs, CLE40, promotes differentiation of the columella stem cells (CSCs) at the distal root meristem (RM). CLE40 functions via suppressing the expression of WOX5 (WUSCHEL-related homeobox5), a WUS homolog (Sarkar et al. 2007; Stahl et al. 2009). The cle40 mutant displays a waved root phenotype, with one extra layer of CSC cells (Hobe et al. 2003).

Tracheary element differentiation inhibitory factor (TDIF) is a CLE peptide isolated from a xylogenic Zinnia (Zinnia elegans) cell culture (Ito et al. 2006). Encoded by CLE41 and CLE44 in Arabidopsis, TDIF stimulates cell proliferation and suppresses differentiation of vascular elements (Ito et al. 2006). Treatments with TDIF or overexpression of CLE41 and CLE44 genes resulted in enlarged vascular bundles due to procambium cell proliferation (Hirakawa et al. 2010). TDIF peptides also showed axillary bud-promoting activity (Yaginuma et al. 2011). TDIF functions through another WUS homolog, WOX4, which is preferentially expressed in the vascular meristem tissues procambium and cambium. The expression level of WOX4 was upregulated upon application of TDIF peptides in a receptor-dependent manner (Hirakawa et al. 2010).

CLE8, which is exclusively expressed in young embryos and endosperm, positively regulates expression of WOX8. The CLE8-WOX8 signaling module functions to promote seed growth through regulating basal embryonic cell division patterns, endosperm proliferation, and the timing of endosperm differentiation (Fiume and Fletcher 2012).

In legumes, the symbiotic nodules are formed as a result of dedifferentiation and reactivation of cortical root cells. CLE peptides have been shown to be involved in the regulation of nodule development and autoregulation of nodulation (AON), which controls the number of nodules produced (Okamoto
et al. 2009
; Mortier et al. 2010). A number of CLE genes have been shown to be upregulated during nodulation (Okamoto
et al. 2009
; Mortier et al. 2010; Lim et al. 2011). Simultaneous knockdown of MtCLE12 and MtCLE13, two of the nodulation-associated CLE genes in Medicago truncatula, resulted in an increase in nodule numbers (Mortier et al. 2012). Overexpression of GmRIC1 and GmRIC2, two soybean CLE genes, strongly suppressed root nodulation in a receptor-dependent manner (Lim et al. 2011). It is interesting to note that these CLEs function in legume leaves to activate a signal that eventually reaches the roots to control nodule number (Kosslak and Bohlool 1984; Okamoto et al. 2009; Mortier et al. 2010). CLE-mediated regulation of cell proliferation and differentiation in nodule meristems also involves the WOX family transcription factors (Osipova et al. 2012).

The only group of CLE domain proteins known to be present in non-plant species are those identified in parasitic nematodes, including the soybean cyst nematode Heterodera glycines (Wang et al. 2001), the beet cyst nematode Heterodera schachtii, (Wang et al. 2011) and the potato cyst nematode Globodera rostochiensis (Lu et al. 2009). These types of nematodes parasitize roots by modulating host developmental programs to form enlarged, multinucleate feeding cells called syncytia, which serve as the nutritive source. Several studies have suggested that nematode-encoded CLEs function to facilitate the formation of feeding structures in host roots through the molecular mimicking of plant CLE ligands (Guo et al. 2011; Replogle et al. 2011; Wang et al. 2011). G. rostochiensis-secreted GrCLE1 can be correctly processed to an active form by host-plant proteases, and the processed GrCLE1 peptides bind directly to plant CLE receptors (Guo et al. 2011).

This review attempts to summarize recent progress in the field of CLE ligand signaling. Focuses of this article include proteolytic processing of precursor proteins in releasing CLE peptides, molecular structure of mature CLE ligands in relation to bioactivities, and ligand-receptor interactions.

Proteolytic Processing of CLE Proteins

All CLE precursors have a similar structure: small proteins (usually less than 150 amino acids) consisting of an N-terminal signal peptide, followed by a variable domain with significant sequence diversity, and a conserved C-terminal CLE motif (Cock and McCormick 2001; Figure  1A). In vitro and in vivo data indicate that the 14-amino acid CLE motif is the only functional region of CLV3 protein (Fiers et al. 2005, 2006; Ni and Clark 2006). Even if the entire sequence upstream of the CLE domain of CLV3 is replaced with unrelated sequences, the resulting chimeric protein is still fully functional (Ni and Clark 2006). Another study showed that most of the sequences flanking the CLE motif of CLV3 can be deleted without affecting its function (Fiers et al. 2006). Furthermore, treatment of Arabidopsis seedlings with synthetic peptides corresponding to the CLE motif of CLV3, and many other CLE proteins, mimics the overexpression phenotype of these CLE genes (Fiers et al.
2005
; Fiers et al. 2006; Strabala et al. 2006b; Kinoshita
et al. 2007
; Meng and Feldman 2010). This suggests that CLE peptides are likely the active form of CLV3 and other CLE family proteins.

Figure 1.

The CLV3 protein and MCLV3 peptide. (A) Structure of the CLV3 protein.(B) Proteolytic processing of CLE proteins. Proteolytic cleavage of GST-tagged CLV3 (no signal peptide) proteins was detected via western blot analysis. The cleaved products were analyzed via MS/MS, revealing N-terminal processing at the position of arginine 1 and C-terminal trimming activity (Ni et al. 2011).(C) Contribution of individual amino acids within and around MCLV3 (Song et al. 2012).

A putative endogenous mature CLV3 peptide (MCLV3) was identified by in situ matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) analysis (Kondo et al. 2006). In the study, Kondo et al. isolated MCLV3 from CLV3-overexpressing Arabidopsis calli as a dodecapeptide containing the 12 amino acid residues from Arg70 to His81 in CLV3 located within the CLE domain. Two of the three proline residues within MCLV3 were modified to hydroxyproline (Ph) (Kondo et al. 2006). The study also suggested that the 12-amino-acid RTVPhSGPhDPLHH is the minimal unit required for CLV3 signaling, since synthetic peptides lacking the N-terminal arginine or C-terminal histidine are not biologically active (Kondo et al. 2006). Using nano-LC-MS/MS analysis of apoplastic peptides of Arabidopsis plants overexpressing CLV3, a subsequent study identified MCLV3 as a 13-mer with arabinosylations (Ohyama et al. 2009). None of these putative MCLV3s, however, was identified within the SAM, where CLV3 functions.

At the same time, the TDIF peptide was isolated from cultured Zinnia mesophyll cells via high-performance liquid chromatography in conjunction with a bioassay. Similar to MCLV3, TDIF is a dodecapeptide with Pro4 and Pro7 being hydroxylated (Ito et al. 2006).

These findings suggest that CLE proteins are synthesized as inactive protein precursors, from which the active 12/13-amino-acid peptide ligands are liberated by proteolysis. Using the soluble fraction of cauliflower (Brassica oleracea) protein extracts as the source of meristem-derived proteolytic activity, Ni and Clark (2006) treated E. coli-expressed, glutathione S-transferase (GST)-tagged mature CLV3 (no signal peptide) for proteolysis in vitro. Two smaller fragments were detected by GST antibodies after incubation with cauliflower extracts, indicating proteolytic cleavage of GST-CLV3 at the C-terminus (Ni and Clark 2006; Figure  1B). Arabidopsis CLE1 and soybean protein GmCLE23 were cleaved by the same protein extracts (Ni et al. 2011), suggesting that in addition to CLV3, other CLEs could also be processed by the same proteolytic enzyme(s) from cauliflower. Furthermore, the CLE protein processing activities are observed in multiple tissues and various plant species (Ni and Clark 2006; Guo et al. 2011; Ni et al.
2011
).

Subsequent mass spectrum analyses on the digested fragments suggested that the recombinant CLV3 was processed by two separate activities: an N-terminal cleavage before the residue Arg70, matching with the in vivo maturation, and a C-terminal trimming (Ni et al. 2011; Figure  1B). A substitution study indicated that efficient recognition by the processing protease can occur with as little as four residues upstream of the CLE domain, and that the conserved arginine at position +1 (R1) and conserved acidic residues at positions −2 and/or −3 are required for efficient cleavage. When tobacco BY-2 suspension cell cultures were used for CLV3 processing, the processing activity was found to be present in cultured media, but not in cell extracts, suggesting that the protease(s) responsible is only active when secreted. Further analysis of the CLE processing activity using various protease inhibitors suggested that a serine protease seems to be responsible for the N-terminal processing of CLV3 (Ni et al.
2011
).

Successively smaller C-terminal tails were detected in mass spectrum analyses on the in vitro assays (Ni et al. 2011), suggesting that the C-terminal trimming may occur by action of a progressive carboxypeptidase (Figure  1B). Whether or not this reflects processing in vivo is currently unknown. It is interesting to note that SOL1, which encodes a putative Zn2+-carboxypeptidase, has been identified as a suppressor of the short-root phenotype generated from CLE19 overexpression (Casamitjana-Martinez et al. 2003).

Structural Properties of Mature CLE Peptides

The 14 amino acids of the CLE motif are highly conserved among CLE proteins (Cock and McCormick 2001). Synthetic peptides corresponding to the 14-amino-acid motif are functional in rescuing the clv3–2 phenotype (Fiers et al. 2006). Results from both in vitro and in vivo studies, however, suggest a shorter mature/active form of CLE peptides after proteolytic processing (Ito et al. 2006; Kondo et al. 2006; Kondo et al. 2008; Ogawa et al. 2008; Ohyama et al. 2009; Song et al. 2012). Both MCLV3 and TDIF were first isolated as dodecapeptides with two hydroxyproline residues (Ito et al. 2006; Kondo et al.
2006
). Since synthetic MCLV3 peptides lacking either the R1 or H12 residue are not biologically active (Kondo et al. 2006), the 12-amino-acid RTVPSGPDPLHH is the shortest active form of CLV3 protein. Although Ohyama et al. (2009) have isolated MCLV3 as a 13-amino-acid glycopeptide, the H13 residue does not seem to be critical for its function (Ohyama et al. 2009). In the same study, MCLE2 was isolated as a 12-amino-acid glycopeptide, simply because the full-length CLE2 protein ends at H12 (Ohyama et al. 2009). Adding one amino acid to the N-terminus of synthetic peptides, either an alanine residue or the original leucine from the precursor protein, seems to decrease the bioactivities of MCLV3 (Kondo et al. 2008). Adding an arginine to the C-terminus of TDIF, however, did not affect its bioactivities (Ito et al. 2006). MCLV3's bioactivities were not affected by adding one amino acid to its C-terminus (Kondo et al. 2008). Bioactivities of MCLV3 were only partially reduced when the 15 C-terminal amino acids from CLV3 protein were added to its C-terminus (Kondo
et al. 2006
). These data suggest that the 12-amino-acid CLE peptides are both necessary and sufficient for the biological functions of their precursor proteins. Isolation of 13-amino-acid glycopeptides from CLV3-overexpressing Arabidopsis (Ohyama et al. 2009) could be a result of partial C-terminal
trimming.

Within the dodecapeptides, each amino acid contributes differently to CLE function. Substitution of each of the 12 residues with alanine affects CLE activities differently. When used in treating Arabidopsis seedlings, synthetic MCLV3 and a number of other CLE peptides cause consumption of the RM and inhibit root growth (Fiers et al. 2005; Kondo et al. 2006; Meng and Feldman 2010). Using this root inhibition assay, Kondo et al. (2008) showed that MCLV3 residues R1, P9, H11, and H12 are critical, while T2, S5, P7, and L10 are the least important for its activity (Kondo et al. 2008). These authors also showed that residues R1, H12, G6, P4, and V3, are important, while residues T2, S5, P7, and L10 are not important for receptor binding of MCLV3 (Kondo et al. 2008). The significant overlap between results from these two assays suggests that receptor binding is critical in functionality of CLV3.

In a recent study, Song et al. (2012) took an in vivo approach in evaluating the contribution of individual residues to CLV3's function in the SAM. Point mutations were introduced to replace CLE residues with alanine in the peptide-coding region of CLV3. Transgenes carrying such substitutions were evaluated for their ability to rescue the clv3–2 mutation in Arabidopsis. Phenotypic analysis based on floral organ numbers, SAM size and WUS expression revealed that CLE residues D8, H11, G6, P4, R1, and P9, arranged in an order of importance, are critical, while T2, V3, S5, and P7 are trivial for the endogenous CLV3 function in SAM maintenance (Song et al. 2012; Figure  1C). One of the clv3 mutant alleles, clv3–1, carries a G to R point mutation at the G6 position of MCLV3 (Fletcher et al. 1999; Ni et al. 2011), supporting the importance of G6 in CLV3's function in vivo.

Similar alanine substitution analyses on TDIF have identified H1, V3, G6, N8, P9 and N12 as important residues, and E2, S5, P7, I10 and S11 as non-important ones for TDIF's function in suppressing tracheary element differentiation (Ito et al. 2006).

The consistency in identifying CLE residues in positions (R/H)1, G6, and P9 as being important and (T/E)2, S5 and P7 as non-important suggests a functional conservation in the peptide-coding motif. Interestingly, most of the residues that are highly-conserved within the CLE family seem to have a significant contribution to CLE function, with the exception of P7, which is one of the most highly-conserved residues but has been identified as unimportant in CLE function (Oelkers et al. 2008; Song et al. 2012).

The same P7 has been reported to carry one of the two hydroxylation modifications (the other one is P4) in MCLV3 (Kondo et al. 2006) and TDIF (Ito et al. 2006), and is the only residue post-translationally modified with three L-arabinose in the mature forms of CLV3 and CLE2 (Ohyama et al. 2009). Hydroxylations of these two proline residues were not necessary for the bioactivities of either MCLV3 (Kondo et al. 2006) or TDIF (Ito et al. 2006). Arabinosylation of P7, however, has been reported to be required for high-affinity receptor binding of MCLV3 and MCLE2 peptides (Ohyama et al. 2009). When used to treat clv3–2 mutant plants with enlarged SAMs, P7-arabinosylated CLV3 peptides showed higher rescuing activity than non-arabinosylated ones (Ohyama et al. 2009). This is inconsistent with the results from multiple studies showing that P7A mutations have little effect on bioactivities of CLE peptides (Ito et al. 2006; Kondo et al. 2008; Song et al. 2012), including an in vivo study carried out by Song et al. (2012) that evaluated the ability of mutant CLV3 proteins to rescue clv3–2 phenotypes (Figure  1C). This discrepancy may be due to the different experimental systems used in these studies.

Receptor Activation by CLE Peptides

The functionality of a CLE peptide is closely related to its ligand-binding ability (Kondo et al. 2008). CLE genes have been demonstrated to be expressed in diverse tissues (Sharma
et al. 2003
; Jun et al. 2010), where they presumably function in different pathways by interacting with different receptor kinases.

The SAM-regulating CLV pathway currently has the best-studied CLE signaling components. The MCLV3-mediated regulation of stem cell homeostasis at the SAM acts through a unique mode of receptor activation, whereby two receptor complexes bind independently to the same ligand. CLV1 is a leucine-rich repeat (LRR) receptor-like kinase (RLK), which possesses 21 extracellular LRRs, a transmembrane domain, and a cytoplasmic ser/thr kinase domain (Clark et al. 1997). BAM1 and BAM2 are CLV1 homologs, which act redundantly with CLV1 in the SAM center (DeYoung et al. 2006). CLV2 is a receptor-like protein composed of 21 LRRs in the predicted extracellular portion, and a short cytoplasmic tail (Jeong et al. 1999). CORYNE (CRN) is a transmembrane kinase lacking any extracellular domain (Muller et al. 2008;
Figure  2).

Figure 2.

Current model of the CLAVATA (CLV) signaling pathway.In perceiving the CLV3 signal released through proteolytic processing of the precursor protein, the CLV receptors function through two distinct complexes: a CLV2/CRN multimer and a CLV1/BAM multimer. The CLV3 signal is further relayed through the type 2C protein phosphatases POLTERGEIST (POL) and PLL1, leading to the suppression of WUSCHEL (WUS) expression and stem cell differentiation at the shoot apical meristem (SAM).

It has been shown that CLV1 binds to MCLV3 through its extracellular domain (Ogawa et al. 2008). Guo et al. (2010) showed that CLV2 and the BAM receptors also bind to MCLV3 with a similar affinity. A recent study has identified the LRR6–LRR8 region as the CLE-interacting part of the BAM1 protein (Shinohara et al. 2012). These ligand-binding receptors seem to function through two distinct complexes: a CLV2/CRN multimer and a CLV1/BAM multimer (Bleckmann et al. 2010; Guo et al. 2010; Zhu et al. 2010). Consistent with the model that the two ligand-binding receptor complexes function independently, Guo et al. showed that BAM and CLV1 can replace CLV2 function in vivo when overexpressed (Guo et al. 2010). A third receptor complex formed by RECEPTOR-LIKE PROTEIN KINASE 2 (RPK2) oligomerization has been suggested to function in parallel with CLV1/BAM and CLV2/CRN (Kinoshita et al. 2010). Recent studies suggest that CRN lacks kinase catalytic activity (Nimchuk et al. 2011a), and that the CLV1 protein, which has kinase activity, migrates from plasma membranes to lytic vacuoles in an MCLV3-dependent manner (Nimchuk et al. 2011b). The CLV3 signal is further relayed through the type 2C protein phosphatases POLTERGEIST (POL) and PLL1 (Song et al. 2006; Gagne and Clark 2010), leading to the suppression of the transcription of WUS, the homeodomain transcription factor that functions to promote stem cell identity (Mayer et al. 1998; Figure  2).

Receptor-like kinases have also been identified to function in several other CLE signaling systems. Within the RM, CLE40 suppresses WOX5 expression and columella stem cell differentiation through promoting expression of the receptor kinase ARABIDOPSIS CRINKLY4 (ACR4) (Stahl et al. 2009). TDIF/CLE41/CLE44 is recognized by TDR/PXY (TDIF receptor/Phloem intercalated with XYLEM), an LRR receptor kinase located in the plasma membrane of procambial cells (Hirakawa et al. 2008; Etchells and Turner 2010). Hirakawa et al. (2008) showed that TDIF binds to TDR in a highly specific manner. Upregulation of the WOX4 transcription factor in the procambium and cambium by TDIF is TDR-dependent (Hirakawa et al.
2010
). A number of CLV1 homologous receptor kinases in legume plants, including HYPERNODULATION ABERRANT ROOT (HAR) 1 from Lotus japonicus (Okamoto et al. 2009), GmNARK (Glycine max nodule autoregulation receptor kinase) from Glycine max (Lim et al. 2011), and SUPER NUMERIC NODULES (SUNN) from Medicago truncatula (Mortier et al. 2010), have been shown to be involved in CLE regulation of nodulation and AON, suggesting putative ligand-receptor interactions.

Some of the RLKs, especially those with broad expression patterns, might be involved in multiple CLE signaling pathways. CLV2, for example, is not only involved in CLE regulation of meristem activities at both the SAM and the root apical meristem (RAM) (Fiers et al. 2005; Miwa et al. 2008; Meng and Feldman 2010; Song et al. 2010), it has also been reported to be involved in CLE-mediated regulation of nodulation (Krusell et al. 2011) and ligand mimicry by CLEs encoded by plant parasitic nematodes (Guo et al. 2011). CLV2, however, is not involved in the CLE40-WOX5 signaling module that regulates differentiation of columella stem cells at the distal RM (Stahl et al. 2009). In addition, the BAM receptors are also involved in CLE signaling in the RAM (Guo et al. 2011). TDR seems to function to perceive CLE signals during both vascular development and axillary bud formation (Hirakawa et al. 2008; Yaginuma et al. 2011).

Perspectives

During the last couple of years, significant progress has been made in understanding the functions of CLE ligands in plants. With the first endogenous mature CLE peptide isolated, proteolytic processing activities of CLE precursor proteins characterized, a number of CLE receptors identified and direct ligand-receptor interactions established, CLE signaling in plants is a rapidly developing area of research. Furthermore, the use of synthetic peptides to create gain-of-function phenotypes allows rapid genetic screens and easy access to CLE-induced cellular and biochemical changes. The presence of multiple CLEs in a single tissue type (Jun et al. 2010), and the fact that multiple CLEs can replace CLV3 when expressed at the SAM (Ni and Clark 2006), however, make studies on the specificity of CLE genes very challenging.

Several approaches could be taken to study the specificity of plant CLEs. First, cell-type specific determination of CLE gene expression would provide key information about putative functions. With GUS reporter lines available for most of the 32 Arabidopsis CLE genes (Jun et al. 2010), fine dissection of the expression patterns of multiple CLE genes within certain groups of cells is feasible. Comparison of spatial as well as temporal expression patterns between different CLE genes would help our understanding of how functional specificity is determined. Second, the specific interactions between CLE ligands and receptors may also be critical for functional specificity. Although some CLE receptors have been shown to be able to bind multiple CLE peptides (Guo et al. 2010), a study on binding kinetics and activation mechanisms of individual CLEs to target receptor proteins could reveal another layer of regulation on specificity determination. Third, proteolytic processing and maturation of CLE peptides are critical for their function. In order to test the role of CLE maturation as a layer of regulation for functional specificity, proteolytic processing as well as post-translational modification of CLE peptides in specific cell types need to be determined.

(Co-Editor: Chun-Ming Liu)

Acknowledgements

We thank Dr. Steven Clark of the University of Michigan for his critical reading of the manuscript. Research in our lab has been supported by a startup fund from the Tobacco Research Institute, Chinese Academy of Agricultural Sciences.

Ancillary