A basic/helix–loop–helix transcription factor controls leaf shape by regulating auxin signaling in apple
Summary
- Climate-driven phenological change across local spatial gradients leads to leaf shape variation. At higher elevations, leaves of broadleaf species tend to become narrower, but the underlying molecular mechanism is largely unknown.
- In this study, a series of morphometric analyses and biochemical assays, combined with functional identification in apple, were performed.
- We show that the decrease in apple leaf width with increasing altitude is controlled by a basic/helix–loop–helix transcription factor (bHLH TF), MdbHLH3. The MdbHLH3-overexpressing lines have a lower transcript abundance of MdPIN1 encoding an auxin efflux carrier but a higher transcript abundance of MdGH3-2 encoding a putative auxin amido conjugate synthase, resulting in a lower free auxin concentration; feeding the transgenic leaves with exogenous auxin partially restores leaf width. MdbHLH3 transcriptionally suppresses and activates MdPIN1 and MdGH3-2, respectively, by specifically binding to their promoters. This alters auxin homeostasis and transport, consequently leading to changes in leaf shape.
- These findings suggest that the bHLH TF MdbHLH3 directly modulates auxin signaling in controlling leaf shape in response to local spatial gradients in apple.
Introduction
Plant leaves are produced from the shoot apical meristem (SAM) and its growth feature dictates the diversity of shape (Efroni et al., 2010). Leaf shape diversity falls into two broad categories: leaf shape diversity among species caused by genetic variation (Tsukaya, 2004) and diversity within species caused by environmental factors and developmental cues (Tsukaya, 2006). Different altitude or latitude drives phenological change across local spatial gradients; this spatial variation refers to complicated environmental factors such as temperature and ultraviolet light and is critical for determining diversity of leaf shape variation within species (Gurevitch, 1988; Hovenden & Vander Schoor, 2006; Rafferty et al., 2020). There has been considerable recent debate concerning the use of such leaf characters as a proxy for palaeoclimates, especially as environmental control of leaf form is poorly understood for the taxa being used for climatic reconstructions (Hovenden & Vander Schoor, 2006). Therefore, elucidating the molecular mechanisms underlying leaf shape variation within species as a result of spatial variation seems very important for understanding the drivers of diversification in leaf shape (Chitwood et al., 2016).
The leaf shape variation within species as a result of spatial variation refers to many genes that control indeterminacy of SAM. For example, engineered alterations in the expression of a homeobox gene LeT6 and maize homeobox-containing Knotted-1 (Kn1) gene cause dramatic changes in leaf morphology in tomatoes (Hareven et al., 1996; Janssen et al., 1998), while an enhancer element in the homeobox gene REDUCED COMPLEXITY (RCO) altered leaf shape by changing gene expression from the distal leaf blade to its base (Vlad et al., 2014; Vuolo et al., 2016), suggesting that the homeobox genes are important regulators of leaf development. Class 1 KNOTTED1-like (KNOX1) genes seem to be key regulators during the development of compound leaves in almost all vascular plants (Champagne et al., 2007; Hay & Tsiantis, 2010); however, KNOX gene expression in leaves is not sufficient to promote leaf dissection during secondary morphogenesis (Bharathan et al., 2002; Hay & Tsiantis, 2010), suggesting that other factors regulate differential leaf growth at this stage.
The phytohormone auxin is a signal in the pattern of organ initiation from the SAM, the generation of leaf vascular tissue and the development of leaf serrations (Mattsson et al., 2003; Heisler et al., 2010; Kasprzewska et al., 2015). It is considered to be a crucial determinant of leaf shape variation within species resulting from spatial variation; these leaf shape variations are regulated by many genes, such as the auxin receptor gene TIR1, the pin-formed (PIN) protein family of auxin efflux carriers, as well as auxin importers (AUXs/LAXs) and auxin response factors (ARFs) (Hay et al., 2006; Bilsborough et al., 2011; Ren et al., 2011; Kasprzewska et al., 2015; Ben-Gera et al., 2016; Wu et al., 2018). These findings suggest a conserved role for auxin in the specification of developmental pattern, and the resemblance between leaf and leaflet initiation makes auxin an excellent candidate in the regulation of leaf morphology. In addition to auxin-associated genes in the regulation of leaf shape, transcription factors (TFs) constitute a large family of proteins, such as MYB, AP2 and NAC, as well as basic/helix–loop–helix (bHLH) TFs that have been implicated in the control of leaf shape in response to changes in environmental signals in plants (Kim et al., 2003; Berger et al., 2009; Jiang et al., 2012). Among these TFs, the bHLH superfamily, which is named for its highly conserved alkaline/helix–loop–helix domains, has been extensively studied in higher plants and animals (Duek & Fankhauser, 2005; Feller et al., 2011). The typical bHLH TF consists of two conserved motifs: a basic region and an HLH region. The basic region is responsible for DNA recognition and binding, while the HLH region, composed of hydrophobic residues, is involved in dimerization (Murre et al., 1989). Interestingly, the plant-specific bHLH-containing DNA-binding proteins TCP14 and TCP15 affect internode length and leaf shape by using SRDX fusions and reporter gene analysis in Arabidopsis (Kieffer et al., 2011).
Multiple environmental factors at different altitudes can affect the optimal shape of leaves (Gurevitch, 1988; Hovenden & Vander Schoor, 2006), but the underlying molecular mechanism involved in the alteration of leaf shape remains largely unexplored. In this work, we show that an apple bHLH TF, MdbHLH3, is expressed at a higher level at higher altitude and controls leaf shape by specifically suppressing auxin efflux carrier MdPIN1, and activating auxin amido conjugate synthase gene MdGH3-2.
Materials and Methods
Plant materials and growth conditions
Apple cv ‘Red Delicious’ trees were planted on Maiji mountain (Tianshui, Gansu, China) at altitudes of between 800 and 2000 m in early April 2013, with 20 trees planted at each latitude. Apple leaves from the trees grown at different altitudes were gathered for 3 yr during 2015–2017 and used for morphological and RNA-seq analysis.
Three MdbHLH3 overexpression lines, MdbHLH3-36, MdbHLH3-27 and MdbHLH3-44, and wild-type (WT) control trees were planted in the field at an experimental farm in April 2011. At least 30 apple leaves were taken from each of the three MdbHLH3 overexpression lines and WT control trees for morphological characterization and RNA-seq analysis.
The IAA-treated leaves of WT and MdbHLH3 transgenic lines were sprayed with exogenous 100 mg l−1 IAA from the inception of the leaf buds. They were sprayed with this application once a week for a month, after which the leaf morphological indexes were again measured.
In all, 22 apple cultivars were planted in the field at an experimental farm of Shandong Agricultural University in April 2012, with at least six trees planted for each cultivar. For analysis of the leaf variation among these 22 cultivars, we selected between 54 and 70 leaves for each cultivar. Apple leaves from these different apple cultivars were gathered for 3 yr (from 2016 to 2018) and then used for morphological analysis.
Apple calli were induced from the young embryos of ‘Orin’ apple (Malus domestica Borkh.). They were grown on MS medium supplemented with 0.5 mg l−1 IAA and 1.5 mg l−1 6-benzyladenine at room temperature in the dark. The calli were subcultured three times every 15 d before being used for genetic transformation and other assays.
Morphometric analyses
For analyses of leaf shape along altitude, we selected at least three trees at each of the altitudes 800, 1200, 1600 and 2000 m, with 10 mature leaves sampled from each tree. For leaf comparison between the WT and the three mutants (MdbHLH3-36, MdbHLH3-27 and MdbHLH3-44), we also selected three individuals for each and 10 leaves per individual. For analysis of the leaf variation among the 22 cultivars, we selected between 54 and 70 leaves for each cultivar. A total of 1535 leaves were scanned using a Hewlett-Packard printer (LaserJet Pro MFP M128fn; Hewlett-Packard, Palo Alto, CA, USA) with a resolution of 600 dpi. Twelve landmarks were selected from each scanned leaf as previously described (Viscosi et al., 2009; Y. N. Hu et al., 2019), consisting of: the junction between the petiole and the leaf blade; the first serration on the right-hand side of the leaf; the right-hand point of one-fifth of the leaf from the lower side; the right-hand point of the widest part of leaf; the right-hand point of three-fifths of the leaf from the lower side; the right-hand point of four-fifths of each leaf from the lower side; the leaf tip; the left-hand point of four-fifths of the leaf from the lower side; the left-hand point of three-fifths of the leaf from the lower side; the left-hand point of the widest part of the leaf; the left-hand point of one-fifth of the leaf from the lower side; and the first serration on the left-hand side of the leaf. The x and y coordinates of each landmark were recorded on digitized leaves using the program ImageJ (Abràmoff et al., 2004). The 12 landmarks were converted to a configuration of 24 cartesian coordinates, which were stored in a ‘.txt’ file. Fourteen outliers were detected using Morphj (Klingenberg, 2011) and were removed from subsequent analyses. A generalized Procrustes analysis (GPA) was performed using the R package shapes (Dryden, 2013). For the 12 landmarks in two dimensions (x, y coordinates), the procGPA function was used to perform GPA. Eigenleaves were visualized using the ‘shapepca’ function and principal component (PC) scores, percentage variance explained by each PC and Procrustes-adjusted coordinates were obtained from procGPA object values.
RNA extraction, RT-PCR and qRT-PCR assays
RNA extraction, reverse transcription polymerase chain reaction (RT-PCR) and quantitative RT-PCR (qRT-PCR) assays were conducted as previously described by Hu et al. (2016b). All of the primers used in this study are listed in Supporting Information Table S1.
Plasmid construction and genetic transformation
The sense full-length sequence of MdbHLH3 was amplified to construct sense overexpression vectors. The resulting PCR products were inserted into the PBI121 vector under the control of the 35S promoter. This vector was genetically transformed into tissue cultures of apple cv ‘Gala’ using Agrobacterium strain LBA4404 as described previously (Xie et al., 2012).
RNA-seq analysis
Total RNAs were extracted from leaf buds of apple cv ‘Red Delicious’ at different altitudes (800, 1200, 1600, 2000 m), and apple cv ‘Gala’ of WT and MdbHLH3 transgenic apple trees. Subsequently, the RNAs were used to construct libraries for high-throughput parallel sequencing using an Illumina genome analyser II. A rigorous algorithm was used to identify the differentially expressed genes (DEGs) in these samples. The false discovery rate (FDR) was set at 1% to determine the threshold of the P-value in multiple tests and analyses by manipulating the FDR value (Audic & Claverie, 1997). P < 0.001 and the absolute value of log2ratio > 1.5 were used as the threshold to determine the significance of the gene expression differences according to Audic & Claverie (1997). A gene ontology (GO) analysis was conducted to predict gene function and calculate the functional category distribution frequency.
Chromatin immunoprecipitation (ChIP) qRT-PCR and electrophoretic mobility shift assay (EMSA)
35S::MdbHLH3-Myc and 35S::Myc stable transgenic apple calli were used for the ChIP-qRT-PCR analysis. The anti-Myc antibody (Beyotime, Nanjing, China) was used for ChIP as described by Hu et al. (2016a). The resultant samples were used as templates for qRT-PCR. The primers used for ChIP-PCR are listed in Table S1. EMSA was conducted according to Hu et al. (2016b). MdbHLH3 was cloned into the expression vector PET-32a-c. The MdbHLH3-His recombinant protein was expressed in Escherichia coli strain BL21 and purified using glutathione sepharose beads (Thermo Scientific, San Jose, CA, USA). The oligonucleotide probe of the MdPIN1, MdGH3-2 and MdFLS promoters were labelled with an EMSA probe biotin labelling kit (Beyotime) according to the manufacturer’s instructions. The recombined protein of MdbHLH3-His was incubated with 10 × binding buffer, 1 μg μl−1 poly (dI-dC), and 400 fmol biotin-labelled double-stranded binding consensus oligonucleotides (total volume 20 μl) using a LightShift Chemiluminescent EMSA Kit (Thermo Fisher Scientific, Rockford, IL, USA). The binding reaction was performed at room temperature for 20 min. The DNA–protein complexes were separated on 6.5% nondenaturing polyacrylamide gels, electrotransferred and detected following the manufacturer’s instructions. The binding specificity was also examined by competition with a fold excess of unlabelled oligonucleotides. The primers used for EMSA are listed in Table S1.
Histochemical staining of β-glucuronidase (GUS) activity
Reporter constructs containing the MdFLS promoter sequence was prepared as previously described (Hu et al., 2017).
Transient expression assay in Nicotiana benthamiana leaves
Transient expression assays in N. benthamiana leaves were performed as previously described by D. G. Hu et al. (2019). The MdPIN1 and MdGH3-2 promoters were amplified and cloned into pGreenII 0800-LUC vectors, which generated the reporter constructs MdPIN1pro::Luc and MdGH3-2pro::Luc, respectively. The effector (35Spro::MdbHLH3) was constructed by cloning the open reading frame of the MdbHLH3 into the pGreenII 62-SK vector. A charge-coupled imaging apparatus (NightOWL II LB983 in conjunction with Indigo software; Berthold Technologies, Bad Wildbad, Germany) was used to collect the LUC images and quantify luminescence intensity. Transformed leaves were sprayed with and soaked in 100 mM luciferin, after which they were placed in darkness for 6 min before luminescence examination.
Flavonoid fluorescence staining
To measure flavonoid accumulation, we incubated apple leaves in 2-aminoethyl diphenylborinate (DPBA) staining solution containing 0.25% (w/v) DPBA and 0.005% (v/v) Triton X-100 for 5 min as described by Murphy et al. (2000). The leaves were then washed for 5 min with 50 mM sodium phosphate buffer (plus 0.005% (v/v) Triton X-100, pH 7.0). After excitation with 488 nm (argon) laser, the DPBA emission was collected at 570–650 nm using LSM710 (Carl Zeiss confocal fluorescence microscope) (Silva-Navas et al., 2016).
Determination of free IAA content
Samples (c. 0.15 g FW) were collected and vacuum-dried at low temperature (−30°C) for 12 h. The samples were then treated and extracted according to the method described by Sun et al. (2018). After extraction, the free IAA was measured by high-performance liquid chromatography.
In situ hybridization
The SAMs of WT and three MdbHLH3 transgenic lines were fixed in FAA overnight at 4°C followed by dehydration through a series of graded ethanol concentrations and a xylene series, then embedded in Paraplast Plus (Sigma) and sliced into sections (8 mm). Digoxigenin-labelled RNA probes of MdbHLH3, MdPIN1 and MdGH3-2 were synthesized in vitro using SP6 and T7 RNA polymerase (Roche). The hybridization and immunological detection of the hybridized probes were conducted according to the method described by Chen et al. (2015). The hybridized probes were detected using anti-digoxigenin-Ap antibody and then visualized with 4-nitroblue tetrazolium chloride and 5-bromo-4-chloro-3-indolyl-phosphate (NBT/BCIP). Photographs were taken using an Imager D2 (Carl Zeiss).
IAA immunohistochemical localization
Excised tissue samples were prefixed immediately in a 2% (w/v) aqueous solution of l-ethyl-3-(3-dimethyl-aminopropyl)-carbodiimide hydrochloride (EDAC; Sigma) and postfixed in 4% paraformaldehyde and 2.5% glutaraldehyde overnight at 4°C, dehydrated with a graded ethanol series, embedded in paraffin, and sectioned into 8 µm slices. Then, the slides were reacted with the anti-IAA antibody followed by anti-mouse IgG secondary antibody, and then immunodetected using a chemiluminescent method according to Hou et al. (2004).
Statistical analysis
Samples were analysed in triplicate, and the data are expressed as the means ± standard deviation unless noted otherwise. Statistical significance was determined using Student’s t-test. Differences with P ≤ 0.01 were considered to be statistically significant.
Accession numbers
Sequence data from this article can be found in the Apple Genome and Epigenome (https://iris.angers.inra.fr/gddh13/) under accession nos. MdbHLH3 (MD11G1286900), MdPIN1 (MD14G1236300), MdPIN2 (MD16G1244500), MdPIN4 (MDP0000144113), MdAUX1 (MD12G1162400), MdTAA1 (MD12G1162400), MdIAA27 (MD02G1057200), SAUR family protein (MD07G1156400), SAUR family protein (MD10G1061400), GH3-1 (MD05G1092300), GH3-2 (MD05G1092900), GH3-3 (MD09G1091000), GH3-4 (MD09G1202300), ARF (MD05G1309400), ARF (MD07G1152100), ARF (MD08G1015500), Auxin efflux carrier (MD13G1025300), Auxin efflux carrier (MD16G1027900), YUCCA (MD15G1098700), YUCCA (MD15G1184800), Phospholipase D1/2 (MD04G1197500).
Results
Apple leaf width decreases with altitude
All landmarks were aligned using a GPA, and a principal component analysis (PCA) was performed to visualize the major sources of shape variance of leaves of the apple variety ‘Red Delicious’ obtained at altitudes between 800 and 2000 m. The first four PCs summarized 68.8% of shape variance among leaves from different altitudes (Fig. 1a). PC1 and PC2 produce largely overlapping clusters but leaves from an altitude of 800 and 2000 m overlapped to a lesser extent (Fig. 1b). The shape variance, represented by PC1 and PC2, is mainly influenced by leaf width and marginally influenced by leaf length (Fig. 1a). This suggests that altitude affects leaf shape in apple trees. No significant difference in leaf length was detected between altitudes, but both leaf width and leaf area decreased with altitude (Fig. 1c–e).

Genome-wide RNA sequencing (RNA-seq) analysis shows that altitude-mediated alteration in leaf width is highly correlated with MdbHLH3
To elucidate the likely mechanism of altitude-mediated alteration in leaf shape, total RNAs were extracted from the leaf buds obtained at different altitudes (800, 1200, 1600 and 2000 m) for RNA-seq analysis. At least 50 million raw reads were generated and the average length was about 120 bp, encompassing at least 6.52 Gb of sequence data, which was sufficient to analyse gene expression quantitatively. Subsequently, we performed pairwise comparisons of transcript abundance to determine DEGs (−2 > log(fold-change) > 2; FDR < 0.01) in these four types of apple leaf samples. The total number of DEGs were 1463 at altitudes between 1200 m and 800 m (847 up; 616 down), 1490 between 1600 and 1200 m (1081 up; 409 down), and 1584 between 2000 and 1600 m (604 up; 980 down) (Fig. 2a; Dataset S1). Among these DEGs, 95 genes were shared between all three contrasts (Fig. 2a).

To facilitate identification of regulatory genes for leaf shape, we performed GO enrichment analysis by subjecting the sequences to InterPro and GO annotations. Interestingly, ‘plant hormone signal transduction’ was abundant in the Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis and enriched bubble diagram of the downregulated unigenes in the 1200 vs 800 m (Fig. 2b), 1600 vs 1200 m (Fig. 2c) and 2000 vs 1600 m (Fig. 2d) leaf samples. Furthermore, a group of potential candidate unigenes from the DEGs associated with auxin biosynthetic and signalling, as well as auxin-responsive genes were selected for further investigation. The corresponding ID numbers and log2(fold-change) values of these unigenes are listed in Dataset S1. The heatmaps display the average absolute expression values after log2 transformation in these apple leaf samples. Among them, the expression of some auxin transport unigenes (PIN3/10, LAX2, and ABC transporter F family member 4-like) and a range of auxin-responsive protein genes (ARF2/3/4/5/8 and AFB2), as well as auxin biosynthetic genes (YUCCA6/8) was significantly reduced, whereas the expression of auxin negative regulators, including GH3.1/3.6 and SAUR32/36/71, was significantly upregulated (Fig. 2e).
Remarkably, the RNA-seq analysis showed that the expression of a bHLH TF MdbHLH3 was notably elevated in the 1200 vs 800 m, 1600 vs 1200 m and 2000 vs 1600 m leaf samples (Fig. 2e). Subsequently, the MdbHLH3 expression pattern was further validated by real-time qRT-PCR assay. Indeed, the expression of MdbHLH3 increased linearly as altitude increased (y = 0.0052x – 3.3587, R2 = 0.9899) with those at 2000 m having the highest expression level (Figs 2f, S1).
Taken together, these results suggest that altitude-mediated alteration in leaf shape is highly correlated with the MdbHLH3 expression pattern.
MdbHLH3 overexpression lines exhibit altered leaf morphology in apple
To determine the physiological role of MdbHLH3 in plant development, we transformed apple with the 35S::MdbHLH3-GFP construct using Agrobacterium strain LBA4404. Three of the 35S::MdbHLH3-GFP transgenic apple lines, MdbHLH3-36, MdbHLH3-27 and MdbHLH3-44, with different expression levels, MdbHLH3 (Fig. 3a), were selected from eight independent transformed lines. These 35S::MdbHLH3-GFP transgenic apple lines displayed a strikingly visible alteration in leaf morphology (Fig. 3b). PCA analysis separated the WT from a cluster of MdbHLH3-36, MdbHLH3-27 and MdbHLH3-44, with PC1 and PC2 explaining 51.9% and 16.6% of variation, respectively (Fig. 3c). The cluster of MdbHLH3-36 overlapped partially with a largely overlapping cluster of MdbHLH3-27 and MdbHLH3-44 (Fig. 3c). The shape variance, represented by PC1 and PC2, is mainly influenced by leaf width and marginally influenced by leaf length (Fig. 3d). No significant difference was detected in leaf length between WT and the three 35S::MdbHLH3-GFP transgenic apple lines (Fig. 3e); however, the ratio of leaf length to width of the three MdbHLH3 transgenic apple leaves was significantly larger than that of the WT apple leaves (Fig. 3f). These results suggest that MdbHLH3 modulates leaf development and leaf shape in apple.

In addition, we also observed abnormal leaf shapes in some of the MdbHLH3 transgenic lines (Fig. 3g). These abnormal leaves accounted for 31.7%, 22.1% and 41.4% of the leaves in MdbHLH3-36, MdbHLH3-27 and MdbHLH3-44 transgenic lines, respectively, whereas most of the WT apple leaves are normal (Fig. 3h). These data indicated that MdbHLH3 is involved in controlling leaf morphology in apple.
Leaf shape is highly correlated with the expression of auxin-associated genes
To elucidate the potential mechanism of MdbHLH3 in regulating leaf shape, we performed high-throughput RNA-seq on the leaf buds obtained from the three MdbHLH3 transgenic lines. Over 50 million raw reads were generated with an average length of reads about 125 bp. Subsequently, we conducted pairwise comparisons of transcript abundance to determine DEGs (−2 > log(fold-change) > 2; FDR < 0.01) in these four types of apple leaves. The total number of DEGs were 3635 between MdbHLH3-27 and the WT control (880 up; 2755 down), 3057 between MdbHLH3-36 and the WT control (750 up; 2307 down), and 4186 between MdbHLH3-44 and the WT control (1055 up; 3131 down) (Fig. 4a; Dataset S2). Among these DEGs, 165 genes were upregulated and 1371 genes were downregulated in all three MdbHLH3-overexpressing apple leaves compared with the WT control (Fig. 4b,c).

Gene ontology and KEGG enrichment analyses were performed by subjecting the sequences to InterPro and GO annotations, as well as KEGG pathway map. In addition to confirming the overexpression levels of MdbHLH3 in these three MdbHLH3-overexpressing lines (MdbHLH3-36, MdbHLH3-27 and MdbHLH3-44) (Dataset S2), these annotations revealed that the expression levels of several genes, such as AUX and ARF, in the auxin signalling pathway, were significantly downregulated, whereas the putative auxin amido conjugate synthase MdGH3 was upregulated in the three MdbHLH3 transgenic lines (Fig. 4d). Furthermore, GO functional classification analysis of auxin biosynthetic and signalling, as well as auxin-responsive genes, showed that most of these genes were downregulated, but the auxin-responsive genes, including SAUR36 and SAUR71, were upregulated in the three MdbHLH3 transgenic lines (Fig. 4e). These top gene hits, especially genes in the auxin signalling pathway, are largely shared with the altitude RNA-seq datasets. These results suggest that MdbHLH3-mediated leaf shape is correlated with the expression of auxin-associated genes in apple.
Overexpression of MdbHLH3 decreases auxin accumulation in apple leaves
To validate RNA-seq data, some auxin biosynthetic and transport genes were selected for qRT-PCR assays. qRT-PCR analysis demonstrated that the expression levels of auxin biosynthetic genes (MdYUC2, MdYUC6 and MdTAA1) and auxin influx carrier MdAUX1, as well as auxin efflux carriers (MdPIN1, MdPIN2 and MdPIN4), were significantly reduced in the three MdbHLH3-overexpressing apple leaves compared with the WT (Fig. 5a).

As overexpression of MdbHLH3 altered the expression of auxin-associated genes in apple, we measured the endogenous free IAA concentration in the leaves of the WT and MdbHLH3 transgenic plants. As observed with the downregulated expression of auxin transport genes, such as MdAUX1, MdPIN1, MdPIN2 and MdPIN4 (Fig. 5a), the free IAA concentration in the leaf buds of MdbHLH3-transgenic apple was decreased by 13.74%, 9.43% and 16.55%, respectively, compared with the WT control (Fig. 5b). Considering that MdbHLH3-mediated leaf shape is correlated with auxin-associated gene expression, the lower auxin concentration in the transgenic lines suggests that MdbHLH3 modulates leaf shape by regulating auxin homestasis and transport. In addition, the free IAA content in the shoots and roots of MdbHLH3-transgenic apple was also decreased compared with the WT control (Fig. S2). Meanwhile, overexpression of MdbHLH3 led to auxin-induced root development abnormalities, further supporting the role of MdbHLH3 in regulating auxin accumulation (Fig. S3).
The availability of monoclonal antibodies highly specific for IAA made it possible to localize and evaluate endogenous IAA concentrations in plants (Avsian-Kretchmer et al., 2002). We prefixed leaf tissue samples of the WT and MdbHLH3 transgenic apple plants with EDAC, which cross-links the carboxyl group of IAA to structural proteins in the plant tissues, creating the epitope recognized by this anti-IAA monoclonal antibody. As shown in Fig. 5(c), lower red IAA signals were detected in the leaf tissue of MdbHLH3 transgenic apple plants than in the WT control, confirming that MdbHLH3 transgene decreases auxin accumulation in apple leaves. In addition, we examined the expression pattern of MdbHLH3 in SAMs by in situ hybridization. Fig. 5(d) shows the expression pattern of MdbHLH3, indicating an increased expression in these three MdbHLH3 transgenic lines. These results together suggest that MdbHLH3 plays a crucial role in determining auxin accumulation.
If a lower IAA concentration is responsible for the alteration in leaf shape in the MdbHLH3-overexpression lines, exogenous application of IAA from the early stage of leaf development should partially restore the leaf shape of these transgenic lines to the WT control. Indeed, exogenous 100 mg l−1 IAA treatment from the inception of leaf buds either partially (MdbHLH3-44) or completely (both MdbHLH3-36 and MdbHLH3-27) restored the leaf shape (Fig. 5e,f). Interestingly, the proportion of abnormal leaves in the MdbHLH3 transgenic lines was also noticeably reduced by the IAA treatment (Fig. 5g). These results suggest that MdbHLH3 overexpression leads to leaf shape changes in an auxin-dependent pathway.
MdbHLH3 specifically binds to the promoters of MdPIN1 and MdGH3-2, suppressing and activating their transcriptional expression, respectively
Basic/helix–loop–helix TFs recognize the E-box (5′-CANNTG-3′) or G-box (5′-CACGTG-3′) cis-element in the promoters of their target genes (Fisher & Goding, 1992). To determine if MdbHLH3 directly regulates the expression of auxin-associated genes, we analysed the promoters of auxin-associated genes. This analysis revealed at least one typical bHLH-binding E-box or G-box cis-element in the promoters of auxin efflux carriers (MdPIN1, MdPIN2, MdPIN4 etc.), and auxin biosynthetic genes (MdYUCCAs and MdTAA1), as well as auxin-responsive genes (MdARFs, MdIAAs, MdSAURs, MdGH3s and phospholipase D1/2) (Fig. S4); however, there was no typical bHLH-binding E-box or G-box cis-element in the promoter of auxin influx carrier MdAUX1 (Fig. S4). To verify in vivo binding of MdbHLH3 to the promoters of these genes, ChIP-PCR assays were performed. 35S::MdbHLH3-Myc and 35S::Myc transgenic apple calli were used. The bHLH cis-element-containing promoter regions of MdPIN1 and MdGH3-2 were found to be enriched in 35S::MdbHLH3-Myc transgenic calli compared with the 35S::Myc control, but the promoters of the other genes were not (Fig. 6a,b). These results provide in vivo evidence for binding of MdbHLH3 to the promoters of MdPIN1 and MdGH3-2.

To further confirm the binding of MdbHLH3 to the promoters of MdPIN1 and MdGH3-2 in vitro, an EMSA was carried out using a prokaryote-expressed and purified MdbHLH3-His fusion protein. When a bHLH-binding cis-element-containing oligonucleotide was used as a labelled probe, a specific DNA-MdbHLH3 protein complex was strongly detected. Meanwhile, the formation of these complexes decreased with increasing amounts of the unlabelled bHLH competitor probe with the same sequence, while the bHLH cis-elements mutants did not (Fig. 6c,d). These results suggest that MdbHLH3 binds directly to the promoters of MdPIN1 and MdGH3-2.
Transient luciferase (Luc) imaging assays were used to verify the activation of MdPIN1 and MdGH3-2 expression by MdbHLH3. Two recombinant plasmids (MdPIN1pro::Luc and MdGH3-2pro::Luc) containing the promoter of MdPIN1 and MdGH3-2 fused to the reporter gene luciferase, respectively, were combined with 35Spro::MdbHLH3 and coinfiltrated into tobacco leaves. Strong luminescence signal was detected in the coexpression regions of 35Spro + MdPIN1pro::Luc, but much weaker and no luminescence signals were seen in the 35Spro::MdbHLH3 + MdPIN1pro::Luc and negative controls, respectively (Fig. 6e,f). However, coexpression of 35Spro::MdbHLH3 + MdGH3-2pro::Luc exhibited noticeably greater luminescence signals than the 35Spro + MdGH3-2pro::Luc and negative controls (Fig. 6e,f). These results indicate that MdbHLH3 negatively regulates the MdPIN1 expression, but positively regulates the MdGH3-2 expression.
In addition, qRT-PCR assays were performed to assess the expression of MdPIN1 and MdGH3-2 in WT and three MdbHLH3 transgenic apple leaves (MdbHLH3-36, MdbHLH3-27, MdbHLH3-44). The results showed that all three transgenic lines had significantly lower transcript abundances of MdPIN1, but significantly higher expression levels of MdGH3-2 than the WT control (Fig. 6g). Combined with the Luc imaging assays, this result demonstrates that MdbHLH3 suppresses and activates the transcription of MdPIN1 and MdGH3-2, respectively. In line with these gene expression profiles, in situ hybridization also showed that the expression of MdPIN1 in SAMs was decreased, whereas the expression of MdGH3-2 in SAMs was increased, in these three MdbHLH3 transgenic lines in relation to the WT control (Fig. 6h).
As flavonoids act as negative regulators of auxin transport in plants (Brown et al., 2001; Peer et al., 2004; Besseau et al., 2007), we performed a series of physiological and biochemical assays to determine if MdbHLH3 also regulates some of the key genes involved in flavonol synthesis. All three MdbHLH3 transgenic lines (MdbHLH3-36, MdbHLH3-27, MdbHLH3-44) had higher flavonoid concentrations and higher transcript abundances of flavonol synthase gene MdFLS (Fig. S5). ChIP-PCR, EMSA, and GUS activity assays showed that MdbHLH3 binds directly to the promoter of MdFLS. These findings demonstrate that MdbHLH3 also controls flavonoid biosynthesis to modulate auxin signalling by directly targeting flavonol synthase gene MdFLS in apple.
The mechanism of MdbHLH3-mediated leaf shape also operates in different apple genotypes
To evaluate whether the mechanism of MdbHLH3-mediated leaf shape also operates in different apple varieties, we characterized leaf shape in 22 apple genotypes at the same altitude. PCA analysis showed distinct clusters among some genotypes as well as some overlapping clusters among other genotypes. The shape variance, represented by PC1 and PC2, is mainly influenced by leaf width and marginally influenced by leaf length (Fig. 7a,b). The ratio of leaf length to width varied among these genotypes, with ‘Gala’ apple having the largest ratio and ‘Fuji × QinGuan-911’ having the smallest ratio (Fig. 7c). Interestingly, the total free IAA concentrations and relative MdbHLH3 expression vary with the ratio of leaf length to width in these different apple varieties (Fig. 7d,e). Across all the genotypes, the ratio of leaf length to width was positively correlated with the relative expression level of MdbHLH3, but negatively correlated with total free IAA content (Fig. 7f). These relationships are consistent with those found on ‘Red Delicious’ apple trees grown at different altitudes, suggesting that the mechanism of MdbHLH3-mediated leaf shape also operates in different apple genotypes.

Discussion
Leaf shape is considered as a major determinant of plant architecture, and strongly affects plant performance (Tsukaya, 2004, 2006). Although genetic control of leaf development and leaf shape in some model plants has been well understood (Tsukaya, 2004, 2006), and the link between auxin and leaf width has been reported (Hendelman et al., 2012; Ben-Gera et al., 2016), the mechanism by which the multiple environmental factors associated with altitude affect leaf shape remains elusive (Gurevitch, 1988; Hovenden & Vander Schoor, 2006). Here, we show that a bHLH TF, MdbHLH3, acts as a key regulator in controlling apple leaf shape via auxin signalling in response to altitude.
MdbHLH3 modulates leaf shape via auxin signalling in response to altitude
Genome-wide transcriptome analysis of apple leaves collected at different altitudes allowed us to identify MdbHLH3 as a potential regulator for controlling leaf width and auxin signalling as a possible pathway for MdbHLH3’s action in response to altitude (Fig. 2). Interestingly, overexpression of MdbHLH3 in apple significantly decreased leaf width, confirming its role in controlling leaf shape in this study (Fig. 3). The lower free auxin concentration detected in the MdbHLH3-overexpressing lines was closely tied to a lower transcript abundance of MdPIN1 encoding an auxin efflux carrier and a higher transcript abundance of MdGH3-2 encoding a putative auxin amido conjugate synthase, and promoter binding assays demonstrated transcriptional suppression and activation of MdPIN1 and MdGH3-2, respectively, by MdbHLH3 (Figs 5, 6). In other words, MdbHLH3 is a transcription activator of MdGH3-2 but a repressor of MdPIN1 (Fig. 6). The decreased leaf width is probably the cause of the decrease in cell size, but not the cell number, in those MdbHLH3 transgenic lines (Fig. S6). In addition, the expression of MdbHLH3 increased along with decrease of auxin content linearly as altitude increased, with those at 2000 m having the highest expression level (Figs 2f, S7). Partial or complete restoration of leaf width of the transgenic leaves by exogenous auxin feeding confirmed that MdbHLH3 regulates apple leaf width via auxin signalling (Fig. 5e–g). As flavonoids can act as negative regulators of auxin transport in plants (Brown et al., 2001; Peer et al., 2004; Besseau et al., 2007), transcriptional activation of flavonol synthase gene MdFLS by MdbHLH3 may also contribute to auxin signalling (Fig. S5). Notably, bHLHs are known to physically interact with ARF proteins (Varaud et al., 2011; Oh et al., 2014). These studies imply that bHLHs regulate auxin signal through both transcriptional regulation and protein interactions.
Our finding that MdbHLH3 modulates apple leaf shape via auxin signalling is consistent with the crucial role of auxin in leaf development. Leaf initiation from the SAM involves a balance between primordia formation and cell proliferation. Several genes involved in either process, such as CLAVATA1/3 (Clark et al., 1997; Fletcher et al., 1999), WUSCHEL (Schoof et al., 2000), KNOTTED1 (Smith et al., 1995), and PHANTASTICA (Kim et al., 2003), have been found to play key roles in affecting leaf shape and size. Overexpression of a novel small peptide ROTUNDIFOLIA4 decreases cell proliferation and alters leaf shape in Arabidopsis (Narita et al., 2004). In rice, a mutant gene termed narrow leaf 7 (nal7) causes a significant decrease in the width of the leaf blade (Fujino et al., 2008). In addition, TCP14 and TCP15, the plant-specific bHLH-containing TCP TFs, affected internode length and leaf shape in Arabidopsis (Kieffer et al., 2011). Most of these genes act by affecting the phytohormone auxin which acted as a crucial signal in morphological aspects of leaf development.
Although MdbHLH3 expression level increases with altitude, it is most likely a response to lower temperature and higher ultraviolet light intensity associated with higher latitude rather than the higher altitude itself. This was supported by recent studies showing that flowering times shifted at different rates across elevations probably because of elevation-specific changes in temperature (Rafferty et al., 2020). Indeed, the expression of MdbHLH3 has been found to increase as temperature decreases and ultraviolet intensity increases (Xie et al., 2012). This suggests that the MdbHLH3 expression level integrates various environmental signals associated with altitude to regulate leaf development. This is consistent with the sessile nature of plants as they have to evolve a series of signal transduction pathways to cope with adverse environmental conditions.
The finding that the mechanism of MdbHLH3-mediated leaf development also operates in different apple genotypes suggests that bHLH3 may play a role in the evolution of apple leaf shape. The earliest terrestrial plants have no leaves. During the Devonian period about 300 Ma, plants use rod-like branches for photosynthesis. A flat and wide blade obviously has a larger surface area than a rod, and thus can be more efficient in photosynthesis. As a result, angiosperm leaves exhibit a great diversity of shapes that range from developmental events within apical organ response to microenvironment to variation among species within and between communities and among orders or families at present (Tsukaya, 2004) to expand the environmental adaptability of plants from an evolutionary point of view. In addition, as photosynthetic leaves are critical to plant growth and survival, the variation in their shapes may reflect natural selection on functions (Tsukaya, 2004; Nicotra et al., 2011). A closer look at the relationship between MdbHLH3 and its homologs in plant species residing at various positions in the evolutionary tree will help to infer its role in the evolution of leaf shapes.
Jack of all trades: MdbHLH3 is a multifunctional regulator in apple
In addition to being a major regulator for leaf shape in this work, MdbHLH3 has been found to play a key role in regulating many other processes in apple. MdbHLH3 was first found by Xie et al. (2012) to interact with MYB TF MdMYB1 by directly activating its downstream anthocyanin biosynthesis genes, MdDFR and MdUFGT, for promoting anthocyanin accumulation and fruit coloration in response to low temperature. Subsequently, the glucose sensor MdHXK1 was found to directly interact with and phosphorylate MdbHLH3, which stabilizes MdbHLH3 protein and enhances its transcription of the anthocyanin biosynthesis genes for anthocyanin biosynthesis (Hu et al., 2016a). MdbHLH3 is also a crucial component of the MYB-bHLH-WDR (MBW) complex that controls not only anthocyanin synthesis but also the activities of V-ATPase and V-PPase and transport of anthocyanins into the vacuole (Hu et al., 2016b). Recently, the glucose-inhibited ubiquitin E3 ligase MdPUB29 was found to interact with and ubiquitinate MdbHLH3 in regulating ethylene biosynthesis and fruit ripening (D. G. Hu et al., 2019). The bHLH TF MdbHLH3 also modulates leaf senescence in apple via the regulation of dehydratase-enolase-phosphatase complex 1 (Hu et al., 2020). Thus, MdbHLH3 TF is a multifunctional regulator for diverse processes including anthocyanin synthesis, vacuolar acidification, ethylene biosynthesis, leaf senescence and auxin signalling by integrating environmental factors and developmental cues such as low temperature and sugar concentrations, as well as hormonal signalling. Considering that leaf shape links to photosynthesis and mineral nutrition as well as temperature optima in various plant species (Njoku, 1957; Nicotra et al., 2008) and variations in sugars and hormone concentrations (ABA, gibberellins, etc.) are concomitant with leaf shape changes in higher plants (Gray, 1957; Le Hir et al., 2006). Our work clearly demonstrates that the master regulator MdbHLH3 modulates leaf shape by altering auxin signalling in apple (Fig. 8).
The decrease in apple leaf width with increasing altitude or morphological differences of apple cultivars is controlled by the bHLH TF MdbHLH3. MdbHLH3 transcriptionally suppresses and activates MdPIN1 and MdGH3-2, respectively, by specifically binding to their promoters. This alters auxin homeostasis and transport, consequently leading to changes in leaf shape. In addition, as flavonoids can act as negative regulators of auxin transport in plants (Brown et al., 2001; Peer et al., 2004; Besseau et al., 2007), transcriptional activation of the flavonol synthase gene MdFLS by MdbHLH3 may also contribute to auxin signalling and change the leaf shape. Exogenous application of auxin obviously influences the leaf shape (Fig. 5e). After uncovering the mysteries of leaf shape control, we can change the leaf shape of plants through molecular design to expand the environmental adaptability of plants. Designing leaf shape has enormous importance in improving the environment and increasing crop yield. Therefore, these findings may be useful for developing biotechnological strategies as well as for informing traditional breeding programmes for generating new cultivars to expand the environmental adaptability with improved yield and quality of crops.

Acknowledgements
We would like to thank Dr Takaya Moriguchi of National Institute of Fruit Tree Science in Japan for ‘Orin’ apple calli, Dr Nan Ma at China Agricultural University for critical reading of the manuscript. This work was supported by grants from the National Key Research and Development Program (2018YFD1000200), National Natural Science Foundation of China (31972375, 31770230), Ministry of Agriculture (CRAS-27), Transgenic Research Program (2016ZX08010-001) and Nanjing Agricultural University (ZW201805).
Author contributions
Y-JH and D-GH planned and designed the research. D-GH, NW, D-HW, Y-XW, Y-WZ, J-YD, K-DG and XX performed experiments, conducted fieldwork, analysed data etc. D-GH, LC, NW and Y-JH wrote the manuscript. D-GH, NW and D-HW contributed equally to this work.