Whole exome sequencing identifies an AMBN missense mutation causing severe autosomal-dominant amelogenesis imperfecta and dentin disorders

Tooth development is a complex process that involves precise and time-dependent orchestration of multiple genetic, molecular, and cellular interactions. Ameloblastin (AMBN, also named “amelin” or “sheathlin”) is the second most abundant enamel matrix protein known to have a key role in amelogenesis. Amelogenesis imperfecta (AI [MIM: 104500]) refers to a genetically and phenotypically heterogeneous group of conditions characterized by inherited developmental enamel defects. The hereditary dentin disorders comprise a variety of autosomal-dominant genetic symptoms characterized by abnormal dentin structure affecting either the primary or both the primary and secondary teeth. The vital role of Ambn in amelogenesis has been confirmed experimentally using mouse models. Only two cases have been reported of mutations of AMBN associated with non-syndromic human AI. However, no AMBN missense mutations have been reported to be associated with both human AI and dentin disorders. We recruited one kindred with autosomal-dominant amelogenesis imperfecta (ADAI) and dentinogenesis imperfecta/dysplasia characterized by generalized severe enamel and dentin defects. Whole exome sequencing of the proband identified a novel heterozygous C-T point mutation at nucleotide position 1069 of the AMBN gene, causing a Pro to Ser mutation at the conserved amino acid position 357 of the protein. Exfoliated third molar teeth from the affected family members were found to have enamel and dentin of lower mineral density than control teeth, with thinner and easily fractured enamel, short and thick roots, and pulp obliteration. This study demonstrates, for the first time, that an AMBN missense mutation causes non-syndromic human AI and dentin disorders.

Dentinogenesis is a highly ordered process in which the organic predentin matrix is progressively mineralized by ectomesenchymally derived cells called odontoblasts. 35 The odontoblasts differentiate at the bell stage of tooth development, forming a single layer of cells lining the pulp cavity, where they secrete the organic predentin matrix into the underlying space. 36 AMBN expression is also detected in pulpal mesenchymal cells, including preodontoblasts and young odontoblasts. 37 AMBN is known to be involved in the mesenchymal-ectodermal interaction that precedes dentin and enamel secretion, and the sequential expression pattern of AMBN also acts as a signalling molecule. 38 Currently, hereditary dentin disorders are divided into dentinogenesis imperfecta (DGI) and dentin dysplasia (DD). 39 Recent genetic studies have shown that mutations of collagen type I alpha 1 chain (COL1A1), collagen type I alpha 2 chain (COL1A2), DSPP, secreted modular calcium-binding protein 2 (SMOC2), vacuolar protein sorting 4 homologue B (VPS4B), and ssu-2 homologue (SSUH2) cause dentin disorders. [39][40][41][42][43][44][45][46][47][48][49] AMBN can enhance reparative dentin formation, which confirms that this protein has an important role in odontoblast differentiation and dentinogenesis. 50 The enamel defects confirm the observations from mouse models and show that human AI is associated with disrupted Ambn function. However, to date, mutations of AMBN have not been associated with the phenotypes of hereditary dentin disorders. In this study, we report, for the first time, that AMBN mutations cause both non-syndromic human AI and hereditary dentin disorders.

Tooth phenotype
We identified a Chinese family with generalized enamel and dentin defects involving the primary and secondary dentitions. On clinical examination of the teeth (Fig. 1), 11 individuals (II:1, II:2,  III:1, III:3, III:6, III:7, III:8, IV:2, IV:3, IV:7, and IV:9) were affected, exhibiting hypomineralized AI and hereditary dentin disorders, but were in good general health and exhibited no abnormalities upon extraoral examination. Unaffected individuals (III:2, III:5, III:9, III:10, IV:1, IV:4, IV:5, IV:6, IV:8, and IV:10) did not carry the mutation and exhibited normal dentition. The crowns of the affected family members appeared grey or brownish-blue (Fig. 2c, e-h, i). The enamel was thinner and often dislodged because of alterations to the enamel-dentin junction. Consequently, the exposed hypomineralized dentin was quickly worn away by attrition. The radiographic aspect showed bulbous crowns because of an important cervical constriction, short and thick roots, and pulp obliteration with periapical radiolucencies in non-carious teeth (Fig. 2d, j, l). With development of the disease, the absorption of periapical bone gradually increased, which affected the tooth-supporting tissues, leading to loosening of the teeth and eventually early tooth loss. Two teeth of the secondary dentition of individual family member IV:9 were in a high grade of mobility due to circumferential alveolar bone resorption, although this individual was only 21 years old (Fig. 2d). There were six secondary teeth missing in individual IV:2 and four loosened secondary teeth with large-area periapical bone defects in non-carious condition (Fig. 2j), although this individual was 22 years old. All the other family members who were heterozygous for the mutation had similar experiences: detached enamel, circumferential alveolar bone resorption in non-carious teeth, tooth loosening, and eventual tooth loss. By the age of approximately 40, they lost almost all of their teeth or had only a few loose roots remaining ( Fig. 2k-n). To maintain the physical occlusion position and improve masticatory function, the patients received a long-span fixed partial denture repair ( Fig. 2k-n), removable partial denture ( Fig. 2k-i), implant prosthodontics ( Fig. 2k-i), or removable complete denture (Fig. 2o, p) according to their family economic situation.
Enamel thickness also appeared to be reduced. The pulp chambers were narrower with pulp chamber calcification (Fig. 3b). The roots were shorter, with small or obliterated root canals (Fig. 3b).
When examined with scanning electron microscopy (SEM), the normal control teeth exhibited the prismatic structure characteristic of normal human enamel and dentin (Fig. 3e, g, i). SEM revealed that the mutation had little effect on the prismatic structure, but the numbers of enamel prisms of affected teeth were obviously reduced (Fig. 3f); the enamel also displayed distinct regions where the enamel structure was particularly disturbed (Fig. 3h). The outline of the dentin-enamel junction (DEJ) was changed: the width of the DEJ was increased and the structure of sigmoidal transitions within the DEJ zone, which firmly combine enamel and dentin together, became a straight line in appearance (Fig. 3h). Examination of the ultrastructure of dentin of affected teeth showed that the distribution of dentin tubules was not even and that the dentin tubular diameter was not consistent (Fig. 3j). This verified that there was an increase in the thickness of major peritubular dentin in mutant teeth, which reduced the size of or even obliterated the dentin tubules compared to control teeth (Fig. 3j).

Genetic analysis
To identify the pathogenic variant responsible for the phenotype, we first screened for mutations in the coding and splice regions of the DSPP gene (exons 1-5). However, no mutations were identified. To confirm this result, long-range PCR was conducted with primers designed to amplify all exons, including the highly repetitive sequence in exon 5 of DSPP. Sequencing of the resulting product revealed that there were no mutations present within the DSPP exons or the flanking splice regions. In addition, we also analysed the copy number variations (CNVs) in the DSPP gene via array comparative genomic hybridization. Data analysis revealed no pathogenic CNVs in the DSPP gene in affected individuals in this family. Next, we performed exome sequencing with genomic DNA for affected individuals IV:2 and IV:9, and control IV:1 in collaboration with Genesky-Shanghai (China). Sequence analysis of all exons showed that 47 potentially pathogenic variants were shared by the two affected individuals and not by the unaffected individual. None of these variants were within genes known to be involved in AI, except for AMBN, or in DGI as identified by function, expression or animal model studies, including AMELX, ENAM, KLK4, MMP20, DSPP, SMOC2, VPS4B, and SSUH2. However, a novel heterozygous missense mutation (c.1069C > T) was identified in exon 13 of AMBN (NM: 016519, chr4:71472172), causing a proline to serine mutation at the conserved amino acid position 357 of the protein (Fig. 4a). The pathogenic variant of AMBN was validated by Sanger sequencing followed by co-segregation analysis in all affected and unaffected family members (Fig. 4b).
The candidate region was confirmed by additional genotype data from microsatellite markers on 4q, and linkage analysis gave a maximum logarithm of odds score of 4.515 between markers D4S1541 and D4S1558 (Figs. 1 and 4a). Further, this nucleotide change was not detected in DNA samples from 185 normal unrelated controls and 15 dominant AI patients matched for Chinese ethnicity.
Functional analysis of mutant AMBN The c.1069C > T mutation in AMBN changes the three-dimensional (3D) structure of the AMBN protein predicted by I-TASSER, leading to a loss-of-function protein estimated by Polyphen-2 ( Fig. 5b). To evaluate the effect of the mutation, the cellular localization of AMBN was analysed using wild type (WT) and mutant constructs in human embryonic kidney (HEK) 293 cells. The result showed that the mutant AMBN was localized in the cytoplasm, which was the same as WT AMBN (Fig. 5c). However, the mutant protein exhibited aggregation in transfected HEK293 cells (Fig. 5c). Moreover, we made two WT and mutant constructs that were transiently expressed in human gingival fibroblasts (HGFs) and HEK293 cells. There was no obvious difference in mRNA expression between cells transfected with the mutated transgenic vector (ΔcAMBN -GFP) or normal transgenic vector (cAMBN -GFP) in either HGF (P > 0.05) or HEK293 cells (P > 0.05) (Fig. 5d). However, we found that expression of the aberrant protein was significantly increased compared with the WT AMBN protein in both HGF (P < 0.001) and HEK293 cells (P < 0.01) (Fig. 5e, f).

DISCUSSION
We have identified a large Chinese pedigree with segregating severe hypoplastic AI and dentin disorders in an autosomaldominant manner. To predict the affected codon, we used whole exome sequencing and identified a rare variant at exon 13 of the AMBN gene on chromosome 4q shared by all affected individuals. Further investigation confirmed that this was a heterozygous missense mutation at position 1069 (c.1069C > T) of the coding region. Human AMBN is composed of 447 residues, many of which are unchanged or only substituted with a similar residue during more than 200 Ma of mammalian evolution. 56 It is worth noting that the average percentage of proline is 14.09% in the full-length AMBN sequence and 23.78% for the proline-rich region, compared with the 5.1% encountered in most proteins. 56 Estimation of the codon conservation index along the human AMBN sequence indicated that position P375 of exon 13 is highly conserved. 56 The evolutionary analyses performed on various secretory calciumbinding phosphoproteins, including AMELX, ENAM, matrix extracellular phosphoglycoprotein, AMTN, and dentin matrix acidic phosphoprotein 1 proteins, which are involved in tooth development, have shown that the patterns of long-term evolutionary conservation have a decisive influence on validating the potential pathogenicity of variants identified in human genetic diseases. 57 The conserved position sites are very important and such variants usually result in a genetic disorder. 56 In the present report, our results show that mutation of conserved position P375 of AMBN also causes severe dental defects.
In our study, we report an AMBN missense mutation causing severe autosomal-dominant AI. AMBN is an enamel matrix protein (EMP) that could determine the supramolecular enamel matrix assembly by the proline-rich region and is located on chromosome 4 in a region containing the locus for the autosomal hypoplastic form of AI (AIH2). 58 EMPs consist principally of AMELX, AMBN, and ENAM. 59 EMPs have a role in maintaining the prismatic structure of growing enamel crystals at the rod and inter-rod boundaries. 60,61 Heterozygous Ambn +5,6/−5,6 mice had significantly less well-mineralized prismatic enamel of near-normal thickness compared with WT. 62 Upon SEM examination, affected teeth exhibited a phenotype in which an ordered thinner prismatic structured layer of enamel was formed with fewer enamel rods and wider inter-rod distances (Fig. 3f). This patient phenotype is similar to that of heterozygous Ambn +5,6 /−5,6 mice. In addition, the mineral density of enamel was significantly decreased in affected individuals via deletion of AMBN exon 6. The mean mineral density of enamel is approximately 2.47 g/cm 3 . In our study, the mean mineral density of enamel was 2.45 g . cm -3 in individuals IV:2 and IV:9, which is consistent with the enamel density values obtained from affected individuals with genomic deletion of AMBN exon 6.
We report the association of a genetic change in AMBN with dentin defects in the family identified in this study. In our case, some of the dentin features characteristic of DGI-II and DD-I were observed in the affected individuals (Fig. 2). They appeared similar to those associated with DGI-II, such as bulbous crowns with cervical constriction, mild discoloration, and pulp obliteration. Radiologically, permanent teeth of DD-I have short roots with a crescent-shaped pulpal remnant parallel to the cemento-enamel junction in the permanent dentition and total pulpal obliteration in the dentition. 48,49 There are usually numerous periapical radiolucencies in non-carious teeth. 48,49 Compared with DD-I, the clinical crowns of affected teeth in our study had similar pulp and root phenotypes, especially the phenotype of periapical radiolucencies in non-carious teeth (Fig. 2). The clinical phenotypes of teeth from patients in our research ranged from mild to severe with age. Because pulp chamber and canal obliteration are often progressive, radiographic follow-up at various ages is recommended to determine diagnosis. 63 It has also been reported that ACP4 mutations caused recessive hypoplastic AI, in which the enamel layer in the affected teeth was thinner and dentin was mildly hypermineralized. 28 The dentin mineralization in the affected individual was different from most dentin defect cases. 28 However, in our study, we found that the dentin of teeth of affected individuals was hypomineralized with lower mineral density, which was consistent with most dentin defect cases.
AMBN has been reported to play crucial roles in bone development and remodelling/repair. 64 However, the roles of AMBN in bone are unconfirmed. There were no obvious bone abnormalities identified by detailed clinical evaluations in the family with genomic deletions in AMBN exon 6. However, mouse models with loss of Ambn function did not have obvious bone defects. 32,65 In our study, we also did not observe any obvious bone abnormalities in the family, which is consistent with the aforementioned studies.
AMBN is believed to participate in ameloblast attachment to the underlying enamel matrix and in modulation of enamel crystal growth. 66 However, no mutations of AMBN have previously been associated with dentin defects. In this report, we found that the genetic change of AMBN leads to dentin disorders. Teeth are ectomesenchymal organs that develop via a series of signalling interactions between cells of the oral epithelium and ectomesenchyme. 67 It has been reported that the AMBN gene contains two runt-related transcription factor 2 (RUNX2) binding sites, and AMBN transcription is regulated by the interaction of the transcription factor RUNX2 and the AMBN promoter. 68,69 On the other hand, DSPP is one of the major pathogenic genes involved in dentin disorders and is also transiently expressed in Fig. 2 Dental phenotype and panoramic radiograph of the patients. a Normal permanent teeth; b panoramic radiograph of normal teeth. Dental phenotype of IV:9 c,e-h. An intraoral view of a male patient at the age of 21 shows brown crowns with thin and chipping enamel, especially in the cervical part of all teeth (black arrows), and short crowns of the permanent maxillary central incisors worn out by attrition (black arrow). d,j,l Panoramic radiographs of patients show bulbous crowns covered with thin enamel, cervical constriction (asterisk), short constricted roots, and apical radiolucencies in the second molars (white arrows). i An intraoral view of a patient at the age of 22 shows grey crowns with enamel chipped and absent in places. k,m Above the age of approximately 40, patients had lost almost all of their teeth or had several loose roots. l,n Patients were treated by long-span fixed partial denture repair, l implant prosthodontics, k removable partial denture, or o removable complete denture according to their family economic situation. Finally, o,p all patients lost all their teeth at an early age (generally by their 50s).
Whole exome sequencing identifies an AMBN missense mutation... T Lu et al. ameloblasts. 70 DSPP analysis revealed three binding sites for RUNX2. RUNX2 has been shown to be involved in multiple signalling pathways and to control the differentiation of dental pulp cells to odontoblasts during dentinogenesis. 71,72 RUNX2 expression in dental pulp cells and odontoblasts was altered in DSPP mutants. 73 Overall, AMBN, RUNX2 and DSPP were expressed in ameloblasts and odontoblasts during amelogenesis and dentinogenesis. 38,68,69 These genes have certain inherent connections that may interact with and restrict each other. Additional mutations in DSPP are always associated with dentin defects and have been reported in two families with hypoplastic enamel. 30 It has been observed that Ambn expression precedes traumainduced reparative dentin formation in rat molars. 74 The exposed pulp cavities were effectively closed by thick dentin bridges in recombinant Ambn-treated teeth compared with calcium hydroxide-treated teeth, which were covered only by a thin dentin bridge. This result indicated that AMBN plays an important role in the formation of reparative dentin during pulpal healing and links it firmly to odontoblast differentiation and function. 50 Therefore, we deduced that AMBN is involved in odontoblast differentiation and physiology via cell signalling. The interaction between AMBN, RUNX2, and DSPP may underlie the pathogenic mechanism of AMBN mutations associated with dentin defects. ADAI typically affects one or more individuals in each generation of a family. 75 The clinical manifestations in affected individuals may be consistent or show variable expression, resulting in subtle to substantial differences between all affected individuals in the same family. 76 The phenotype in ADAI may be predominantly or exclusively hypoplastic, manifesting as thin enamel and spacing between the teeth, crowns reduced in size ab initio with thinner enamel, and crowns that are yellowish-brown in hue. 76 The most abundant enamel protein is amelogenin, which is expressed primarily from a gene on the X-chromosome (AMELX), mutations of which have been shown to cause X-linked AI. 77 Ambn −/− mutant mouse models develop autosomaldominant AI in a dose-dependent manner, which is consistent with the traits of our study of the Chinese kindred, whereas a homozygous 2347 bp genomic deletion and a novel homozygous splice-site mutation (c.532-1G > C) in AMBN have been shown to cause autosomal recessive hypoplastic human AI. 8,34 The enamelin gene, ENAM, is implicated in the pathogenesis of the dominant forms of AI. 78 However, autosomal recessive inheritance has also been recorded for ENAM mutations in homozygotes, where heterozygotes presented with a severe and a milder, local form. 18 The hereditary dentin disorders were documented as autosomal-dominant genetic diseases characterized by abnormal Fig. 3 Tooth ultrastructural analyses. a-d High-resolution X-ray CT analysis of exfoliated teeth from control individual IV:8 and affected individuals IV:2 and IV:9. a 3D reconstruction of the tooth CT data: individual IV:8 and individuals IV:2 and IV:9. b 3D reconstruction of pulp chambers. Teeth of IV:2 and IV:9 exhibit thistle-shaped pulp chambers. c Typical CT sections through the teeth are presented using false colour calibrated with respect to mineral density to generate mineral density maps. Scale bar is marked in g/cm 3 . d Mean enamel and dentin mineral density for each tooth is also shown graphically. The control tooth of IV:8 exhibits enamel and dentin apparently normal in structure and density. Affected teeth of IV:2 and IV:9 exhibit enamel and dentin significantly reduced in mineral density compared with the normal tooth (P < 0.001). *P < 0.05, **P < 0.01, or ***P < 0.001. Videos of 3D-rendered CT data showing surface detail and the internal structure of all teeth are available as Supplementary Material. e-j SEM of representative exfoliated teeth. e Tooth of individual IV:8 exhibits normal enamel architecture comprising prisms (rods) of individual enamel crystallites. f Enamel of patient IV:9 tooth is in the normal range in terms of the order but characterized by fewer enamel prisms (rods) and wider inter-rod distances, h with occasional areas exhibiting a disturbed structure (asterisk). h The width of the DEJ of patient IV:9 is increased and the shapes of transitions within the DEJ zone are straight instead of sigmoidal. i SEM of the control tooth of individual IV:8 shows normal dentin structure. j The distribution of dentin tubules was not even and the size of the dentin tubular diameter was not consistent with the tooth of patient IV:9. j Most peritubular dentin is thicker and dentin tubes are smaller or obliterated completely compared to the control tooth.
Whole exome sequencing identifies an AMBN missense mutation... T Lu et al. dentin structure affecting either the primary or secondary dentitions. 48,49 In conclusion, the identification of a heterozygous genomic missense mutation in exon 13 (c.1069C > T) of AMBN in a Chinese family confirms, for the first time, that AMBN mutations cause autosomal-dominant hypoplastic human AI and dentin defects. This is the third report of an AMBN mutation being associated with AI. There is no obvious co-segregating bone or other clinical features associated with this mutation in the affected or unaffected members. In fact, the primary and secondary dentitions are affected by the mutation. The youngest affected individual was entering adulthood when assessed and any effect that the AMBN mutation may have on the primary teeth is thus unknown. The mechanisms causing AI and dentin defects in this case are unclear but might be reflected by the nature of the mutation identified, i.e., that this mutation is a missense with a potential gain of function, whereas the mutations previously identified are potentially loss of function. However, further studies reporting on other human patients and AMBN variants will be required to confirm this.

MATERIALS AND METHODS
Patients and clinical examination A Han family with hypoplastic AI was identified when the proband, whose teeth had yellowish-brown enamel that broke off in large pieces, came to Nanfang Hospital (Guangdong province, China) for tooth root canal treatment. All the patients in the family received X-rays as part of a clinical examination. All procedures in this study were approved by the institutional review board and ethics committee of Nanfang Hospital, an affiliate of Southern Medical University.
Tooth ultrastructure analysis With their informed consent, two wisdom teeth extracted from two patients (IV:2 and IV:9) and a normal control wisdom tooth obtained from the unaffected family member IV:8 were subjected to high-resolution CT using a Skyscan 1176 (Bruker-microCT, Kontich, Belgium) with the following settings: 90 kV, 278 µA, an isotropic resolution of 17.4 µm, and projection rotation steps of 0.50 to 360°. 3D models of the teeth were reconstructed with NRecon v.1.6.3 software (Bruker-microCT). The CT images were calibrated using hydroxyapatite mineral of known densities [0.25 g·cm -3 and 0.75 g·cm -3 (Bruker)]. Measurement of the mean mineral density of the enamel of each tooth was carried out using ImageJ software (http://imagej.nih.gov/ij/). The mean mineral densities of the enamel and the dentin were determined by measuring the mineral density of enamel or dentin on every tenth CT slice taken through each tooth. Calibrated colour contour maps of mineral density were also generated using ImageJ.
For SEM, all extracted teeth were cut into slices using a lowspeed hard tissue cutting machine (Buehler, Lake Bluff, IL, USA). The tooth slices were polished using carborundum papers (3 M, Maplewood, MN, USA), followed by a nail buffer. Sections were etched by immersion in 30% phosphoric acid for 30 s, rinsed using excess distilled water for 5 min and dehydrated using ascending concentrations of ethyl alcohol overnight under vacuum. Specimens were fixed and sprayed with gold using an auto sputter coater (Agar Scientific, Elektron Technology, Stansted, UK). The microstructure of the samples was observed with a Hitachi S-3400N scanning electron microscope (Hitachi, Tokyo, Japan) using a 123 eV Nano XFlash R Detector 5010 (Bruker) and an accelerating voltage of 20 kV.
Whole exome sequencing Three micrograms of genomic DNA isolated from each of three individuals (IV:1, IV:2, and IV:9) were processed according to the TruSeq DNA Sample Prep Kit protocol (Illumina, San Diego, CA, USA). Agilent SureSelect XT Human All Exon V5 (Agilent Technologies, Santa Clara, CA, USA) was used as the capture reagent. Subsequently, sequencing was performed using a 300 bp paired-end protocol on a Genome Analyzer IIx (Illumina). The read alignment and variant calling were performed by the BaseSpace BWA Enrichment v1.0 App. The sequence was aligned with BWA Genome Alignment Software, and variant calling was performed with GATK using the human reference sequence hg19/GRCh37. Available genomic databases (dbSNP, 1000 Genomes Project, Exome Variant Server, Exome Aggregation Consortium and a local Paris Descartes Bioinformatics platform database) were used to filter exome variants and exclude variants with a frequency > 1%. We performed a number of steps to filter out nucleotide false variants: first, we filed the intronic variants and synonymous single-nucleotide variants; next, we removed the variants  Protein expression HEK293 cells and HGF cells were cultured in six-well dishes at 37°C in a 5% CO 2 atmosphere in a sterile incubator (Thermo Fisher Scientific). DMEM (Invitrogen) supplemented with 10% FBS (GIBCO, Thermo Fisher Scientific) was used for cell culture. To investigate differences in AMBN expression, cells were transfected with 2 μg of WT or mutant AMBN recombinant plasmid (ΔcAMBN-FLAG) or empty vector (negative control). After a further 24 h of culture, cells were collected and homogenized in radio-immunoprecipitation assay lysis buffer (Merck-Millipore, Darmstadt, Germany) with the protease inhibitor phenyl methyl sulfonyl fluoride (Beyotime Institute of Biotechnology, Jiangsu, China). Total protein was extracted and assessed using a standard bicinchoninic acid assay (Beyotime). Loading buffer was added before boiling for 5 min. For each sample, 20 μg of protein was loaded onto a 10% SDSpolyacrylamide gel electrophoresis gel (Bio-Rad, Hercules, CA, USA), transferred onto a polyvinylidene difluoride membrane (Millipore), and blocked with 5% skimmed milk (Thermo Fisher Scientific) and 0.1% Tris-buffered saline-Tween 20 (TBST) at room temperature for 1.5 h. The membranes were incubated overnight at 4°C with mouse anti-FLAG (F1804, Sigma, USA). The membranes were then washed with TBST and incubated with goat anti-mouse IgG (F9006, Sigma) at room temperature for 2 h. SuperSignal West Pico ECL (Thermo Fisher Scientific) was freshly prepared just before detection of proteins. Protein expression was detected by exposure to an automatic chemiluminescence image analysis system (Tanon, Shanghai, China) and evaluated using SuperSignal™ West Femto Maximum Sensitivity Substrate (Thermo Fisher Scientific).

Statistical analyses
The significance of differences between two groups was determined using Student's t-test followed by Bernoulli correction and adjusted P-values were used to identify significant differences.