ALKBH3-dependent m1A demethylation of Aurora A mRNA inhibits ciliogenesis

Primary cilia are antenna-like subcellular structures to act as signaling platforms to regulate many cellular processes and embryonic development. m1A RNA modification plays key roles in RNA metabolism and gene expression; however, the physiological function of m1A modification remains largely unknown. Here we find that the m1A demethylase ALKBH3 significantly inhibits ciliogenesis in mammalian cells by its demethylation activity. Mechanistically, ALKBH3 removes m1A sites on mRNA of Aurora A, a master suppressor of ciliogenesis. Depletion of ALKBH3 enhances Aurora A mRNA decay and inhibits its translation. Moreover, alkbh3 morphants exhibit ciliary defects, including curved body, pericardial edema, abnormal otoliths, and dilation in pronephric ducts in zebrafish embryos, which are significantly rescued by wild-type alkbh3, but not by its catalytically inactive mutant. The ciliary defects caused by ALKBH3 depletion in both vertebrate cells and embryos are also significantly reversed by ectopic expression of Aurora A mRNA. Together, our data indicate that ALKBH3-dependent m1A demethylation has a crucial role in the regulation of Aurora A mRNA, which is essential for ciliogenesis and cilia-associated developmental events in vertebrates.

Introduction m 1 A RNA modification is an important type of reversible methylation with the addition of a methyl group and a positive charge at the N 1 position of adenosine [1][2][3] . m 1 A RNA modification is able to block the Watson-Crick interface, and thus influences RNA secondary structures and protein-RNA interactions 1,2 . m 1 A has been identified in tRNAs, rRNAs, mRNAs, and mitochondrial RNAs 2,4-9 . The dynamic methylation on RNA is mediated by effector proteins, which are termed writers (methyltransferases), erasers (demethylases) and readers (RNA binding proteins) 10 . The writers of m 1 A modification include TRMT6 (tRNA methyltransferase 6), TRMT61A, TRMT61B, and TRMT10C, while ALKBH1 (Alpha-ketoglutarate-dependent dioxygenase AlkB homolog 1) and ALKBH3 are demethylases 1 . Emerging evidence shows that m 1 A RNA modification plays a key role in RNA metabolism and gene expression 1,2,[4][5][6][7][8][9] . However, the physiological function of m 1 A modification remains unclear.
Cilia are microtubule-based hair-like structures protruding from vertebrate cell surface, which transduce extracellular signals into cellular responses [11][12][13] . Cilia have been well demonstrated to be essential for many developmental processes 14 . Disruption of cilia structure or function causes numerous human diseases, termed ciliopathies, including polycystic kidney disease, Bardet-Biedl syndrome, and nephronophthisis [15][16][17] . Nevertheless, the role of m 1 A RNA modification in ciliogenesis remains unexplored.
Primary cilia occur in G0 or early G1 phase and disappear before cells enter mitosis 18 . Primary cilia assembly involves several successive stages, including maturation of the mother centriole into a basal body, formation of ciliary vesicle, recruitment of tau-tubulin kinase 2, removal of centriolar coiled-coil protein 110, assembly of transition zone, and elongation of microtubule axoneme [19][20][21] . Cilia disassembly is mainly regulated by the Aurora A-HDAC6 (histone deacetylase 6) pathway and the NEK2 (NIMArelated kinase 2)-KIF24 (kinesin family member 24) pathway. At early G1 phase, Aurora A is activated and recruited at basal body, and then phosphorylates and activates HDAC6, which deacetylates and destabilizes axonemal tubulins, to facilitate ciliary disassembly 22 . In S/G2 phases, NEK2 phosphorylates and activates the microtubule depolymerizing kinesin KIF24 to block re-ciliation 23 .
Here, we find that the m 1 A demethylase ALKBH3 functions as a negative regulator of ciliogenesis by removing m 1 A sites on Aurora A mRNA, a key regulator of cilia disassembly. Moreover, we also identify that Alkbh3 is essential for cilia-associated developmental processes in zebrafish by regulating Aurora A mRNA stability. Collectively, these data suggest that the ALKBH3-Aurora A axis plays a crucial role in regulating ciliogenesis and cilia-related embryonic events in vertebrates.

ALKBH3 inhibits ciliogenesis in an m 1 A-dependent manner
To investigate the role of m 1 A RNA modification in ciliogenesis, we first performed a functional screening targeting genes encoding m 1 A-regulating proteins by using small interfering RNAs (siRNAs) in human RPE-1 (retinal pigmented epithelial 1) cells in normal conditions (DMEM/F12 medium with 10% serum). Depletion of TRMT6, TRMT61A, or TRMT61B had no significant effect on ciliogenesis in RPE-1 cells ( Supplementary Fig.  S1). Silencing of ALKBH1 or TRMT10C with one of the two siRNAs showed an effect on cilia ratio. Importantly, knockdown of ALKBH3 with two different siRNAs significantly increased the percentage of ciliated cells, but not influenced the length of cilia (Fig. 1a-d). Furthermore, the enhanced ciliation in ALKBH3-depleted cells was significantly reversed by ectopic expression of ALKBH3-FLAG proteins ( Fig. 1e-g). To confirm the function of ALKBH3 in ciliation, we exogenously expressed ALKBH3-FLAG and found that forced expression of ALKBH3-FLAG significantly reduced ciliogenesis in RPE-1 cells treated with serum starvation (Fig. 1h-j).
Since the demethylation activity of ALKBH3 on mRNA is inhibited by the mutation of D193A 7,24 , we generated a catalytically inactive mutant of ALKBH3 (ALKBH3-D193A) to examine whether the regulation of ALKBH3 in ciliogenesis is depend on its demethylation activity. The LC-MS/MS quantification of m 1 A mRNA modification in RPE-1 cells confirmed that the ALKBH3-D193A construct we made was indeed catalytically inactive ( Supplementary  Fig. S2). The ciliary phenotype in ALKBH3-depleted cells was significantly restored by ectopic expression of wildtype ALKBH3-FLAG, but not by ALKBH3-D193A-FLAG mutant in RPE-1 cells (Fig. 1k-m). In addition, overexpression of ALKBH3-D193A mutant had no significant effect on ciliation in RPE-1 cells with serum deprivation (Supplementary Fig. S3). Together, these data suggest that the m 1 A RNA demethylase ALKBH3 acts as a negative regulator of ciliogenesis in mammalian cells.
ALKBH3 knockdown leads to increased m 1 A level on Aurora A mRNA To explore the molecular mechanism of ALKBH3 in ciliogenesis, we tried to identify possible mRNAs that are targeted by ALKBH3. First, we systematically analyzed the published datasets from m 1 A-ID-seq 2 and m 1 A-quantseq 25 in HEK-293T cells and acquired 38 potential m 1 A-modified transcripts. These potential m 1 A-modified transcripts were further overlapped with cilia-associated genes from CiliaCarta 26 and differentially expressed genes of RNA-seq in ALKBH3-depleted RPE-1 cells. As a result, only Aurora A, a master negative regulator of ciliation, was overlapped among these four datasets (Fig. 2a). To investigate if there exist m 1 A sites on Aurora A mRNA, a gene-specific methylated RNA immunoprecipitation-qPCR assay was performed. The data showed that the m 1 A abundance of Aurora A mRNA in ALKBH3-depleted cells was significantly increased when compared to that of control cells (Fig. 2b), indicating that Aurora A mRNAs may contain m 1 A modification.
To identify the precise m 1 A sites on Aurora A mRNA, we carefully analyzed the m 1 A-ID-seq dataset from wildtype and ALKBH3 knockout HEK-293T cells 2 , and discovered there were potential m 1 A peaks located near the translation initiation region of Aurora A mRNA (Fig. 2c). Given that RT-1306 is an evolved HIV reverse transcriptase that able to induce mutations at m 1 A sites on RNAs 25 , we employed RT-1306 to identify the m 1 A sites on Aurora A mRNA. Our sequencing data from the cDNAs of control cells and ALKBH3-depleted cells revealed that there existed three m 1 A sites (+13, +33, +40) in the coding sequence region near the translation initiation site of Aurora A mRNA (Fig. 2d). Moreover, we also performed TA cloning from the above cDNAs amplified by RT-1306 and picked single colony for sequencing analysis. The results displayed that the mutation rate of these m 1 A sites was significantly increased in ALKBH3-depleted cells (Fig. 2e). Taken together, these observations suggest that ALKBH3 catalyzes the demethylation of m 1 A sites on Aurora A mRNA.

ALKBH3 facilitates the stability and translation of Aurora A mRNA
Since the m 1 A modification of mRNA has been reported to influence mRNA metabolism processes 27 , including . Cilia length was determined by Image J software (d). e-g, k-m RPE-1 cells treated with the indicated siRNAs for 24 h were transfected with control, ALKBH3-FLAG, or ALKBH3-D193A-FLAG plasmid for another 24 h in normal conditions, and then subjected to western analysis or immunofluorescence. h-j RPE-1 cells transfected with the indicated plasmids for 24 h in normal conditions were treated with serum starvation for another 24 h and then applied for western analysis or immunofluorescence. Western blotting of FLAG and ALKBH3 protein was presented (e, h, k). Representative confocal images of RPE-1 cells with anti-Arl13b (green) and anti-FLAG (red) antibodies were shown (f, i, l). The percentage of ciliated cells in control or FLAG-positive group was analyzed (g, j, m). Actin was served as a loading control. DNA was stained by DAPI (blue). Cilia are indicated by white arrows. Scale bars, 10 μm. n, the number of total cells calculated. Data are presented as the means ± SD from at least three independent experiments. Student's t-test; ns not significant; *P < 0.05, **P < 0.01, ***P < 0.001. mRNA stability and translation, we first examined whether ALKBH3 regulates the mRNA levels of Aurora A. The results revealed that knockdown of ALKBH3 significantly decreased Aurora A mRNA levels in RPE-1 cells (Fig. 3a,  b). Further experiments showed that the decay rate of Aurora A mRNA was also significantly enhanced in cells depleted of ALKBH3 when the transcription was halted with actinomycin D in cells (Fig. 3c). Next, we determined if the RNA translation efficiency of Aurora A mRNA is affected by ALKBH3. Western analysis showed that ALKBH3 depletion significantly decreased Aurora A protein levels, which was effectively reversed by ectopic expression of ALKBH3 but not ALKBH3-D193A mutant ( Fig. 3d-g). Moreover, polysome profiling in RPE-1 cells displayed that stably knockdown of ALKBH3 resulted in reduced Aurora A mRNA abundance in high-molecularweight (HMW) polysome portions, which generally have high translation efficiency 28 (Fig. 3h, i), implying that depletion of ALKBH3 may suppress the translation of Aurora A mRNA. Collectively, these data suggest that ALKBH3 is essential for the regulation of the stability and translation of Aurora A mRNA.

ALKBH3 inhibits ciliogenesis through Aurora A
Given that Aurora A is a key regulator to facilitate cilia disassembly 19,21,[29][30][31] , we reasoned that ALKBH3 may affect cilia disassembly. To verify this, RPE-1 cells were treated with serum starvation for 24 h to induce ciliation, and then stimulated with serum for another 24 h to trigger cilia disassembly (Fig. 4a). Under this condition, we observed a dramatic decrease of cilia in control RPE-1 cells after serum stimulation for 24 h (Fig. 4b-d). However, ALKBH3-depleted cells were significantly resistant to serum-induced cilia disassembly. Thus, these data suggest that ALKBH3 functions to regulate cilia disassembly.
Since the protein levels of Aurora A are highly dynamic during ciliary cycle 29 , we then observed the profiles of ALKBH3 and Aurora A during serum starvation/re-stimulation. Our data showed that total levels of Aurora A were slightly increased at 2 h, and peaked at 18 h and 24 h after serum stimulation (Fig. 4e), which was consistent with the previous report 29 . We also found the similar Ectopic expression of Aurora A reverses the ciliary defects in ALKBH3-depleted cells. a-f RPE-1 cells transfected with the indicated siRNAs or not were treated with serum starvation for 24 h and serum stimulation for another 24 h, and then subjected to the following analyses at the indicated time points. Schematic illustration of experimental strategy used for the cilia assembly and disassembly experiments (a). Representative confocal images of RPE-1 cells with anti-Arl13b (green) and γ-tubulin (red) antibodies (b). Cilia are indicated by white arrows. Quantification analysis of the percentage of ciliated cells (c). Cilia length was determined by Image J software (d). Western blotting analysis of Aurora A and ALKBH3 proteins at the indicated time points (e, f). Actin was served as a loading control. g-i RPE-1 cells treated with the indicated siRNAs for 24 h, followed by infection with lentivirus carrying GFP-control or GFP-Aurora A for another 48 h in normal conditions, and then subjected to western analysis or immunofluorescence. Western blotting of GFP-Aurora A and ALKBH3 protein. Actin, a loading control (g). Representative confocal images of RPE-1 cells with anti-GFP (green) and anti-acetylated-α-tubulin (red) antibodies (h). The percentage of ciliated cells in GFP-control or GFP-Aurora A-positive group (i). DNA was stained by DAPI (blue). Scale bars, 10 μm. n, the number of total cells calculated. Data are shown as the means ± SD from at least three independent experiments. Student's t-test; ns, not significant; *P < 0.05, **P < 0.01, ***P < 0.001. expression pattern of ALKBH3 as Aurora A during serum starvation/re-stimulation (Fig. 4e). Of note, Aurora A protein levels in ALKBH3-depleted RPE-1 cells were obviously reduced throughout serum stimulation (Fig. 4f), implying that the regulation of Aurora A expression by ALKBH3 may happen throughout ciliary cycle.
Given that the phosphorylation of Aurora A at the centrosomes is sufficiently to induce disassembly of cilia 21 , we performed immunofluorescence experiments to examine the intensity change of centrosomal Aurora A or phosphorylated Aurora A in ALKBH3 knockdown RPE-1 cells (Supplementary Fig. S4). Our data revealed that both Aurora A and phosphorylated Aurora A levels were evidently reduced in ALKBH3-depleted cells ( Supplementary  Fig. S4a, e). Further results showed that the fluorescence intensities of Aurora A and phosphorylated Aurora A at centrosomes were significantly decreased in cells depleted of ALKBH3 ( Supplementary Fig. S4c, g). The ratio of cells with centrosomal Aurora A or phosphorylated Aurora A was also significantly reduced ( Supplementary Fig. S4d, h). These observations suggest that ALKBH3 depletion attenuates the levels of centrosomal Aurora A and phosphorylated Aurora A.
In addition, silencing of Aurora A significantly promoted ciliogenesis in RPE-1 cells (Supplementary Fig. S5), which was consistent with previous report 31 . Moreover, rescue experiments showed that exogenous expression of GFP-Aurora A protein was able to significantly reverse the enhanced ciliation induced by ALKBH3 knockdown (Fig. 4g-i). These data together indicate that ALKBH3 inhibits ciliation via promoting Aurora A expression.

ALKBH3 deficiency promotes ciliation to induce cell cycle arrest
To explore whether the enhanced ciliation in ALKBH3-depleted cells is associated with cell cycle arrest, we first examined the effect of ALKBH3 deficiency on cell cycle progression in RPE-1 cells without serum starvation. FACS analysis revealed that ALKBH3 depletion increased the percentage of cells at G0/G1 phase ( Supplementary Fig. S6a, b). Immunoblotting analyses displayed that cell cycle marker cyclin A was reduced in ALKBH3-depleted cells, indicating a decrease of S or G2/M phase cells in response to ALKBH3 knockdown ( Supplementary Fig. S6c) 31 . Further EdU-labeling experiments showed a significant decrease in the percentage of EdU-incorporated cells when ALKBH3 was depleted (Supplementary Fig. S6d, e). These data suggest that ALKBH3 depletion causes cell cycle arrest at G0/G1 phase.
Then, we asked whether the cell cycle arrest in G0/G1 phase induced by ALKBH3 depletion is mediated by its enhanced ciliation. Since IFT20 is a key intraflagellar transport protein required for cilia formation and has no significant effect on cell cycle progression 29,32 , we depleted IFT20 to abolish cilia growth in ALKBH3 knockdown cells to observe cell cycle progression. The data from FACS profiles, cell cycle marker, and EdU incorporation showed that cilia abrogation induced by IFT20 depletion significantly reversed the G0/G1 arrest in ALKBH3-depleted cells (Supplementary Fig. S6), indicating that ALKBH3 depletion induces G0/G1 arrest in a cilia-dependent manner.
Next, we examined whether Aurora A knockdown also led to a cilia-dependent cell cycle arrest in RPE-1 cells. Silencing of Aurora A induced cell cycle arrest at G0/G1 phase, which was also significantly reversed by simultaneous depletion of IFT20 (Supplementary Fig. S7). Collectively, these results suggest that depletion of either ALKBH3 or Aurora A causes enhanced ciliation in RPE-1, which may subsequently induce cell cycle arrest at G0/ G1 phase.

Molecular characterization of zebrafish alkbh3
To explore the role of ALKBH3 in ciliogenesis during vertebrate development, we first cloned zebrafish alkbh3 (GenBank, NM_001003511.2) (Fig. 5a). Bioinformatics analysis revealed that alkbh3 encoded a deduced 282 aa protein with a conserved 2OG-Fe (II) oxygenase superfamily domain 33 , implying that zebrafish Alkbh3 may also have demethylation activity (Fig. 5b). The amino acid residue D193 is critical for the demethylation activity of ALKBH3 on mRNA in homo sapiens 7,24 , while the residue D185 of Alkbh3 in Danio rerio was predicted to be evolutionarily conserved with D193. Alignment analysis showed that zebrafish Alkbh3 shared high homology with human ALKBH3 (identity, 56.7%; similarity, 98%) (Fig. 5c).
Then, we examined the temporal expression of zebrafish alkbh3 during embryo development, and discovered that alkbh3 mRNA was maternally deposited and ubiquitously expressed throughout early zebrafish embryonic stages (Fig. 5d). Whole-mount in situ hybridization revealed that alkbh3 mRNA was maternally provided, enriched in brain at 48 hpf (hours post fertilization) to 72 hpf, and decreased to low levels at 96 hpf (Fig. 5e).

alkbh3 morphants display ciliary phenotypes in zebrafish embryos
To investigate whether Alkbh3 influences ciliogenesis and embryonic development in zebrafish, we designed antisense morpholino oligonucleotides (alkbh3 MO) to block the translation of alkbh3 mRNA. The efficiency of alkbh3 MO to inhibit the expression of Alkbh3-EGFP fusion protein was validated by using a reporter construct consisting of alkbh3 MO targeting sequence fused to EGFP (Supplementary Fig. S8). alkbh3 morphants displayed ciliary phenotypes in zebrafish embryos, including curved body, pericardial edema, abnormal otoliths in otic  vesicles and dilation in pronephric ducts (Fig. 6a-d). Importantly, these defects in alkbh3 morphants were significantly reversed by co-injection with alkbh3 mRNA, but not alkbh3-D185A mRNA that was predicted to a catalytically inactive mutant of Alkbh3 based on the evolutionary conservation analysis of amino acid sequence (Fig. 5b). Further observations showed that cilia in the pronephric ducts of alkbh3 morphants were visibly defective compared with that of control morphants at 28 hpf, which were also significantly rescued by ectopic expression of alkbh3 mRNA rather than alkbh3-D185A mRNA (Fig. 6e, f). Together, these data imply that Alkbh3 may play a crucial role in cilia-associated developmental processes possibly through its demethylation activity in zebrafish.

Ectopic expression of Aurora A reverses the ciliary defects in alkbh3 morphants
To define if Aurora A is involved in the regulation of cilia-related events by Alkbh3 during zebrafish embryogenesis, we examined the expression levels of aurora A mRNAs in alkbh3 morphants, and found that aurora A mRNA was significantly decreased in alkbh3-depleted embryos (Fig. 7a), which was consistent with our data from mammalian cells (Fig. 3b). Moreover, co-injection of aurora A mRNA was able to significantly reverse the ciliary phenotypes in alkbh3 morphants (Fig. 7b-e). Additional data showed that ectopic expression of aurora A mRNA significantly rescued the abnormal ciliation in pronephric ducts in alkbh3 knockdown embryos (Fig. 7f Fig. 6 Zebrafish alkbh3 morphants exhibit ciliary phenotypes. Zebrafish embryos were injected with control morpholino (ctrl MO) or alkbh3 translation-block morpholino (alkbh3 MO) at 1-cell stage and subjected to the following analyses. a, b Bright-field micrographs and statistical analysis of the ciliary phenotypes in alkbh3 morphants. c, d Representative images, and quantification of pronephric ducts in the indicated groups. The lumen of pronephric ducts is indicated by white dashed lines. Scale bar, 10 μm. The widths of pronephric ducts in the indicated groups are measured and the median width of control morphants' ducts was set at 100%. e Immunostaining of cilia in pronephric ducts in morphants. Cilia were stained by anti-acetylated-α-tubulin antibody (green). The lumen of pronephric ducts is indicated by white dashed lines. Scale bar, 10 μm. f Quantification of embryos with abnormal pronephric cilia. n, the total number of embryos in all experiments. Data are shown as the means ± SD from at least three independent experiments. Student's t-test; ns, not significant; *P < 0.05, **P < 0.01, ***P < 0.001. g). Collectively, these results suggest that Alkbh3 influences cilia-associated developmental events possibly by regulating aurora A mRNA in vertebrate embryogenesis.

Discussion
In this manuscript, our study reveals a previously uncharacterized role of the m 1 A demethylase ALKBH3 in vertebrate ciliogenesis and embryonic development. ALKBH3 removes the m 1 A modification of Aurora A mRNA in the coding sequence region near the translation initiation site to inhibit Aurora A mRNA decay and promote its mRNA translation, which consequently maintains the abundance of Aurora A protein to suppress ciliogenesis (Fig. 7h). m 1 A RNA modification occurs in diverse RNA species and impacts RNA metabolism and gene regulation; 7,8,34 however, its physiological consequences remain unclear. In this study, we find that depletion of alkbh3 causes several developmental defects, including curved body, pericardial edema, abnormal otoliths, and dilation of pronephric ducts in zebrafish embryos. Moreover, these phenotypes are significantly rescued by wild-type alkbh3, but not its catalytically inactive mutant. Thus, our data for the first time show the biological function of m 1 A RNA modification during vertebrate embryonic development.
Few studies on the roles of RNA modification in cilia structure and function have been reported until Narry Kim's group in 2021 showed that the m 6 A demethylase FTO (fat mass and obesity-associated protein) targets FOXJ1 (forkhead box protein J1) mRNA to regulate proper motile ciliogenesis 35 . However, the function of m 1 A RNA modification in ciliation is still unknown. Here we identify that ALKBH3 is a negative regulator of ciliogenesis in mammalian cells. Knockdown of ALKBH3 remarkably promotes ciliation in mammalian cells, which is significantly reversed by wild-type ALKBH3, but not its catalytically inactive mutant. Collectively, these observations indicate an important role of m 1 A RNA modification in ciliation, providing a new perspective for the study of cilia regulation.
ALKBH3 is a member of Alpha-ketoglutaratedependent dioxygenase AlkB homology family, which is able to remove alkyl adducts from nucleobases by oxidative dealkylation 36 . Previous studies have shown that ALKBH3 is an alkylation damage repair enzyme to repair m 1 A and m 3 C (3-methylcytosine) alkylation damage in RNA and ssDNA 33,37 . Emerging data demonstrate that ALKBH3 can remove the methyl group of m 1 A in mRNA and tRNA 2,7,38,39 . However, little is known about the role of ALKBH3-mediated RNA demethylation in vertebrate ciliation and development. Herein, we find that ALKBH3 influences ciliogenesis and cilia-associated developmental events by targeting aurora A mRNA in an m 1 A-dependent manner.
Aurora A has been demonstrated as a key protein to promote cilia disassembly 24,25 . In the initiation of ciliary disassembly, Aurora A is phosphorylated and activated at the basal body, and then phosphorylates and activates HDAC6 (histone deacetylase 6), which in turn deacetylates and destabilizes axonemal tubulin to further facilitate ciliary disassembly 21 . Human enhancer of filamentation 1, trichoplein and pitchfork have been shown to interact with Aurora A at the ciliary base and increase the level of phosphorylated Aurora A during cilia disassembly 21,31,40 . However, the post-transcriptional regulation of Aurora A mRNA remains unclear. In this manuscript, our data show that ALKBH3 demethylates m 1 A sites in the coding sequence region of Aurora A mRNA to promote its stability and translation. Future studies are clearly required to explore the molecular mechanism underlying the epigenetic regulation of Aurora A kinase.
While our study shows that ALKBH3-dependent m 1 A demethylation of Aurora A mRNA regulates ciliogenesis and vertebrate development, two limitations of this work should be addressed. (1) The published datasets of m 1 A-seq we employed are from 293 T cells, while RPE-1 cells are used to investigate the role of ALKBH3 in ciliogenesis in our experiments. (2) We overlapped ciliaassociated genes from CiliaCarta with m 1 A-seq data to detect the potential targets that mediate the function of ALKBH3 in ciliogenesis, which may preclude finding unknown genes that regulate ciliation. Therefore, future works on systematically screening the substrates mediating the regulation of ALKBH3 in ciliogenesis are clearly needed.

Materials and methods
Cell culture and transfection RPE-1 cells were maintained in DMEM/F12 (Corning, USA) supplemented with 10% FBS (fetal bovine serum, ExCell Bio, China) and antibiotics at 37°C in 5% CO 2 . The indicated plasmids were transfected into RPE-1 cells with Lipofectamine 3000 (Invitrogen, USA) according to the manufacturer's instructions.

Plasmid constructions
Full-length human ALKBH3 and ALKBH3-FLAG was amplified from cDNAs and cloned into pCS2+ vector. The mutation of ALKBH3-D193A and ALKBH3-D193A-FLAG was produced by PCR and cloned into a pCS2+ vector. Full-length human Aurora A was amplified from cDNAs and cloned into pLVX-GFP vector to generate GFP-Aurora A. Full-length zebrafish alkbh3-flag or alkbh3-D185A-flag was inserted into pCS2+ to generate morpholino-resistant alkbh3-flag or alkbh3-D185A-flag plasmids. Full-length zebrafish aurora A-flag was also inserted into pCS2+ to generate pCS2-aurora A-flag plasmid. The reporter alkbh3-5′ATG-EGFP was constructed by inserting EGFP downstream of alkbh3-5′ATG, which harbors the morpholino-targeting sequence into pCS2+ vector under the control of CMV promoter.

TA cloning assay
Aurora A PCR products amplified from cDNAs that were reverse-transcripted by RT-1306 in the control and ALKBH3-depleted cells were cloned into the pEASY-Blunt Zero vector according to the manufacturer's instructions (TransGen Biotech, China).

Western blot analysis
Western blot assays were performed as described previously 41 . Cells were lyzed on ice in TBSN buffer (20 mM Tris, pH 8.0, 150 mM NaCl, 0.5% NP-40, 5 mM EGTA, 1.5 mM EDTA, 0.5 mM Na 3 VO 4 , 20 mM p-nitrophenyl phosphate) with protease inhibitor cocktails (Roche, Switzerland) for 30 min. The protein samples were separated by SDS-PAGE, and then transferred to polyvinylidene fluoride membranes (Millipore, USA). The membranes were blocked with 5% non-fat milk for 1 h at room temperature and then incubated with the corresponding primary antibodies overnight at 4°C. After washing, the membranes were incubated with HRPconjugated secondary antibodies for 1 h at room temperature. Protein bands were detected using ChemiDoc Touch Imaging System (Bio-Rad, USA).

Immunofluorescence analysis
Immunofluorescence assays were carried out as described previously 42,43 . Cells grown on coverslips were fixed with 100% methanol for 10 min at -20°C and blocked with 5% BSA in 0.1% PBST (0.1% TritonX-100 in phosphate-buffered saline buffer) for 1 h. Cells were then incubated with the indicated primary antibodies overnight at 4°C, followed by species-specific Alexa Fluor 488-, 555-, and 647-conjugated secondary antibodies (Invitrogen, USA) for 1 h. Nuclei were stained with DAPI (Sigma-Aldrich, USA). Images were acquired by using a laser scanning confocal microscope (OLYMPUS FV3000 OSR).

RNA-seq
RPE-1 cells were transfected with control or ALKBH3 siRNA for 48 h in DMEM/F12 medium with 10% serum. Total RNAs were extracted from control and ALKBH3-depleted cells and then subjected to library construction and sequencing (Novagene company, China).

m 1 A RNA immunoprecipitation
Total RNAs were extracted from RPE-1 cells with TRIzol reagent followed by an additional DNase I treatment to avoid DNA contamination. 40 µg total RNAs were then incubated with 1 μg anti-m 1 A antibody in RIP buffer (150 mM NaCl, 0.1% NP-40, 10 mM Tris, pH 7.4) at 4°C overnight. After incubation, 40 μl Protein A/G beads rinsed with RIP buffer were added to the mixture of RNA and antibody and incubated for an additional 4 h at 4°C. Beads were washed five times with IPP buffer and precipitated RNAs were further purified with TRIzol according to the manufacturer's instructions (Invitrogen, USA). Input and immunoprecipitated RNAs were reverse transcribed into cDNAs and quantified by qRT-PCR. mRNA decay assay RPE-1 cells were treated with 5 μg/mL actinomycin D (MedChemExpress, USA) to suppress global mRNA transcription. Cells were harvested at different time points and total RNAs were extracted for reverse transcription. The levels of Aurora A mRNA after transcription inhibition were detected by qRT-PCR.

Polysome fractionation
Polysome fractionation was performed as described previously 28,44 . Briefly, ALKBH3-depleted or control RPE-1 cells (four 150-mm culture dishes) were treated with 10 μg/ mL cycloheximide (MedChemExpress, USA) for 5 min at 37°C. Then, cells were harvested and 500 μl of cytoplasmic extract was layered onto 11 mL of 5%-50% sucrose gradient and centrifuged at 36,000 rpm in a Beckman SW-41Ti rotor for 2 h at 4°C. Gradients were fractionated and monitored at absorbance 254 nm (Brandel, USA). Polysome-associated and cytosolic RNAs of each fraction were isolated by using TRIzol and analyzed by qRT-PCR.

Flow cytometry analysis
For flow cytometry analysis, RPE-1 cells were collected, washed once with cold phosphate-buffered saline, and then fixed in 70% ethanol. DNA was stained with 100 μg/ mL propidium iodide and 50 μg/mL RNase A for 30 min at 37°C. The samples were analyzed using an FC 500 MCL Flow Cytometer (Beckman Coulter).
EdU assay RPE-1 cells incubated with 50 μM EdU (5-Ethynyl-2′deoxyuridine) for 2 h were fixed for 30 min in 4% PFA. EdU staining was then performed using the Cell-Light EdU Apollo567 In Vitro Kit according to the manufacturer's instructions (RIBOBIO, China).

Whole-mount in situ hybridization
Whole-mount in situ hybridization was carried out as described previously 45 . Antisense RNA probes for alkbh3 mRNA were labeled with digoxigenin (Roche, Switzerland). Zebrafish embryos at different stages were fixed in 4% paraformaldehyde and incubated with digoxigeninlabeled RNA probes. Alkaline phosphatase-coupled antidigoxigenin antibody (Roche, Switzerland) was used to detect hybridized probes, and 5-bromo-4-chloro-3-indolyl phosphate (BCIP)/nitro blue tetrazolium (NBT) solution (Sigma, USA) was used as the chromogenic substrate.

Whole-mount immunofluorescence microscopy
Whole-mount immunofluorescence was performed as described previously 45 . Zebrafish embryos were fixed in 4% paraformaldehyde overnight at 4°C. Fixed embryos were washed with washing buffer (0.1% Triton X-100, 0.2% DMSO, and 5% bovine serum albumin in PBS buffer) for 15 min, and then blocked with blocking buffer (0.1% Triton X-100, 0.2% DMSO, 10% goat serum and 5% bovine serum albumin in PBS buffer) for 1 h at room temperature. Embryos were then incubated with primary antibodies diluted in blocking buffer overnight at 4°C, followed by secondary antibodies (Invitrogen, USA) overnight at 4°C. Nuclei were stained with DAPI (Sigma-Aldrich, USA). Samples were rinsed with washing buffer, placed in PBS in a 35 mm petri dish, and examined using 40× water-immersion objective lenses by confocal laser scanning microscopy (Olympus BX61W1-FV1000).

Statistics
Data are representative of at least three independent experiments. Means and standard deviations (SD) were calculated and shown in the graphs. Student's t-test was performed using GraphPad Prism software. A value of P < 0.05 was considered statistically significant.