N6‐methyladenosine demethylase ALKBH5 suppresses colorectal cancer progression potentially by decreasing PHF20 mRNA methylation

CRC tumour tissues and tumour‐adjacent normal tissues were collected from Peking University People's Hospital. All colorectal tissues were pathologically confirmed.

Six human colon cancer cell lines (RKO, SW480, HCT116, HCT8, LS174T, LOVO) were purchased from the Cell Resource Center of Peking Union Medical College (China). RKO, SW480, HCT116, HCT‐8, LS174T and LOVO were cultured in MEM, IMDM, IMDM, RPMI 1640, MEM and F12K, respectively. All the cell lines were cultivated in the corresponding medium containing 10% FBS (Gibco, USA) in 5% CO₂ environment at 37°C.

Three siRNAs of ALKBH5 (siALKBH5‐1: ACAAGTACTTCTTCGGCGA, siALKBH5‐2: GCGCCGTCATCAACGACTA, siALKBH5‐3: CTGAGAACTACTGGCGCAA) and two siRNAs of human PHF20 (siPHF20‐1: CCCGAGAAATACACCTGTTAT, siPHF20‐2: ATTGTGCCACTGATGATAAAC) were synthesized by RiboBio (China). In addition, the two most efficient sequences for ALKBH5 were used to construct lentiviral shRNA plasmids. The lentiviral plasmid expressing shALKBH5 or shNC, overexpressing ALKBH5 or an empty vector were purchased from GeneCopoeia (USA). The packaging plasmid, envelope plasmid and target plasmid were transfected to the 293T cells to obtain the lentivirus using the Lenti‐Pac HIV Expression Packaging Kit (GeneCopoeia, USA). Then, the lentivirus was used to infect LOVO and RKO.

Immunohistochemical staining (IHC) was performed, as previously described.13 ALKBH5 staining index score (0–12) was calculated using the staining intensity multiplying by the percentage of ALKBH5 positive staining. The staining intensity is divided into four grades (negative: 0; weak: 1; moderate: 2; strong: 3), and the percentage of positive staining is divided into five grades (<5%: 0; 5%–25%: 1; 26%–50%: 2; 51%–75%: 3; >75%: 4). Anti‐ALKBH5 (Sigma, USA) was used. All scores were independently evaluated by two pathologists.

RT‐qPCR was performed, as previously described.13 The primer sequences are listed in Table S1.

Western blot assays were performed, as previously described.13 Primary antibodies included anti‐ALKBH5 (Sigma, USA), anti‐PHF20 (CST, USA) and anti‐GAPDH (CST, USA).

To perform the cell proliferation assays, we seeded 3 × 10³ cells into a 96‐well plate. Next, 10‐μl CCK8 solution was added into each well of the plate. After incubation for 2 h, the absorbance at 450 nm was detected to assess cell proliferation. Moreover, cell proliferation was assessed for five consecutive days after cells seeded.

To conduct the colony formation assays, we seeded 5 × 10² cells into a 6‐well plate. After cultivated for 10 days, the clones were fixed in 4% paraformaldehyde solution for 25 min. Then, the clones were stained with .1% crystal violet solution for 25 min. Last, the clones were counted after washed with PBS three times.

To perform the migration assays, we seeded 1 × 10⁵ cells into the upper chambers (Corning, USA) and added medium with 10% FBS to the lower chambers. To perform the invasion assays, the upper chambers were coated with Matrigel (Sigma, USA). After cultivated for 48 h, the cells migrating below the chambers were fixed, stained and counted under a microscope.

Female Balb/c nude mice (6–8 weeks) were purchased from the Experimental Animal Center of Military Medical Sciences (China). We injected 1 × 10⁶ cells subcutaneously in the flanks of mice and measured the length (L) and width (W) of tumour every 2 days. Tumour volume (V) was calculated as follows: V=12L·W2. The mice were euthanized at 16th days after the injection of cancer cells. Then, the subcutaneous tumours were removed and weighted.

Tumour cells were treated with actinomycin D (Sigma, USA) at 5 μg/ml. The cells were collected after incubated for 0, 3 or 6 h, and then RNA was extracted to perform RT‐qPCR assays as described earlier. The half‐life of mRNA was calculated to assess its stability.

The mRNA was purified from total RNA using the GenElute mRNA Miniprep Kit (Sigma, USA). The m⁶A modification level of mRNA was evaluated utilizing the EpiQuik m⁶A RNA methylation quantification kit (EpiGentek, USA). In brief, 200‐ng mRNA was incubated with capture antibody and then incubated with detection antibody. The m⁶A modification level of mRNA was quantified colourimetrically by measuring the absorbance at 450 nm.

The wild type and m⁶A sites mutated PHF20 were constructed into the luciferase reporter vector (GeneCopoeia, USA). The sequences of the wild type and m⁶A sites mutated PHF20 are listed in Table S2. Cells were transfected with siNC or siALKBH5. The cells were re‐seeded into a 24‐well plate after incubated with siNRA for 48 h and then were transfected with PHF20 3′UTR‐WT or PHF20 3′UTR‐MUT. The cells were lysed after cultivated for 24 h. Firefly and Renilla luciferase activities in cell lysates were analysed utilizing the Dual‐Glo Luciferase Assay system (Promega, USA).

Total RNA was isolated with a TRIzol reagent (Invitrogen, CA). A total of 1‐μg RNA was used to perform RNA sequencing. Sequencing libraries were generated using the NEBNext Ultra RNA Library Prep Kit for Illumina (NEB, USA). Sequencing was performed based on the Illumina NovaSeq platform. Differentially expressed genes were identified utilizing the DESeq2 R package (|log₂(fold change)|>0 and p < 0.05).

The m⁶A modification of mRNA was quantified using the Magna MeRIP m⁶A Kit (Millipore, Germany). In brief, 10% of mRNA purified from total RNA was saved as inputs. Magna ChIP Protein A/G Magnetic Beads were washed and then incubated with 5‐μg anti‐m⁶A antibody with rotation for 30 min at room temperature. The antibody‐beads mixed with mRNA were incubated at 4°C with RNase inhibitors for 2 h. Then, the methylated mRNAs were eluted and purified utilizing the RNeasy mini kit (QIAGEN, Germany). RT‐qPCR was used to determine m⁶A enrichment by normalizing to the input.

All results are presented as mean ± SD. The SPSS 17.0 was used for data analysis. The χ² test was utilized to analyse the correlation between ALKBH5 and clinicopathological features of patients. Kaplan–Meier analysis was used to construct survival curves, and the differences of different groups were estimated by log‐rank test. For all continuous variables, Student's t‐test or ANOVA was performed to evaluate the differences. A p value of .05 was considered statistically significant.

To evaluate the expression of ALKBH5 in cancers, we initially detected mRNA expression of ALKBH5 in 23 solid cancers in TCGA (The Cancer Genome Atlas) datasets. In 12 solid cancers, ALKBH5 was significantly dysregulated in comparison to adjacent normal tissues (Figure 1A). As shown in Figure 1A, compared to adjacent normal tissues, ALKBH5 was significantly downregulated in seven solid cancers (CRC, GBM, KICH, KIRP, PRAD, THCA and UCEC) and upregulated in five solid cancers (CHOL, HNSC, KIRC, LIHC and LUSC). We discovered that, compared with adjacent normal tissues, ALKBH5 was significantly reduced in tumour tissues in TCGA‐CRC cohort. We also downloaded the RNA sequencing data of CRC from GEO (Gene Expression Omnibus) datasets. The results confirmed the reduction of ALKBH5 mRNA in CRC tissues (GSE39582↗ and GSE87211↗; Figure 1B). Furthermore, RT‐qPCR showed that ALKBH5 mRNA expression in CRC tissues was significantly lower than that in adjacent normal tissues (Figure 1C). We also evaluated ALKBH5 protein expression in CRC. IHC assays indicated that ALKBH5 protein expression was significantly decreased in CRC tissues (Figure 1D,E). What is more, the m⁶A quantification assays of mRNA implied that m⁶A modification levels of mRNAs were higher in CRC tissues (Figure 1F).

In order to explore the clinical value of ALKBH5, we analysed the correlation between ALKBH5 and clinicopathological features of CRC patients at the mRNA and protein levels. The results indicated that, at the mRNA levels, downregulated ALKBH5 was significantly associated with distant metastasis (p = .025) and late clinical stage (p = .034) (Table 1; Figure 2A,B). Consistently, at the protein levels, downregulated ALKBH5 was significantly related to distant metastasis (p = .042) and late clinical stage (p = .024) (Table 2; Figure 2D,E). The Kaplan–Meier analysis indicated that CRC patients with a loss of ALKHB5 had shorter overall survival both at the mRNA (Figure 2C, p = .026) and protein (Figure 2F, p = .048) levels.

Next, we explored the effect of ALKBH5 on colon cancer cell function. We initially identified ALKBH5 mRNA expression in five colon cancer cells, and noted that ALKBH5 was elevated in LOVO and RKO cells, compared to other colon cancer cell lines (Figure 3A). Three specific siRNAs were utilized to knock‐down ALKBH5 in LOVO and RKO cells (Figure 3B,C). Furthermore, two siRNAs (siALKBH5‐2, siALKBH5‐3) with high knock‐down efficiency were utilized to construct knock‐down plasmids (shALKBH5‐2, shALKBH5‐3) for subsequent experiments. We constructed two stable cell lines using two independent ALKBH5‐knock‐down lentiviruses. The knock‐down efficiency was validated both at the mRNA and protein levels (Figure 3D–F). Subsequently, the CCK‐8 assays revealed that the knock‐down of ALKBH5 markedly enhanced cellular growth, as well as the viability of LOVO and RKO cells (Figure 3G). It was confirmed that the knock‐down of ALKBH5 significantly increased the colony formation abilities of LOVO and RKO cells through the colony formation assays (Figure 3H). Additionally, we assessed migration and invasion abilities of LOVO and RKO cells using the transwell assays. The results indicated that the knock‐down of ALKBH5 drastically elevated cell migration and invasion abilities (Figure 3I,J).

To further determine whether ALKBH5 affects colon cancer cell function in vivo, we established tumour xenograft models. We found that tumour growth rate was faster (Figure 4C), and tumour volume and weight were increased (Figure 4A,B,D) when ALKBH5‐knock‐down LOVO cells were implanted (n = 5). Moreover, the expression of Ki‐67 was also increased in the shALKBH5 groups in comparison to the shNC groups (Figure S1).

We also constructed stably transfected cell lines with ALKBH5 overexpression utilizing a lentivirus vector, and RT‐qPCR and Western blot assays validated the overexpression efficiency (Figure 5A–C). The CCK‐8 assays revealed that the overexpression of ALKBH5 markedly repressed cellular growth and the viability of LOVO and RKO cells (Figure 5D). The colony formation assays confirmed that the overexpression of ALKBH5 resulted in a significant decrease of the colony formation abilities of LOVO and RKO cells (Figure 5E). Additionally, the overexpression of ALKBH5 drastically weakened cell migration and invasion by the transwell assays (Figure 5F,G). Then, our results of tumour xenograft models indicated that tumour growth rate was slower (Figure 6C), and that tumour volume and weight were decreased (Figure 6A,B,D) when LOVO cells overexpressing ALKBH5 were implanted (n = 5).

To identify the m⁶A‐modified targets, we collected tumour tissues and normal tissues from six CRC patients to perform MeRIP‐seq.14 We identified 1343 dysregulated m⁶A peaks in six pairs of tissues by MeRIP‐seEq (Figure 7A). Among them, 625 m⁶A peaks were hypermethylated in tumour tissues, whereas 718 m⁶A peaks were hypomethylated in tumour tissues (Figure 7A). Next, 625 hypermethylated m⁶A peaks were distributed among the transcripts of 265 genes, and 718 hypomethylated m⁶A peaks were distributed among the transcripts of 311 genes. Interestingly, we discovered that in the six pairs of tissues for MeRIP‐seq, the expression of ALKBH5 in tumour tissues was significantly lower than that in normal tissues (Figure 7B). We hypothesized that the transcripts of these 265 genes were regulated by ALKBH5. In order to further explore m⁶A modification targets regulated by ALKBH5, we performed RNA‐seq in ALKBH5‐knock‐down LOVO cells and control cells. The results demonstrated that after ALKBH5 knocked down, 679 genes were differentially expressed, including 362 upregulated genes and 317 downregulated genes (Figure 7C,D). Combining MeRIP‐seq and RNA‐seq, we found 19 genes regulated potentially by ALKBH5 through m⁶A modification, including 11 upregulated genes (PCDH7, QSOX1, LAMA5, CCDC69, LITAF, MDK, TGFB1I1, ATP2B4, DUSP3, PHF20 and PLIN3) and 8 downregulated genes (NPNT, KRT20, PCSK7, SAMD4A, NEDD4L, SLC35A3, SPTLC2 and DNM1L) after ALKBH5 knock‐down (Figure 7E).

Then, we examined the expression of these 19 genes after ALKBH5 knock‐down. The results of RT‐qPCR displayed that among the 11 upregulated genes, 5 genes (CCDC69, LITAF, TGFB1I1, PHF20 and PLIN3) were upregulated both in ALKBH5‐knock‐down LOVO and RKO cells (Figure 7F). In addition, we discovered that among the eight downregulated genes, only PCSK7 was downregulated both in ALKBH5‐knock‐down LOVO and RKO cells (Figure 7G). Next, we examined the correlation between ALKBH5 and the six candidate genes in the TCGA database. Our results indicated that only PHF20 was negatively correlated with ALKBH5 in CRC (Figure 7H). We also verified elevated protein expression of PHF20 in ALKBH5‐knock‐down cells compared to control cells by Western blot assays (Figure 7I). Moreover, visualization analysis showed that compared to normal tissues, the m⁶A peaks of PHF20 mRNA were more abundant in tumour tissues (Figure 7J). Thus, PHF20 may be a key molecule regulated by ALKBH5 through m⁶A modification.

As a demethylase, ALKBH5 is able to remove m⁶A modification from mRNA. The m⁶A quantification assays of mRNA indicated that ALKBH5 loss significantly increased the m⁶A modification level of mRNAs (Figure 8A). The MeRIP‐qPCR assays demonstrated that ALKBH5 loss significantly increased the m⁶A modification level of PHF20 mRNA (Figure 8B). Next, we predicted the m⁶A sites of PHF20 mRNA utilizing the SRAMP database (http://www.cuilab.cn/sramp↗) and BERMP database (http://www.bioinfogo.org/bermp↗). We discovered that the m⁶A methylation sites on PHF20 mRNA were mainly located in the 3′UTR, and that there were eight possible m⁶A methylation sites. According to the m⁶A methylation sites on PHF20 mRNA, we constructed a luciferase reporter vector (Figure 8C). The luciferase reporter assays illustrated that ALKBH5 loss significantly enhanced the expression of the wild‐type PHF20 3′UTR plasmid, but that there was no significant effect on PHF20 3′UTR plasmid with m⁶A mutation sites (Figure 8D).

To assess the effect of ALKBH5 on PHF20 mRNA stability, we treated colon cancer cells with actinomycin D and found that ALKBH5 loss improved stability of PHF20 mRNA and prolonged its half‐life (Figure 8E). Furthermore, our data showed that PHF20 knock‐down contributed to a significant inhibition of the proliferation, colony formation, migration and invasion of LOVO (Figure S2). Then we carried out rescue experiments to determine whether ALKBH5 had an effect on the biological function of colon cancer cells by regulating PHF20. It showed that PHF20 knoc‐kdown inhibited the shALKBH5‐mediated enhancement of the proliferation, colony formation, migration and invasion of LOVO and RKO (Figure 8F–I).