Restriction enzymes, PCR polymerases, and T4 DNA Ligase were from New England BioLabs (Ipswich, MA, United States). pRL-CMV, the control, Renilla luciferase expressing plasmid, was from Promega (Madison, WI, United States). The single guide RNA (sgRNA) targeting the human miR-210 gene was designed using the website https://www.benchling.com/crispr/, and the corresponding oligos were 5′-CAC​CGA​GGG​GCT​GCC​CTG​CGC​CTG​G-3′ and 5′-AAA​CCC​AGG​CGC​AGG​GCA​GCC​CCT​C-3’. The oligos were synthesized by Sangon (Shanghai, China) and cloned into the BsmbI sites in LentiCRISPR (Sanjana et al., 2014) to obtain LentiCRISPR-miR-210, which was verified by Sanger sequencing (Sangon). Coding sequences of the full length BNIP3L (NM_004331) and its miR-210 binding site mutant were synthesized and cloned into the p3xFLAG-CMV 10 vector (General Biology, Anhui, China). To construct a firefly luciferase (luc) reporter with BNIP3L in the 3′ untranslated region, BNIP3L sequence including the putative miR-210 binding site was amplified from the above p3xFLAG-CMV-BNIP3L plasmid with primers 5′-GCG​CTA​GCG​GAA​AAT​GAG​CAG​TCT​CT-3′ and 5′-GCC​TCG​AGA​TTC​ATG​TTG​TGC​A-3′, digested with NheI and XhoI, and cloned into pCMV-luc (Ghosh et al., 2010). The BNIP3L mutant lucierase reporter with the miR-210 binding seed sequence 5′-ACACGTAC-3′ deleted was constructed using the Quikchange method (Stratagene, La Jolla, CA, USA) with primers 5′-AAA​TGG​GGG​GCT​GGA​CAT​CCT​CAT​CCT​CCA-3′ and 5′- TGG​AGG​ATG​AGG​ATG​TCC​AGC​CCC​CCA​TTT -3’. Plamsids were verified by Sanger sequencing.

The human 293T cell line was acquired from Sangon. Cells were cultured in the Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum and 2 mM L-glutamine (Invitrogen, Waltham, MA, United States) and maintained at 37°C with 5% CO₂ in a cell culture incubator. Cells were transfected with plasmid DNAs (and the control or miR-210 mimic RNA (Sangon), if necessary), with 400 ng of total nucleic acids per 24-well, using Lipofectamine 2000 (Invitrogen). For hypoxia testing, cells were placed in a modular incubator chamber (Billups-Rothenberg, San Diego, CA, United States) and flushed with 95% N₂/5% CO₂, which reduced the oxygen level to approximately 3% (Ghosh et al., 2010). Or more commonly, cells were treated with the control DMSO (Sangon) or 0.5 mM dimethyloxalylglycine (DMOG, catalog number 71210, Cayman Chemical, Ann Arbor, MI, USA) for 24 h, unless indicated otherwise.

LentiCRISPR-miR-210 was transfected into 293T cells, and the wild-type (WT) control transfected with the empty LentiCRISPR vector. After 48 h cells were transferred to 10 cm culture dishes (Sangon) and selected with 2 μg/mL puromycin (catalog number ST551, Beyotime, Shanghai, China). Single clones were picked using cloning rings (Corning, Corning, NY, United States), expanded, and their miR-210 status determined by PCR and sequencing. For PCR, genomic DNA was isolated from 293T cells using the FastPure Cell/Tissue DNA Isolation Mini Kit (Vazyme, Nanjing, China), and amplified with primers 5′-GCA​AGC​TTC​GGG​GGG​TCG​GGC​T-3′ and 5′-GGG​TAT​CTG​GCC​CAG​CCT-3’. PCR products were inserted into the pMD19-T vector (Takara, Beijing, China), and approximately 10 isolates for each 293T clone examined by Sanger sequencing to identify the miR-210 KO cells. To assess the off-target effects of the sgRNA, potential off-target sites were predicted using the website https://www.benchling.com/crispr/, primers designed (Supplementary Table S1) to amplify the genomic DNA, and the PCR products analyzed by Sanger sequencing.

Total RNA was isolated from 293T cells using Trizol (Invitrogen). For miR-210 expression analysis, the miRNA 1st Strand cDNA Synthesis Kit (Vazyme) was used to reverse transcribe RNA with the miR-210-specific primer 5′-GTC​GTA​TCC​AGT​GCA​GGG​TCC​GAG​GTA​TTC​GCA​CTG​GAT​ACG​AC-3’. For mRNA expression analyses, the SuperScript III First-Strand Synthesis SuperMix (Invitrogen) was used to reverse transcribe RNA. qPCR reactions were carried out on the QuantStudio 6 Flex Real-Time PCR System (Thermo Fisher Scientific, Waltham, MA, United States) using the SYBR Green qPCR Master Mix (Thermo Fisher Scientific). The relative expression of target genes was calculated with the 2^−ΔΔCT method normalized to that of the U6 or Actin mRNA (Zhang et al., 2013). qPCR primers were synthesized by Sangon: miR-210: 5′-CGC​TGT​GCG​TGT​GAC​AGC-3′ and 5′-AGT​GCA​GGG​TCC​GAG​GTA​TT-3′, U6: 5′-GCT​TCG​GCA​GCA​CAT​ATA​CTA​AAA​T-3′ and 5′-CGC​TTC​ACG​AAT​TTG​CGT​GTC​AT-3′, Actin: 5′-GGA​CTT​CGA​GCA​AGA​GAT​GG-3′ and 5′-AGC​ACT​GTG​TTG​GCG​TAC​AG-3′, E2F3: 5′-TGG​TAC​CAT​TGA​GTT​GCT​GCT​ATT-3′ and 5′-AGC​TCA​TGT​GTT​GCC​CTT​TAT​ACA-3′, HOXA1: 5′-CAG​CGC​AGA​CTT​TTG​ACT​GGA​TG-3′ and 5′-TCC​TTC​TCC​AGT​TCC​GTG​AGC​T-3′, ISCU: 5′-GGG​TCC​CTT​GAC​AAG​ACA​TCT-3′ and 5′-CCT​TTC​ACC​CAT​TCA​GTG​GCT​A-3′, MNT: 5′-ACG​TAC​TGG​AGA​TTG​ACC​GCG​T-3′ and 5′-CCT​CGT​CTA​TGT​TGT​CCT​CAC​C-3′, and BNIP3L: 5′-AAT​GTC​GTC​CCA​CCT​AGT​CG -3′ and 5′-CCC​CCA​TTT​TTC​CCA​TTG​CC-3’.

All experiments had at least three biological replicates. The MTT (thiazolyl blue tetrazolium bromide, catalog number A600799, Sangon) assay to measure cell proliferation and the scratch assay to examine cell migration had been described (Zhang et al., 2013). For MTT assay 3,000 cells were seeded per well in a 96-well plate, and cell numbers measured once every 24 h. For time points longer than 3 days, cells were split 1:8 on day 3 for continuous studies. For scratch assay, cells were grown in a 24-well plate to over 80% confluency, then a pipet tip was used to make a straight line scratch across the cell monolayer surface. After floating cells were removed, the remaining cells were cultured in serum-free media for the next 24 h. At 0, 12, and 24 h the scratch or the gap it introduced was monitored under a light microscope, with cells entering the gap region considered migration.

The Cell Cycle and Apoptosis Analysis Kit (YEASEN, Shanghai, China) was used to study the cell cycle profiles. Approximately 2 × 10⁵ cells were washed and collected by centrifugation at 1,000 g for 5 min. The cells were fixed in cold 70% ethanol, washed, and stained with a propidium iodide (PI) solution containing RNase A. The stained cells were filtered through a 400-mesh sieve and analyzed using a BD Accuri C6 flow cytometer and the FlowJo v10.6 software (Becton Dickinson, Franklin Lakes, NJ, United States). Cells at different phases of the cell cycle were determined automatically using the default settings. The Annexin V-Alexa Fluor 647/PI Apoptosis Detection Kit (YEASEN) was used to study apoptosis. Harvested cells were stained with Annexin V-Alexa Fluor 647 and PI and analyzed using a BD Accuri C6 flow cytometer and the FlowJo v10.6 software as described above.

Reactive oxygen species (ROS) levels were measured using the Meilun Reactive Oxygen Species Assay Kit (Meilunbio, Dalian, Liaoning, China). The kit uses 2,7-Dichlorodi-hydrofluorescein diacetate, a dye that fluoresces when oxidized by ROS. Approximately 1 × 10⁵ cells were stained with the dye for flow cytometry studies, and cells were excited at 488 nm and detected at 525 nm. Fluorescence signals were quantified using the FlowJo v10.6 software. Alternatively, cells grown in a 6-well plate were stained with the dye and monitored under a microscope. The mitochondrial membrane potential was measured using the JC-1 Mitochondrial Membrane Potential Assay from Sciben (Nanjing, China), per manufacture’s instructions. JC-1 is a dye that accumulates in the mitochondria. When the membrane potential is low, JC-1 exists as a monomer and produces green fluorescence. When the membrane potential is high, JC-1 aggregates and yields red fluorescence. Hence the ratio of red vs. green fluorescence can be an index of mitochondrial membrane potential changes. Approximately 1-6 × 10⁵ cells were stained with JC-1 solution, washed, and then analyzed using a BD Accuri C6 flow cytometer. Cells were excited at 490 nm and 525 nm and detected at 530 and 590 nm, respectively. The FlowJo v10.6 software was used to quantify red and green fluorescence signals automatically.

For RNA-seq (Azenta Life Sciences, Burlington, MA, Unites States), a cDNA library was prepared from 1 µg of 293T total RNA. Briefly, poly(A) mRNA was isolated by oligo (dT) beads and fragmented with divalent cations at high temperature, before cDNA synthesis. cDNA was tailed with dA, followed by T-A ligation to add adapters, which was then amplified by PCR and loaded onto an Illumina HiSeq/Illumina Novaseq/MGI2000 instrument for sequencing with a 2 × 150 paired-end configuration, according to the manufacturer’s instructions (Illumina, San Diego, CA, United States). Raw sequence data were processed by Cutadapt (V1.9.1), and the clean data aligned to the human reference genome GRCh38 with Hisat2 (v2.0.1). Gene expression levels were obtained using HTSeq (v0.6.1). RNA-seq data have been deposited in the Gene Expression Omnibus under the accession number GSE243063. Student’s t-test was performed to identify differentially expressed genes from three biological replicates (p < 0.05), and DAVID (https://david.ncifcrf.gov) (Sherman et al., 2022) used to analyze differences in GO (Gene Ontology) TERM-BP (biological processes)-Direct (GO_BP_Direct for short) and KEGG (Kyoto Encyclopedia of Genes and Genomes) pathways.

Cell lysis, protein extraction, and Western blotting were performed as described (Zhang et al., 2013). The primary FLAG antibody was from Sigma (dilution 1:5,000, catalog number F3165, St Louis, MO, United States), GAPDH antibody from Solarbio (dilution 1:2000, catalog number K200057M, Beijing, China), and BNIP3L antibody from Beyotime (dilution 1:2000, catalog number AF63333). For reporter assays, 293T cells were transfected with approximately 10 ng of a firefly luc reporter plasmid, 5 ng of pRL-CMV, and 400 ng of a control or miR-210 mimic RNA. Two days later, the cells were lyzed, and firefly and Renilla luc activities measured using the Dual-Luciferase Assay System (Promega).

GraphPad Prism 7.0 (GraphPad Software, San Diego, CA, United States) was used to perform the two-sided, two-sample unequal variance Student’s t-test and Dunnett’s test, with p-value <0.05 considered statistically significant.

To target and disrupt the human miR-210 gene, we designed an sgRNA as depicted in Figure 1A and constructed the vector LentiCRISPR-miR-210 (MATERIALS AND METHODS). We then transfected 293T cells with LentiCRISPR-miR-210, which co-expressed the sgRNA and Cas9 protein. Controls (WT) were transfected with the empty vector LentiCRISPR DNA. Clonal cells were selected, and one WT clone and three KO clones (#1, 2, and 3) were acquired and amplified. The loss of miR-210 was identified and confirmed by PCR, DNA sequencing, and qPCR. At the miR-210 gene locus KO#1 and 2 had slight differences in the DNA deletion, while KO#3 was the same as KO#1 (Figure 1B). And qPCR results verified all three had no discernible miR-210 RNA expression (Figure 1C).

The CRISPR/Cas9 system can induce off-target mutations. To further characterize our KO cells, we obtained the top 11 potential off-target sites in the human genome based on the sgRNA sequence (Supplementary Table S1). Genomic DNA for these sites was amplified and sequenced. The results revealed that none of the KO cell lines contained mutations in those loci (Supplementary Table S2).

Considering the relationship between miR-210 and hypoxia response, we next measured miR-210 expression under hypoxia. Hypoxia response can be induced by placing cells in a low oxygen chamber or mimicked by treating cells with DMOG, a classic drug that stabilizes HIF-1ɑ (Jaakkola et al., 2001). Both methods significantly elevated miR-210 levels in the WT 293T cells (Figure 1D). For simplicity we would use DMOG to mimic hypoxia thereafter. In the KO cells, as expected, miR-210 expression is undetectable under both the normoxic and hypoxic conditions (Figures 1C,D). Since all three KO cell lines (#1, 2, and 3) exhibited consistent results (Figures 1B–D), we would use two of them (#1 and 2) for subsequent experiments.

To understand the functions of miR-210, we then investigated the impact of miR-210 KO on 293T cell behaviors, including growth, migration, apoptosis, and metabolism. We first used MTT assays to measure cell number increase, or cell growth. Under normal oxygen conditions, KO cells exhibited slower growth than the WT cells (Figure 2A). DMOG-induced hypoxia responses hindered cell proliferation in both the WT and KO cells (Figures 2B, C). We also examined their cell cycle profiles by flow cytometry. Compared to the WT cells, more KO cells were at the S phase but fewer at the G2/M phase (30.7% G2/M for the WT, 26.8% and 25.7% for KO, Figure 3A), while DMOG appreciably reduced WT cells in the G2/M phase (Figure 3B). These results were consistent with the MTT data suggesting that KO and DMOG both reduced cell numbers.

Next we used scratch assays to study cell migration, which is important for biological processes such as development, wound healing, angiogenesis, and metastasis. Under serum-free growth conditions, KO cells migrated slower into the gap region than the WT cells with both DMSO treatment (Figure 4A) and DMOG treatment (Figure 4B). Comparisons of the quantifications are presented in Figure 4C, also showing that DMOG/hypoxia stimulated migration in both the WT and KO cells.

We then examined apoptosis in 293T cells by Annexin V staining and flow cytometry. High Annexin V signals identified apoptotic cells, and we found that miR-210 KO cells had a higher apoptotic ratio compared to the WT cells (Figure 5A), and DMOG increased apoptosis in both the WT and KO cells (Figure 5B). As mitochondrial functions and metabolism are closely linked to oxygen and hypoxia as well as to apoptosis, we also measured the intracellular ROS levels and relative mitochondrial membrane potential. We found that KO cells had higher ROS content (Figure 6A), whereas DMOG treatment enhanced ROS in both the WT and KO cells (Figures 6B, C). Compared to WT cells, KO cells had lower mitochondrial membrane potential, and DMOG led to a further drop in the membrane potential in all cells (Figure 6D).

As shown above, miR-210 KO perturbed a wide range of phenotypes in 293T cells. To begin to unravel the underlying mechanisms, we performed poly(A) RNA-seq on miR-210 WT and KO cells under control and DMOG treatment conditions. Differentially expressed genes were obtained from three independent experiments and then subject to GO and KEGG pathway analyses. We first compared the KO vs. WT cells under normal growth conditions (Table 1). The top GO term differences prominently featured transcription and DNA replication, and the top KEGG pathway differences included metabolic pathways and DNA replication. The DNA replication and metabolic pathways findings were in line with the cell growth and cell cycle phenotypes we observed in Figures 2, 3 as well as the putative link between miR-210/hypoxia response and metabolism.

Secondly, we compared normal vs. DMOG, i.e., hypoxia-mimicking conditions. Data from the WT cells yielded GO terms implicating the development of multiple organs and wound healing, all known associations with hypoxia responses (Table 2), but few KEGG terms, likely because there were only three WT replicate samples. Thus we also looked at the more numerous KO samples, since hypoxia response was maintained in KO cells as well (see below and DISCUSSION). This analysis demonstrated that hypoxia led to changes in GO terms such as translation, metabolism, and DNA replication initiation (organ development also present but outside of the top 10), and KEGG including numerous metabolic pathways, the HIF-1 signaling pathway, and the cell cycle (Table 2).

Lastly, we examined changes in hypoxia responses, defined as the expression of a gene under DMOG treatment divided by that under DMSO treatment, in KO cells vs. WT cells. GO term differences included mitochondrial respiratory chain complex IV assembly, morphogenesis, response to wounding, and kidney development, and KEGG included a large number of metabolic pathways and apoptosis (Table 3). These processes are likely the ones impacted by miR-210 during hypoxia.

Animal miRNAs bind to complementary sequences in target mRNAs to repress the latter’s expression (Bartel, 2018), so miR-210 KO was expected to result in increased expression of miR-210 target genes and mRNAs. To identify these target genes we focused on mRNAs that were expressed higher in the KO cells than WT cells. There were 955 such genes (p < 0.05; Supplementary Table S3). We then compared them with the miRNA target prediction site miRWalk (Sticht et al., 2018) and the experimentally validated miRNA target site miRTarBase (Huang et al., 2022). Of the 955 genes, 714 were predicted by miRWalk (Figure 7A; Supplementary Table S3). GO terms of these genes included cell migration, angiogenesis, negative regulation of cell proliferation, and response to wounding, and many KEGG pathways were related to adhesion, human cancers and diabetes (Table 4), again consistent with KO cell phenotypes (Figures 2–4) and the presumed functions of miR-210 in disease and metastasis. On the other hand, there were 156 validated mouse and human miR-210 targets in miRTarBase, but only 9 were upregulated in KO cells (Figure 7A; Supplementary Table S3). Since many reported miR-210 target genes, such as EFNA3, HOXA1, PTP1B, PLK1, CDC25B, CCNF, BUB1B, MNT, FAM83D, E2F3, CASP8AP2, AIFM3, BNIP3, SIN3A, and ISCU have important functions (INTRODUCTION), we checked their changes individually and found that only PTP1B, CDC25B, and FAM83D significantly increased mRNAs upon miR-210 KO in 293T cells (Supplementary Table S3). For the other genes we selected E2F3, HOXA1, ISCU, and MNT for additional qPCR verification and confirmed they were not upregulated (Figure 7B). Thus, miR-210 KO did not enhance mRNA levels of the vast majority of reported miR-210 target genes, at least in 293T cells.

The major phenotypes of miR-210 KO cells were reduced cell proliferation and enhanced apoptosis (Figures 2, 3, 5), with ROS and mitochondrial membrane potential changes potentially related to or associated with such cellular phenotypes as well (Figure 6). Gene expression, GO, and KEGG pathway analyses supported these observations, although the mechanisms required experimental validation. As the first step we focused on understanding how miR-210 suppressed apoptosis in 293T cells. We noticed that rat BNIP3, a pro-apoptotic member of the Bcl-2 family (Field and Gordon, 2022), is induced by hypoxia and has been reported to be a miR-210 target (Wang et al., 2013). BNIP3 mRNA was increased in our KO cells, although the change was not statistically significant (p > 0.05), but the mRNA increase of BNIP3L or Nix, a BNIP3 relative (Field and Gordon, 2022), was significant, and BNIP3L is predicted by miRWalk as a miR-210 target (Supplementary Table S3). So we performed qPCR experiments, which confirmed that BNIP3L mRNA was indeed elevated in KO cells (Figure 8A). DMOG treatment increased BNIP3L mRNA levels in both the WT and KO cells (Figure 8A), consistent with the report of BNIP3L responding to hypoxia (Bruick, 2000).

We next examined BNIP3L protein expression by Western blotting (Figure 8B). BNIP3L ran as a 40kD protein on the blot, with a 30kD band probably representing either an isoform or BNIP3 (Field and Gordon, 2022). KO#1 and KO#2 cells expressed a higher level of BNIP3L protein, compared to the WT cells (Figure 8B, lanes 1-3). Figure 8B further shows that in WT cells, DMOG increased BNIP3L protein levels (compare lanes 1 and 4), and under these same conditions and compared to WT cells, KO cells still had higher BNIP3L protein expression (compare lanes 5 and 6 with 2 and 3). Unusually, the predicted miR-210 binding site is in the coding region of BNIP3L mRNA (Figure 8C). To test if miR-210 can inhibit BNIP3L expression, we first employed a conventional reporter assay. As shown in Figure 8D, the expression of a luc reporter containing WT BNIP3L sequence at its 3′ untranslated region was reduced approximately 60% by miR-210 in co-transfected 293T cells, while a mutant (MT) BNIP3L reporter was not significantly repressed by miR-210. Next, to further confirm that miR-210 was able to directly suppress BNIP3L expression, we constructed a FLAG-tagged BNIP3L and its mutant version with three nucleotide mutations that code for the same amino acids but should weaken the binding to miR-210 (Figure 8C). Cotransfecting miR-210 with the WT plasmid led to approximately 50% reduction in FLAG-BNIP3L protein expression (Figure 8E, compare lanes 2 and 3). miR-210 had no effect on the MT protein expression (Figure 8E, compare lanes 4 and 5). A model was proposed and presented in Figure 8F to illustrate how miR-210 through BNIP3L might counteract the effects of hypoxia and HIF-1α on apoptosis.