Deletion of the fih gene encoding an inhibitor of hypoxia-inducible factors increases hypoxia tolerance in zebrafish

Many aerobic organisms have developed molecular mechanism to tolerate hypoxia, but the specifics of these mechanisms remain poorly understood. It is important to develop genetic methods that confer increased hypoxia tolerance to intensively farmed aquatic species, as these are maintained in environments with limited available oxygen. As an asparaginyl hydroxylase of hypoxia-inducible factors (HIFs), factor inhibiting HIF (FIH) inhibits transcriptional activation of hypoxia-inducible genes by blocking the association of HIFs with the transcriptional coactivators CREB-binding protein (CBP) and p300. Therefore, here we sought to test whether fih is involved in regulating hypoxia tolerance in the commonly used zebrafish model. Overexpressing the zebrafish fih gene in epithelioma papulosum cyprini (EPC) cells and embryos, we found that fih inhibits the transcriptional activation of zebrafish HIF-α proteins. Using CRISPR/Cas9 to obtain fih-null zebrafish mutants, we noted that the fih deletion makes zebrafish more tolerant of hypoxic conditions than their WT siblings, but does not result in oxygen consumption rates that significantly differ from those of WT fish. Of note, we identified fewer apoptotic cells in adult fih-null zebrafish brains and in fih-null embryos, possibly explaining why the fih-null mutant had greater hypoxia tolerance than the WT. Moreover, the fih deletion up-regulated several hypoxia-inducible genes in fih-null zebrafish exposed to hypoxia. The findings of our study suggest that fih plays a role in hypoxia tolerance by affecting the rate of cellular apoptosis in zebrafish.

Aerobic organisms use oxygen (O 2 ) to survive. Over evolutionary time, aerobic organisms have developed sophisticated cellular mechanisms that sense and respond to O 2 gradients, as well as physiological systems that adapt to changes in these gradients (1)(2)(3)(4)(5)(6). As consistent oxygen supply affects evolution, studies of hypoxia adaptation (chronic hypoxia) are common (5,7,8). However, organisms often encounter acute hypoxic conditions in addition to chronic hypoxia (5,(7)(8)(9), yet organis-mal mechanisms of adaptation to chronic hypoxia (i.e. hypoxia tolerance) remain largely unknown. Recent work in our laboratory has shown that hypoxia tolerance involves the hypoxiainducible factor (HIF) 2 signaling pathway, but the molecular mechanisms involved require further exploration (10,11).
FIH is a second category of hydroxylase that catalyzes asparagine hydroxylation, blocking association of the HIF-1/2␣ transcriptional factors and the CBP/p300 transcriptional coactivators, and thus results in inhibition of HIF-1/2␣ transcriptional activity (17,20,21). Intriguingly, the loss of the Fih gene in mice has little or no discernable impact on the classical aspects of HIF function, but does cause changes to glucose and lipid homeostasis mechanisms (22). Whether FIH is involved in hypoxia adaptation or tolerance remains unclear.
The effects of chronic hypoxia are of great interest to the aquaculture industry because intensive cultivation often leads to water pollution, limiting available oxygen (23,24). Therefore, the development of fish strains tolerant of hypoxia conditions, through either hybridization or genetic manipulation, is critical (25). Here, we aimed to test whether fih was involved in the tolerance of hypoxic conditions by zebrafish (Danio rerio). We first tested the effects of fih overexpression on the transcriptional activity of zebrafish hif genes. We then knocked out fih in zebrafish using CRISPR/Cas9. We found that fih-null zebrafish were more tolerant to hypoxia than their WT siblings, indicating that fih is critical for adaptation to acute hypoxia.

fih-null zebrafish are more tolerant of hypoxia
Using CRISPR/Cas9, we knocked out fih in zebrafish to obtain two mutant lines (M1, fih ihb20170818/ihb20170818 ; M2, fih ihb20170819/ihb20170819 (Fig. S5, A and B). Our semiquantitative RT-PCR assays showed that fih mRNA was reduced in both mutant lines as compared with the WT (Fig. S5, C and D), indicating that fih had been successfully knocked out in the two mutant lines. Our observations indicated that fih-null zebrafish developed normally and were generally indistinguishable from

Deletion of fih increases hypoxia tolerance
their WT siblings. Moreover, the protein levels of hif-1a and hif-2a were enhanced in fih-null zebrafish brains under hypoxia (5% O 2 ), as revealed by Western blotting assays (Fig. S6), further validating loss of fih function in fih-null zebrafish.
We next tested whether adult mutant zebrafish (3 months postfertilization (mpf)) were affected by hypoxia differently from their WT siblings. After a 4-h exposure to 5% O 2 , WT zebrafish were concentrated near the surface of the water, whereas fih-null zebrafish were distributed throughout the water volume (Fig. 4A). After 6 h, the WT zebrafish began dying, but no obvious abnormalities were observed in fih-null zebrafish (Fig. 4B). Between 8 and 14 h of hypoxia, all WT zebrafish died (Fig. 4, C-F). At 14 h, all fih-null zebrafish remained alive; also, they were swimming closer to the surface of the water (Fig. 4, C-F). To quantify these observations, we calculated a survival curve for zebrafish larvae (3 days postfertilization (dpf)) and found that under hypoxic conditions (2% O 2 ), fewer fih-null zebrafish larvae died than their WT siblings (Fig. 4G).
We then determined whether the observed differences between WT and fih-null zebrafish were caused by differences in rate of oxygen consumption by comparing the oxygen consumption rate between adult WT and fih-null zebrafish. We found no significant differences in oxygen consumption (Fig.  S7). Our data therefore suggest that fih-null zebrafish are more tolerant of hypoxic conditions than WT zebrafish.

fih-null zebrafish have fewer apoptotic cells under hypoxia
To investigate why fih-null zebrafish larvae were more tolerant of hypoxia than their WT siblings, we used Acridine orange staining to detect apoptotic cells in zebrafish larvae. Under normoxia, there were no differences in the numbers of apoptotic cells between WT and fih-null zebrafish larvae (Fig. 5A). However, under hypoxic conditions (2% O 2 ), significantly more apoptotic cells were observed in the WT zebrafish larvae as compared with the fih-null larvae (p Ͻ 0.005 for 12-60 h of exposure; p Ͻ 0.05 for 72 h of exposure; Fig. 5, B and C).
We also counted the apoptotic brain cells in adult (3 mpf) fih-null zebrafish and their WT siblings subjected to hypoxic conditions (5% O 2 ) using terminal deoxynucleotidyltransferase dUTP nick-end labeling (TUNEL) assays. After 6 h of hypoxia (5% O 2 ), there were significantly more apoptotic cells in the WT zebrafish brains than in the fih-null zebrafish brains (Fig.  6). Therefore, these data suggested that the hypoxic resistance of the fih-null zebrafish might be due to lower rates of hypoxiainduced apoptosis.

Deletion of fih increases hypoxia tolerance
Taken together, these data suggest that the hypoxia-resistant phenotype exhibited in fih-null zebrafish might be due to their resistance to hypoxia-induced cell apoptosis or autophagy.

Loss of fih in zebrafish leads to the up-regulation of hypoxia-inducible genes under hypoxia
We examined the expression of six well-defined hypoxiainducible genes (vegf, cited2, pail, pdk1, il11a, and ldha) (10,11,26,37) in adult (3 mpf) WT and fih-null zebrafish under hypoxia (5% O 2 ). Our semiquantitative RT-PCR assay showed that all tested genes were up-regulated in fih-null zebrafish (M1) as compared with WT zebrafish under hypoxia (Fig. 7, A-E). Our tests with M2 returned similar results (Fig. S9), eliminating the possibility that our results were affected by CRISPR/ Cas9 off-targeting. Our data therefore suggested that the knockout of fih increased the transcription of HIF, possibly accounting for the observed increased tolerance of hypoxia observed in fih-null zebrafish.

Loss of fih in zebrafish causes improved glucose homeostasis
To determine the effect of fih deletion on glucose physiology, blood glucose was measured (Fig. S10). Loss of fih had an effect on glucose clearance, similar to what was exhibited in fih-null mice (22). These data suggest that fih plays a role in glycolysis through modulating HIF-␣'s function.

Discussion
Fish tend to have multiple copies of genes that are present as single copies in mammals, possibly due to the complexities of adapting to various aquatic environments (38). For example, zebrafish have two copies of each of the homologs to mammalian HIF-1␣ and HIF-2␣ (39). Consistent with previous

Deletion of fih increases hypoxia tolerance
research, we found that these copies (hif-1␣a/b and hif-2␣a/b) do indeed have some redundant functions; all four genes induce hypoxia-inducible promoter activity (39). Surprisingly, however, we found that the overexpression of fih had different effects on the copies; overexpression of fih down-regulated hif-1␣a (hif-1␣a N694A) and hif-2␣a (hif-2␣a N811A), but not hif-1␣b (hif-1␣b N754A) and hif-2␣b (hif-2␣b N786A). It is possible that one copy of hif-␣ (hif-1␣b or hif-2␣b) is strictly regulated by fih hydroxylation, similar to mammals (17), but the other copy (hif-1␣a or hif-2␣a) is not. Alternatively, asparagine sites other than the one that is evolutionarily conserved in hif-1␣a and hif-2␣a might be hydroxylated by fih. Further investigation of the differences between these gene copies is necessary to better understand the physiological functions of zebrafish hif-1␣ and hif-2␣.
Our results indicated that fih-null zebrafish appeared to develop and reproduce normally but had an increased tolerance to hypoxia compared with WT zebrafish. Therefore, fih might be a good candidate for gene editing in intensively farmed economically important fish species.
Previously, we found that hypoxia tolerance was reduced in foxo3b-null and tet1-null zebrafish, as compared with WT zebrafish (10,11). However, the two cellular mechanisms varied dramatically with respect to hypoxia signaling; foxo3b inhibits hypoxia signaling by up-regulating vhl expression, whereas tet1 enhances hypoxia signaling by stabilizing HIF-␣ proteins (10,11). Here, fih suppressed the hypoxia signaling pathway, and hypoxia tolerance was increased in fih-null zebrafish as compared with WT zebrafish. Indeed, it is difficult to predict which gene deletions will increase hypoxia tolerance in zebrafish based on whether or not a given gene suppresses hypoxia signaling. However, in all cases, it appears that sufficient HIF-␣ is required for hypoxia tolerance.
Significantly, more apoptotic cells were detected in WT zebrafish than in the fih-null zebrafish after exposure to

Deletion of fih increases hypoxia tolerance
hypoxic conditions, possibly explaining the greater hypoxic sensitivity of WT zebrafish. It has been shown that hypoxia activates several genes that induce apoptosis or autophagy (31,40,41). Here, although several hypoxia-inducible genes were up-regulated by hypoxia in fih-null zebrafish, this did not lead to reduced hypoxia-induced apoptosis. Indeed, the mechanisms causing the decreased rate of apoptosis in the fih-null zebrafish remain unclear. Interestingly, the autophagy-related genes were diminished in fih-null zebrafish after exposure to hypoxic conditions. Therefore, it appears that decreased expression of autophagy-related genes might also contribute to hypoxia tolerance of fih-null zebrafish, at least in part. Further investigation of molecules downstream of HIF-␣ that are involved in hypoxia tolerance are necessary to understand the mechanisms underlying the hypoxia signaling pathway in zebrafish.
Apart from HIF-␣ proteins, other substrates of FIH have been reported (42)(43)(44), which might also contribute to hypoxia tolerance of fih-null zebrafish. Notably, as a downstream target of HIF-␣, bnip3 was reduced instead of increased in fih-null zebrafish, implying that substrates of fih other than HIF-␣ might be involved in regulating bnip3 expression under hypoxia. Identification of other substrates that could potentially contribute to the phenotypes displayed in fih-null zebrafish might help to understand the function of fih in vivo thoroughly.

Cell culture and transfection
EPC cells (originally obtained from the American Type Culture Collection) were cultured in 24-well plates in medium 199 (Invitrogen) supplemented with 10% fetal bovine serum and maintained at 28°C in a humidified incubator containing 5% CO 2 . EPC cells were transfected with the constructed plasmids using VigoFect (Vigorous Biotechnology, Beijing, China) following the manufacturer's instructions. pTK-Renilla (Promega) was used as an internal control. After transfection, luciferase activity was measured with the Dual-Luciferase reporter assay system (Promega).

Embryo preparation and luciferase reporter assays in embryos
Zebrafish were maintained in a recirculating water system under standard conditions, and spawning was inducted following standard protocols. All procedures involving zebrafish were approved by the institutional animal care and use committee of the Institute of Hydrobiology, Chinese Academy of Sciences (Wuhan, Hubei, China).

Deletion of fih increases hypoxia tolerance
We injected one-cell stage zebrafish embryos (generated as described above) with 0.75-1.25 ng of Cas9 RNA and 0.075 ng of sgRNA per embryo. The mutations were initially detected using a heteroduplex mobility assay (HMA) as described previously (45). If the HMA results were positive, the remaining embryos were raised up to adulthood as the F0 generation and were then back-crossed with WT zebrafish (strain AB) to generate the F1 generation. F1s were initially genotyped with HMA. Genotypes were confirmed by sequencing target sites. Heterozygous F1s were back-crossed with WT zebrafish (strain AB; disallowing offspring-parent matings) to generate the F2 generation. F2 adults carrying the target mutation were intercrossed to generate F3 offspring. The F3 generation contained WT (ϩ/ϩ), heterozygote (ϩ/Ϫ), and homozygote (Ϫ/Ϫ) individuals. The primers used to identify mutants were 5Ј-TCAC-GCTGCTTGTATGACTG-3Ј and 5Ј-CGGTATTTGTC-GGGTTGGAA-3Ј.

Hypoxia tolerance of zebrafish adults and larvae
A Ruskinn Invivo2 I-400 work station was used to induce hypoxia. During preliminary experiments, we noticed that the zebrafish body weight significantly affected hypoxia tolerance. Therefore, we selected adult zebrafish (3 mpf) with similar body weights (0.26 Ϯ 0.02 g) for tests of hypoxia tolerance. We added three fih-null zebrafish and three WT sibling controls into separate 250-ml Erlenmeyer flasks containing 250 ml of water. The oxygen concentration in the Ruskinn Invivo2 I-400 work station was set to 5% (5.06 kilopascals). Both flasks were then placed in the Ruskinn Invivo2 I-400 work station simultaneously. The behavior of the zebrafish in each flask was closely monitored over 14 h. This experiment was performed three times.
To obtain the survival curve of zebrafish larvae (3 dpf) subjected to hypoxia, the oxygen concentration in the Ruskinn Invivo2 I-400 work station was set to 2% instead of 5%. During preliminary experiments, we found that zebrafish larvae (3 dpf) were more tolerant of hypoxic conditions than adults. Under 5% oxygen, zebrafish larvae began to die after staying more than 36 h. If zebrafish larvae were to stay under hypoxic conditions for too long, this might cause some nonspecific phenotypes unrelated to hypoxia tolerance. Therefore, we used 2% oxygen to treat zebrafish larvae. We added fih-null zebrafish larvae (3 dpf) and WT sibling control (3 dpf) into the Ruskinn Invivo2 I-400 work station simultaneously. We counted the numbers of dead larvae every 12 h. This experiment was performed three times. The oxygen concentrations in water were 7.36 Ϯ 0.02, 5.78 Ϯ 0.06, or 4.38 Ϯ 0.07 mg/liter when the oxygen concentrations of the work station were set to 21, 5, or 2% (28°C), respectively.

Oxygen consumption of adult zebrafish
We measured zebrafish oxygen consumption in 12 250-ml flasks, each containing 250 ml of water. The initial oxygen concentration of the water in each flask was measured with an LDO101 probe (HQ30d, HACH) (7.70 Ϯ 0.07 mg/liter). We selected 12 adult zebrafish (six fih-null and six WT siblings) of approximately the same weight for this experiment. We placed one fih-null or one WT zebrafish in each of the 12 flasks. The flasks were then tightly sealed. After 4 h, we measured the oxygen concentration in three flasks containing fih-null zebrafish and three flasks containing WT zebrafish siblings with the LDO101 probe. After an additional 4 h, we measured the oxygen concentration in the remaining six flasks with the LDO101 probe individually.

Apoptotic cell detection in embryos
We used Acridine orange staining to detect dying cells in embryos (48 hpf) under hypoxia (2% O 2 ) and normoxic conditions, as described previously (40). We photographed the stained embryos using a Leica M205 FA fluorescent dissection microscope and used ImageJ software to quantify the number of dying cells in the photographs. We counted all dying cells within the same square for each group of embryos, and three equal-sized squares were counted per group. This experiment was performed three times.

TUNEL staining
We counted the apoptotic cells in the brains of fih-null zebrafish and their WT siblings subjected to hypoxia (5% O 2 ) with a TUNEL assay using the Apoptag Peroxidase In Situ Apoptosis Detection Kit (Millipore) following the manufacturer's instructions. We compared the apoptotic cell ratio (number of apoptotic brain cells/total number of brain cells) between fihnull and WT zebrafish. We performed three independent assays.

Blood glucose measurement
Fish were anesthetized, and whole blood was collected as described previously (46). Blood glucose was measured by a One Touch Ultra Glucose Meter (Life Scan). For the postprandial blood glucose measurement, fish were fully fed for 15 min and then were put into freshwater for 5 h.

Statistical analysis
We compared mean relative luciferase activity, mean gene expression level, and mean apoptotic cell ratios using unpaired Student's t tests in GraphPad Prism version 5. We considered p Ͻ 0.05 statistically significant.
Author contributions-W. X. designed the study and wrote the manuscript. X. C. and D. Z. designed the study, conducted the experiments and analyzed the data. J. W., X. L., and G. O. contributed the reagents. All authors analyzed the results and approved the final version of the manuscript.