Kinetic Investigations of the Role of Factor Inhibiting Hypoxia-inducible Factor (FIH) as an Oxygen Sensor

Background: The hypoxia-inducible factor (HIF) hydroxylases (FIH and PHDs) regulate hypoxia sensing in animals. Results: FIH·HIF-α reacts faster with O2 than PHD2·HIF-α but slower than FIH·ankyrin complexes. Conclusion: The kinetics of catalysis by isolated FIH and PHD2 reflect their cellular activities. Significance: We provide a kinetic rationale for cellular observations that FIH activity is more hypoxia-tolerant than that of the PHDs.

The chronic hypoxic response in animals is mediated by the heterodimeric (␣/␤) hypoxia-inducible transcriptional factor (HIF), 3 accumulation of which triggers up-regulation of an array of genes, including those encoding for erythropoietin, vascular endothelial growth factor (VEGF), vasomotor regulators, and glycolytic enzymes (reviewed in Refs. [1][2][3][4]. The levels and activity of HIF isoforms in cells are regulated by post-translational hydroxylation of their HIF-␣ domains, as catalyzed by prolyl hydroxylase domain enzymes (PHD1-3, also known as EGLN1-3) and factor inhibiting HIF (FIH), collectively termed the HIF hydroxylases (5). The PHDs catalyze hydroxylation of one or both of two proline residues in the O 2 -dependent degradation region of HIF-␣ (C-terminal oxygen-dependent degradation domain (CODD) and N-terminal oxygen-dependent degradation domain (NODD)), resulting in its targeting to the proteasome (2,6). FIH catalyzes hydroxylation of an asparaginyl residue in the HIF-␣ C-terminal transactivation domain (CAD), which leads to reduced interaction of HIF with the transcriptional co-activator proteins CBP/p300 (CREB-binding protein/p300) (7)(8)(9). Together with the PHDs, FIH is proposed to act as an O 2 sensor in cells, providing a direct link between cellular O 2 concentrations and the hypoxic response (5,10,11).
Although FIH was first identified as a HIF hydroxylase, it has subsequently been shown to hydroxylate semiconserved Asn residues in many proteins containing ankyrin repeat domains (ARDs) (18). ARDs comprise a conserved sequence of 33 residues and are extensively involved in protein-protein interactions (reviewed in Ref. 19). The range of ARD-containing substrates for FIH is apparently very diverse (19), including tankyrase, NF-B (nuclear factor -light chain enhancer of activated B cells, a protein complex that controls transcription of DNA), Notch-1 (neurogenic locus notch homolog protein, a transmembrane protein involved in signaling), MYPT-1 (the myosin phosphatase targeting protein, a regulatory subunit of protein phosphatase 1), and vanilloid 3 channel (hydroxylation of which abolishes activity (20)). In contrast to the "switchlike" signaling effect of CAD hydroxylation on HIF activity, no such well defined role for FIH-catalyzed ARD hydroxylation has yet been defined. In some but probably not all cases, Asn hydroxylation has been observed to stabilize the ARD protein fold (21)(22)(23)(24). Competition between ARDs and HIF-␣ isoforms for FIH is proposed to tune the hypoxic response and has the potential to enable memory of hypoxic events (25). Recombinant FIH has been shown to be capable of catalyzing hydroxylation of residues besides Asn in the context of ARDs (e.g. His, Leu, or Ser residues) (26), with some of these reactions having been shown to occur in human cells (27), demonstrating that FIH is a highly promiscuous enzyme.
The HIF hydroxylases are Fe(II)/2-oxoglutarate (2OG)-dependent oxygenases, a family of enzymes with wide-ranging biological roles (28). The 2OG oxygenases have a conserved double-stranded ␤-helix fold that supports an Fe(II)-coordinating HX(D/E)X n H catalytic triad and employ a conserved consensus oxidation mechanism (29,30): formation of an enzyme⅐ Fe(II)⅐2OG⅐substrate ternary complex allows O 2 activation followed by 2-electron oxidation of 2OG coupled to oxidation of the substrate (Fig. 1). A substantial body of structural, kinetic, and spectroscopic evidence supports this consensus mechanism (29,(31)(32)(33)(34). Transient kinetics combined with spectroscopic techniques have provided evidence for intermediates (including the reactive Fe(IV)-oxo species (32)) and have been used to measure the kinetics of events following O 2 binding. For some 2OG oxygenases, O 2 binding is followed by rapid formation of the reactive intermediates (ϳϽ1 s) (e.g. for TauD (taurine dioxygenase) (32,35), vCPH (a viral collagen prolyl hydroxylase) (36), and DAOCS (deacetoxycephalosporin C synthase) (37). However, for PHD2, the most important of the human PHDs, we found that the rate of reaction subsequent to O 2 exposure was considerably slower than for TauD (ϳ100-fold) (38). Combined with the graded response of PHD2 to increasing O 2 concentrations in steady-state kinetic assays and also in cells (11,39), we have proposed that the apparently slow reaction of PHD2 with O 2 reflects its role as an O 2 sensor (38).
Although FIH is proposed to be an O 2 -sensing HIF hydroxylase, cellular and kinetic data suggest that FIH is less efficient at detecting deviations from normoxia than the PHDs. In cells, FIH is still substantially active under hypoxic conditions, whereas PHD activity drops significantly (39). The differences are apparently reflected in the lower reported K m app (O 2 ) values for FIH (ϳ90 -200 M for FIH compared with ϳ250 M to 1.7 mM for PHD2 (10, 40 -42)). Interestingly, the K m app (O 2 ) for FIH has also been reported to be substrate-dependent (43), with a lower K m app for O 2 for Notch-1 (ARD) hydroxylation than for HIF-1␣ hydroxylation, suggesting that FIH may be more active toward at least some ARDs than HIF-␣ substrates in hypoxia.
Because targeting the HIF hydroxylases for the treatment of hypoxia-related diseases is of medicinal interest (PHD inhibitors are presently undergoing large scale clinical trials), a detailed understanding of the kinetics of these enzymes with respect to O 2 is important to compare and rationalize the sensitivity of the HIF hydroxylases to hypoxia. Following our observation of the slow reaction of PHD2 with O 2 , we therefore conducted a detailed investigation of the kinetics of the reaction of FIH with O 2 . First, we investigated whether the slow reaction with O 2 (as observed for a PHD2⅐HIF-1␣ complex (38,42)) was conserved among O 2 -sensing HIF hydroxylases. Second, we investigated the effect of different substrates on the kinetics of the reactions of PHD2 and FIH with O 2 , in terms of HIF-1/2␣ isoforms, HIF hydroxylation sites for PHD2, and FIH with HIF versus ARD substrates. We describe spectroscopic, kinetic, and mechanistic evidence revealing that the reaction of PHD2 with O 2 is notably slow in the presence of all tested HIF substrates. We also report that the reaction of FIH with O 2 is significantly faster than for PHD2 in the presence of all HIF-1/2␣ substrates and to an even greater degree in the presence of all tested ARD substrates. The results are consistent with studies indicating that FIH is more active than the PHDs under hypoxic conditions (39) as well as the reported ability of FIH to preferentially hydroxylate ARD rather than HIF substrates (19,44). The work also demonstrates that, at least for PHD2 and FIH, the kinetic properties of these enzymes with respect to O 2 reflect their cellular hypoxia-sensing ability.
Enzyme Activity Assays and Studies of O 2 Dependence-Enzymatic assays were conducted as described (46,47). Unless otherwise stated, the assay mixture typically contained 5 M FIH, 500 M substrate, 500 M 2OG, 1 mM L-ascorbate, and 50 M Fe(NH 4 ) 2 (SO 4 ) 2 or 4 M PHD2, 100 M substrate, 300 M 2OG, 4 mM L-ascorbate, and 50 M Fe(NH 4 ) 2 (SO 4 ) 2 . For HIF-1␣ CODD 19-mer , the background levels of methionine oxidation resulting from assay conditions were subtracted based on non-enzymatic controls (48). Determination of initial velocities in steady-state kinetic experiments was performed varying the (co-)substrate concentrations and maintaining concentrations of components at levels higher than 2 ϫ K m app to ensure saturation conditions.
The O 2 dependence of FIH-catalyzed reactions was determined by conducting enzymatic assays in an "In vivo 500 Hypoxia Workstation" (Ruskinn) using buffers pre-equilibrated under different pO 2 values for 24 h. The O 2 dependence of PHD2-catalyzed reactions was determined by conducting enzymatic assays in sealed glass vials with variable N 2 /O 2 contents (42). Reactions were stopped by spotting samples onto a matrix-assisted laser desorption/ionization-time of flight mass spectrometry (MALDI-TOF-MS) target plate and immediately mixing with ␣-cyano-4-hydroxycinnamic acid matrix. Hydroxylation levels were assessed by MALDI-TOF-MS as described (46).
Stopped-flow UV-visible Spectroscopy Experiments-Stoppedflow UV-visible spectroscopy experiments were performed as described previously (32,42). Anerobically prepared assay mixtures containing 0.5 mM apo-FIH, 1 mM substrate, 5 mM 2OG, 0.4 mM Fe(II) or 0.8 mM apo-PHD2, 1 mM substrate, 5 mM 2OG, 0.5 mM Fe(II) were rapidly mixed with O 2 -free/O 2 -saturated buffer. The reaction was observed over 1000 s using a photodiode array detector, and kinetic traces were analyzed using Origin version 8.51 software. Data were fitted with a double exponential function (y ϭ A 1 ϫ exp(Ϫk rise t) ϩ A 2 ϫ exp(Ϫkfall t) ϩ y 0 ). The A 1 , A 2 , and y 0 parameters were restricted with upper limits to match the observed amplitude/offset changes.
Rapid Quench-Flow Experiments-Rapid quench-flow experiments were performed using the same conditions and preparation of the assay mixture as used in stopped-flow experiments. Rapid quench-flow equipment (TgK Scientific Ltd.) in an anaerobic glove box was used for the experiments. Samples were quenched with 1% CF 3 CO 2 H and analyzed using MALDI-TOF-MS (peptide hydroxylation) as described above or liquid chromatography-mass spectrometry (LC-MS, for succinate production). For LC-MS, chromatographic separation was performed at 50°C using a Waters ACQUITY BEH Amide 1.7-m, 2.1 ϫ 100-mm column on a Waters ACQUITY TM ultraperformance liquid chromatography system (Waters Corp., Milford, MA). The following eluents were used: mobile phase A: 10% H 2 O, 90% acetonitrile (v/v), 10 mM ammonium formate; mobile phase B: 50% H 2 O, 50% acetonitrile (v/v), 10 mM ammonium formate. The elution gradient was 0 -7.0 min linear from 10 to 30% B, and 7.0 -9.0 min at 10% B for re-equilibration of the column. A constant flow rate of 0.4 ml/min was used. Analytes were detected in negative ionization mode using single reaction monitoring on a Quattro triple quadrupole mass spectrometer (Waters) with a cone voltage of 15 V and a capillary voltage of 3.0 kV. The desolvation temperature was set to 250°C, and the source temperature was set to 120°C.
Intact protein LC-MS-Samples obtained from FIH kinetic assays with 2CA and 3CA were diluted to a substrate concentration of 5 M and then analyzed using a Waters LCT Classic mass spectrometer by electrospray ionization in positive ion mode. An Agilent HP1100 LC system was used, and samples were run through a Phenomenex Aeris Widepore 3.6-m XB-C8 150 ϫ 2.1-mm column at 0.5 ml/min in a gradient of phase A (95% H 2 O, 5% CH 3 CN, 0.01% formic acid) to phase B (95% CH 3 CN, 5% H 2 O, 0.01% formic acid). The LC/MS used the following method: 0 -1.5 min, 100% A and then 1.5-12 min linear gradient from 0 to 100% B, 12-14 min 100% B, and 15.0 -19.0 min 100% A for re-equilibration of the column. Data were collected and analyzed using MassLynx version 4.0.
CD Spectroscopy-CD spectra were recorded using a Chirascan CD/fluorimeter (Applied Photophysics) spectrometer. Measurements were carried out at 20°C in a 0.1 cm path length quartz cuvette, with protein concentration of 0.125 mg/ml in 5 mM sodium phosphate buffer (pH 8.0). Spectra were recorded in the wavelength range of 260 -190 nm at 0.5-nm intervals, and each spectrum represents an average of three scans. Spectra were baseline-corrected by subtraction of the buffer spectrum. Data were expressed as the mean residue ellipticity ([], degrees⅐cm 2 ⅐dmol Ϫ1 ). Thermal denaturation experiments were performed under the same conditions. The ellipticity at 220 nm ( 220 ) was monitored from 10 to 90°C.
MALDI-TOF/TOF-MS/MS-Hydroxylation sites were assigned unambiguously by MALDI-TOF/TOF mass spectrometry (Fig. 2) using a Bruker Ultraflex instrument, essentially as described previously (25). Thus, peptide samples were spotted onto an AnchorChip target plate with an ␣-cyano-4-hydroxycinnamic acid matrix using the dried droplet method. MS and MS/MS spectra were acquired by manual operation in reflectron mode, and the instrument was calibrated directly prior to data acquisition using monoisotopic peptide masses with Peptide Calibration Standard II (Bruker Daltonics, Coventry, UK). Sample ionization was achieved with a nitrogen laser (337 nm), and MS/MS spectra were generated by laser-induced dissociation (49). Assignment of hydroxylation sites in other peptides used in this study was performed previously (8,21,22,26,50).

Steady-state Kinetic Studies of FIH-and PHD2-catalyzed
Hydroxylation of HIF-1/2␣-As a prelude to studies on the reaction of PHD2 and FIH with O 2 , (apparent) K m values were determined for 2OG and HIF peptide substrates (HIF-1/2␣ CODD/NODD/CAD; Table 1) under standard conditions. Enzyme assays were conducted as described under "Experimental Procedures," and levels of peptide hydroxylation were analyzed using MALDI-TOF-MS. Differences were observed in the relative efficiencies of catalysis in overall time course assays; consistent with a previous report (46), PHD2 showed a higher level of activity toward CODD (70 -90% substrate hydroxylation after 15 min for both HIF-1␣ and HIF-2␣, respectively) than NODD (50 and 20% substrate hydroxylation for HIF-1␣ and HIF-2␣, respectively) ( Fig. 3, A and B). Under the standard conditions, FIH only hydroxylated 10 -20% of HIF-␣ 19-mer fragments; therefore, for further studies, 35-mer HIF-␣ CAD peptides were used. Requirements of longer peptides to observe more efficient FIH-catalyzed hydroxylation has been reported and is likely related to insufficient binding of short CAD peptides (10,40,51). Increasing the length of the HIF-1␣ CAD peptide resulted in nearly complete substrate hydroxylation, whereas analyses increasing the length of the HIF-2␣ CAD showed only a 10% increase in activity and Ͻ30% substrate hydroxylation (Fig. 3, C and D).
To subsequently study the O 2 dependence of the PHD2-and FIH-catalyzed reaction, the initial rates of hydroxylation were determined either in a hypoxic work station or by performing the reaction in sealed glass vials under N 2 /O 2 gas mixes with variable O 2 concentration (42). Half-saturation of O 2 for FIHcatalyzed hydroxylation of HIF-1␣ CAD peptides was achieved at 9 -11% O 2 , giving K m app (O 2 ) as 110 Ϯ 30 M, which was much lower than for PHD2-catalyzed hydroxylation of CODD/ NODD HIF-1/2␣ (K m app (O 2 ) Ͼ 400 M in all cases), consistent with previous reports (Fig. 4) (10,11,42). Notably, these results are consistent with previous K m app (O 2 ) estimates for FIH and are also supportive of the reported ability of FIH to retain activity to a greater degree of hypoxia than the PHDs in cells (39). Interestingly, despite the difference in final hydroxylation levels and steady-state product accumulation time courses, there was no difference in K m app (O 2 ) for HIF-1␣ CAD 19-and 35-mer peptides, suggesting no influence of substrate binding on the affinities of O 2 to an enzyme⅐substrate complex when saturating concentrations of the peptides are used (supplemental Fig. S4).

Pre-steady-state Kinetic Studies of the Reactions of PHD2⅐ Fe(II)⅐2OG⅐HIF-␣ CODD/NODD Substrates with O 2 -Previ-
ously reported pre-steady-state kinetic studies on PHD2 focused on its reaction with CODD HIF-1␣ 19-mer peptide (38,42); there is no reported information on the rates of PHD2 reaction with O 2 in the presence of NODD HIF-1␣ or HIF-2␣ substrates. To ascertain whether a slow reaction with O 2 under pre-steady-state conditions is common to PHD2-catalyzed hydroxylation of both CODD and NODD, we investigated the reaction of PHD2 with NODD substrates in comparison with CODD for both HIF-1␣ and HIF-2␣. Pre-steady-state reactions of PHD2 were initiated by rapid mixing of anaerobic PHD2⅐ Fe(II)⅐2OG⅐HIF-␣ complexes with O 2 -saturated buffer at 5°C. Reactions were initially observed using a photodiode array detector, and difference spectra (in comparison with the reaction with O 2 -free buffer) were subsequently analyzed.
The formation of an enzyme⅐substrate complex for 2OG oxygenases is characterized by a weak absorbance feature at 520 -530 nm due to metal ligand charge transfer interactions of Fe(II) and 2OG (32-34, 38, 53). Consistent with the consumption of the enzyme⅐substrate complex, and dependent on active PHD2⅐Fe(II)⅐2OG⅐substrate complex, a decay in absorbance at 520 nm was observed during the PHD2-catalyzed reaction, with the minimum reached at ϳ20 s, after which an absorbance increase was observed (Fig. 5, A and B). Although  To investigate the rate of the PHD-catalyzed reaction with longer HIF substrates, equivalent experiments were conducted with recombinant HIF-1␣(344 -503) NODD protein; these showed the same kinetics of PHD2⅐Fe(II)⅐2OG⅐NODD complex consumption as was observed in the presence of the 19-mer peptide (Fig. 5C), confirming that peptide binding is not ratelimiting. Consistent with a previous report (42), the formation and degradation of a transient species absorbing at max ϭ 310 nm was observed (in this and other pre-steady-state assays in this study; Fig. 5D); however, the slow rates of its formation suggest that (consistent with our previous observations (38,42)) an Fe(IV)-oxo intermediate ( max ϳ320 nm (32)) species probably does not accumulate during PHD2 catalysis.
To investigate correlation of the spectroscopic observations with rates of product accumulation, equivalent reactions were analyzed by rapid quench-flow methodology. Anerobically prepared PHD2⅐Fe(II)⅐2OG⅐peptide substrate complexes were mixed with O 2 -saturated buffer in a 1:1 ratio; reactions were quenched at defined time points with 1% aqueous CF 3 CO 2 H. The level of peptide hydroxylation was assessed by MALDI-TOF-MS (42), and succinate accumulation was analyzed by LC-MS. To obtain rate constants for product accumulation, the results were fitted with a single exponential function (Fig. 5E and Table 3). Besides NODD HIF-2␣ being hydroxylated at a faster rate than NODD HIF-1␣, our results showed no significant difference in the kinetics of hydroxylated product formation for PHD2 in the presence of any of the tested NODD and CODD substrates. These results suggest that the more rapid enzyme⅐substrate complex depletion observed for CODD HIF-2␣ did not affect the rate-limiting step in product formation. The time courses of succinate accumulation (reflecting 2OG turnover) were coupled to peptide hydroxylation within the limits of detection (Table 3 and supplemental Fig. S5).
Pre-steady-state Kinetic Studies of the Reactions of FIH⅐Fe(II)⅐ 2OG⅐HIF-1/2␣ CAD Substrates with O 2 -Although the presteady-state reaction of PHD2 with O 2 has been previously investigated (38,42), to date, no studies on the pre-steady-state kinetics of FIH have been reported. We carried out stoppedflow and rapid quench-flow experiments to compare the reaction of FIH with O 2 with that of PHD2 and to relate the results to the differences reported in K m app (O 2 ) values. Upon rapid mixing of anaerobic FIH⅐Fe(II)⅐2OG⅐HIF-␣ complexes with O 2 -saturated buffer at 5°C, a decay in absorbance in the 500 nm region and subsequent increase resulting in a minimum at ϳ0.3-3 s was observed in stopped-flow UV-visible absorbance experiments, as reported for PHD2 and other 2OG oxygenases ( Fig. 6A and Table 3) (32-34, 36, 53). Kinetic traces were fitted using a double exponential function; the fitting parameters of intermediate formation and decay are summarized in Table 3. The results reveal that the 500 nm decay occurs much more rapidly for FIH compared with PHD2: 4.5 Ϯ 0.3 s Ϫ1 for FIH with CAD HIF-1␣ 35 PHD2 with CODD HIF-1␣, suggesting a faster O 2 -initiated reaction for FIH (Fig. 6A).
The kinetic constant of product formation for the FIH⅐ Fe(II)⅐2OG⅐HIF-1␣ CAD 35-mer complex upon reaction with O 2 was determined to be 0.33 Ϯ 0.07 s Ϫ1 (Table 3 and Fig. 6B). This was faster than the kinetic constant of product formation for an FIH⅐Fe(II)⅐2OG⅐HIF-2␣ CAD 35-mer complex upon reaction with O 2 (0.12 Ϯ 0.01 s Ϫ1 (Table 3 and Table 2 and supplemental Fig. S3). Notably, the O 2 -initiated rates of product accumulation were significantly faster in the case of FIH compared with PHD2 ( Fig. 6B; rate constants, 0.33 Ϯ 0.07 s Ϫ1 for FIH with CAD HIF-1␣ 35-mer and 0.0155 Ϯ 0.0012 s Ϫ1 for PHD2 with CODD HIF-1␣), consistent with the differences in 500 nm decay rates observed by stoppedflow analyses and potentially related to the reduced K m app (O 2 ) values for FIH compared with PHD2.
During the stopped-flow analyses, we observed late formation of a transient species with maximum absorbance at 310 nm and a maximum accumulation time at 50 -100 s, as observed for PHD2 (Fig. 6C). Comparison with the O 2 -activated product formation studied by rapid quench-flow experiments revealed that this species is formed after the hydroxylation of the peptide substrates was complete, suggesting that a detectable Fe(IV)oxo intermediate does not accumulate during the FIH-catalyzed reaction. The observed accumulation of a 310-nm transient species can be rationalized by the formation of a transient iron species as a result of substrate-uncoupled 2OG turnover subsequent to consumption of all prime substrate and/or an oxidation of bound or unbound Fe(II) (42). Thus, similar to the observations for PHD2 (38,42), accumulation of an Fe(IV)-oxo intermediate was not observed (at 320 nm) in FIH catalysis, at least with our tested substrates.
Steady-state Kinetic Studies of FIH-catalyzed Hydroxylation of ARD Substrates-FIH catalyzes hydroxylation of multiple ARD substrates (18,19,21,23,43,44). It has been reported that the FIH K m app (O 2 ) value for ARD-containing proteins Notch1-3 is 10-fold lower than for HIF-1␣ (43), suggesting a relatively higher affinity of FIH for O 2 when catalyzing ARD hydroxylation, compared with HIF-␣ hydroxylation. We therefore wished to further investigate the kinetic dependence of FIH on O 2 with respect to ARD and HIF substrates. We compared FIH hydroxylation of a consensus ankyrin repeat 20-mer peptide (1CA 20-mer , designed based on ARD sequence analyses (21,54,55)) and a 20-mer fragment of tankyrase-1, tnkrs-1 20-mer (a multifunctional poly(ADP-ribose) polymerase containing 24 ARDs) to HIF-␣ isoforms by determining steady-state kinetic parameters for these substrates as a prelude to investigating O 2 dependence (43). Overall, the K m app values for ankyrin repeat peptide FIH substrates are all lower than for either HIF-1␣ or HIF-2␣ CAD peptide substrates (ϳ180 -300 M for HIF-1/2␣ CAD and Ͻ100 M for ankyrin substrates (Table 4); kinetic constants derived from data in supplemental Fig. S7), supporting the previously reported stronger binding of ARDs to FIH (43). At the same time, k cat app values for ankyrin peptides were higher than for HIF-␣ (1.6 Ϯ 0.1 and 0.56 Ϯ 0.04 s Ϫ1 for 1CA 20-mer and HIF-1␣ CAD 35-mer , respectively), implying more efficient turnover of the ankyrin substrates (Table 4 and supplemental Fig. S7), again consistent with previous reports (43).
ARD proteins probably must undergo at least partial unfolding from their consensus fold prior to productive FIH-catalyzed hydroxylation (21,44). To investigate the effect of ankyrin repeat folding on the efficiency of hydroxylation, 1CA 20-mer was compared with two-and three-consensus ankyrin repeat proteins (2CA and 3CA). The CA repeat sequence was designed previously based on bioinformatics studies of multiple natural ARD sequences (21,54,55). Circular dichroism experiments showed that whereas 2CA does not possess significant secondary structure (2CA is unfolded at Ͼ10°C), 3CA is folded in solution ( Fig. 7 and Refs. 21,22). The kinetic parameters for 1CA 20-mer , 2CA, and 3CA were determined. They showed a decrease in substrate and co-substrate K m app values for 2CA compared with 1CA 20-mer (K m app (substrate) ϭ 25 Ϯ 7 and 16 Ϯ 3 M, for 1CA and 2CA, respectively) in line with increased efficiency of longer substrates for FIH but an increase in the K m app value for the 3CA substrate (60 Ϯ 15 M (Table 4); kinetic constants derived from data in supplemental Fig. S8). Notably, there was a reduction in turnover rate for 3CA substrate. These results suggest that the requirement for at least partial unfolding of the ankyrin repeat prior to FIH-catalyzed hydroxylation affects catalytic efficiency (21,44).

Pre-steady-state Studies of the Reactions of FIH⅐Fe(II)⅐2OG⅐ ARD complexes with O 2 -
To further investigate the difference in the O 2 dependence of the FIH-catalyzed hydroxylation of HIF-␣ and ARD substrates, stopped-flow and rapid quenchflow experiments were performed. Interestingly, the determined pre-steady-state kinetic parameters suggested markedly faster O 2 -initiated reaction rates in the presence of ankyrin rather than HIF-␣ CAD substrates. Upon reaction of FIH⅐Fe(II)⅐2OG⅐1CA 20-mer and FIH⅐Fe(II)⅐2OG⅐tnkrs-1 20-mer complexes with O 2 , rate constants of product accumulation were determined to be 1.3 Ϯ 0.2 and 1.5 Ϯ 0.1 s Ϫ1 , respectively (i.e. ϳ10-fold faster than the reaction with HIF 35-mer substrates) (Fig. 9). Rates of succinate production were also determined and showed similar relative changes (determined rate constants are summarized in Table 5 and supplemental Fig. S6); for both HIF-␣ isoforms as well as ankyrin substrates, the reaction was fully coupled. In previously reported steady-state kinetic studies (43), when the length of the peptide substrate was increased from 20-mer to 34-mer for ankyrin peptides, a ϳ100-fold increase in substrate affinity was observed. O 2 initiation of the FIH reaction with a 35-mer tankyrase-1 peptide resulted in product formation at a rate ϳ7-fold faster than for the 20-mer (determined rate constants 1.5 Ϯ 0.1 s Ϫ1 for tnkrs-1 20-mer hydroxylation, 11 Ϯ 3 s Ϫ1 for tnkrs-1 35-mer hydroxylation). Consistent with faster hydroxylation, stopped-flow UVvisible absorbance experiments revealed that an absorbance minimum at 500 nm was reached faster for ankyrin substrates compared with HIF-␣ substrates (Fig. 10A). As for FIH-cata-lyzed HIF-␣ CAD hydroxylation, a transient species absorbing at 310 nm was observed subsequent to FIH-catalyzed ARD hydroxylation (Fig. 10B) (i.e. accumulation of an Fe(IV)-oxo intermediate was not observed (at 320 nm) in FIH catalysis).
To investigate the effects of folding of ankyrin repeats on the pre-steady-state kinetics of the reaction with O 2 , the presteady-state parameters of the reaction with 1CA were compared with those with 2CA and 3CA (Table 5). In contrast to the effect of lengthening of the tnkrs-1 substrate, increasing the length of the consensus ankyrin repeat substrate reduced the rate constant of product accumulation, particularly for the 3CA folded substrate (0.051 Ϯ 0.001 s Ϫ1 for 3CA compared with 1.3 Ϯ 0.2 s Ϫ1 for 1CA), implying that ankyrin repeat folding/unfolding may have an effect on the rate of the reaction with O 2 (Fig. 9 and Table 5). It is interesting to note that the 3CA substrate has both a significantly decreased rate of product accumulation and a significantly increased K m app (O 2 ) compared with the shorter ankyrin substrates.
UV-visible stopped-flow analyses showed that the rate of the absorbance decay at 500 nm decreased with increased number of consensus ankyrin repeats (Fig. 10C, 17.6 Ϯ 0.1, 11.9 Ϯ 0.2, and 10.5 Ϯ 0.1 s Ϫ1 for 1CA, 2CA, and 3CA, respectively). It is possible that the differences in FIH kinetics for the longer 3CA substrate may be due, at least in part, to the requirement for 3CA to unfold for effective catalysis (21,44). This unfolding may be necessary for productive access of O 2 to the active site.
It is notable that in the presence of the 3CA substrate, the rise in the absorbance at 500 nm was relatively fast (1.7 s Ϫ1 ), whereas the rate of product accumulation was slow (0.051 s Ϫ1 ). Whether or not this absorbance at 500 nm reflects reformation of an enzyme⅐substrate complex (32) or off-pathway accumulation of Fe(III) (42), this result raises the possibility that the reaction of FIH⅐Fe(II)⅐2OG⅐3CA with O 2 is rapid but that the hydroxylated product needs to (partially) fold before it is released. This proposal is consistent with the observation that the 2CA and 1CA substrates do not fold to form the consensus ankyrin repeat structure in solution (22).

Discussion
Hypoxia and the resultant response mediated by HIF are characteristic of multiple diseases, including some cancers (56,57). The available evidence to date indicates that the number of direct and physiologically relevant interfaces between O 2 and the HIF system that act in a chronic hypoxia-sensing capacity may be rather limited, with clear evidence currently only for the HIF hydroxylases themselves acting as hypoxia sensors, at least with respect to the HIF system itself. Large scale clinical trials are ongoing with PHD inhibitors for the treatment of anemia and hypoxia-related disorders (57,58). Thus, understanding the context-dependent roles of the HIF hydroxylases is of current clinical relevance. Our results clearly reveal striking differences between the reactivity of two of the HIF hydroxylases with respect to O 2 : PHD2, the most important of the human PHDs, and FIH. The data are consistent with prior less detailed kinetic analyses (10,43) and, importantly, with cel- lular observations (39) that FIH is more active at lower O 2 concentrations than the PHDs. The complexities of the HIF hypoxic sensing system, as are likely present in any pleiotropic eukaryotic sensing mechanism acting at transcriptional and other levels (e.g. variations in protein levels and location), mean that we are very far from quantitative correlations between kinetic studies with isolated components and the "in cell" sensors. Nonetheless, the qualitative correlation between the relative sensitivities of PHD2 and FIH catalysis to changes in O 2 availability with isolated components and in cells is striking. Previous studies of the pre-steady-state kinetics of PHD2 showed very slow O 2 -mediated initiation of CODD HIF-1␣ hydroxylation (38,42). In addition to the slower rate of PHD-dependent HIF hydroxylation in response to changed O 2 concentrations, this observation was proposed to be related to the biological role of PHD2 as an O 2 sensor in cells (39). Steadystate kinetic studies of O 2 dependence have shown that K m app (O 2 ) values for FIH are lower compared with PHD2 (10, 43); our results support these findings. Of note, we also observed a significantly faster O 2 -initiated reaction of FIH than of PHD2 under pre-steady-state conditions, consistent with our proposal that the slow reaction of PHD2 with O 2 is linked to its ability to act as an O 2 sensor, as reflected by its K m app (O 2 ) value (10, 11, 40 -42).
The results also reveal that FIH manifests mechanistic similarity to PHD2 in its lack of accumulation of the reactive Fe(IV)-oxo intermediate, which possesses distinctive spectroscopic features absorbing at 320 nm (32). This observation suggests that in terms of the rate-limiting steps of catalysis, FIH is similar to PHD2, rather than other 2OG oxygenases, such as TauD and vCPH (32,36). Further, solvent isotope effect studies of FIH-catalyzed HIF hydroxylation have indicated that the rate-limiting step occurs after dissociation of the iron-bound water molecule from the active site and probably includes the O 2 activation step (59), similar to that observed for PHD2 (60). This probably contributes to a degree of restraint with respect to O 2 activation subsequent to substrate binding (binding of an Fe(II)-coor-   Fig. 1). We also investigated the HIF substrate dependence of O 2 kinetics for both PHD2 and FIH. PHD2 has been shown to express a preference for hydroxylation of the CODD site, compared with NODD (46,61). However, we found in both presteady-state and steady-state experiments that O 2 -initiated reaction rates are independent of the nature of the prime substrate. Faster O 2 -initiated decay of the 520 nm-absorbing enzyme⅐Fe(II)⅐2OG⅐substrate complex was observed in our stopped-flow experiments for HIF-2␣ CODD hydroxylation compared with HIF-1␣ CODD, suggesting faster O 2 activation for the former. However, we did not observe any difference in the pre-steady-state time course of product accumulation for HIF-1/2␣ CODD, and the determined k cat app values for both substrates were the same within error (0.060 Ϯ 0.006 and 0.069 Ϯ 0.006 s Ϫ1 for HIF-1␣ and HIF-2␣ CODD, respectively). Thus, although the initial step of O 2 -initiated decay of the 520 nm-absorbing species is faster for the reaction with HIF-2␣ CODD, the overall effect is not rate-limiting for the reaction, and steady-state parameters with respect to O 2 are not affected.
FIH is seemingly a remarkably promiscuous oxygenase, being able to catalyze hydroxylation of Asn residues in both HIF and ARDs (19). Whereas the O 2 dependence of HIF-1␣ versus HIF-2␣ was not significantly different for FIH catalysis, our results provide evidence that K m app (O 2 ) values of FIH are lower with respect to ankyrin substrates, compared with HIF substrates, consistent with reported studies on Notch1-3 (43). Notably, the pre-steady-state rate of reaction of FIH with O 2 was much faster in the presence of (most) ankyrin substrates than HIF substrates (e.g. 0.33 and 11 s Ϫ1 for HIF-1␣ CAD 35 FIGURE 10. Stopped-flow UV-visible absorbance changes upon FIH-catalyzed hydroxylation of HIF-1/2␣ CAD and ARD substrates. Reaction mixtures containing anaerobic 0.5 mM apo-FIH, 0.4 mM Fe(II), 5 mM 2OG, and 1 mM peptide in HEPES 50 mM (pH 7.5) were rapidly mixed with O 2 -saturated buffer at 5°C in 1:1 ratio and analyzed using a photodiode array detector. The data were fitted with a double exponential function. A, decay at 500 nm (representing enzyme⅐substrate complex depletion) observed upon FIH-catalyzed hydroxylation of ARD substrates compared with HIF-1/2␣ CAD. B, formation and degradation of 310 nm species observed upon FIH-catalyzed HIF-1/2␣ CAD and ARD hydroxylation. C, decay at 500 nm observed upon FIH-catalyzed hydroxylation of 1CA, 2CA, and 3CA substrates. reaction under pre-steady-state conditions correlates with the steady-state K m app (O 2 ) parameter. Whereas the 1CA and 2CA FIH substrates are not folded in solution, 3CA possesses a defined secondary structure (22) (note that these are consensus (i.e. averaged) ARD sequences rather than specific natural ARD substrates). The results imply that the secondary structure of FIH ankyrin repeat substrates can affect the rate of reaction with O 2 . Previous studies have suggested that whereas for HIF-␣ CAD, the secondary structure does not affect the efficiency of the FIH-catalyzed hydroxylation, for ankyrin repeat substrates, unfolding of the substrate is required prior to hydroxylation (44,62). It has also been proposed that the difference in the efficiency of FIH-catalyzed ARD hydroxylation is related to the stability of the ARD protein fold (21). A protein with four contiguous consensus ankyrin repeats was found not to be hydroxylated by FIH, whereas 3CA is a substrate (21). The natural ARD proteins, Notch and IB␣, contain more than six ARDs and are known to be relatively efficient substrates for FIH (43,62), including in cells (18). Thus, the number of contiguous ARDs in a protein does not necessarily correlate with the hydroxylation efficiency, which is also related to local stability and sequence of the hydroxylated region (21,22). (Note also that FIH probably interacts with ARDs that it does not hydroxylate (22).) The observation that ARD folding affects FIH catalysis is in agreement with our observations, where a slower reaction with a folded ankyrin substrate was observed. It suggests that the unfolding step, which occurs prior to hydroxylation, affects the kinetics of ARD hydroxylation. If O 2 reactivity is limited by access to the active site, this may rationalize the observed slower rate for the 3CA substrate. This may also rationalize the overall observation of faster kinetics with respect to O 2 for FIH than for PHD2; PHD2 structural studies revealed a "tight" active site relative to that of FIH (45,63). Further, if the nature of the enzyme-substrate interaction affects O 2 access to the active site, this raises the possibility that alternative substrates for the HIF hydroxylases may have different O 2 sensitivity profiles.
The influence of enzyme-substrate interactions on the reaction of FIH with O 2 is of interest in terms of understanding the prime biological function of FIH. Although FIH is involved in hypoxia sensing in cells (25), the precise importance of the role of ARD hydroxylation in the hypoxic response is not well understood. Taking into account the abundance of ARD substrates in cells, our results indicate that FIH has the potential to hydroxylate ARD substrates under conditions of low O 2 availability, where its ability to hydroxylate HIF substrates may be substantially diminished. Thus, at least on a short time scale (i.e. tens of minutes) FIH-catalyzed ARD hydroxylation may not directly contribute in a sensing capacity to the hypoxic response. However, it has the possibility of contributing to the context-dependent regulation of the hypoxic response on a longer time scale (including by enabling memory effects (64)) by altering the kinetics of FIH HIF-␣ catalysis in cells by regulating the fraction of FIH free to hydroxylate HIF-␣, because (at least some) hydroxylated ARDs bind less tightly to FIH than unhydroxylated ARDs (19,44).