Many Amino Acid Substitutions in a Hypoxia-inducible Transcription Factor (HIF)-1α-like Peptide Cause Only Minor Changes in Its Hydroxylation by the HIF Prolyl 4-Hydroxylases

Three human prolyl 4-hydroxylases (P4Hs) regulate the hypoxia-inducible transcription factors (HIFs) by hydroxylating a Leu-Xaa-Xaa-Leu-Ala-Pro motif. We report here that the two leucines in the Leu-Glu-Met-Leu-Ala-Pro core motif of a 20-residue peptide corresponding to the sequence around Pro564 in HIF-1α can be replaced by many residues with no or only a modest decrease in its substrate properties or in some cases even a slight increase. The glutamate and methionine could be substituted by almost any residue, eight amino acids in the former position and four in the latter being even better for HIF-P4H-3 than the wild-type residues. Alanine was by far the strictest requirement, because no residue could fully substitute for it in the case of HIF-P4H-1, and only serine or isoleucine, valine, and serine did this in the cases of HIF-P4Hs 2 and 3. Peptides with more than one substitution, having the core sequences Trp-Glu-Met-Val-Ala-Pro, Tyr-Glu-Met-Ile-Ala-Pro, Ile-Glu-Met-Ile-Ala-Pro, Trp-Glu-Met-Val-Ser-Pro, and Trp-Glu-Ala-Val-Ser-Pro were in most cases equally as good or almost as good substrates as the wild-type peptide. The acidic residues present in the 20-residue peptide also played a distinct role, but alanine substitution for any six of them, and in some combinations even three of them, had no negative effects. Substitution of the proline by 3,4-dehydroproline or l-azetidine-2-carboxylic acid, but not any other residue, led to a high rate of uncoupled 2-oxoglutarate decarboxylation with no hydroxylation. The data obtained for the three HIF-P4Hs in various experiments were in most cases similar, but in some cases HIF-P4H-3 showed distinctly different properties.

The hypoxia-inducible transcription factors (HIFs), 1 which are essential for the regulation of cellular and systemic oxygen homeostasis, are ␣␤ heterodimers in which both types of sub-unit are basic helix-loop-helix Per-Arnt-Sim proteins. The human ␣ subunit has three isoforms, HIF-1␣ to HIF-3␣, of which HIF-1␣ and HIF-2␣ are expressed constitutively but are rapidly degraded under normoxic conditions (for reviews see Refs. [1][2][3][4]. This degradation is mediated by the oxygen-sensitive degradation domain, which contains two critical proline residues. Hydroxylation of at least one of these to 4-hydroxyproline is essential for the binding of HIF␣ to the von Hippel-Lindau (VHL) E3 ubiquitin ligase complex and for the subsequent rapid proteasomal degradation (5)(6)(7)(8). This hydroxylation is catalyzed in humans by three recently identified cytoplasmic and nuclear HIF prolyl 4-hydroxylases (HIF-P4Hs) (9 -11), which are distinct from the well characterized endoplasmic reticulum lumenal collagen prolyl 4-hydroxylases (C-P4Hs) (12)(13)(14)(15)(16). All P4Hs belong to the family of 2-oxoglutarate dioxygenases and require Fe 2ϩ , 2-oxoglutarate, O 2 , and ascorbate (12)(13)(14). The K m values of the three human HIF-P4Hs for O 2 are slightly above the concentration of dissolved O 2 in the air (17), and thus even a small decrease in the O 2 concentration will inhibit their activities, leading to stabilization of HIF␣ and dimer formation with HIF␤. The dimer is then translocated to the nucleus, where it becomes bound to the HIF-responsive elements in a number of hypoxia-inducible genes, such as those for erythropoietin, vascular endothelial growth factor, glycolytic enzymes, and glucose transporters (1)(2)(3)(4).
The two critical proline residues in human HIF-1␣, Pro 402 and Pro 564 , are located in the Leu-Thr-Leu-Leu-Ala-Pro-Ala and Leu-Glu-Met-Leu-Ala-Pro-Tyr sequences, respectively. Pro 564 is the principal hydroxylation site (8,10), the K m values of HIF-P4Hs 1 and 2 for a 19-residue peptide corresponding to the N-terminal hydroxylation site being ϳ20 -50 times higher than those for a peptide corresponding to the C-terminal site, whereas HIF-P4H-3 did not hydroxylate the N-terminal peptide at all (17). All three HIF-P4Hs also hydroxylated 19residue peptides corresponding to the C-and N-terminal hydroxylation sites in HIF-2␣ and one in HIF-3␣, with the sequences Leu-Glu-Thr-Leu-Ala-Pro-Tyr, Leu-Ala-Gln-Leu-Ala-Pro-Thr, and Leu-Glu-Met-Leu-Ala-Pro-Tyr, respectively, although the second of these had distinctly higher K m values than the other two (17). These data are in agreement with the suggestion that the hydroxylation may involve a conserved core sequence Leu-Xaa-Xaa-Leu-Ala-Pro (8,10). Initial mutagenesis experiments in fact indicated that substitution of Leu 562 by Ala or Ala 563 by Gly may prevent any hydroxylation (5,9). A more recent study has demonstrated, however, that Leu 562 3 Val, Leu 562 3 Ala, and Ala 563 3 Ser mutants were utilized by the three HIF-P4Hs only marginally less effectively than a wild-type substrate, whereas Leu 559 3 Val was utilized somewhat less effectively (18). Leu 574 has recently been identified as an additional important residue for hydroxylation (19), which agrees with the previous finding that the deletion of two residues, Gln 573 and Leu 574 , from the C terminus of a 19-residue HIF-1␣ C-terminal peptide increased the K m values for HIF-P4Hs 1 and 2 ϳ9and 7-fold, respectively, although the K m for HIF-P4H-3 was unaltered, and the V max values for all three isoenzymes were likewise unaffected (17).
Pro 402 and Pro 564 have several acidic residues in their vicinity, but mutation of Asp 556 , Asp 558 , Glu 560 , Asp 569 , or Asp 570 to asparagine appeared to have little effect on the hydroxylation (6). An Asp 558 3 Ala and Glu 560 3 Gln double mutant also served as a substrate, although less efficiently than the wildtype substrate, whereas an Asp 569 3 Asn, Asp 570 3 Asn, and Asp 571 3 Asn triple mutant failed to act as a substrate at all (18).
The C-P4Hs catalyze an uncoupled decarboxylation of 2-oxoglutarate in the presence of all cosubstrates but in the absence of any peptide substrate at a rate that is ϳ0.5-1% of that of the hydroxylation reaction observed in the presence of a saturating concentration of the peptide substrate (12)(13)(14)20). Some peptides that become bound to the C-P4Hs but do not act as substrates enhance the rate of the uncoupled decarboxylation, but it still remains below ϳ2% of that of the complete reaction (12)(13)(14). However, substitution of the hydroxylatable proline in a peptide substrate by 3,4-dehydroproline gave an uncoupled decarboxylation rate similar to the hydroxylation rate with the nonmodified peptide (21). No data are currently available to indicate whether substitution of the hydroxylatable proline in HIF␣ model peptides would lead to a similar high rate of uncoupled decarboxylation.
We studied here the sequence requirements of the three human HIF-P4Hs by systematically varying each residue preceding the proline in the Leu-Glu-Met-Leu-Ala-Pro sequence of a 20residue peptide corresponding to the C-terminal hydroxylation site in HIF-1␣. In addition, we studied in detail the effects of substitution of any of six of the seven acidic residues in this peptide by alanine, either alone or in combination with other acidic residues and substitution of the hydroxylatable proline by proline analogues. Our data indicate that the two leucines can be replaced by many residues with either no significant decrease or only a modest decrease in substrate properties and that the glutamate and methionine can well be replaced by most, although not all, amino acids, whereas the alanine is a relatively strict requirement. Any of the six acidic residues studied could be substituted by alanine with no decrease in the substrate properties, and even three of them could be substituted in some combinations with no negative effects. Peptides in which the proline was replaced by either 3,4-dehydroproline or L-azetidine-2-carboxylic acid gave a very high rate of uncoupled 2-oxoglutarate decarboxylation. The data obtained with the three HIF-P4Hs were in most cases similar, but in some cases HIF-P4H-3 showed distinctly different properties.

Generation of Recombinant Baculoviruses for the Expression of
FLAGHis-tagged HIF-P4H Isoenzymes-The FLAGHis tag was amplified by PCR from the plasmid d28e6 (Fibrogen Inc.) and cloned into the 3Ј ends before the stop codons of the recombinant human HIF-P4H isoenzymes 1, 2, and 3 in the pVL baculovirus expression vector (17). The recombinant baculovirus expression vectors were cotransfected into Spodoptera frugiperda Sf9 cells with BaculoGold DNA (Pharmingen) by calcium phosphate transfection, and the recombinant baculoviruses were amplified (22).
Expression of Recombinant HIF-P4H Isoenzymes in Insect Cells and Purification of the Enzymes-Nontagged HIF-P4Hs 1 and 2 (17), glutathione S-transferase-tagged HIF-P4H-3 (17), or FLAGHis-tagged HIF-P4Hs 1-3 were expressed in Sf9 or H5 cells cultured in suspension in Sf900IISFM serum-free medium (Invitrogen). The cells, seeded at a density of 1 ϫ 10 6 /ml, were infected with the corresponding viruses at a multiplicity of 5, harvested 72 h after infection, washed with a solution of 0.15 M NaCl and 0.02 M phosphate, pH 7.4, and homogenized in a 0.1 M NaCl, 0.1 M glycine, 10 M dithiothreitol, 0.1% Triton X-100, and 0.01 M Tris buffer, pH 7.8. The soluble fractions were used directly for enzyme activity assays or subjected to further purification.
Nontagged HIF-P4H-2 was partially purified by SP Sepharose chromatography. The cells were homogenized in 120 mM NaCl, 20 mM HEPES buffer, pH 7.5, followed by an addition of 0.25% Triton X-100 and 2.5 mM dithiothreitol. The lysate was incubated at 4°C for 1 h, centrifuged at 20,000 ϫ g for 20 min, and diluted 1:4 in 20 mM HEPES buffer, pH 6.5. The sample was loaded into an SP Fast Flow column (Amersham Biosciences) equilibrated with 20 mM HEPES buffer, pH 6.5. The column was washed with 50 mM NaCl, 20 mM HEPES buffer, pH 6.5, and eluted with a linear gradient of NaCl from 50 to 333 mM in HEPES buffer, pH 6.5, followed by elution with 500 mM NaCl, HEPES buffer, pH 6.5. HIF-P4H-2 was eluted at a 200 mM NaCl concentration. Recombinant HIF-P4H isoenzymes with C-terminal FLAGHis tags were purified to homogeneity with an anti-FLAG M2 affinity gel (Sigma) (23). 2 P4H Activity Assays-HIF-P4H activity was assayed by a method based on the hydroxylation-coupled decarboxylation of 2-oxoglutarate (17,24). The synthetic HIF-1␣ peptide 554 DTDLDLEMLAPYIPMD-DDFQ 573 used as a substrate, and its substituted variants were obtained from SynPep and Mimotopes. The nonpurified peptides used in the initial experiments were obtained from Mimotopes and were used at a concentration of 200 M, whereas those used in the measurements of K m and V max values had a purity of more than 85%. K m and V max values for the various HIF peptide substrates were determined at saturating concentrations of the other components, except that the O 2 concentration was that of air and thus nonsaturating (17).
Liquid Chromatography/Mass Spectrometry Analysis of Peptide Hydroxylation-The enzyme activity assays used in this set of experiments were conducted with partially purified HIF-P4H-2, following the standard protocol (17,24), except that the concentration of 2-oxoglutarate was 100 M, and that of the HIF-1␣ peptide 554 DTDLDLEALA-PYIPADDDFQ 573 or peptides in which the hydroxylatable Pro 564 was substituted by 3,4-dehydroproline or L-azetidine-2-carboxylic acid was likewise 100 M. The two methionines were substituted by alanines in all of these peptides to eliminate the possible formation of auto-oxidation products. The reaction was carried out at 37°C for 4 h and stopped by the addition of 1 mM EDTA. Hydroxylation of the peptides was analyzed by liquid chromatography with mass spectrometric detection using a Finnigan LCQ TM DUO LC/MS instrument (Thermo Electron) with electrospray ionization in negative mode. Chromatography was carried out using a YMC Pro Pack C18 (150 ϫ 2.0 mm, 3 , 120 Å) column with a gradient of 1-75% acetonitrile in 0.1% formic acid at a flow rate of 0.2 ml/min.

RESULTS
Leu 559 and Leu 562 Can Be Substituted by Many Other Residues with Relatively Small Effects-The 20-residue control peptide used as a substrate in the present study corresponds to the hydroxylation site around Pro 564 in HIF-1␣ and has the sequence 554 DTDLDLEMLAPYIPMDDDFQ 573 . It differs from the 19-residue peptide used previously (17) in that it contains two additional residues in its N terminus and lacks one, Leu 574 , in its C terminus. The K m values of a purified version of this peptide for HIF-P4Hs 1, 2, and 3 are 15, 25, and 5 M, respectively (see below). Because of the high cost of synthetic peptides, all of the initial experiments were performed using crude peptide preparations of less than 50% purity. The nominal peptide concentrations used in the initial experiments were 200 M, but because of the low degree of purity, the true concentrations were less than 100 M. Thus the concentration of the control peptide used for HIF-P4Hs 1 and 2 may have been less than 4 -6.5 times the K m , and that for HIF-P4H-3 may have been less than 20 times the K m . Consequently, the assays may have been insensitive for substitutions that cause only relatively small K m effects with no changes in V max , especially in the case of HIF-P4H-3. Enzyme activity was assayed based on measurement of the radioactivity of the 14 CO 2 formed during the hydroxylation-coupled stoichiometric decarboxylation of 2-oxo[1-14 C]glutarate using Triton X-100 buffer extracts of homogenates of insect cells expressing recombinant HIF-P4Hs as sources of the enzymes (17), or in the case of HIF-P4H-2, preparations that had been partially purified from such extracts by one chromatography step. Because the purity of the peptides was low, the reaction rates obtained in the initial experiments are given in Tables I, II, and IV only as Ͼ120, 80 -120, 50 -80, 30 -50, 10-30, and Ͻ10% relative to that obtained with the control peptide of a similar low degree of purity.
Substitution of Leu 559 by other residues gave identical data for HIF-P4Hs 1 and 2 (Table I). Its replacement by tryptophan under the conditions used gave a reaction rate for both isoenzymes that was significantly higher than that obtained with leucine (120 -130%), whereas the rates obtained with tyrosine and methionine were similar to that with leucine. Five additional amino acids, isoleucine, phenylalanine, alanine, glutamate and aspartate, gave rates that were 50 -80% of that with leucine, whereas the lowest rates, ϳ10 -30%, were obtained with asparagine, threonine, histidine, arginine, and lysine, thus including all of the basic amino acids (Table I). In the case of HIF-P4H-3, there were two residues, isoleucine and phenylalanine, that gave significantly higher rates than leucine (ϳ130%), whereas five additional residues, tryptophan, tyrosine, methionine, alanine, and valine, gave rates similar to that with leucine, seven other residues gave rates that were 50 -80% of that with leucine, and five residues gave rates that were less than 50% (Table I). It may be noted that the ranking of the peptides was not identical for HIF-P4H-3 and the other two isoenzymes ( Table I).
Substitution of Leu 562 again gave identical data for HIF-P4Hs 1 and 2, and very similar, although not identical, data for HIF-P4H-3 (Table I). No residue was significantly better than leucine in this position, but four, valine, isoleucine, phenylalanine, and arginine, gave rates similar to that with leucine for isoenzymes 1 and 2, and five, valine, isoleucine, phenylalanine, tyrosine, and methionine, for isoenzyme 3 (Table I). For all three HIF-P4Hs, there were eight residues altogether that gave rates that were at least 50% of that with leucine, seven of them being the same for all three isoenzymes. The lowest rates for all three isoenzymes were obtained with glutamate, glycine, proline, and aspartate (Table I). Interestingly, these residues include the two acidic amino acids, whereas in the Leu 559 position the least effective residues were histidine, arginine, and lysine.
Glu 560 and Met 561 Can Be Substituted by Most Amino Acids, whereas Ala 563 Is a Relatively Strict Requirement-The requirements for Glu 560 , Met 561 , and Ala 563 were studied by means of experiments similar to those above. Glu 560 could be substituted by most amino acids in the case of HIF-P4H-2 with no significant change in the reaction rate (Table II), with seven additional residues giving rates of 50 -80%, whereas lysine and proline were the only ones that gave a rate less than 50% (ϳ40%). The data obtained with HIF-P4H-1 were very similar, although not identical, to those with HIF-P4H-2 (details not shown). In the case of HIF-P4H-3 any amino acid in this position gave a rate that was similar to or even slightly higher (up to 140%) than that obtained with glutamate (Table II). Interestingly, arginine and lysine were among the best substitutes for HIF-P4H-3, whereas they were among the least effective ones for HIF-P4Hs 1 and 2 (Table II).
Met 561 could be replaced by almost any amino acid with no significant change in the reaction rate for HIF-P4H-2, proline being the only residue that gave a rate less than 80% (ϳ50%) ( Table II). The data for HIF-P4H-1 were again similar, although not identical, to those for HIF-P4H-2 (details not shown). Most amino acids also gave rates similar to or slightly higher (up to ϳ130%) than that with methionine for HIF-P4H-3, whereas three residues, aspartate, glutamate, and proline, gave rates below 80%, with proline giving the lowest rate (ϳ60%) ( Table II).
Ala 563 could be substituted only by serine with no significant change in the reaction rate for HIF-P4H-2, whereas threonine and lysine gave rates of 50 -80% (Table II). Valine and glutamine gave rates of 30 -50%, and four other amino acids gave rates of 10 -30%, whereas all of the others gave rates less than 10% (Table II). The ranking of residues was similar for HIF-P4H-1, but the percentages obtained with some amino acids were even lower than for HIF-P4H-2, the rate obtained with serine being only ϳ60% and that with valine 20%, whereas the rate obtained with lysine was the same as for HIF-P4H-2 (details not shown). In the case of HIF-P4H-3, isoleucine, valine, and serine gave rates similar to that with alanine, six others gave rates of 50 -80%, five 30 -50%, and four gave rates of less than 30% (Table II).
Analysis of Purified Peptides, Including Those with More than One Substituted Residue, Confirms Data Obtained in the Initial Experiments-To study whether the data obtained with peptides of low purity can be confirmed with those of a high degree of purity, many purified peptides were studied as substrates of the three HIF-P4Hs using either enzymes similar to those in the initial experiments or enzymes purified to homogeneity by an affinity chromatography based on their FLAG tags. Identical K m and relative V max values were obtained with both enzyme sources.
Substitution of Leu 559 by tryptophan gave for all three HIF-P4Hs a V max of 120% with essentially unchanged K m values relative to the control peptide (Table III). These data are in agreement with the increased reaction rates found for isoen- zymes 1 and 2 with the nonpurified peptide, whereas the values obtained for isoenzyme 3 were better than those obtained in the initial experiments, in which the rate was below the limit of 120% (Table I). Substitution of Leu 559 by phenylalanine likewise gave a V max of 120% for isoenzyme 3 with no change in K m , whereas the values obtained for isoenzyme 2 were similar to those of the control peptide and thus slightly better than those suggested by the reaction rate determined in the initial experiments, the V max for isoenzyme 1 being only 60%, in agreement with the decreased reaction rate in the initial experiments (Tables I and III). Substitution of Leu 559 by isoleucine, glutamate, or lysine gave changes in the K m and V max values that were in a good agreement with those observed in the initial experiments, with isoleucine giving a V max of 120% for isoenzyme 3 but 60 and 70% for isoenzymes 1 and 2 and glutamate giving a V max of 60 -70% for all three isoenzymes (Tables I and  III). The very low reaction rate found with the Leu 559 3 Lys substitution in the initial experiments was evidently due to the combined effect of an increased K m and a decreased V max (Table III). Substitution of Met 561 by alanine had no V max effect with any of the isoenzymes but slightly increased the K m values (Table  III). Substitution of Leu 562 by phenylalanine, arginine, glycine, aspartate, or proline gave K m and V max effects for all three HIF-P4Hs that were mostly in a good agreement with the changes in the reaction rates seen in the initial experiments (Tables I and III). The phenylalanine and arginine substitutions had only minor effects on the K m values in the cases of HIF-P4Hs 1 and 2, whereas those for HIF-P4H-3 were 2-and 4-fold, respectively. The phenylalanine substitution slightly decreased the V max for isoenzymes 1 and 3, whereas the arginine substitution had no effect on the V max with any of the HIF-P4Hs. The other three substitutions led to marked increases in K m and decreases in V max for all three isoenzymes.
Ala 563 could be substituted by serine in the cases of all three HIF-P4Hs with relatively minor changes in the K m values and no change in the V max for HIF-P4Hs 2 and 3, whereas the V max for HIF-P4H-1 was reduced by 30% (Table III). Substitution by valine likewise had only minor effects on the K m values for the three isoenzymes and no effect on the V max for isoenzyme 3, whereas the V max values for isoenzymes 1 and 2 were reduced by 80 and 60%, respectively (Table III). Substitution of the alanine by glycine or proline led to markedly increased K m and decreased V max values for all three isoenzymes (Table III). All of the effects of the Ala 563 substitutions are in perfect agreement with the changes found in the reaction rates in the initial experiments (Tables II and III).
b The V max values are expressed relative to that obtained for each enzyme with the standard 20-residue peptide (taken as 100%). c ND, not determined.
V max values identical to that of the control, one giving a V max of 70%, whereas in the case of HIF-P4H-1 two peptides gave V max values of 80 -90% and three of 60 -70% (Table III). Our data thus indicate that the three HIF-P4Hs, and especially HIF-P4H-3, act well on sequences with cores distinctly different from the conserved Leu-Xaa-Xaa-Leu-Ala-Pro core motif (8,10).

Substitutions of Alanine for Acidic Residues Cause Either No Effect Or Only Minor Effects in Many
Cases-The 20-residue peptide contains seven acidic residues, six of which, excluding the N-terminal aspartate, were subjected to alanine substitution studies with peptides of a similar low degree of purity as in the other initial experiments. As indicated in Table IV, three of the acidic residues, Asp 556 , Asp 558 , and Glu 560 were on the N-terminal side of the Pro 564 to be hydroxylated, and three, Asp 569 , Asp 570 , and Asp 571 , were on the C-terminal side. The data are shown only for HIF-P4H-2, the results obtained with the other two isoenzymes being very similar.
Any single acidic residue could be substituted by alanine with no decrease in the substrate properties of the peptide under the conditions used (Table IV). Also, any two acidic residues could be substituted with no decrease in the reaction rate unless one of them was Asp 569 , i.e. the first of the three consecutive aspartates in the PYIPMDDD sequence, in which case the rate was 50 -80% (Table IV). Interestingly, even three acidic residues could be substituted with no significant decrease in the reaction rate when two of them were on the N-terminal side of Pro 564 and one on the C-terminal side, but not Asp 569 (Table IV), whereas substitution of three residues, of which two were on the C-terminal side and one on the N-terminal side, gave rates of 50 -80%, unless the pair on the C-terminal side was Asp 569 and Asp 570 , in which case the rate was lower (Table IV). Substitution of all three acidic residues on the N-terminal side, and some other combinations of three and even four residues, gave rates that were 30 -50% of the control, whereas some combinations of four residues gave rates that were only 10 -30% (Table IV). Substitution of all three residues on the C-terminal side or substitution of four residues, of which three were on the C-terminal side, gave the lowest rates, less than 10% (Table IV).

Substitution of Pro 564 by 3,4-Dehydroproline or L-Azetidine-2-Carboxylic Acid, but Not by Other Proline Analogues or Other Amino Acids, Leads to a High Rate of Uncoupled 2-Oxoglutarate Decarboxylation-Peptides in which
Pro 564 was substituted by 3,4-dehydroproline or L-azetidine-2-carboxylic acid (Fig. 1), gave rates for the decarboxylation of 2-oxoglutarate in the initial experiments with all three HIF-P4Hs that were similar to or even higher than the hydroxylation-coupled decarboxylation rate obtained with the control peptide (details not shown). These effects were highly specific, because the decarboxylation rates found with peptides in which the proline had been replaced by ␤-thioproline, L-homoproline (L-pipecolic acid), L-trans-4-hydroxyproline, 3-azetidine-carboxylic acid, nipecolic acid, or L-tetrahydroisoquinoline carboxylic acid were less than 5% (details not shown). Also, replacement of the proline by serine, cysteine, methionine, asparagine, glutamine, or histidine caused no significant decarboxylation (details not shown).
Further experiments were carried out with highly purified peptides to determine the K m and V max values and to assess whether the high 2-oxoglutarate decarboxylation rates ob- served with the 3,4-dehydroproline and azetidine-2-carboxylic acid peptides indeed represented uncoupled decarboxylation.
To avoid the formation of auto-oxidation products of methionine, the two methionines present in the control peptide were substituted by alanines in all of these peptides, including the control peptide itself.
The V max values of the 3,4-dehydroproline peptide for all three HIF-P4Hs were 120 -140%, whereas the K m values were similar to that of the control peptide or slightly lower ( Table V). The K m values of the azetidine-2-carboxylic acid peptide for the HIF-P4Hs were 3-6-fold relative to that of the control peptide, the V max values for HIF-P4Hs 1 and 2 being 70%, whereas that for HIF-P4H-3 was 110% (Table V).
Possible hydroxylation of the peptides was studied by performing enzyme reactions in which the concentrations of the peptide and 2-oxoglutarate were each 100 M, and the reaction was run for 4 h, in which time most of the 2-oxoglutarate had been consumed. Liquid chromatography with mass spectrometric detection was then used to separate the peptide from the other reaction components and analyze its mass spectral features. In agreement with our previous data (details not shown), the elution position of the hydroxylated control peptide (retention time, 40.92 min; Fig. 2A) differed distinctly from that of the nonhydroxylated peptide (retention time, 41.40 min; Fig. 2B), whereas the mobilities of the 3,4-dehydroproline (Fig. 2, C and D) and azetidine-2-carboxylic acid (details not shown) peptides were identical with and without enzyme incubation. The corresponding mass spectral data, which represent the Ϫ2 charge state of the peptides, are shown as insets in the four panels in Fig. 2. In such analyses the mass to charge ratio m/z of the nonhydroxylated control peptide should be 1137.2, and that of the hydroxylated control peptide should be 1145.2. The values 1137.92 and 1145.95 obtained here (Fig. 2, A and B) are identical with the theoretical values within the bounds of experimental error and indicate that the control peptide had become hydroxylated. The corresponding theoretical values for the nonhydroxylated and hydroxylated 3,4-dehydroproline peptide are 1136.2 and 1144.2, respectively. The values 1136.84 and 1136.67 obtained here (Fig. 2, C and D) are identical with each other and with the theoretical value for the nonhydroxylated peptide within the bounds of experimental error. Mass spectral analysis of the azetidine-2-carboxylic acid peptide likewise showed that it was not chemically altered by the enzyme incubation (details not shown). It is thus evident that the high 2-oxoglutarate decarboxylation rates observed with these two peptides were not due to any hydroxylation-coupled decarboxylation.
Experiments similar to the above were also carried out to study whether all of the 2-oxoglutarate decarboxylation found with peptides in which residues other than the proline had been modified indeed represented hydroxylation-coupled decarboxylation. Two peptides with sequences DTDLDWEALA-PYIPADDDFQ and DTDLDLEALKPYIPADDDFQ were selected for these experiments. The mobilities of both peptides were altered by the enzyme incubation, as in the case of the control peptide (above), and mass spectrometry indicated that they had become hydroxylated (details not shown).

DISCUSSION
Our data clearly indicate that the two leucines in the Leu-Xaa-Xaa-Leu-Ala-Pro core motif can be substituted by many amino acids with essentially no decrease or only a modest decrease in the properties of the 20-residue HIF-1␣-like peptide as a substrate for the three HIF-P4Hs. Furthermore, substitution of Leu 559 in the cases of isoenzymes 1 and 2 by one residue, tryptophan, and in the case of isoenzyme 3 by three residues, isoleucine, phenylalanine, or tryptophan, even slightly improved the substrate properties of the peptide. As expected, the two Xaa positions could be occupied by almost any amino acid, but a surprising finding was that in the case of Glu 560 eight amino acids were even better for HIF-P4H-3 than the authentic residue. These amino acids included arginine and lysine, whereas in the cases of HIF-P4Hs 1 and 2 arginine, histidine and lysine were among the four least effective substitutes. The position of Met 561 could likewise be occupied by residues that improved the substrate properties for isoenzyme 3, namely tyrosine, tryptophan, leucine, and valine, but surprisingly, three residues, aspartate, glutamate, and proline, were less effective than methionine for isoenzyme 3, whereas only proline was less effective than methionine for isoenzymes 1 and 2. Ala 563 was found to be by far the strictest requirement, as it could be fully substituted by no residue in the case of HIF-P4H-1, by only one residue, serine, in the case of HIF-P4H-2, and by three residues, isoleucine, valine, and serine, in the case of isoenzyme 3, whereas only four additional residues gave a rate of more than 30% with isoenzyme 2 under the conditions tested.
Hydroxylation by a C-P4H requires an Xaa-Pro-Gly triplet, in which Xaa can be any amino acid and the glycine can in some cases be replaced by alanine or glutamine (13). A tripeptide with this sequence fulfills the minimum requirement, but longer peptides are much better substrates, because amino acids in other parts of the peptide also affect enzyme-substrate interaction. Furthermore, the conformation of the peptide has a major effect in that the triple-helical conformation of collagenlike peptides completely prevents hydroxylation (13). Our present data indicate that the HIF-P4Hs require an Ala-Pro sequence, in which no other residue can fully replace the alanine in the case of isoenzyme 1, and only one or three residues may do so in the cases of isoenzymes 2 and 3. Previous work has demonstrated that the HIF-P4Hs require even longer peptide substrates than the C-P4Hs, the minimum requirement for any hydroxylation being a peptide of more than eight residues (17). The present and previous (18) data suggest that the requirement for a long peptide may be due to the marked redundancy in the substrate sequence, which has preferences for certain positions such as those occupied by Leu 559 and Leu 562 , those occupied by acidic residues, and that occupied by Leu 574 (19),  and consequently the HIF-P4Hs may have a long peptidesubstrate-binding site with multiple interactions (17). A specific peptide conformation may not be a requirement for binding, because NMR studies have shown that a 19-amino acid peptide corresponding to the HIF-1␣ residues Asp 556 -Leu 574 , which is efficiently hydroxylated by all three HIF-P4H isoenzymes (17), does not have a structured conformation in solution (25). This was confirmed here because HIF-P4H-2 was found to hydroxylate equally effectively peptides heated to 100°C for 10 min to destroy any possible structural conformation followed by rapid cooling to 0°C before use as the corresponding nonheated peptides (data not shown).
The three-dimensional structures of the HIF-P4Hs and the full-length HIF-1␣ are currently unknown, but the structure of a HIF-1␣-peptide-pVHL complex has been solved (25,26). Residues 560 -577 of the hydroxylated peptide were found to become bound to one of the ␤ sheets of the pVHL in an extended ␤ strand-like conformation as if the peptide was a complementary ␤ strand, although there is no extensive main chain hydrogen bonding to define it as such (25,26). The ␤ sheet-like interactions contribute to the stability of the complex, but the optimized hydrogen bonding to the hydroxyproline discriminates between binding of hydroxylated and unmodified HIF-1␣ to pVHL (25,26). Crystal structures of various 2-oxoglutarate dioxygenases, including the HIF asparaginyl hydroxylase have shown that their catalytic sites are all located in a jelly roll ␤ barrel structure formed by eight ␤ strands, the cosubstrates and the peptide substrate becoming bound in a cavity that is located between the two sheets of the jelly roll and exposed at one side of the ␤ barrel (14,(27)(28)(29). It therefore seems very likely that the catalytic site of the HIF-P4Hs has a similar structure (10) and that HIF-1␣ may become bound to these enzymes by forming an intermolecular ␤ sheet-like structure and inserting the proline to be hydroxylated into the ␤ barrel cavity, the binding being thus analogous to the interaction of HIF-1␣ with pVHL (26).
Previous studies have demonstrated clear differences between the three HIF-P4Hs in their substrate requirements, especially between isoenzyme 3 and the two others. One such difference is that shortening of a 19-residue HIF-1␣-like peptide at its N or C terminus influences the K m and V max values for the three isoenzymes differently, isoenzyme 3 being the least sensitive to this (17). On the other hand, isoenzyme 3 failed to hydroxylate a 19-residue peptide corresponding to the N-terminal hydroxylation site around Pro 402 in HIF-1␣, whereas this peptide was hydroxylated by HIF-P4Hs 1 and 2, although with K m values ϳ20 -50 times higher than those for a peptide corresponding to the C-terminal site (17). This difference could not be explained by the residue following the proline being alanine in the N-terminal peptide but tyrosine in the C-terminal one, because a Tyr 565 3 Ala substitution in the C-terminal peptide had only very minor effects on the K m values and no effect on the V max with any of the three isoenzymes (17). A Tyr 565 3 Gly substitution had a slightly larger effect on K m , but again none on V max with any of the isoenzymes (17). The present data further indicate that isoenzyme 3 differed very distinctly in its sequence requirements from isoenzymes 1 and 2 in some cases, the latter two having identical or very similar requirements in most cases but not all.
Because many of the observations reported here are based on experiments with peptides of low purity, the possibility cannot be excluded that some of the findings with respect to a specific amino acid substitution may be slightly erroneous and that the true values obtained with a purified peptide would be slightly higher or lower than those shown in Tables I, II, and IV. It should be noted, however, that the data obtained with various peptides for a single position showed a remarkable consistency with respect to the nature of the amino acid, in that the three basic amino acids histidine, arginine, and lysine were the least effective substitutes for Leu 559 , for instance, aspartate and glutamate were among the least effective substitutes for Leu 562 , and arginine, histidine, and lysine were among the four least effective substitutes for Glu 560 in the cases of HIF-P4Hs 1 and 2. Also, the seven Leu 559 substitutes that gave a rate for isoenzyme 3 that was similar to or higher than that with leucine were all hydrophobic amino acids, as also were the four Met 561 substitutes that gave a rate for isoenzyme 3 that was higher than that with methionine. Such a consistency would be unlikely if the purities of the individual peptides had varied substantially.
The data obtained on the effects of the various amino acid substitutions were confirmed and extended in experiments with purified peptides. Excellent agreement was found between the data obtained with the purified and impure peptides, with only a few minor exceptions. Some substitutions were found to influence only the K m or V max values, but those having the strongest effects influenced both K m and V max . Interestingly, peptides with more than one amino acid substitution having the core sequences Trp-Glu-Met-Val-Ala-Pro, Tyr-Glu-Met-Ile-Ala-Pro, Trp-Glu-Met-Val-Ser-Pro, and Trp-Glu-Ala-Val-Ser-Pro were essentially as good or even better substrates for HIF-P4Hs 2 and 3 than the control peptide with the Leu-Glu-Met-Leu-Ala-Pro motif, the V max values of the first two of these peptides for HIF-P4H-1 being 90 and 80% of that of the control, whereas those of the latter two were ϳ60 and 70%, evidently because of the unfavorable effect of the alanine to serine substitution in the case of this isoenzyme. The peptide Ile-Glu-Met-Ile-Ala-Pro was a slightly worse substrate than the control peptide for all three isoenzymes, its V max values for HIF-P4Hs 3, 2 and 1 being 90, 70, and 60%, respectively, with only a minor effect on the K m for any of the isoenzymes. It is possible that the HIF-P4Hs may have additional in vivo substrates that are regulated by oxygen-dependent proline hydroxylation and subsequent binding to pVHL, ubiquitination, and proteasomal degradation, with one suggested substrate being the large subunit of RNA polymerase II (30). Previous data reporting that a conserved Leu-Xaa-Xaa-Leu-Ala-Pro core motif may be an important requirement for hydroxylation (8,10) suggest that it might be possible to find additional HIF-P4H substrates by means of computerized protein sequence searches and verification of the results by hydroxylation experiments with synthetic peptides. Our present data indicate, however, that such an approach is not feasible because of the lack of specificity of the sequence of the core motif.
None of the six acidic residues studied was found to be critical for the substrate properties of the 20-residue peptide, thus confirming data obtained previously for five of these residues (6). Acidic residues were nevertheless found to play a distinct role in the substrate properties of the peptide, although as many as three of them could be substituted simultaneously by alanines in some combinations with no adverse effects. The data further indicate that Asp 569 may be more critical than the other acidic residues studied and that acidic residues on the C-terminal side of the hydroxylatable proline may be more critical than those on the N-terminal side. Although confirmatory experiments using purified peptides having alanine substitutions for the acidic residues were not undertaken, remarkable consistency was found in the data with respect to the effects of mutations of more than one residue, suggesting that purified peptides would have given at least very similar, if not identical results.
Substitution of the hydroxylatable proline in a peptide substrate for the C-P4Hs by 3,4-dehydroproline has been found to lead to uncoupled decarboxylation of 2-oxoglutarate at a rate similar to the rate of hydroxylation with the nonmodified peptide (21). Our present data indicate that the HIF-P4Hs are similar to the C-P4Hs with respect to the effect of such a substitution and extend those reported for the C-P4Hs by demonstrating that replacement of the proline by azetidine-2-carboxylic acid also leads to a high uncoupled decarboxylation rate, the effects of the two proline analogues being highly specific in that no other replacement of the proline residue studied caused any significant uncoupled decarboxylation.
Ascorbate is not needed in the hydroxylation reaction catalyzed by the C-P4Hs, but it is an absolute requirement for the C-P4Hs both in vitro and in vivo (12)(13)(14). The reaction requiring ascorbate has been shown to be the uncoupled 2-oxoglutarate decarboxylation, in which ascorbate is consumed stoichiometrically with the decarboxylation (31,32). The C-P4Hs catalyze this uncoupled decarboxylation in the absence of any peptide substrate at a rate that is ϳ0.5-1% of the rate of the hydroxylation reaction (13,20). In the uncoupled decarboxylation the highly reactive ferryl ion formed at the catalytic site during the first half-reaction of the catalytic cycle is probably converted to Fe 3ϩ⅐ O Ϫ , making the enzyme unavailable for new catalytic cycles until reduced by ascorbate (13,31,32). The C-P4Hs catalyze occasional uncoupled decarboxylation cycles even in the presence of a saturating concentration of their peptide substrates, and the main biological function of ascorbate in their reactions both in vitro and in vivo is believed to be that of serving as an alternative oxidizable substrate in the uncoupled decarboxylation cycles (12-14, 31, 32). The HIF-P4Hs also catalyze an uncoupled 2-oxoglutarate decarboxylation in the absence of their peptide substrates, at ϳ3% of the rate of the hydroxylation-coupled decarboxylation, 2 and the presence of a 3,4-dehydroproline or azetidine-2-carboxylic acid residue at the catalytic site markedly increases this rate, as found here. It therefore seems highly likely that the main function of ascorbate in the HIF-P4H reaction is similar to that in the C-P4H reaction, i.e. to serve as an alternative oxygen acceptor in the uncoupled decarboxylation cycles.