Biochemical Evidence for Heme Linkage through Esters with Asp-93 and Glu-241 in Human Eosinophil Peroxidase

The covalent heme attachment has been extensively studied by spectroscopic methods in myeloperoxidase and lactoperoxidase (LPO) but not in eosinophil peroxidase (EPO). We show that heme linkage to the heavy chain is invariably present, whereas heme linkage to the light chain of EPO is present in less than one-third of EPO molecules. Mass analysis of isolated heme bispeptides supports the hypothesis of a heme b linked through two esters to the polypeptide. Mass analysis of heme monopeptides reveals that >90% have a nonderivatized methyl group at the position of the light chain linkage. Apparently, an ester had not been formed during biosynthesis. The light chain linkage could be formed by incubation with hydrogen peroxide, in accordance with a recent hypothesis of autocatalytic heme attachment based on studies with LPO (DePillis, G. D., Ozaki, S., Kuo, J. M., Maltby, D. A., and Ortiz de Montellano P. R. (1997) J. Biol. Chem. 272, 8857–8860). By sequence analysis of isolated heme peptides after aminolysis, we unambiguously identified the acidic residues, Asp-93 of the light chain and Glu-241 of the heavy chain, that form esters with the heme group. This is the first biochemical support for ester linkage to two specific residues in eosinophil peroxidase. From a parallel study with LPO, we show that Asp-125 and Glu-275 are engaged in ester linkage. The species with a nonderivatized methyl group was not found among LPO peptides.

MPO, EPO, and LPO are primarily found in granules of neutrophil and eosinophil leukocytes and secretions of exocrine glands, respectively. Their oxidation of halide and pseudohalide is part of the defense system against bacteria and parasites. Although several substrates have been found in vitro, the physiologically relevant substrates are believed to be chloride (MPO) and thiocyanate (EPO and LPO), which are oxidized to the toxic products hypochlorite and hypothiocyanite (7)(8)(9). Thyroid peroxidase functions in biosynthesis of thyroid hormones by oxidation of iodide (10). LPO and thyroid peroxidase are single chain enzymes of about 75 and 100 kDa, respectively (10,11). In contrast, the polypeptide chains of MPO and EPO are cleaved after synthesis to form a heavy chain of approximately 55 kDa and a light chain of approximately 15 kDa that remain associated. EPO is a monomer with the expected size of 70 kDa, whereas MPO is a disulfide-linked dimer of 150 kDa (12)(13)(14)(15)(16).
The mammalian peroxidases are distinguished from other heme-containing peroxidases by tight binding of the heme group to the apoprotein. Early experiments with LPO demonstrated that the heme group could not be extracted with acidified acetone and, furthermore, that it was not linked through thioethers as found in cytochromes. Based on this and other lines of evidence, heme linkage through ester(s) was proposed (17). But the heme group has also been suggested to be a thioderivative disulfide-linked to the protein (18). Results supporting and in conflict with both of these hypotheses followed (19 -24).
Structurally, MPO is the best characterized of the mammalian peroxidases (5,25). A 2.28-Å crystal structure has been obtained that allows evaluation of possible interactions between amino acid side chains of MPO and the heme group. Based on distances in the crystal structure, ester linkages from hydroxylated methyl groups on pyrrole rings A and C to Glu-242 in the heavy chain and to Asp-94 in the light chain, respectively, were proposed. It was further suggested that the vinyl groups of pyrrole ring A formed a sulfonium ion linkage with Met-243 (23,25,26). Unlike the two acidic residues, this methionine has no obvious equivalent in LPO or EPO (11,25). It has also been suggested that Asp-94 and Glu-242 exist as protonatable residues capable of influencing the absorption spectrum of the heme group (27,28). In short, however, after many years of investigation with mixed results, linkage of the heme group through two esters seems to be accepted for LPO and MPO, although for LPO, the residue that corresponds to Asp-94 in MPO has not been unambiguously identified (24).
Due to its limited availability, EPO is the least studied of the four peroxidases, and compared with LPO and MPO, insuffi-cient evidence has been generated to allow generalization within the subfamily. EPO has primarily been the subject of comparative spectroscopic studies (23, 29 -31), and no direct biochemical evidence has been reported.
Biochemical studies on mammalian peroxidases have previously been carried out on the isolated heme group as released with proteolytic enzymes, such as Pronase, that also cleaves ester bonds. Exceptions are recent spectroscopic studies on LPO (20,24). In this work, we present data based on analysis of the isolated heme group with covalently bound EPO peptides. We chemically identify the bonds as esters, and we identify the residues that are engaged in this linkage. Furthermore, we demonstrate that in vivo one of the two esters is only partly formed but that autocatalysis in vitro results in formation of an enzyme with two esters. LPO was analyzed in parallel.

EXPERIMENTAL PROCEDURES
Proteins-Human EPO was purified using a previously published procedure (32) with modifications. Briefly, isolated eosinophil granules from single donors with marked eosinophilia were extracted with 0.2 M sodium acetate, pH 4.0, and the supernatant was loaded onto a Sephadex G-75 SF (Amersham Pharmacia Biotech) equilibrated and eluted with 50 mM sodium acetate, pH 4.3. EPO eluted as a reasonably isolated peak. As judged by SDS-PAGE, the material was more than 90% pure (see Fig. 1). The original procedure included an additional ion-exchange chromatography step on CM-Sephadex in 0.1 M CAPS buffer, pH 10.0. In our hands, polymerization through cysteine residues is minimized to a negligible level if the pH is not elevated. For this study, nonpolymerized protein was important, and the purity obtained directly from the gel filtration column was sufficient. Therefore, the ion-exchange step was omitted. Different preparations of EPO were used. The functionality of EPO from patients with marked eosinophilia is the same as EPO from normal donors (33). Partly purified bovine LPO was obtained as a gift from Svenska Mejeriernes Riksförening, Sweden. The lyophilized LPO was further purified before use by ion-exchange chromatography on a UNO-S12 column (Bio-Rad). The enzyme was dissolved in 35 mM sodium phosphate, pH 6.0, applied to the column equilibrated with the same buffer, and eluted by a linear gradient of 0 -1000 mM sodium chloride in the equilibration buffer. Pooled fractions were concentrated by ultrafiltration. The purity of the protein (Ͼ95%), finally in 35 mM sodium phosphate, pH 6.0, was assessed by SDS-PAGE and capillary electrophoresis. The A 412 /A 280 ratio of the purified protein was 0.92.
Miscellaneous Procedures-SDS-PAGE was performed in 16% Tristricine gels (34). Samples (2 g) were diluted 10-fold in 20 mM sodium phosphate, 150 mM sodium chloride (phosphate-buffered saline), pH 7.4 (some samples were adjusted as specified in the text), and prepared for loading by boiling in sample buffer for 1 min or longer (up to 10 min) with or without 10 mM dithiothreitol. Prior to addition of loading buffer, some samples were incubated with hydrogen peroxide at different concentrations for 5 min at room temperature. The concentration of the 30% (v/v) hydrogen peroxide (Merck) stock solution was verified gravimetrically (expected density of the stock solution is 1.11 kg/liter), and dilutions made in phosphate-buffered saline. Digests were made of EPO and LPO with 1 ⁄100-1 ⁄25 (w/w) thermolysin (Type X protease, Sigma), trypsin (Worthington), and chymotrypsin (Roche Molecular Biochemicals) alone or in different combinations for 1-4 h at 50 or 37°C. Prior to addition of enzymes, the pH was adjusted to 7.2-8.0 with 1 M Tris. Incubation of isolated heme-containing peptides with ammonia was performed in a capped 1.5-ml tube containing 50 l of 25% (w/w) ammonia. Peptides, immobilized on glass filters (for the Applied Biosystems 477A sequencer, see below), were inserted into the tube without making contact to the liquid and incubated for 20 h at room temperature. After the incubation, traces of ammonia were removed in an exsiccator.
Column Chromatography-Gel filtration on Superose 12 HR 10/30 (Amersham Pharmacia Biotech) was performed in 6 M guanidine hydrochloride, 20 mM sodium phosphate, pH 7.2. The flow rate was 0.2 ml/min, and the eluate was monitored at 280 and 398 nm via serial connection of variable wavelength monitors. Samples were loaded after addition of solid guanidine hydrochloride to a final concentration of 6 M. Peptides were separated by reversed-phase high pressure liquid chromatography on a 4 ϫ 250-mm Nucleosil C18 column (Macherey-Nagel) eluted with a flow rate of 1 ml/min at 50°C. A gradient was formed from 0.1% trifluoroacetic acid (Rathburn) (solvent A) and 90% acetonitrile (Rathburn) containing 0.075% trifluoroacetic acid (solvent B), increas-ing the amount of solvent B linearly by 1%/min. The eluate was monitored at 226 and 398 nm via serial connection of variable wavelength monitors. In Figs. 2 and 3, the resulting chromatograms are overlaid and aligned.
Sequence Analysis-Edman degradation was performed on an Applied Biosystems 477A sequencer equipped with an on-line high pressure liquid chromatograph (35). For sample loading, isolated peptides (20 -200 pmol) were pipetted onto polybrene-coated glass filters, and proteins separated by SDS-PAGE were blotted onto a ProBlott membrane (Applied Biosystems), Coomassie-stained, and excised.
Mass Spectrometry-Mass spectra were acquired with a Bruker BI-FLEX matrix-assisted laser desorption ionization-time of flight instrument (Bruker-Franzen, Bremen, Germany) equipped with a 1-m flight tube, a reflector, a 337-nm nitrogen laser, and a 500-MHz digitizer. Thin film matrix surfaces were prepared using the fast evaporation technique (36) from ␣-cyano-4-hydroxycinnamic acid (Sigma) dissolved in acetone/water (99:1) to 30 g/l. A 0.5-l volume of the analyte (0.1-10 pmol/l) was deposited on the matrix surface and allowed to dry onto the crystals. Spectra were obtained by averaging 20 -50 single-shot spectra. Spectra were calibrated internally by co-crystallizing small amounts of angiotensin II (Sigma) and adrenocorticotropic hormone, fragment 18 -39 (Sigma), with the analyte and by using the calibration constants of well known matrix ions. Nonheme peptides were observed as MH ϩ species, but heme peptides were observed as M ϩ species, with the charge resulting from the Fe 3ϩ ion of the heme group and the formal charge of Ϫ2 of the four nitrogen atoms. To confirm this, spectra of hemin (C 34  , 0.3%) and with the absence of a 55Fe species. Furthermore, a cytochrome c heme peptide, with the heme group bound via two thioethers, was also observed as an M ϩ species (not shown). In spectra of EPO and LPO heme peptides, the intact heme peptide was often observed along with free peptide(s) and the free heme group resulting from cleavage of the peptide-heme linkage(s) in the instrument. The free peptides were observed as MH ϩ species, and the heme group as an M ϩ species, but with a mass reduction of 1 mass unit compared with hemin, because the heme macrocycle is capable of stabilizing a -CH 2 radical.

RESULTS
Preparation and Characterization of Nonpolymerized Human EPO-A preparation of human EPO that showed a minimum of polymerization through cysteine residues in nonreduced SDS-PAGE was made. In our hands, the traditional final step of ion-exchange chromatography at pH 10 (32) led to extensive polymerization of the protein. Because the purity of the protein prior to this chromatographic step was above 90%, we did not carry out further purification. Three bands appear in nonreduced SDS-PAGE of the preparation (Fig. 1, lane 2). Sequence analysis of protein blotted onto a polyvinylidene difluoride membrane confirmed the identity of the top band (70 kDa) as two-chained EPO with equimolar amounts of EPO heavy and light chain, the middle band (55 kDa) as heavy chain, and the bottom band (15 kDa) as light chain. As judged from this gel, we estimate that less than one-third of the EPO molecules have covalently linked light and heavy chains. Native EPO contains two free sulfhydryl groups, 2 consistent with the finding that it readily polymerizes at elevated pH. But the partial separation of the heavy and light chain cannot be explained by interchange of disulfide bonds, because in the native molecule, the two cysteine residues of the light chain are paired. 2 Thus, it appears that the EPO exists in two forms. In one, the heavy and light chains are covalently bound, whereas in the other, the two chains are noncovalently associated. The length of sample boiling prior to loading the gel did not affect the extent of chain separation (not shown). When incubated with reductant, however, the separation of EPO into heavy and light chain was complete (Fig. 1, lane 1). Incubation at pH 2 with SDS also resulted in complete chain separation (not shown).
The Heme Group Connects the EPO Heavy and Light Chain or Is Bound to the Heavy Chain Only-In nondenaturing gel filtration, EPO elutes as one peak (not shown), but as in SDS-PAGE, the partial chain separation is evident in denaturing gel filtration (Fig. 2). Furthermore, the recorded absorption at 398 nm shows that when the heavy and light chains of EPO are separated by denaturation alone, the heme group is linked to the heavy chain. Following incubation with reductant, the heme group elutes as a third peak, decreased in absorption and separate from the heavy and the light chain peaks (not shown). Although this would be consistent with heme linkage through disulfides, as proposed earlier for LPO (18), recent results with different mammalian peroxidases suggests linkage by ester bonds (20,(23)(24)(25). Curiously, incubation of EPO for 30 min at room temperature with 18 mM dithiothreitol in 6 M guanidine hydrochloride at pH 7.2 decreased the absorption at 398 nm more than 50% (not shown), suggesting modification or precipitation of the heme.
Autocatalysis of Covalent Heme Attachment in EPO-Recently, it was shown that covalent attachment of the heme group to the single polypeptide of LPO can occur by an autocatalytic process in the presence of hydrogen peroxide (22). With EPO, we have independently confirmed and visualized this process directly by SDS-PAGE (Fig. 1, lane 3). The intensity of the upper band increased with increasing concentrations of hydrogen peroxide (not shown). To bring Ͼ90% of the protein to the 70-kDa EPO form, a concentration of 30 M hydrogen peroxide was required, corresponding to a 10-fold molar excess over EPO (Fig. 1, lane 3). Changing the pH in the reaction buffer to 6.0 or 8.5 did not affect the result. As expected, when the modified EPO was incubated with reductant, only heavy and light chain were seen in SDS-PAGE (not shown).
Isolation and Analysis of EPO Peptides with Covalently Bound Heme-Previously, indirect evidence for heme linkage through esters in the mammalian peroxidases, mainly LPO, was provided through investigations on isolated heme groups and by crystallography of MPO (25). Recent studies with LPO argue for ester linkage based on studies with isolated heme peptides, but complete assignment of amino acid residues could not be made (20,24). In general, direct peptide evidence has been lacking for all mammalian peroxidases.
EPO has not previously been investigated at the peptide level. To study the heme linkage in this protein and to study the basis for the apparently different attachment to the EPO light and heavy chains, peptides from a thermolytic digest were separated by reversed-phase high pressure liquid chromatography (Fig. 3A). The majority of peptides (90% based on peak heights) that show absorption at both 226 and 398 nm, thus 2 A. R. Thomsen., C. Oxvig., and L. Sottrup-Jensen, unpublished observation.  (Tables I and III). containing the heme group, were further analyzed.
N-terminal sequence analysis revealed that some were heme bispeptides (E-TL-2, E-TL-3), referred to below as bispeptides, and some were heme monopeptides (E-TL-1, E-TL-4, E-TL-5), referred to as monopeptides (Table I). 3 The bispeptides contained equimolar amounts of peptides derived from two regions in the EPO light and heavy chains, respectively. All monopeptides contained peptides derived from the heavy chain. This result, including the relative abundance of bispeptides (about 30%), is in fine agreement with the results from SDS-PAGE and gel filtration detailed above. Concordant results were obtained from several other digests with thermolysin, trypsin, and chymotrypsin alone or in different combinations (not shown and Table I). We never observed a monopeptide with polypeptide originating from the EPO light chain.
To substantiate the covalent heme linkages as esters and to identify the acidic residues putatively engaged herein, peptides E-TCT-101 and E-TL-3 (Table I) were immobilized on glass filters and incubated in an atmosphere of ammonia overnight at room temperature to convert the carboxyl moiety of the putative esters into amides by aminolysis. Sequence analysis of the peptides after incubation unambiguously identified the positions of heme linkage (Table II). In cycle 3 of peptide E-TCT-101 ( 239 STETPK), partial conversion of Glu-241 to Gln was observed. Likewise, in cycle 3 of peptide E-TL-3 ( 91 FIDHD and 232 FLAGDTRSTETPK), partial conversion of Asp-93 to Asn was observed. Importantly, in cycle 5, no conversion of Asp-95 to Asn was seen. When the same peptides were subjected to sequencing without prior incubation with ammonia, very little Gln and Asn was observed in the respective positions (Table II). We did not see a complete conversion of the expected acidic residues to the corresponding amides; thus, to some extent, hydrolysis occurred during the incubation. However, the residues of EPO engaged in heme linkage have now been identified as Asp-93 of the light chain and Glu-241 of the EPO heavy chain.
Analysis of Heme Peptides Derived from LPO-Because LPO is a single-chain protein, partial heme linkage to one of two amino acid side chains would not be evident from SDS-PAGE. We thus repeated the peptide isolation procedure with purified LPO. Again, several heme-containing peptides were found (Fig.  3B) and subsequently identified by sequence analysis (Table  III). In striking contrast to EPO, 80% of the observed LPO peptides were bispeptides, and the remaining 20% were monopeptides (Table III). But like EPO, the polypeptides of the monopeptides were variants of the same peptide stretch. In a sequence alignment, this region corresponds to the region observed for EPO monopeptides. When the LPO peptide L-TL-2 ( 123 IVDHD and 273 ASEQ) was subjected to sequence analysis following exposure to ammonia, partial conversion of Asp-125 (but not Asp-127) to Asn and of Glu-275 to Gln was seen (not shown). Therefore, the residues engaged in ester linkage of the heme group in LPO are Asp-125 and Glu-275. Fig. 4 shows the heme group as derivatized with polypeptides from EPO or LPO in heme bispeptides. 4 Two Different Forms of the Heme in the Monopeptide of EPO but Not LPO-Mass spectrometry was used to confirm the identity of all analyzed peptides from EPO and LPO. Regardless of origin, the contribution of the heme group to the total mass of bispeptides was the same, 612.6 Ϯ 0.2 Da (Table IV) 3A), except E-TCT-101, which was isolated from a tryptic/chymotryptic digest. This peptide was the most abundant heme peptide of this digest.
b For amino acid position in sequence, see Fig. 4. Sequences were determined by N-terminal sequence analysis and verified by mass spectrometry.
c Heme monopeptides contain one polypeptide fragment covalently bound to the heme group, and heme bispeptides contain two.

TABLE II
Sequence analysis of EPO peptides exposed to ammonia About 90% pure peptides were exposed to an atmosphere of ammonia for 20 h at room temperature prior to sequence analysis. Cycle  a For identity of peptides and position in the amino acid sequence, see Table I and Fig. 4. Here, residues of cycles 1 are numbered. Values are yield of the expected amino acid from successive cycles of Edman degradation. Where acidic residues occur, the yields of the corresponding amides are also shown. Yields from sequence analysis of the same peptides not treated with ammonia are given in parentheses.
b In cycles 1 and 5 of peptide E-TL-3, where the same amino acid was found in both sequences of the bispeptide, 50% of the actual value is ascribed to each sequence.  IVDHDLD and FRASEQI Bis (7%) a All peptides were isolated from a thermolytic digest of LPO (Fig.  3B).
b For amino acid position in sequence, see Fig. 4. Sequences were determined by N-terminal sequence analysis and verified by mass spectrometry.
c Heme monopeptides contain one polypeptide fragments bound to the heme group, and heme bispeptides contain two. The relative abundances among analyzed heme peptides are estimated from peak heights on the chromatogram. pected heme contribution to the total mass in a heme bispeptide with two esters, formally (HC-C-1-heme b core-C-5-CH) (Fig. 4), is 4.03 mass units lower, i.e. 612.48 Da, in perfect agreement with the observed masses. Hence, this result further supports that the heme of EPO is heme b and that it is linked via esters.
In contrast, the contribution of the heme group to the total mass of monopeptides was not constant, but rather appeared in two groups separated by 16 mass units (Table IV). A heme monopeptide could arise from hydrolysis of the ester at C-5 (Fig. 4). Hydrolysis would leave a hydroxymethyl group on C-5, and the expected heme contribution to the total mass would be 18 mass units higher than in the bispeptides (C-5-CH 2 OH rather than C-5-CH), i.e. 630.51 Da. This mass was observed for all LPO monopeptides, but rarely for EPO monopeptides. Of the EPO peptides presented here (Fig. 3A and Table I), only one, E-TL-1, conformed to this mass. The majority of peptides, constituting 90% of all EPO monopeptides, had a heme contribution to the total mass that was 2 mass units higher than for the bispeptides. This mass is consistent with the presence of a methyl group at C-5 (C-5-CH 3 ). As discussed below, we suggest that an ester with the EPO light chain was never formed in the majority of EPO molecules and that the original methyl group is present at C-5. The distribution of monopeptides described here was also seen in digests of other EPO preparations (not shown).

DISCUSSION
By analysis of intact human EPO and proteolytic peptides thereof, we show that 1) the heme group is covalently bound to both the light and heavy chain of the EPO polypeptide in less than one-third of EPO molecules, 2) the heme group is bound only to the heavy chain in the majority of EPO molecules, 3) incubation of EPO with excess hydrogen peroxide attaches the unbound EPO light chain to the heme group in an autocatalytic reaction, 4) the two acidic residues that are engaged in binding are Asp-93 and Glu-241, and as they can be converted into the corresponding amides by aminolysis, linkage by ester bonds is confirmed, and 5) in molecules where the heme group is attached to the heavy chain only, the site of possible light chain attachment is a nonderivatized methyl group that cannot result from ester hydrolysis. Parallel studies on LPO demonstrate that 6) the two acidic residues in the LPO polypeptide that bind the heme group are Asp-125 and Glu-275, and 7) both of these residues are bound to the heme group in the majority of LPO molecules. Our findings represent the first biochemical data on heme attachmement in EPO, and our data on EPO and LPO are relevant to a long standing controversy on the nature of the heme attachment in mammalian peroxidases.
With isolated EPO bispeptides (Table I), we provide evidence for heme attachment through esters. First, the peptide masses are in perfect agreement with linkage of a heme b prosthetic group by two esters (Table IV). Second, the bonds are susceptible to cleavage by ammonia (Table II). Concordant results were obtained with bispeptides derived from LPO (Tables III  and IV). Previously, based on biophysical methods, ester bonds have been proposed for the heme linkage in MPO and LPO. Principally from a high resolution MPO crystal structure (25), and from spectroscopy of LPO peptides (24), two ester bonds were proposed. Recent infrared difference spectra of MPO, LPO, and EPO also point toward ester linkage (23).
The sites of heme attachment in EPO were unambiguously identified by sequence analysis of peptides converted by aminolysis as Asp-93 of the light and Glu-241 of the heavy chain (Table II), and in a similar experiment, the corresponding residues of LPO were identified as Asp-125 and Glu-275. This is the first biochemical identification of specific residues engaged in heme linkage for any of the peroxidases. Prior to this study, only LPO had been studied at the peptide level, and the peptides 273 ASEQIL and 121 GQIVDHDLDFAPETEL could be released by alkaline hydrolysis from an LPO bispeptide (24). The first of those peptides suggests that Glu-275 is engaged in heme binding; the second leaves several candidate residues, including the Asp-125 identified here. With MPO, two candidate heme binding acidic residues were pointed out from the crystallographic structure of this protein, and for the first time convincing evidence for binding by esters was presented (25). Neither LPO nor EPO contains a residue equivalent to the methionine of MPO that is engaged in sulfonium ion linkage of the heme group (25). In conclusion, two esters are common to heme linkage in MPO, LPO, and EPO. Iron-protoporphyrin IX, heme b, has methyl groups at C-1 and C-5. Here, heme b is shown with ester linkages through hydroxylated methyl groups at C-1 and C-5, as proposed for MPO (25). Fragments of the polypeptide chains of EPO and LPO are shown above and below the heme structure with indication of the Glu and Asp residues engaged in ester linkages (Table II) (Table IV).  Tables I and II for identity of peptides. b Calculated peptide masses. All Glu and Asp residues are considered as uncharged acids.
c Observed mass minus peptide mass. The two variants of the heme monopeptides are mono-CH 2 OH, which results from hydrolysis of the ester at C-5, and mono-CH 3 , which originates from an EPO molecule where this ester had never been formed. Bis peptides, with both esters formed, are indicated.
Unreduced SDS-PAGE of EPO reveals that the majority of EPO exists in a form in which the light and heavy chain are not covalently linked (Fig. 1, lane 2). We are not aware of any publication that shows the result of unreduced SDS-PAGE of purified EPO. However, chain separation under these conditions was mentioned in one report (16). After incubation with reductant, the chain separation was complete (Fig. 1, lane 1), compatible with the earlier hypothesis, now rejected, of heme linkage by disulfide bond(s) (18). In denaturing gel filtration, the partial chain separation was also evident, and this experiment further demonstrated that the heme group is never bound to the EPO light chain alone (Fig. 2). After incubation with reductant, the protein chains are fully separated, and the heme group elutes separately (not shown). We conclude that the esters can be broken by incubation with reductant, possibly in a relatively rapid reaction of trans-esterification resulting in thioesters. However, the heme group itself also seems to be modified over time, but less rapidly, because its absorption at 398 nm decreased gradually in the presence of reductant. This might also be due to precipitation of the heme group.
Following incubation with hydrogen peroxide the protein appeared intact with both esters formed in SDS-PAGE (Fig. 1,  lane 3). This experiment was prompted by an elegant, recent study (22) showing that heme attachment in LPO can occur by an autocatalytic reaction in the presence of hydrogen peroxide. With the two-chained EPO, this can be directly visualized in SDS-PAGE. Because the reaction product was still intact in unreduced SDS-PAGE, even under conditions of prolonged sample boiling, the chain separation described above was not a result of sample preparation. As expected, incubation with reductant rapidly caused complete chain separation (not shown).
A priori, the EPO monopeptides, invariably containing a peptide fragment derived from the heavy chain, would be expected to result from selective hydrolysis of the ester to the light chain. But of five thermolytic peptides from one digest ( Fig. 3A and Table I), only one (E-TL-1, less than 10% abundance) has the expected mass of a peptide with a hydroxymethyl group at C-5 of the heme group. The other EPO monopeptides differ in mass by 16 units, corresponding to the presence of a nonderivatized methyl group on C-5 (Table IV and Fig. 4). An obvious interpretation of this finding is that the ester to the light chain of EPO had never been formed during biosynthesis. Because a methyl group is known to function as a substrate in the autocatalytic reaction (22), the fact that this process can occur with native EPO further supports this interpretation.
The likely presence of a methyl group on C-5 seems to be unique to EPO (Table IV). EPO is also different with regard to the efficiency of the autocatalytic reaction. Compared with the reaction in LPO, more hydrogen peroxide was required to drive the reaction to completion. With LPO, the amount of hydrogen peroxide was four times higher than theoretically predicted (one equivalent of hydrogen peroxide is formally required to produce each ester link), demonstrating that the process is inefficient (22). With EPO, a 10-fold molar excess of hydrogen peroxide was required to form the remaining esters between the light chain and the heme group. Thus, it seems that two factors, efficiency of the reaction and availability of hydrogen peroxide, determine the extent to which the heme group is attached, and it seems that the autocatalytic reaction with EPO is even more inefficient than with LPO. In agreement with this, the majority of heme peptides isolated from LPO were bispeptides (Table III).
The finding that the EPO heavy chain is always ester-bonded and that the light chain is bound to a much lower extent indicates that the formation of an ester at C-1 of the heme group occurs more readily than at C-5, and possibly that esterification at C-5 requires that the ester at C-1 is already formed. But that does not explain why the limited ester hydrolysis seems to occur only at C-5 in both EPO and LPO. We speculate that the propionic acid at C-6 is able to break the Asp-esters with the formation of a lactone. The hydrolytic equilibrium of this seven-member ring would greatly favor the hydroxy acid, which is the species we observed.