Chapter 49

Biosynthesis of Hemes

SAMUEL I. BEALE

INTRODUCTION

Hemes are members of the tetrapyrrole family of biomolecules, which also includes chlorophylls, bilins, and corrins. Hemes are nearly ubiquitous cell components that play essential roles in energy metabolism and oxidative catalysis. In Escherichia coli and Salmonella typhimurium (official designation, Salmonella enterica serovar Typhimurium), hemes are key components of the electron transfer apparatus through which these organisms gain energy. In addition, hemes play important roles as enzyme prosthetic groups in mineral nutrition and oxidative catalysis.

General Outline of Tetrapyrrole Biosynthesis

All biological tetrapyrroles can be arranged as products of a single, branched biosynthetic pathway (Fig. 1). The biosynthetic steps from early precursors to protoporphyrin-based hemes constitute the major, common portion of the pathway, and other steps leading to specific groups of products can be considered branches off the main axis. The pathway is a highly conserved one, and with few exceptions, the biosynthetic intermediates and enzyme-catalyzed reactions are very similar or identical in all organisms in which they have been studied. The existence of a branched pathway with several end products implies a need for regulation to ensure that the products are synthesized in appropriate proportions in response to changing environmental and growth conditions. Some aspects of the regulation of heme synthesis in E. coli and S. typhimurium have been grouped for discussion near the end of this chapter.

Fig. 1General outline of tetrapyrrole biosynthesis. End products that are synthesized by E. coli and/or S. typhimurium are enclosed in boxes.

Source:

Knowledge of the genetics and biochemistry of heme formation in E. coli and S. typhimurium has advanced rapidly in the past few years. Virtually all of the enzymes for heme biosynthesis and their encoding genes have been identified (Table 1), most of the genes have been cloned, their sequences have been determined, and at least one of the enzymes has been crystallized. Preliminary progress toward understanding the regulation of heme biosynthesis has also been made, and this area will undoubtedly yield major advances in the near future.

Table 1Enzymes and genes of tetrapyrrole biosynthesis in E. coli and S. typhimurium

The Hemes of E. coli and S. typhimurium

In contrast to other groups of organisms, E. coli and S. typhimurium each contain a relatively small repertoire of tetrapyrrole end products (Fig. 2): hemes b (protoheme), c, d, and o, which function as the prosthetic groups of respiratory cytochromes and several enzymes, and siroheme, the prosthetic group of sulfite and nitrite reductases. In addition, anaerobically growing cells of S. typhimurium, and possibly E. coli as well, are capable of synthesizing coenzyme B₁₂, a tetrapyrrole-based cofactor that is the subject of chapter 47 of this volume. The present chapter is concerned specifically with the structures and biosynthesis of hemes in E. coli and S. typhimurium. However, inasmuch as all tetrapyrroles share a common biosynthetic pathway, much of the material covered here is applicable to tetrapyrrole biosynthesis in other organisms. Conversely, much of the available information about tetrapyrrole biosynthesis has been gained from studies of other organisms, such as plants, algae, cyanobacteria, and anoxygenic phototrophs, which synthesize large quantities of these compounds. This information is applicable to E. coli and S. typhimurium and is included in the following discussion where appropriate.

Fig. 2Hemes of E. coli and S. typhimurium. The Fischer scheme for the numbering of porphyrin substituents and the conventional labeling system used for pyrrole rings and meso bridge positions are illustrated for protoheme.

Source:

TETRAPYRROLE PRECURSOR FORMATION: ALA BIOSYNTHESIS

δ-Aminolevulinic acid (ALA) can be considered the first universal, committed tetrapyrrole precursor (Fig. 1). Discussion of ALA formation is complicated by the fact that two different biosynthetic routes to ALA exist. The first route of ALA formation to be described is condensation of glycine with succinyl-coenzyme A (CoA), a reaction catalyzed by the pyridoxal phosphate-requiring enzyme ALA synthase (succinyl-CoA:glycine C-succinyltransferase [decarboxylating]; EC 2.3.1.37). Although this pathway was originally assumed to be universal, it now is thought to be confined to eukaryotes that do not contain plastids (e.g., animals, yeasts, fungi) and the α-proteobacteria, a large bacterial group that includes facultative phototrophs in the genera Rhodopseudomonas, Rhodobacter, and Rhodospirillum and other species in the genera Erythrobacter, Methylobacterium, Agrobacterium, Rhizobium, Azorhizobium, and Bradyrhizobium (8, 168).

The second route of ALA formation occurs in plants, algae, and most groups of bacteria, including the enteric bacteria. This pathway is called the five-carbon pathway because ALA is formed from the intact five-carbon skeleton of glutamate (Fig. 3). The currently accepted model for the transformation of glutamate to ALA consists of three enzyme-catalyzed steps. In the first step, glutamate is activated by ligation to tRNA in a reaction identical to the tRNA-charging reaction in protein biosynthesis. Like aminoacyl-tRNA formation in general, this reaction requires ATP and Mg²⁺. Next, the tRNA-bound glutamate is converted to a reduced form in a reaction that requires a reduced pyridine nucleotide. The product of this reduction has been variously characterized as glutamate 1-semialdehyde (GSA) (81), the hydrated hemiacetal form of GSA (81), or a cyclized form of GSA (109). Finally, the positions of the nitrogen and oxo atoms of the reduced five-carbon intermediate are interchanged to form ALA. In contrast to the glycine/succinyl-CoA pathway, which requires one unique reaction, the five-carbon pathway requires two unique reactions in addition to the glutamyl-tRNA synthetase reaction, which also occurs in protein synthesis.

Fig. 3ALA biosynthesis from glutamate via the tRNA-dependent five-carbon pathway, which occurs in plants, algae, and most bacteria, including E. coli and S. typhimurium. The intermediate, GSA, is shown in linear, hydrated, and cyclic forms, which probably coexist as an equilibrium mixture in solution.

Source:

tRNAGlu

ALA formation in extracts of organisms that form ALA from glutamate is blocked by preincubation with RNase A (8, 25, 85, 118, 189, 237). Addition of the RNase inhibitor RNasin plus low-molecular-weight RNA from the same species restores activity. In all species that have been examined, the RNA required for ALA formation is tRNA^Glu(UUC) (204, 206). The same tRNA is used for both ALA and protein synthesis (205). The first anticodon base of tRNA^Glu from E. coli cells and barley chloroplasts is modified to 5-methylaminomethyl-2-thiouridine (205). In E. coli, modification of this base is important for efficient charging with glutamate by glutamyl-tRNA synthetase (220).

Glutamyl-tRNA Synthetase

The same glutamyl-tRNA synthetase (EC 6.1.1.17) is used to charge tRNA^Glu for both protein and ALA synthesis (29, 187). Glutamyl-tRNA synthetase from E. coli is a monomeric enzyme of 56,000 molecular weight (122, 183) that is encoded by the gltX gene (24, 194). Glutamyl-tRNA synthetases from E. coli and Bacillus subtilis contain a tightly bound Zn²⁺ atom that is required for native conformation and catalytic activity (134). The aminoacylation reaction requires ATP and Mg²⁺.

Under some conditions, E. coli glutamyl-tRNA synthetase is copurified in a 1:1 ratio with a 46,000-molecular-weight "regulatory" polypeptide (129) that increases the thermal stability of the synthetase and its affinity for glutamate and ATP (130). The interaction between the two proteins is weak, and some isolations have yielded only the monomeric enzyme (183, 242). The 65,500-molecular-weight B. subtilis glutamyl-tRNA synthetase also is copurified with a 46,000-molecular-weight regulatory subunit (184), but in this case, the protein has been identified as adenylosuccinate lyase (EC 4.3.2.2), an enzyme involved in purine biosynthesis (65). It is of interest that the E. coli regulatory peptide has the same molecular weight as E. coli glutamyl-tRNA reductase, because in Chlamydomonas reinhardtii, glutamyl-tRNA reductase has been reported to form a ternary complex with glutamyl-tRNA synthetase and glutamyl-tRNA (99) (see below). It will be of interest to determine whether the E. coli regulatory peptide has glutamyl-tRNA reductase activity.

Glutamyl-tRNA

The requirement for a low-molecular-weight RNA to support ALA synthesis from glutamate was first shown for extracts of plant chloroplasts (118) and algae (85, 237). The requirements for ATP and Mg²⁺, as well as for RNA, for ALA synthesis from glutamate strongly suggested that glutamyl-tRNA formation is a required step. This conclusion was supported by the finding that for barley chloroplast tRNA^Glu, the 3'-terminal CCA is required for activity in the ALA-forming assay (206). To show that glutamyl-tRNA is a true intermediate, Chlorella vulgaris extracts were incubated with [¹⁴C]glutamate and tRNA^Glu in the presence of ATP and Mg2+ to form [¹⁴C]glutamyl-tRNA. Next, the [¹⁴C]glutamyl-tRNA was isolated and purified. Finally, the purified [¹⁴C]glutamyl-tRNA was incubated with NADPH (which is required for the glutamyl-tRNA reductase reaction) and a partially purified cell extract from which glutamyl-tRNA synthetase was removed. This second incubation yielded [¹⁴C]ALA (4). Under the conditions of the second incubation, free [¹⁴C]glutamate plus uncharged tRNAGlu did not produce [¹⁴C]ALA. Also, as shown initially with Synechocystis sp. strain PCC 6803 extracts, ALA formation from glutamyl-tRNA does not require ATP (205). These results clearly demonstrated that glutamyl-tRNA is a true precursor to ALA and a substrate for glutamyl-tRNA reductase and that the essential role of tRNA^Glu is in the formation of glutamyl-tRNA through the reaction catalyzed by glutamyl-tRNA synthetase.

Glutamyl-tRNA Reductase

Glutamyl-tRNA reductase catalyzes NADPH-linked reduction of tRNA-activated glutamate to form GSA, which is the first committed step unique to ALA formation from glutamate. Affinity-purified glutamyl-tRNA reductase is active in GSA formation from glutamyl-tRNA in the absence of glutamyl-tRNA synthetase and GSA aminotransferase (186).

Under some conditions, glutamyl-tRNA reductase appears to form a complex with glutamyl-tRNA synthetase. In the presence of glutamyl-tRNA, Chlamydomonas reinhardtii glutamyl-tRNA synthetase and glutamyl-tRNA reductase migrate as a single entity on glycerol gradient centrifugation (99). A complex of the two enzymes may facilitate the channeling of glutamyl-tRNA toward ALA biosynthesis and regulate competition with the protein-synthesizing apparatus for glutamyl-tRNA.

Affinity-purified glutamyl-tRNA reductase from Synechocystis sp. strain PCC 6803 and Chlorella vulgaris requires a divalent metal such as Mg²⁺, Mn²⁺, or Ca²⁺ for activity, with optimum activity occurring at 15 mM Mg²⁺ (144). The metal requirements of glutamyl-tRNA reductases from other organisms have not been reported.

It is of interest that glutamyl-tRNA reductase from a given source can use, as substrate, glutamyl-tRNA formed from tRNAGlu from some sources but not others. For example, the barley and Chlorella vulgaris enzymes can use plant and algal tRNAs but not E. coli, yeast, or animal tRNAs (118, 239). On the other hand, the reductase from Chlamydomonas reinhardtii and Chlorobium vibrioforme can use E. coli tRNA (83, 189). The E. coli reductase has a strong preference for E. coli tRNA over tRNA from Chlorobium vibrioforme, Synechocystis sp. strain PCC 6803, Chlorella vulgaris, Chlamydomonas reinhardtii, B. subtilis, or Euglena gracilis (7, 100).

Glutamyl-tRNA reductase is encoded by the hemA gene in organisms that use the five-carbon pathway. hemA mutants of E. coli and S. typhimurium are dependent on ALA for growth (57, 200, 201, 230). Mutation of the E. coli hemA gene results in a deficiency of glutamyl-tRNA reductase (5). The tRNA substrate specificity of glutamyl-tRNA reductase in hemA strains of E. coli complemented with DNA from Chlorobium vibrioforme resembled that of the Chlorobium vibrioforme enzyme rather than that of the E. coli enzyme. This result indicates that the hemA gene encodes a structural component of glutamyl-tRNA reductase that determines the tRNA specificity (7, 138). Yeast cells, which do not form ALA via the five-carbon pathway, do not contain glutamyl-tRNA reductase activity. E. coli hemA, when expressed in yeast cells, yields glutamyl-tRNA reductase activity (231). This result confirms that the hemA product is sufficient to catalyze glutamyl-tRNA reduction. The hemA product from most sources has a predicted molecular weight of approximately 45,000, although the B. subtilis gene product is somewhat larger at 50,800 molecular weight (174).

Despite the similarity of the hemA genes and their encoded peptides, native glutamyl-tRNA reductases isolated from different sources have widely divergent physical properties. On gel filtration columns, B. subtilis glutamyl-tRNA reductase migrates as an oligomer with an apparent molecular weight of 230,000 (207). In contrast, the Chlamydomonas reinhardtii enzyme was reported to be a monomer with a molecular weight of 130,000 (37; S. Krishnasamy and W.-Y. Wang, Plant Physiol. 93:S62, 1990). Synechocystis sp. strain PCC 6803 glutamyl-tRNA reductase was reported to have a native molecular weight of 350,000 (186), but the entity with this molecular weight was later found to be a complex consisting of glutamyl-tRNA reductase and another enzyme, acetohydroxyacid isomeroreductase, with which it is copurified (188). Extracts of E. coli were reported to contain two glutamyl-tRNA reductases with molecular weights of 45,000 and 85,000 (100). For both enzymes, the native and sodium dodecyl sulfate-denatured molecular weights were equal, a fact that indicates that the enzymes are monomeric. The role of the 85,000-molecular-weight enzyme is uncertain. Extracts of hemA mutant E. coli cells were devoid of the 45,000-molecular-weight enzyme but still had active 85,000-molecular-weight enzyme (231). Nevertheless, hemA cells were dependent on exogenous ALA for growth (57, 200, 201, 230). Mutations causing loss of the 85,000-molecular-weight enzyme have not been reported, and the gene that encodes the 85,000-molecular-weight enzyme has not been identified.

GSA

GSA has been chemically synthesized by several methods for use as a substrate for enzyme-catalyzed conversion to ALA (69, 82, 117). Material identical to chemically synthesized GSA accumulates in greening leaves (119, 232) and algal extracts (25) that have been treated with gabaculine, a mechanism-based suicide inhibitor of ω-aminotransferases which blocks conversion of GSA to ALA (see below).

Jordan et al. (109) investigated the structure of GSA in aqueous solution by nuclear magnetic resonance and mass spectroscopy. They concluded that the compound exists as the cyclic ester of the carboxyl group with the hydrated aldehyde group rather than the free or hydrated aldehyde (Fig. 3). The cyclic structure does not contain free aldehyde or carboxylic acid functions and is more compatible with previously reported properties of the chemically synthesized product (stability in aqueous solution, heat stability) than with those of the free α-aminoaldehyde. The cyclic compound, not GSA, was proposed to be the product of glutamyl-tRNA reductase and the substrate of GSA aminotransferase. It seems likely that the cyclic and linear forms of GSA coexist in solution in dynamic equilibrium, analogously to the aldose sugars.

GSA Aminotransferase

GSA aminotransferase (EC 5.4.3.8) catalyzes the conversion of GSA to ALA. Enzymes capable of converting chemically synthesized GSA to ALA have been purified from several plant, algal, and bacterial sources. The transamination reaction requires no added substrate other than GSA. The enzyme contains bound pyridoxal phosphate (6, 30).

GSA aminotransferase is inhibited by the mechanism-based suicide substrate analog gabaculine (3-amino-2,3-dihydrobenzoic acid) (6, 119). Gabaculine reacts irreversibly with enzyme-bound pyridoxal phosphate to form a secondary amine. Gabaculine-treated Chlorella vulgaris GSA aminotransferase was reactivated by gel filtration (to remove gabaculine-pyridoxal phosphate adducts and excess gabaculine) followed by incubation with pyridoxal phosphate (6). Another inhibitor of tetrapyrrole biosynthesis in plants, acetylenic GABA (4-amino-5-hexynoic acid) (55), is a powerful inhibitor of GSA aminotransferase (142; Y. J. Avissar and S. I. Beale, Plant Physiol. 89:S51, 1989). Acetylenic GABA, unlike gabaculine, inhibits Chlorella vulgaris GSA aminotransferase irreversibly (Avissar and Beale, Plant Physiol., 1989), which suggests that it reacts with the protein rather than the cofactor. A third powerful aminotransferase inhibitor, fluoromethyl GABA (4-amino-5-fluoropentanoic acid), inhibits chlorophyll synthesis, presumably by inhibiting GSA aminotransferase (64). Although gabaculine arrests growth and/or pigment accumulation in many organisms that use the five-carbon ALA biosynthetic pathway, it does not affect the growth of E. coli or B. subtilis, even though it blocks ALA synthesis from glutamate and GSA in cell extracts (169). The ineffectiveness in vivo may be explainable by impermeability of the cells to gabaculine.

A proposed reaction mechanism for GSA aminotransferase involves transfer of NH₂ from enzyme-pyridoxamine phosphate to the terminal aldehyde carbon of GSA to form enzyme-pyridoxal phosphate and 4,5-diaminovaleric acid and then transfer of the NH₂ at position 4 of the intermediate back to the cofactor, thereby forming ALA and regenerating the pyridoxamine phosphate form of the cofactor. Implicit in this mechanism is that the nitrogen atom of ALA is derived from a different precursor molecule than the one that supplies its carbon atoms. This prediction was tested by the use of a mixture of ¹³C- and ¹⁵N-labeled glutamate molecules as substrate for conversion to ALA by algal cell extracts. When the heavy-isotope labels were on separate substrate molecules, a significant proportion of the ALA product molecules contained two heavy atoms, indicating that the conversion occurs by intermolecular nitrogen transfer (141, 143). This result supports the proposed reaction mechanism and indicates that the enzyme catalyzes two sequential aminotransferase reactions rather than an aminomutase reaction, even though the substrate and final product have the same atomic compositions.

GSA synthesized by reductive ozonolysis of vinyl GABA (dl-4-amino-5-hexenoic acid) is racemic. However, GSA that is derived from l-glutamate would be expected to have an S configuration at the NH₂-bearing asymmetric carbon atom. Friedmann et al. (62) synthesized both (S)- and (R)-4,5-diaminovaleric acid and showed that the S isomer is the preferred substrate for Synechococcus sp. strain PCC 6301 GSA aminotransferase. Recombinant Synechococcus sp. strain PCC 6301 GSA aminotransferase showed a nearly complete preference for (S)-GSA as a substrate, although the R isomer was able to bind at the enzyme active site and elicit spectral changes (214).

Native GSA aminotransferases from E. coli and Synechocystis sp. strain PCC 6803 have molecular weights of 80,000 (89) and 99,000 (187), respectively. Purified aminotransferases from E. coli and Synechococcus sp. strain PCC 6301 have molecular weights of 40,000 (89) and 46,000 (71), respectively, on denaturing sodium dodecyl sulfate-polyacrylamide gel electrophoresis. Therefore, the native enzyme appears to be a homodimer, like other members of the aspartate aminotransferase enzyme family (146).

GSA aminotransferase-encoding hemL genes (called gsa in plants and algae) from several plants, algae, and bacteria, including E. coli (70) and S. typhimurium (59), have been cloned, and their sequences have been determined. These genes encode highly conserved peptides that have predicted molecular weights of approximately 46,000 (59, 139). The peptides have recognizable similarity to other members of the aspartate aminotransferase enzyme family. All hemL-gsa-encoded peptides have a conserved putative active site containing an essential lysine (72), which is at position 265 in the E. coli and S. typhimurium enzyme polypeptides (59, 70, 88). It is believed that the pyridoxal phosphate cofactor binds to this lysine. Mutagenesis of the putative pyridoxal phosphate-binding lysine inactivates the enzyme (72, 88). The region surrounding this lysine is recognizably similar to the active-site regions of other pyridoxal phosphate-requiring enzymes, including ALA synthase, ornithine-α-ketoglutarate δ-aminotransferase, 2-amino-3-ketobutyrate-CoA ligase, and S-adenosylmethionine:7-keto-8-aminopelargonate aminotransferase (70, 226). On the basis of its overall peptide sequence similarity to 2,2-dialkylglycine decarboxylase and 2-amino-6-caprolactam racemase, GSA aminotransferase has been placed within subgroup II of evolutionarily related aminotransferases that use ω-amino acids as substrates (148).

A Synechococcus sp. strain PCC 6301 mutant that was selected for resistance to gabaculine has a GSA aminotransferase with a lower specific activity than the wild-type enzyme. The mutation that confers gabaculine resistance is M248F (72). Most GSA aminotransferases contain a methionine at this position (139). In the E. coli and S. typhimurium proteins, the homologous methionine is at position 240 (59, 70). It is of interest that GSA aminotransferase from Propionibacterium freudenreichii contains a leucine at the homologous position (156). It has not been reported whether the P. freudenreichii GSA aminotransferase is resistant to gabaculine.

hemL mutants of E. coli and S. typhimurium, unlike hemA mutants, have been reported to be leaky (60, 88). The leakiness of the hemL mutants may be caused by nonenzymatic conversion of GSA to ALA, a reaction that occurs in vitro at high GSA concentration, especially at pHs above neutrality (69, 81, 144).

TETRAPYRROLE FORMATION FROM ALA

Conversion of ALA to PBG

Formation of the first pyrrole in the pathway, porphobilinogen (PBG) (Fig. 4), by asymmetric condensation of two ALA molecules, is catalyzed by the enzyme PBG synthase (ALA dehydratase) (EC 4.2.1.24). PBG synthases from all sources examined each have an octameric structure, and the native molecular weights range from 250,000 to 320,000 (86, 108, 161, 211). The 270,000-molecular-weight native E. coli enzyme has a subunit molecular weight of 36,500, a K_m of approximately 800 μM, and a pH optimum of 8.5 (217). The hemB gene that encodes PBG synthase in E. coli has been cloned, and its sequence has been determined (53, 131, 132). The encoded peptide has a molecular weight of 35,600, which agrees with the value obtained for the purified enzyme subunit.

Fig. 4Conversion of ALA to uroporphyrinogen III. Eight ALA molecules are shown in the approximate positions that their atoms occupy in the intermediates and the final product. The Fischer scheme for the numbering of porphyrin substituents and the conventional labeling system used for pyrrole rings and meso bridge positions are illustrated for uroporphyrinogen III.

Source:

Single-turnover experiments with PBG synthase from Rhodobacter sphaeroides established that in the reaction, the first bound ALA molecule is the one that contributes the propionic acid side chain of the product (113). The amino group of this ALA molecule is a critical component of the binding site for the other ALA molecule (216). The fact that the pro-S hydrogen atom that is derived from the C-5 hydrogens of ALA is stereospecifically retained in the product indicates that during the formation of PBG, removal of hydrogen to form the aromatic pyrrole ring must occur while the intermediate is attached to the enzyme (1).

An interesting difference among PBG synthase enzymes from different sources is the metal requirement: the enzyme from animals, yeast cells, and bacteria, including the cyanobacterial species Anacystis nidulans R2 as well as E. coli, requires Zn²⁺ for activity(105, 108, 217); the plant enzyme requires Mg²⁺ (133, 211); and the R. sphaeroides enzyme requires K⁺ (31, 161). A comparative study concluded that PBG synthase from all sources has tightly bound Zn²⁺ but that a second, weaker, metal-binding site binds Zn²⁺ in the animal, yeast, and E. coli enzymes and Mg2+ in the plant enzyme (152). The E. coli and plant enzymes were proposed to have a third metal-binding site that stimulates activity approximately twofold when Mg²⁺ is bound (152). The deduced peptide sequence at the putative active site of the enzyme from E. coli (53, 131, 132), B. subtilis (75), A. nidulans R2 (105), the archaebacterial species Methanothermus sociabilis (26), yeast cells (159), and animals (21, 240) contains a consensus Zn²⁺-binding motif, but the homologous site in the enzyme from plants (22, 115, 203) and Chlamydomonas reinhardtii (140) is different and more closely resembles an Mg²⁺-binding motif. Interestingly, the homologous region of the deduced Bradyrhizobium japonicum peptide has a sequence that partially resembles those of both the Mg²⁺- and the Zn²⁺-requiring enzymes (35).

PBG synthase is competitively inhibited by the ALA analogs levulinic acid (162) and 4,6-dioxoheptanoic acid (succinylacetone) (52, 202). The latter compound was reported to react nonenzymatically with ALA to form an even stronger inhibitor, named succinylacetone pyrrole (27). These PBG synthase inhibitors, which inhibit heme and chlorophyll accumulation and cause ALA to accumulate when they are administered to whole cells and tissues, have been useful in physiological studies to show the relationship between ALA formation and end product formation (13, 14) and to permit the isolation of ALA formed from labeled precursors (15, 17, 123, 149).

Conversion of PBG to Uroporphyrinogen III

Uroporphyrinogen III, the last common precursor of all end product tetrapyrroles, is synthesized from PBG by the sequential action of two enzymes, hydroxymethylbilane synthase and uroporphyrinogen III synthase (Fig. 4).

Hydroxymethylbilane Synthase.

Hydroxymethylbilane synthase (PBG deaminase) (EC 4.1.3.8) condenses four PBG molecules to form the first tetrapyrrole, uroporphyrinogen. Hydroxymethylbilane synthase from R. sphaeroides and Euglena gracilis was used to establish that the order of assembly of the four PBG units is ABCD, as they appear in uroporphyrinogen (Fig. 4) (11, 112). The initial product of enzymic catalysis is the linear tetrapyrrole hydroxymethylbilane (preuroporphyrinogen), which in the absence of a second enzyme, uroporphyrinogen III synthase, spontaneously cyclizes to form uroporphyrinogen I. Uroporphyrinogen I is not a normal metabolite, and neither it nor its oxidation product, uroporphyrin I, accumulates except under abnormal conditions. Biosynthesis of the biologically relevant isomer, uroporphyrinogen III, requires the action of uroporphyrinogen III synthase during or immediately after release of hydroxymethylbilane from hydroxymethylbilene synthase (see below).

Hydroxymethylbilane synthase from most sources is a monomer of approximately 35,000 molecular weight, although the Rhodopseudomonas palustris enzyme was reported to have a native molecular weight of 74,000 and may be a dimer (126). The enzyme does not require metal ions for activity (108). The hemC gene, which encodes hydroxymethylbilane synthase, has been cloned, and its sequence has been determined for several bacteria, including E. coli (2, 223); the E. coli enzyme has been crystallized, and its structure has been determined by X-ray crystallography (136). The deduced molecular weight for the E. coli enzyme is 33,900 (223). The K_m for PBG is 5 to 10 μM at pH 7.4 (73).

An initially puzzling report showed that E. coli hemB mutant cells, which were unable to form active PBG synthase, were also deficient in hydroxymethylbilane synthase activity unless PBG was supplied in the medium (228). In a seemingly unrelated observation, developing pea chloroplasts that were incubated with ¹⁴C-labeled ALA accumulated a ¹⁴C-labeled, 43,000-molecular-weight soluble protein (19) that was initially proposed to be a cytochrome c (20). Finally, highly purified recombinant E. coli hydroxymethylbilane synthase was discovered to contain a covalently bound dipyrromethane (dipyrrole) that is an essential cofactor for enzyme activity (77, 114). The cofactor is synthesized from ALA via PBG (77, 114, 233) and is ligated to a cysteine residue (C242 of the E. coli enzyme) (78, 208, 209). Once bound, the dipyrrole cofactor remains permanently attached to the enzyme, and its free α position serves as a ligand for oligomerization of PBG substrate units to form the tetrapyrrole product (233). Scission of the link between the cofactor and the nascent oligopyrrole chain after the hexapyrrole stage is reached releases hydroxymethylbilane and prepares the enzyme-bound dipyrrole cofactor to accept new substrate molecules. The existence of the dipyrrole cofactor explains the previously reported requirement for PBG to obtain active hydroxymethylbilane synthase in hemB mutant E. coli cells (228) as well as the incorporation of label from [¹⁴C]ALA onto a chloroplast protein,which was later identified as hydroxymethylbilane synthase (33).

Uroporphyrinogen III Synthase.

Hydroxymethylbilane is unstable in solution and rapidly cyclizes to form uroporphyrinogen I, a physiologically nonproductive end product. Uroporphyrinogen III synthase (formerly cosynthase) (EC 4.2.1.75) catalyzes the cyclization of hydroxymethylbilane with inversion of ring D to form the type III product. The mechanism of ring inversion has been the subject of intensive investigation and is now generally believed to involve a spiro intermediate (42, 218). Uroporphyrinogen III synthase has been purified from several sources, including E. coli (3). The native enzyme from all sources is a monomer of approximately 28,000 molecular weight and contains no reversibly bound cofactors or metal ions. The E. coli enzyme has a pH optimum of 7.8 and a K_m of 5 μM (3). The hemD gene, which encodes uroporphyrinogen III synthase, has been cloned from E. coli, its sequence has been determined, and the product has been overexpressed (2, 3, 42, 111, 114, 198). The deduced E. coli enzyme polypeptide has a molecular weight of 29,000. The hemC and hemD genes of E. coli and S. typhimurium are adjacent (2, 196) and appear to form an operon (110, 111, 198). Evidence from several organisms suggests that hydroxymethylbilane synthase and uroporphyrinogen III synthase form a complex (12, 79, 190).

THE BRANCH FROM UROPORPHYRINOGEN III TO REDUCED HEMES AND CORRINOIDS

Siroheme is the prosthetic group of many nitrite and sulfite reductases that function in the conversion of the highly oxidized forms of nitrogen and sulfur found in the environment to the fully reduced forms (NH₄ ⁺ and S^2–) that are used in biosynthesis (157, 158, 212). A distinct dissimilatory nitrite reductase, which also contains siroheme in some species, is used in anaerobic nitrite respiration. The two known siroheme-containing enzymes of E. coli and S. typhimurium are NADPH-dependent (assimilatory) sulfite reductase (EC 1.8.1.2), which can also function as an assimilatory nitrite reductase (147), and NADH-dependent (dissimilatory) nitrite reductase (EC 1.6.6.4). Siroheme is structurally and biosynthetically related to the corrin ring of vitamin B₁₂, which can be synthesized by S. typhimurium, and possibly by E. coli as well, under anaerobic conditions.

Formally, siroheme can be derived from uroporphyrinogen III by (i) methylation of the tetrapyrrole ring at positions 1 and 3 to form precorrin 2, (ii) oxidation of precorrin 2 to the tetrahydroporphyrin sirodihydrochlorin by removal of two electrons, and (iii) chelation of Fe²⁺. As its name suggests, the alternative fate for precorrin 2 is conversion to corrins (Fig. 1).

Chemical arguments suggest that the order of the steps of siroheme formation from uroporphyrinogen III is probably that given above: first methylation, then oxidation, and finally Fe²⁺ chelation. Methylation of rings A and B of uroporphyrinogen III to produce precorrin 2 effectively limits subsequent oxidation beyond the tetrahydroporphyrin (dihydrochlorin) state. Oxidation of precorrin 2 to sirodihydrochlorin produces a compound that has the aromaticity and metal-binding properties necessary for efficient chelation of Fe²⁺.

The methyl groups of siroheme are derived from methionine in E. coli cells (213), and precorrin 2 is formed in vitro from S-adenosyl-l-methionine and uroporphyrinogen III in a reaction catalyzed by S-adenosyl-l-methionine:uroporphyrinogen III methyltransferase derived from E. coli (235). Results of experiments with other organisms indicate that the methyl group at position 1 is added before that at position 3 (28, 47).

In Pseudomonas denitrificans, a vitamin B₁₂-producing obligate aerobe, uroporphyrinogen methyltransferase is encoded by the cobA gene (43). A somewhat different methyltransferase is encoded by the cysG genes in E. coli (173, 234) and S. typhimurium (67). It is of interest that cysG is the only known genetic locus specifically associated with siroheme synthesis in E. coli and S. typhimurium. The cysG product is the only uroporphyrinogen methyltransferase in S. typhimurium, as is indicated by the fact that cysG mutant strains are deficient in the synthesis of both siroheme and vitamin B₁₂ (104). cysG encodes a 52,000-molecular-weight peptide. Its COOH-terminal region is similar to that of the smaller, 29,200-molecular-weight peptide encoded by the P. denitrificans cobA gene. The E. coli cysG product is a multifunctional enzyme that catalyzes all steps of the conversion of uroporphyrinogen III to siroheme in the enteric bacteria, and the enzyme has been named siroheme synthase (67, 216). Protoporphyrin ferrochelatase, which is encoded by the hemH gene in E. coli (previously known as visA), is not needed for siroheme biosynthesis (63, 182). hemH mutant strains of S. typhimurium and B. subtilis that are unable to produce protoheme and terminal oxidase hemes can still form siroheme and, in the case of S. typhimurium, vitamin B₁₂ (76, 245).

THE BRANCH FROM UROPORPHYRINOGEN III TO PROTOHEME

Decarboxylation of Uroporphyrinogen III To Form Coproporphyrin III

Uroporphyrinogen decarboxylase (EC 4.1.1.37) catalyzes the decarboxylation of all four of the acetate residues on uroporphyrinogen to yield coproporphyrinogen, which contains methyl groups in place of the acetates. At physiological substrate concentrations, the decarboxylations occur in a specific sequence, beginning at ring D and proceeding clockwise around the macrocycle (137). At higher substrate concentrations, the decarboxylation sequence becomes random (107, 137). Stereochemical analysis of the reaction indicated that the decarboxylations proceed with retention of configuration about the α-carbon atoms; i.e., the lost carboxyl groups are replaced with solvent hydrogens in the same orientation (9).

Uroporphyrinogen decarboxylases from several sources including the photosynthetic bacteria R. sphaeroides and Rhodo-pseudomonas palustris have been purified to homogeneity (107, 125). The enzyme from all sources is a monomer whose molecular weight ranges from 39,000 to 54,000. An apparent exception is the chicken erythrocyte enzyme, which was reported to be a dimer with a native molecular weight of 79,000 (120). Purified R. sphaeroides uroporphyrinogen decarboxylase has a molecular weight of 41,000 and a pH optimum of 6.8 and, like the enzyme from other sources, is capable of decarboxylating both the natural substrate uroporphyrinogen III (K_m = 6 μM) and uroporphyrinogen I (K_m = 1.8 μM) (107). No metal requirements were detected. An extensive study of the effects of point mutations of the yeast uroporphyrinogen decarboxylase-encoding HEM6 (HEM12) gene on the accumulation of decarboxylation intermediates led to the conclusion that a single active site on the enzyme catalyzes all four decarboxylations of uroporphyrinogen (36).

Genes that encode uroporphyrinogen decarboxylase have been cloned from several eukaryotes, including yeast cells and mammals, as well as from E. coli (167), and their sequences have been determined. The E. coli uroporphyrinogen decarboxylase gene, hemE, encodes a 38,800-molecular-weight polypeptide that has regions of high similarity to uroporphyrinogen decarboxylase from other sources. E. coli hemE mutants were initially selected as photoresistant revertants from a visA (hemH) photosensitive mutant that lacks ferrochelatase and accumulates protoporphyrin IX (see below). Although the hemE mutants are less photosensitive than hemH mutants, they do overproduce uroporphyrin III (90).

Oxidative Decarboxylation of Coproporphyrinogen III To Form Protoporphyrinogen IX

Coproporphyrinogen oxidase (EC 1.3.3.3) oxidatively decarboxylates the propionate groups at positions 2 and 4 of coproporphyrinogen III to produce vinyl groups, thereby yielding protoporphyrinogen IX. The enzyme is specific for the III isomer of coproporphyrinogen over the nonphysiological I isomer, although chemically synthesized coproporphyrinogen IV is also decarboxylated by the enzyme (155). Evidence indicating that the 2-propionate is converted before the 4-propionate includes characterization of a 2-monovinyl intermediate in rat liver preparations (54) and the preferential action of the Euglena gracilis enzyme on the chemically synthesized 2-monovinyl porphyrin over that on the ring 4-monovinyl porphyrin (34).

In obligately aerobic organisms, coproporphyrinogen oxidation is an O₂-requiring reaction. Extracts of anaerobically grown R. sphaeroides cells can form protoporphyrinogen anaerobically when they are incubated with coproporphyrinogen, ATP, oxidized pyridine nucleotide, and methionine (221). Similar requirements were reported for anaerobic extracts of yeast cells (180) and B. japonicum (121). Involvement of S-adenosyl-l-methionine in the anaerobic reaction is indicated by its inhibition by S-adenosyl-l-homocysteine (221).

Seehra et al. (210) studied the anaerobic as well as the aerobic coproporphyrinogen oxidase reactions in extracts of R. sphaeroides. The oxidative decarboxylations proceed with specific loss of the pro-S β-protons of the propionate groups and retention of the pro-R protons. A reaction mechanism involving pyrrolic N-assisted removal of single protons as hydride ions from the β-carbons of the propionate groups was proposed. The α-protons of the propionate groups do not appear to be involved: in the reaction catalyzed by an avian blood extract, both of the α-protons of both propionate groups were retained on the terminal carbon atoms of the protoporphyrinogen vinyl groups (248).

The E. coli coproporphyrinogen oxidase-encoding hemF gene has been cloned by complementation of a yeast HEM13 mutant (227). The E. coli hemF product has a predicted molecular weight of 34,300 and is 43% identical to the yeast HEM13 product. S. typhimurium contains two genes, hemF and hemN, either of which can support aerobic heme synthesis (245). The hemF product has a predicted molecular weight of 34,400 and is 44% identical to the yeast HEM13 product and 90% identical to the E. coli hemF product (246). Because hemN mutants accumulated coproporphyrinogen III and were auxotrophic for protoheme only when grown anaerobically, Xu et al. suggested that the hemN product is an anaerobic coproporphyrinogen oxidase (245). The predicted S. typhimurium hemN product is a 52,800-molecular-weight peptide that is 38% identical to the R. sphaeroides hemF product but has little if any similarity to the S. typhimurium hemF product (247). Because R. sphaeroides cells in which hemF was disrupted were unable to form bacteriochlorophyll under anaerobic inducing conditions but were able to grow aerobically, it was proposed that in R. sphaeroides, the hemF product is an anaerobic coproporphyrinogen oxidase dedicated to bacteriochlorophyll synthesis (41). Interestingly, the predicted amino acid sequences of the R. sphaeroides hemF and S. typhimurium hemN products are significantly similar (35% identity) to a portion of the Rhizobium phaseoli and Rhizobium leguminosarum nifD gene product, which is the α-subunit of nitrogenase.

Oxidation of Protoporphyrinogen IX To Form Protoporphyrin IX

Protoporphyrinogen oxidase (EC 1.3.3.4) catalyzes the removal of six electrons from the tetrapyrrole macrocycle to form protoporphyrin IX in the last biosynthetic step that is common to hemes and chlorophylls. In obligately aerobic organisms, O₂ is the electron acceptor and is required for enzyme activity. In contrast, the R. sphaeroides protoporphyrinogen oxidase cannot use O₂ directly as an oxidant; instead, protoporphyrinogen IX oxidation is coupled to the respiratory electron transport chain (98). The same is true for the protoporphyrinogen oxidase reaction in anaerobic E. coli cells, which is coupled to the reduction of nitrate or fumarate (96, 97).

Protoporphyrinogen oxidases from R. sphaeroides and E. coli are associated with membranes (94, 98). Active protoporphyrinogen oxidase from eukaryotic organelles can be made soluble by detergents, but the R. sphaeroides and E. coli enzymes are inactivated by detergents (94). Detergent-treated protoporphyrinogen oxidase from the anaerobe Desulfovibrio gigas was reported to have a native molecular weight of 148,000 and to contain three dissimilar subunits of 12,000, 18,500, and 57,000 molecular weights (124). In contrast to the membrane-associated enzymes discussed above, the B. subtilis protoporphyrinogen oxidase is soluble (45).

The E. coli protoporphyrinogen oxidase-encoding hemG gene has been cloned, its sequence has been determined, and the product has been expressed (197). The cloned gene complements a protoporphyrinogen oxidase-deficient E. coli mutant, and the expressed hemG product has protoporphyrinogen oxidase activity. E. coli hemG encodes a 21,200-molecular-weight membrane-associated peptide that is considerably smaller than the protoporphyrinogen oxidase peptides from plants, yeast cells, and animals, which have molecular weights ranging from 36,000 to 65,000.

Protoporphyrinogen IX is oxidized to protoporphyrin IX nonenzymatically in aerated solutions, and long-term cultures of protoporphyrinogen oxidase-deficient E. coli mutants are also able to form small amounts of protoporphyrin IX and protoheme-containing enzymes (195). Under conditions of catabolite repression, E. coli cells accumulate a pigment that has a light absorption peak at 503 nm. This pigment has been identified as tetrahydroprotoporphyrin IX, and it may be produced by nonenzymatic partial oxidation of protoporphyrinogen IX, which accumulates when protoporphyrinogen oxidase is repressed (181).

The enzymatic removal of the four meso-bridge hydrogens from protoporphyrinogen IX occurs with an interesting stereospecificity. In the transformation of PBG to protoporphyrinogen IX, the meso-bridge methylene carbon atoms are derived from the aminomethyl carbon (C-11) of PBG (Fig. 4). If the protoheme is produced from PBG that has been labeled nonstereospecifically with ³H at C-11, then all four meso positions of the porphyrin moiety are equally labeled with ³H (105). However, if the PBG is labeled stereospecifically at the pro-S hydrogen at C-11, then the product contains ³H only at the β-meso position (the one between rings B and C). Given the fact that the incorporation of the four PBG groups into the original tetrapyrrole, hydroxymethylbilane, is an oligomerization that probably involves a single reaction mechanism, and given that cyclization of the hydroxymethylbilane to form the porphyrinogen macrocycle does not seem to cause loss of label from the γ and δ meso methylene hydrogens, it seems likely that the C-11 pro-S hydrogen of PBG will be found on the same face of the porphyrinogen at all four meso positions. Therefore, the hydrogen that is removed from the β-meso position must be removed from the face of the molecule opposite to that from which the hydrogens at the α, γ, and δ meso positions are removed. It was hypothesized that three of the meso hydrogens are lost by an oxidation process occurring on one face of the molecule and that the fourth proton is lost from the other face by a tautomerization reaction (105). Confirmation of this hypothesis will require determination of whether the uroporphyrinogen III synthase reaction affects the stereochemistry of the γ and δ meso methylene hydrogens.

An interesting difference between E. coli protoporphyrinogen oxidase and the enzyme from eukaryotic sources is the sensitivity to inhibition by diphenyl ethers, including the widely used herbicide acifluorfen-methyl. These compounds achieve their herbicidal action by inhibiting protoporphyrinogen oxidase in the chloroplasts. As a consequence, protoporphyrinogen IX accumulates and diffuses out of the chloroplasts and into the cytoplasmic membrane, where it is nonspecifically oxidized to protoporphyrin IX. The accumulated protoporphyrin IX causes the cells to become photosensitive, and treated plants die when exposed to daylight. Protoporphyrinogen oxidase from E. coli and B. japonicum is not inhibited by acifluorfen-methyl (95), in contrast to the enzymes from animals, yeast cells, and plants, which are inhibited by micromolar concentrations of the herbicide (32). It may be possible to engineer transgenic plants to be resistant to diphenyl ether herbicides by introducing a bacterial protoporphyrinogen oxidase gene (197).

Chelation of Iron To Form Protoheme

The final step in protoheme formation, insertion of Fe²⁺ into protoporphyrin IX, is catalyzed by ferrochelatase (protoheme ferrolyase; EC 4.99.1.1). In addition to its physiological substrates, ferrochelatase can use Zn²⁺ and Co²⁺ as the metal substrate and deuteroporphyrin IX, mesoporphyrin IX, and hematoporphyrin IX as the porphyrin substrate (106, 179). Ferrochelatase from most sources is an intrinsic membrane protein that requires detergents or chaotropic agents to make it soluble (44). An apparent exception is the B. subtilis ferrochelatase, which was reported to be a soluble enzyme (74). Detergent-solubilized ferrochelatase from R. sphaeroides has no detectable chromophoric prosthetic groups, the pH optimum is 7.6 with protoporphyrin IX as the porphyrin substrate, and the K_ms for protoporphyrin IX and Fe²⁺ are 18 and 20 μM, respectively (44). Ferrochelatases from R. sphaeroides (106) and animal cells (191) are inhibited by protoheme at concentrations below 10 μM, an effect that may be physiologically relevant.

Ferrochelatase mutants of E. coli were originally recognized as photosensitive mutants, and the locus was named visA (153). Later, it was determined that the photosensitivity was caused by protoporphyrin IX, which accumulates in response to the absence of ferrochelatase (154, 160). The ferrochelatase-encoding visA gene, renamed hemH, has been cloned from E. coli (63, 153) and S. typhimurium (245). The molecular weight of the polypeptide encoded by the E. coli gene is 38,000, which is slightly less than that of the polypeptides encoded by the homologous animal and yeast genes, which have molecular weights near 41,000, and somewhat more than that of the soluble B. subtilis hemH product, which is 34,000 (74).

It is of interest that ferrochelatase is strongly inhibited by N-alkyl porphyrins such as N-methyl mesoporphyrin IX (49) and N-methyl protoporphyrin IX (171, 172), which act as substrate analogs. N-Alkyl protoporphyrin IX arises naturally in animal liver cells through N methylation of the heme prosthetic group of cytochrome P-450, which occurs as a side reaction during the metabolism of certain drugs. By blocking ferrochelatase, the released N-methyl protoporphyrin IX prevents further heme synthesis and causes severe hepatotoxicity. Although N-alkyl porphyrins have not been reported to occur naturally in other contexts, their ability to block ferrochelatase action in vivo has proven useful in biosynthetic studies (16).

BIOSYNTHESIS OF OTHER PROTOPORPHYRIN IX-BASED HEMES

Protoheme Ligation to Apoproteins To Form Heme c

Cytochromes of the c type, which contain covalently bound heme c (Fig. 2), are widely distributed in organisms, where they function in photosynthetic and respiratory electron transport chains (150). Even though E. coli and S. typhimurium are capable of aerobic respiration, they do not contain soluble cytochrome c or a cytochrome bc ₁ complex analogous to the mitochondrial electron transport chain components. The major cytochrome c in E. coli and S. typhimurium is cytochrome c ₅₅₂, which contains six covalently bound heme groups and is located in the periplasmic space, where it functions as a dissimilatory nitrite reductase (116). Other, poorly characterized c-type cytochromes have been detected spectrophotometrically in E. coli cells growing anaerobically with nitrate or trimethylamine N-oxide as the respiratory substrate (23, 91).

In eukaryotes, apocytochrome c incorporates protoheme by enzyme-catalyzed ligation of the heme vinyl groups to cysteine residues on the protein. A specific ligating enzyme, named cytochrome c heme lyase, has been described in yeast cells and Neurospora crassa (164, 222). In these eukaryotic cells, heme lyase is localized in the inner mitochondrial membrane (61, 163), and ligation requires reduced (ferro)protoheme as a substrate (165). Genes for cytochrome c heme lyase from yeast cells (CYC3) and N. crassa (cyt-2) have been cloned, and their sequences have been determined (50, 51). Although the two encoded peptides are 32% identical, the yeast peptide is considerably smaller than the N. crassa peptide (molecular weights of 29,600 and 38,000, respectively). A second heme lyase, cytochrome c ₁ heme lyase, is responsible for assembly of cytochrome c ₁ in yeast cells (249) and N. crassa (166). Yeast cytochrome c ₁ heme lyase is encoded by the CYT2 gene, which has been cloned and has had its sequence determined. The two yeast heme lyases are 35% identical. The yeast results suggest that each c-type cytochrome may require a different, specific heme lyase for its formation.

It must be stressed that no heme lyase or heme lyase-encoding gene has yet been identified in E. coli or S. typhimurium, and cytochrome c assembly in these species may occur spontaneously. There is a precedent for spontaneous, or autologous, ligation of a tetrapyrrole to an apoprotein cysteinyl sulfur atom: in the biosynthesis of the plant photomorphogenetic pigment phytochrome, ligation of a vinyl group-derived moiety of the tetrapyrrole chromophore phytochromobilin to a specific cysteine residue on the apoprotein occurs spontaneously (48, 56, 128). The phytochrome apoprotein itself may have a chromophore lyase function (56).

Two E. coli genes, cydC and cydD, whose products are homologous to heterodimeric ABC transporter proteins, have been identified (178). Disruption of either of these genes prevents the synthesis of cytochrome bd quinol oxidase (66, 178). The cydC and cydD products are also required for cytochrome c assembly, and these proteins may be involved in heme transport to the periplasmic space, in which both types of cytochrome appear to be assembled (177).

Terminal Oxidase Hemes

E. coli and S. typhimurium contain two O₂-reducing terminal oxidase (quinol oxidase) cytochrome complexes, cytochrome bo and cytochrome bd (127) (Fig. 2). In addition to protoheme, these two terminal oxidases have distinctly different heme prosthetic groups, hemes o and d, respectively. Both heme o and heme d are derived from protoheme.

Heme o has the structure of protoheme to which a 2-farnesylethyl group has been added (185, 244). Because heme o is spectrophotometrically nearly indistinguishable from protoheme, cytochrome o was originally believed to contain only protoheme. However, cytochrome o was later determined to contain one molecule each of protoheme and heme o (192). The cyoE gene of E. coli, when overexpressed, causes conversion of protoheme to heme o in vivo (192), and the overexpressed cyoE gene product catalyzes in vitro condensation of protoheme and farnesyl PP_i to form heme o (193). Moreover, the deduced cyoE product contains a consensus polyprenyl PP_i-binding domain (192). Therefore, the cyoE product is protoheme:farnesyl PP_i farnesyltransferase. In vitro heme o formation requires the presence of reducing agents, a finding that suggests that the reaction requires reduced (ferro)protoheme as a substrate (193).

Heme o is closely related to heme a, which is the prosthetic group of cytochrome aa ₃, the terminal oxidase in animal cells and many bacteria, including B. subtilis. Heme a differs from heme o only in the replacement of the methyl group at position 8 with a formyl group. In B. subtilis, two genes, ctaA and ctaB, are required for heme a synthesis (219). The ctaB gene complements E. coli cyoE mutants. B. subtilis ctaA mutants accumulate heme o instead of heme a, but B. subtilis ctaB mutants accumulate neither heme o nor heme a. Interestingly, the B. subtilis ctaA gene, when expressed in E. coli, causes the accumulation of heme a, even though E. coli normally does not contain this heme (176). These results indicate that ctaB and ctaA encode a farnesyltransferase and a methyl oxidase or oxygenase, respectively. The results also suggest that farnesylation precedes formyl group formation in heme a biosynthesis.

Heme d has a dihydroxychlorin macrocycle structure (215). Although the active form of this heme is believed to be the free dihydroxyl form shown in Fig. 2, the 6-hydroxyl group forms a lactone with the 6-propionate under some isolation conditions. In addition to its presence in cytochrome d, heme d is also found as the prosthetic group of E. coli catalase (EC 1.11.1.6) type HPII. However, heme d from these two sources is subtly different: the two hydroxyl groups on ring C of the cytochrome d prosthetic group are in the relative trans configuration (225), but the hydroxyls are cis in the catalase HPII heme d (40). The absolute configurations of the hydroxyl groups have not been determined for either type of heme d. It can be deduced from the structure of heme d that heme d is derived from protoporphyrin IX or protoheme by hydroxylation. It was proposed (224) and later reported (135) that E. coli catalase HPII catalyzes the formation of its own heme d prosthetic group from protoheme. It is not known whether a similar autologous conversion occurs for the prosthetic group of cytochrome bd. As described above in the discussion of heme c biosynthesis, cytochrome bd formation, like that of cytochrome c, requires several gene products that appear to mediate heme transport across the cytoplasmic membrane into the periplasmic space.

REGULATION OF HEME SYNTHESIS

E. coli and S. typhimurium, like other organisms, are able to regulate the contents and compositions of their hemes, especially those that are components of the respiratory apparatus, in response to environmental signals such as O₂ tension and the presence of other respiratory substrates (10, 80, 92). The regulation of heme biosynthesis has not been as thoroughly studied in E. coli and S. typhimurium as in animals, plants, and photosynthetic bacteria, and the details underlying the regulatory mechanisms remain to be determined. Nevertheless, certain provisional conclusions can be drawn from the available data.

Several lines of evidence converge to support a regulatory model in which the cellular level of available or free protoheme controls the rate of heme synthesis at the level of the first step unique to heme synthesis, the formation of GSA by the action of glutamyl-tRNA reductase. First, E. coli cells that are grown in the presence of ALA accumulate heme (175). This result indicates that ALA formation is the rate-limiting step of heme biosynthesis. Second, recombinant E. coli and S. typhimurium cells that have increased levels of glutamyl-tRNA reductase as a result of overexpression of the hemA gene accumulate porphyrins as well as ALA (heme accumulation was not reported) (38, 79). This result indicates that glutamyl-tRNA reductase is the rate-controlling step of ALA formation. Third, cells that are prevented from synthesizing heme by mutation of the hemH (visA) gene, which encodes ferrochelatase, accumulate protoporphyrin IX and become sensitive to light (160). This result indicates that an interruption of heme formation deregulates heme precursor biosynthesis and that heme, rather than a precursor, is the effector of feedback regulation. Fourth, ALA accumulation in a double-mutant E. coli strain that is heme permeable and lacks PBG synthase is suppressed by the addition of heme to the medium (O. Brathwaite, W. Chen, W. Xiao, C. S. Russell, and S. D. Cosloy, FASEB J. 5:A1543, 1991). The fact that heme reduces ALA formation even when the heme is supplied exogenously corroborates the conclusion that heme, rather than a precursor, is the feedback regulator. Finally, recombinant E. coli cells that have overexpressed levels of the rat heme-degrading enzyme heme oxygenase accumulate sufficient quantities of biliverdin IXα to impart a green color to the cells (93, 241). This result indicates that an increase in the rate of heme catabolism results in a compensatory increase in the rate of heme synthesis.

It is of interest that the cellular response to the presence of multiple copies of the hemA gene introduced on plasmids is strain specific and appears to be related to whether or not the cells are permeable to heme. E. coli, which is normally impermeable to heme (200), was reported to be unable to overexpress hemA more than three- or fourfold, even when the gene was present on a multicopy plasmid (231). More recently, it was found that a heme-permeable E. coli strain produces much higher quantities of ALA and porphyrins than heme-impermeable strains in response to the presence of hemA on a multicopy plasmid (38). The relationship of hemA expression to heme permeability suggests that hemA expression is repressed by high cellular heme concentrations and that cells that are permeable to heme may be able to remove excess heme by excretion or degradation and thereby facilitate hemA overexpression. These results lend further support to the model for feedback regulation of glutamyl-tRNA reductase by heme. However, direct support for the proposed mechanism underlying the relationship of hemA expression to heme permeability awaits measurement of cellular heme levels in heme-permeable and heme-impermeable cells under hemA overexpression conditions.

In agreement with the proposed model for feedback regulation by heme, heme inhibited ALA formation from glutamate in extracts of barley plastids (68), and heme is an allosteric regulator of glutamyl-tRNA reductase activity in algae (84, 236), cyanobacteria (186), and anaerobic phototrophic bacteria (189). In these cases, heme inhibits the enzyme in vitro at micromolar concentrations, which suggests that the inhibition is likely to be significant in regulating the rate of ALA formation in response to the cellular demand for end product tetrapyrroles.

There is conflicting evidence on whether heme inhibits E. coli glutamyl-tRNA reductase. One study reported that 5 μM heme caused nearly complete inhibition of glutamyl-tRNA reductase activity in unfractionated E. coli extract (102). The E. coli strain used in that study was one that accumulates and secretes large amounts of porphyrins when it is grown in the presence of thioglycerol (103) and has correspondingly high glutamyl-tRNA reductase activity (102). In contrast, another study found that heme did not inhibit either of the two glutamyl-tRNA reductases purified from E. coli, even at 100 μM (100). It is of interest that the sensitivity of Chlorella vulgaris glutamyl-tRNA reductase to protoheme inhibition is increased severalfold by physiologically relevant concentrations of glutathione (238). It is possible that heme inhibition of the E. coli enzyme in vitro requires the presence of glutathione or some unidentified factor.

In addition to the effects of heme on ALA synthesis, some preliminary evidence suggests that ALA may repress the ALA-synthesizing system. The growth of an E. coli hemB mutant strain with low but detectable PBG synthase activity was supported by high concentrations of ALA in the medium (170). When these cells were grown on glycerol medium containing ALA, they exhibited little or no ability to form ALA from glutamate in vivo or in cell extracts. Although these results were interpreted by the authors of the report to indicate that the hemB product is somehow required for induction of ALA-synthesizing activity, another interpretation is that the high concentrations of exogenous ALA needed to support growth of the hemB mutant strain repressed the ALA-synthesizing system.

It has been noted that directly upstream from the E. coli hemA gene is a divergently transcribed open reading frame whose product may have a regulatory role in hemA expression (230). This possibility was explored by Murooka and coworkers, who determined the sequence of this region and named it the hemM gene (87). It was initially reported that hemM, and not hemA, encodes an enzyme of the five-carbon pathway and that hemA transformants complemented both hemM and hemL mutations (87). More recently, a somewhat contradictory report concluded that hemA expression is essential for ALA synthesis and that hemM expression alone does not produce ALA (38). The results showed that although hemA expression is sufficient for ALA synthesis, expression of both hemA and hemM produces a higher level of ALA synthesis. The open reading frame of hemM encodes a 23,000-molecular-weight polypeptide that does not appear to be related to either the 45,000- or the 85,000-molecular-weight glutamyl-tRNA reductases described by Söll and coworkers (100, 231). The mechanism underlying the action of hemM has not been determined. Several examples of regulation involving divergent promoters are known (18). In some cases, one transcript influences the expression of the other, and in other cases, the product of one transcript is a transcriptional regulator of the other. It will be of interest to determine whether the effect of the hemM gene requires the translation of its open reading frame.

In addition to regulation of the overall rate of heme synthesis, the relative cellular contents of several different hemes must be regulated. An example of a regulated enzyme catalyzing a biosynthetic step for a specific heme is provided by yeast cytochrome c ₁ heme lyase, which is repressed by glucose (250). The only enzymes that have been identified in E. coli and S. typhimurium that catalyze biosynthetic reactions for specific hemes are cysG-encoded sirohemesynthase, and cyoE-encoded farnesyl PP_i:protoheme farnesyltransferase, which forms heme o.

cysG is the only known gene required for siroheme synthesis in E. coli and S. typhimurium (67). cysG is not a part of the cysteine regulon, which includes the cysJ and cysI genes, which encode NADPH-dependent assimilatory sulfite reductase, but is in an operon with the nirB gene, which encodes NADH-dependent dissimilatory nitrite reductase (173). cysG has a substantial basal level of expression in cells growing on cysteine but is induced severalfold in cells growing anaerobically on nitrite (67). Although the basal expression is apparently sufficient for synthesis of NADPH-dependent sulfite reductase and vitamin B₁₂, the induced level is probably required to support synthesis of the high concentrations of NADH-dependent dissimilatory nitrite reductase in cells adapted to nitrite respiration.

cyoE is located within the cyoABCDE operon that encodes the structural components of the cytochrome bo complex, and its expression is therefore coregulated with that of the apoproteins (39). The entire operon is under the control of a consensus O₂-regulated promoter and is also regulated by catabolite repression (151).

Neither cytochrome c heme lyase nor enzymes capable of catalyzing the formation heme d prosthetic group of cytochrome bd have been identified in E. coli or S. typhimurium. Although such enzymes may be identified in the future, the possibility remains open that both of these hemes are autologously formed from protoheme by the cytochromes themselves. As was discussed above, there are precedents for autologous heme d formation by E. coli catalase HPII (135) and for autologous ligation of a tetrapyrrole to a protein cysteinyl sulfhydryl group in the example of phytochrome biosynthesis (56). Therefore, the formation of hemes c and d may be regulated indirectly as a consequence of the availability of protoheme-accepting apocytochrome molecules.

Although E. coli and S. typhimurium cells contain more heme when growing aerobically than when growing anaerobically (80, 92), there is no evidence that O₂ tension has a direct effect on inducing the formation of heme or ALA. The primary effect of O₂ is probably on the induction of various apocytochromes, and heme synthesis probably increases in response to depletion of the pool of free heme in order to supply prosthetic groups to the apocytochromes. In support of this conclusion, overexpression of a rat cytochrome b ₅ gene in E. coli cells resulted in an increase in total cellular heme content (243). Also, expression of a hemA-lacZ fusion construct was 20-fold higher in a hemA mutant than in wild-type E. coli (46). Although expression of the hemA-lacZ construct was 2.5-fold greater in anaerobic than in aerobic wild-type cells (46), extracts from aerobic and anaerobic cells showed no difference in the rate of ALA formation in assay mixtures that were supplemented with ATP, NADPH, and tRNA (169).

The fact that glutamyl-tRNA synthetase and glutamyl-tRNA reductase in some organisms can form a complex in the presence of glutamyl-tRNA (99) suggests that the complex may function to channel glutamyl-tRNA between the two enzymes and thus direct the activated glutamate toward ALA synthesis. It will be of great interest to determine whether this type of complex is formed by the enzymes of E. coli and S. typhimurium and whether the intracellular heme concentration or other metabolic parameters influence the formation of the complex.

It is of interest that with the exception of hemC and hemD, which appear to form an operon, the genes for the biosynthetic steps from glutamyl-tRNA to protoheme are not grouped on the E. coli or the S. typhimurium chromosome (Table 1). This is in distinct contrast to the situation in B. subtilis, in which all of the known genes for these steps are grouped into two operons, of which one, hemAXCDBL, carries all genes for the steps from glutamyl-tRNA to uroporphyrinogen III, the last common intermediate for reduced and oxidized hemes, and the other, hemEHY, carries all of the identified genes for steps from uroporphyrinogen III to protoheme (74, 75). Why the genes should be arranged in this logical manner in B. subtilis but not in E. coli and S. typhimurium is not understood. Perhaps the separation of the genes in the last two organisms allows a greater degree of flexibility in the regulation of their expression under a wider range of growth conditions, e.g., aerobic versus anaerobic, compared to the simpler regulatory demands in B. subtilis, an obligate aerobe.

UPTAKE OF HEME AND HEME PRECURSORS

Wild-type E. coli and S. typhimurium cells do not appear to have a specific heme uptake mechanism and are impermeable to exogenous heme (199, 200). Heme-permeable mutant strains have been isolated by selecting for growth of heme-deficient strains on nonfermentable substrates in the presence of exogenous heme (101, 145). Heme-permeable strains appear to have defective envelopes, and they exhibit increased sensitivity to antibiotics and detergents. As described above, heme permeability is correlated with higher levels of hemA expression from multicopy plasmids. If the higher expression of hemA is due to relief from heme repression (which has not been directly verified), then it can be concluded that the heme permeability is bidirectional, i.e., that the heme-permeable strains are better able to excrete or degrade endogenous heme than are heme-impermeable cells.

Although wild-type E. coli and S. typhimurium cells are unable to take up heme, they do take up ALA, and ALA can support heme-dependent growth of hemA and hemL strains that are unable to synthesize ALA. Although heme is required for growth on nonfermentable substrates such as glycerol, hemA and hemL mutant E. coli and S. typhimurium strains have been observed to grow slowly in such media, even when ALA is not added (58, 229). This residual growth is probably supported by traces of ALA in the medium. ALA auxotrophic cells that are defective in the periplasmic dipeptide permease system show no growth on glycerol unless the medium is supplemented with higher concentrations of ALA than are needed to support full growth of ALA auxotrophs with intact dipeptide permease systems. Uptake of labeled ALA is increased in a S. typhimurium strain with elevated expression of the dpp operon, which encodes the dipeptide permease system, and l-leucyl-glycine competes with labeled ALA for uptake (58). It can therefore be concluded that ALA is taken up by E. coli and S. typhimurium via the dipeptide permease system. Expression of the dpp operon was not increased in ALA auxotrophs during starvation for ALA (58). The dipeptide permease system may have a physiological role in recovering ALA that is continually lost from cells by leakage.

ACKNOWLEDGMENTS

I gratefully acknowledge research grant support from the National Science Foundation, the Department of Energy, and the U.S. Department of Agriculture. I thank Y. J. Avissar for critically reading the manuscript and making many helpful suggestions.