A purified mutant HemA protein from Salmonella enterica serovar Typhimurium lacks bound heme and is defective for heme-mediated regulation in vivo


  • Editor: Jörg Simon

Correspondence: Thomas Elliott, Department of Microbiology, Immunology and Cell Biology, West Virginia University Health Sciences Center, Morgantown, WV 26506, USA. Tel.: +1 304 293 2676; fax: +1 304 293 7823; e-mail: telliott@hsc.wvu.edu


Archaea, plants, and most bacteria synthesize heme using the C5 pathway, in which the first committed step is catalyzed by the enzyme glutamyl-tRNA reductase (GluTR or HemA). In some cases, an overproduced and purified HemA enzyme contains noncovalently bound heme. The enteric bacteria Salmonella enterica and Escherichia coli also synthesize heme by the C5 pathway, and the HemA protein in these bacteria is regulated by proteolysis. The enzyme is unstable during normal growth due to the action of Lon and ClpAP, but becomes stable when heme is limiting for growth. We describe a method for the overproduction of S. enterica HemA that yields a purified enzyme containing bound heme, identified as a b-type heme by spectroscopy. A mutant of HemA (C170A) does not contain heme when similarly purified. The mutant was used to test whether heme is directly involved in HemA regulation. When expressed from the S. enterica chromosome in a wild-type background, the C170A mutant allele of hemA is shown to confer an unregulated phenotype, with high levels of HemA regardless of the heme status. These results strongly suggest that the presence of bound heme targets the HemA enzyme for degradation and is required for normal regulation.


5-Aminolevulinic acid (ALA) is the product of the first committed step in the heme biosynthetic pathway, which also leads to siroheme and vitamin B12 in Salmonella enterica. Most bacteria, as well as plants and archaea, form ALA in a two-step reaction starting from the C5 skeleton of glutamate charged to glutamyl-tRNA (tRNAGlu). The initial enzyme of the pathway, glutamyl-tRNA reductase (GluTR or HemA), uses NADPH to reduce the tRNA-activated glutamate, forming GSA. GSA is subsequently converted to ALA by GSA-AT, the product of the hemL gene (reviewed in Jahn et al., 1992; Beale, 1996). The latter reaction can proceed slowly in vitro in the absence of enzyme (Hoober et al., 1988), which explains the growth of hemL mutants at about 80% of the wild-type rate in unsupplemented minimal medium (Wang et al., 1997). With ALA supplementation, hemL and hemA mutants grow as well as the wild type.

We use the growth of hemL mutants in the absence or presence of ALA to study the effect of limiting the output of the heme pathway, which then reveals its regulation. Regulation is characterized by a marked instability (half-life≈20 min) of the HemA enzyme during normal growth. Stabilization of the protein occurs in response to heme limitation, and leads to a >10-fold increase in enzyme abundance under these conditions (Wang et al., 1999a). The two ATP-dependent proteases responsible for HemA degradation have been identified as Lon and ClpAP (Wang et al., 1999a). These enzymes are not thought to be limiting when HemA accumulates, and there is no evidence for a protease adaptor acting as RssB does in the RpoS system (Bougdour et al., 2008). This led us to suggest that HemA protein might alternate between protease-sensitive and protease-resistant conformations (Wang et al., 1999b). In one model, cellular redox status would allow the formation of a disulfide bond involving one or more of three cysteine residues in this cytoplasmic enzyme. In the second model, heme would bind directly to the protein. Examples of both mechanisms exist in Alphaproteobacteria and eukaryotic cells (Hou et al., 2006; Landfried et al., 2007). Our objective was to determine whether either of these mechanisms governs HemA regulation in Salmonella.

Here, we demonstrate that purified HemA protein of S. enterica contains noncovalently bound heme. We have also been able to show that a single mutation (C170A) has two effects: it blocks regulation by stabilizing HemA, and it results in the production of protein that does not contain bound heme. We suggest that these effects are related and that they support the regulatory model in which binding of heme to the HemA enzyme in vivo triggers protease attack. Interference with this binding is likely to be part of the mechanism of stabilization.

Materials and methods

Bacterial strains

The strains used in this study are listed in Supporting Information, Table S1; all S. enterica strains are derived from LT2.

Growth of cultures

Cultures were grown in either Luria-Bertani (LB) medium (Chen et al., 1996), modified minimal morpholinepropanesulfonic acid (MOPS) medium (Neidhardt et al., 1974; Bochner & Ames, 1982) containing 0.2% glycerol as the carbon source, or NCE (no citrate E) medium with 0.2% glycerol as the carbon source (Berkowitz et al., 1968). Plates were prepared with nutrient agar (Difco) and 5 g NaCl L−1 or with NCE medium. ALA was used at 2 μM in minimal medium and at 150 μM in a rich medium. Adaptation of hemL mutant strains to growth in the absence of ALA has been described previously (Wang et al., 1997).

Plasmid construction

Techniques for plasmid construction followed standard methods (Maniatis et al., 1982). Mutations and C-terminal truncations were made by PCR and verified by sequencing. Plasmids are also listed in Table S1.

Purification of expressed proteins

Cultures were grown overnight in LB containing ampicillin (100 μg mL−1) and chloramphenicol (20 μg mL−1), diluted 1 : 10 into fresh medium, and incubated at 30 °C for 2 h before induction with isopropyl-β-d-thiogalactopyranoside (IPTG) at a final concentration of 1 mM. After 3 h, cells were harvested by centrifugation. The cell pellet was resuspended in 10-mL lysis buffer [20 mM Tris, pH 8.0, 250 mM NaCl, 10 mM imidazole, and 1 : 100 dilution of Sigma (P8849) protease inhibitor cocktail], and then passed through a French press three times. The extracts were clarified by centrifugation and the supernatants were bound to 2.5 mL nickel-nitrilotriacetic acid (Ni-NTA) resin (Qiagen) on a rocker at 4 °C for 1 h. Ni-NTA was washed twice with buffer containing 5 mM imidazole and then 15 mM imidazole. Protein was eluted in 5 mL of elution buffer (0.5 M imidazole). Purified protein preparations were desalted using PD-10 columns (GE Healthcare).

Immunological detection of proteins

Detection of HemA with the anti-HemA monoclonal antibody by Western blot has been described in detail previously (Wang et al., 1997). The monoclonal anti-FLAG antibody was purchased from Sigma.


The absorption spectra in Fig. 1a were recorded using a DW-2000 UV-Visible spectrophotometer (SLM-Aminco) using the split beam mode, 9.0 nm slit width, and a scan rate of 1.0 nm min−1. The spectra in Fig. 1b were recorded using a Synergy HT Plate Reader (BioTek) measuring absorption at 10-nm intervals from 300 to 650 nm. Cytochrome c (Sigma C7752) and hemin (Sigma H2375) standards were used as controls. Spectra were recorded for undiluted protein, protein diluted 1 : 1 in alkaline pyridine solution (oxidized), and after mixing with a few grains of sodium dithionite (reduced).

Figure 1.

 Purified HemA1−412-His6 exhibits a heme b-type spectrum (a) and the C170A derivative does not (b). The absorption spectrum of purified HemA1−412-His6 (a) recorded before (black line) and after (dark gray line) dithionite addition. The difference spectrum (light gray) is also shown. (b) Purified HemA[C170A] (dashed line) lacks the characteristic heme b peaks exhibited by purified HemA1−412-His6 (solid line).

Pyridine hemochromogen assay

Heme content was determined for purified protein diluted 1 : 1 in an alkaline pyridine solution (0.2 M NaOH, 4.2 M pyridine). A few grains of sodium dithionite were added and the difference in A556 nm and A536 nm of the reduced protein was used to calculate the heme concentration using the emM556−A537 value of 23.4 (Fuhrhop & Smith, 1975). The predicted emM280 for both HemA and HemA1−412-His6 is 30 940 M−1 cm−1 (Pace et al., 1995).

Detection of heme-catalyzed peroxidase activity

Proteins were diluted in duplicate into a standard protein sample buffer with no reducing agent. Beta-mercaptoethanol (β-ME) was added to one of the samples. Samples containing β-ME were boiled for 10 min before loading onto 8% sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) gels. Duplicate gels were loaded with 20 μg of HemA protein or 0.5 μg cytochrome c (Sigma). Following SDS-PAGE, one gel was stained for total protein using Coomassie blue, while proteins in the second gel were transferred to a PVDF membrane for the subsequent detection of peroxidase activity (Dorward, 1993). After transfer, the membrane was rinsed with ∼10 mL PBS for 1 min. Peroxidase activity was detected by covering the membrane with 1 mL each of SuperSignal West Pico (Pierce) reagents for 4–5 min, and then exposing to film.

Transfer of C170A mutation to the chromosome

PCR was performed using the plasmid pTE762 as the template. Integration into the S. enterica chromosome was achieved via linear transformation using a previously published protocol (Wang et al., 1999b) and the results were verified by sequencing.

HemA stability

Cultures grown overnight in minimal glycerol medium at 37 °C were diluted 1 : 50 into the same medium and incubated at 37 °C. At OD600 nm=0.40, protein synthesis was inhibited by addition of chloramphenicol (200 μg mL−1). Aliquots were taken at 0, 30, and 60 min following inhibition and prepared for SDS-PAGE and immunoblot.


HemA Cys→Ala mutants

The HemA protein of S. enterica contains three cysteine residues, C50, C74, and C170, all conserved in Escherichia coli. Of these, only C50, which is essential for catalysis (Moser et al., 1999; Schauer et al., 2002), is conserved among all organisms. Because redox status or disulfide bond formation may be important in HemA regulation, each of the three cysteines of HemA was individually changed to alanine, resulting in the mutants C50A, C74A, and C170A. These were expressed in E. coli from a plasmid bearing the native hemA promoter, but controlled by the lac operator and repressor. Both C74A and C170A complemented an E. coli hemA mutant when expressed at physiological levels, thereby demonstrating that they encode functional proteins. As expected, plasmids encoding either a sequenced amber mutant allele of hemA (Q369Am) or the C50A mutant protein were unable to complement in the same test.

As a first assessment of the regulatory phenotype of the HemA Cys mutants, HemA was analyzed by Western blot of lysates prepared from overnight cultures (Fig. 2a). In a previous report, we observed that HemA protein is undetectable by Western blot in wild-type cultures grown overnight, whereas HemA[KK], a regulatory mutant, is maintained at easily detectable levels (Wang et al., 1999b). HemA[C170A] was nearly as abundant as HemA[KK], whereas HemA levels in C50A, C74A, and wild-type were at or below the limit of detection, suggesting that of the three mutants, C170A alone displays a regulatory defect. To verify this, the Cys→Ala mutants were assessed for correct regulation by comparing HemA levels in the absence and presence of ALA (Fig. 2b). In ALA-supplemented cultures, where the wild type is unstable, HemA levels were much higher in the C170A mutant compared with the wild-type strain and the C74 mutant (Fig. 2b), and slightly higher than HemA[KK] in a similar test (Fig. 2c). We conclude that HemA[C170A] is a regulatory mutant. This effect was further investigated using purified proteins.

Figure 2.

 Western blots of HemA show that HemA[C170A] is a regulatory mutant, with levels of HemA comparable to those of HemA[KK], a known regulatory mutant that was included as a control for stability. (a) An Escherichia coli hemA mutant host strain (TE5812) carrying plasmid pTE740 (hemAp lac o hemA His6), one of its Cys mutant derivatives (C50A, C75A, C170A), or amber mutant Q369Am (lane 369), was grown overnight in minimal glycerol medium containing ampicillin (50 μg mL−1) and ALA (10 μM), and then diluted 1 : 50 into the same medium plus IPTG (35 μM). After 24 h of incubation, samples were prepared for Western blot using anti-HemA (H23) antibody. Strains analyzed: TE7764, TE7767, TE8219, TE8220, TE7929, and TE7884. Lane M represents protein markers. (b) Western blot of plasmid-expressed HemA in adapted (−ALA) and supplemented (+ALA) cultures. (c) Western blot of HemA levels in cultures supplemented with ALA. Strains analyzed: TE7823, TE7824, TE7849, and TE7850. WT, wild type.

Overexpression of HemA

Initial attempts to overexpress either native or His-tagged HemA protein using the standard T7 system were unsuccessful (unpublished data); however, we observed that constructs bearing an amber mutant allele of hemA (Q369Am) allowed overexpression of the truncated protein (Wang et al., 1997), at a high level similar to that observed for other proteins we have purified (e.g. HemL, RpoS). This prompted us to test whether relatively short C-terminal truncations could be overexpressed at high levels as well.

The hemA gene from S. enterica was inserted into a plasmid derived from pET3 under the control of a T7 promoter (Studier & Moffatt, 1986). Various constructs encoded either full-length HemA (amino acids 1–418) or one of several small C-terminal truncations, all bearing a C-terminal His6 tag in addition. Protein overexpression was induced by a standard protocol in E. coli BL21(DE3)/pLysS (Studier & Moffatt, 1986). Analysis of whole-cell lysates from induced cultures showed that while the full-length HemA construct could not be significantly overexpressed, a truncated form of HemA lacking six amino acids from the C-terminus was at least 20-fold more abundant than the native protein when analyzed by Western blot (Fig. 3a).

Figure 3.

 Western bots of HemA show that wild-type HemA cannot be overexpressed, but truncated HemA can be overexpressed and is regulated. (a) Escherichia coli BL21(DE3) carrying plasmids that express either full-length (amino acids 1–418) or C-terminally truncated Salmonella enterica HemA bearing a C-terminal His6 tag were grown in LB medium containing ampicillin (100 μg mL−1) and chloramphenicol (20 μg mL−1) and induced with IPTG (1 mM) 2 h before harvesting cells. Numbers indicate the position of the last amino acid of HemA that is present. An asterisk (*) indicates a 1 : 10 dilution in sample buffer alone, whereas the other samples were diluted 1 : 100 with lysate from the uninduced vector control strain. Strains analyzed: TE7771, TE7772, TE7875, TE7876, and TE7877. (b) Truncated and FLAG-tagged HemA is regulated like the wild type. HemA was expressed from plasmid pTE764 (hemAp lac o hemA FLAG His6) and from plasmids expressing its truncated derivatives in an E. coli hemA+hemL mutant host strain. Cultures were grown in MOPS minimal medium containing ampicillin (50 μg mL−1), IPTG (35 μg mL−1), and 0.2% glycerol as the carbon source in the presence (10 μM) or absence of ALA as described in the text and Wang et al. (1997). HemA was detected by Western blot using the anti-FLAG antibody. A strain with a plasmid encoding HemA truncated at position 408 and lacking the FLAG tag (408-) was included as a negative control. Strains analyzed: TE7910, TE7911, TE7927, and TE7937. Lane M represents protein markers.

It was important to verify that the C-terminal HemA truncations encode functional enzymes and exhibit normal regulation. Plasmid-encoded, truncated, and tagged S. enterica hemA complemented an E. coli hemA mutant. Regulation in response to heme was tested by Western blot (Fig. 3b). To eliminate the possibility that a partial defect in the enzyme activity of the truncated proteins could affect the results of the test, an E. coli host that is wild type for hemA was used, and the plasmid-encoded proteins were specifically detected by an additional C-terminal FLAG tag. Truncated HemA exhibited normal regulation in response to heme limitation.

Purified HemA1−412-His6 contains heme

His-tagged C-terminally truncated HemA was purified by Ni-NTA affinity chromatography. The purified protein was red in color, suggesting the presence of bound heme. The absorption spectrum of purified HemA protein (Fig. 1a) contains features characteristic of heme, including a prominent peak at 424 nm (the Soret band). Upon reduction with Na-dithionite, the peak at 424 nm became sharper and shifted toward a longer wavelength (426 nm), and two other peaks appeared: one at 530 nm and another at 560 nm. The spectrum of reduced heme (hemin), which was used as a control, was very similar to that of the purified protein (data not shown). Three separate protein preparations averaged 0.055 mol heme mol−1 protein monomer as determined by the pyridine hemochromogen assay.

Purified HemA1−412 [C170A]-His6 lacks heme

HemA1−412 [C170A]-His6 was purified according to the same protocol as that used for HemA1−412-His6. The C170A mutant protein was colorless, suggesting that it is unable to bind heme. The absence of heme was also demonstrated by its absorption spectrum, which lacks the peaks characteristic of heme-containing proteins (Fig. 1b).

Treatment with denaturant

The HemA spectrum is that of a b-type heme; this class of molecules is attached noncovalently. Treatment with the strong denaturant, 6 M guanidine-HCl, removed a maximum of 7% of heme from HemA, and in two trials, failed to remove any (Supporting Information). The ability to retain noncovalently bound heme in the presence of strong denaturants has been documented for other proteins (Hargrove & Olson, 1996; Wójtowicz et al., 2009). Although these results demonstrate a strong association between heme and HemA, covalent binding cannot be inferred from this assay. Thiol reagents, which have been used to distinguish covalent heme-protein bonds, are incompatible with Ni-NTA. The nature of the association between heme and HemA was further examined using a different method.

Heme-associated peroxidase activity

Heme-associated peroxidase activity, which can be measured by standard ECL reagents (a Western blot without the antibody; Dorward, 1993), detects heme-binding proteins (such as cytochrome c). Purified proteins were separated by SDS-PAGE and then assessed for heme-associated peroxidase activity. Duplicate gels were stained for protein with Coomassie blue (Fig. 4). Cytochrome c was included as a positive control for a protein with covalent attachment of heme, and its heme-associated peroxidase activity was detected at the expected position based on the mass of the protein (∼13 kDa), independent of treatment with thiol reagents.

Figure 4.

 ECL detection of heme-associated peroxidase activity in purified HemA1−412-His6. Purified HemA proteins (20 μg) and cytochrome c (0.5 μg) were run in duplicate on 8% SDS-PAGE gels; one was stained with Coomassie blue (left panel) and the other was processed for the detection of heme-associated peroxidase activity (right panel). Samples in lanes marked ‘+’ were prepared in sample buffer containing β-ME and boiled, while those marked ‘−’ were unheated and prepared in buffer lacking β-ME. Lane M contains Precision Plus Protein Standards (Bio-Rad), and the molecular masses (kDa) for its component proteins are given for reference.

Both HemA1−412-His6 and HemA1−412 [C170A]-His6 were detected in Coomassie-stained gels at the predicted molecular mass of ∼46 kDa (Fig. 4, left lanes); however, peroxidase activity was only detected for the HemA1−412-His6 and only in unheated samples lacking both dithiothreitol and β-ME. Any one of three treatments, dithiothreitol, β-ME, or boiling, abolished the signal (Fig. 4, and data not shown), indicating that heme is not covalently bound. HemA1−412 [C170A]-His6 failed to produce a detectable signal under any of the conditions tested. Three bands are observed for the untreated wild-type sample. The smallest and most abundant band corresponds to HemA protein. The bands above it are likely aggregates as observed in other studies (Schroder et al., 1992; Verkamp et al., 1992; Schauer et al., 2002).

HemA[C170A] is stable in vivo

According to one model, heme binding to HemA protein sensitizes it to proteolytic attack. The combined observations of a regulatory defect and absence of bound heme in purified C170A led us to predict that the mutant would exhibit increased stability over wild-type HemA. Isogenic strains expressing wild-type and C170A mutant HemA in a single copy from the native locus in the S. enterica chromosome were analyzed by Western blot after inhibition of protein synthesis (Fig. 5a). Wild-type HemA was present at lower levels than the mutants and was detectable only at the initial time point. HemA[KK], included as a positive control, remained stable over the time course of the experiment. In support of the model, the C170A mutant was nearly as stable as HemA[KK] (Fig. 5b).

Figure 5.

 HemA C170A expressed from the native locus is stable and defective for regulation. Wild-type and mutant HemA proteins were expressed from the Salmonella chromosome in single copy. (a) Western blot of HemA levels at 0, 30, and 60 min following inhibition of protein synthesis. (b) Densitometry analysis (imagej) of wild-type (filled circles), HemA[KK] (filled squares) and HemA[C170A] (open circles) HemA. Strains analyzed: TE6825, TE7700, and TE9287.


Our current understanding of heme biosynthesis by the C5 pathway in bacteria involves two different regulatory mechanisms: feedback inhibition of enzyme activity by heme and post-translational control of enzyme abundance by proteolysis (Wang et al., 1997; Wang et al., 1999a). Furthermore, HemA enzyme from Chlorobium vibrioforme, as well as some eukaryotic enzymes, mostly expressed in E. coli, contains tightly bound heme (Vothknecht et al., 1996; Srivastava & Beale, 2005; Srivastava et al., 2005). This work focuses on S. enterica, the species in which regulation by proteolysis was discovered (Wang et al., 1999a).

We were unsuccessful in previous attempts to use the T7 RNA polymerase system to overexpress Salmonella HemA, which is 94% identical to the E. coli enzyme. Jahn's group succeeded with the E. coli HemA enzyme by coexpressing the protein with the chaperones DnaJK and GrpE (Schauer et al., 2002). Analysis of the purified E. coli enzyme showed (1) no copurifying prosthetic group detectable by spectroscopy and (2) no inhibition of enzyme activity by heme in vitro. This apparent difference between the Salmonella and E. coli proteins is currently unexplained, but we suggest that it might be related to the chaperones used in the latter case.

We discovered fortuitously that C-terminally truncated derivatives of HemA can be overexpressed using the T7 system and purified easily. The His6 tag construct used for most of this work is lacking the terminal six amino acids. The truncated derivatives are regulated like the wild type (Fig. 2). We investigated this system further, particularly because the purified preparation of otherwise wild-type protein was red in color, and spectroscopy showed the presence of heme, likely a b-type heme (Fig. 1a).

The second important finding is that C170 is essential both for the tight binding of heme to HemA protein, leading to copurification as observed in the overexpression experiments, but also for correct (i.e. wild type) regulation when the gene is expressed from the native hemA locus in the S. enterica chromosome, with no other differences from the wild type (no truncation). The increased abundance and significantly extended half-life (Figs 3 and 5) clearly establish C170A as a regulatory mutant. These results suggest that the presence of tightly bound heme may tag HemA protein for degradation. Tagging fails in the mutant, and the protein is thereby stabilized.

The crystal structure for HemA from Methanopyrus kandleri, a thermophilic archaeon, has been resolved (Moser et al., 2001). An N-terminal catalytic domain contains the essential conserved cysteine residue (C50 in S. enterica), a second domain binds NADPH, and the extreme C-terminus is implicated in dimer formation (Lüer et al., 2005; Nogaj & Beale, 2005). Among characterized HemA proteins, only E. coli and S. enterica possess a cysteine at position 170; the homologous position in HemA from most other sources contains valine (Brody et al., 1999).

The biochemical characterization of the association of heme with HemA is only preliminary. We observed very tight binding (stable to 6 M guanidine-HCl), and yet it is sensitive to thiol reagents. Heme is bound only to a small fraction of HemA (the heme : protein ratio is ∼1 : 20). The connection between these observations and the stoichiometric (1 : 1) heme present in C. vibrioforme HemA is not clear. Because the residue C170 essential for regulation and heme binding in Salmonella is not conserved in the Chlorobium gene, we suggest that the mechanism of binding might be substantially different in the two proteins.


This work was supported by Public Health Service grants 6M40403 and GM63616. The authors thank Andrew Shiemke and Courtney Williamson for their assistance with absorption spectrometry.