Enzymological and structural characterization of Arabidopsis thaliana heme oxygenase-1.

Arabidopsis thaliana heme oxygenase-1 (AtHO-1), a metabolic enzyme in the heme degradation pathway, serves as a prototype for study of the bilin-related functions in plants. Past biological analyses revealed that AtHO-1 requires ferredoxin-NADP+ reductase (FNR) and ferredoxin for its enzymatic activity. Here, we characterized the binding and degradation of heme by AtHO-1, and found that ferredoxin is a dispensable component of the reducing system that provides electrons for heme oxidation. Furthermore, we reported the crystal structure of heme-bound AtHO-1, which demonstrates both conserved and previously undescribed features of plant heme oxygenases. Finally, the electron transfer pathway from FNR to AtHO-1 is suggested based on the known structural information.

NADPH‐cytochrome P450 reductase

deferoxamine

ferredoxin‐NADP+ reductase

heme oxygenase

isothermal titration calorimetry

size‐exclusion chromatography

Heme oxygenases (HOs; EC1.14.14.18) catalyze oxidative conversion of heme to biliverdin and are widely distributed in plants, animals, and prokaryotes [1, 2]. Biliverdin is the universal precursor for plant phytochrome chromophore, animal bilirubin, and cyanobacterial phycobilins [3, 4, 5]. Heme oxidation catalyzed by HOs also produces iron ion and carbon monoxide (CO), and hence HOs play key roles in iron metabolism and CO signaling [6, 7]. Canonical HOs share a conserved HO fold primarily consisting of α‐helices. The model plant Arabidopsis thaliana has four HOs (AtHO‐1–AtHO‐4), which are clustered into HO1 subfamily (AtHO‐1, AtHO‐3, and AtHO‐4) and HO2 subfamily (AtHO‐2) [8, 9]. The main differences of HO2 from HO1 are the replacement of the ligating His with Arg and the presence of a spacer sequence that is rich in Glu and Asp residues. An in vitro study of AtHO‐1–AtHO‐4 has shown that the HO1 subfamily members are authentic HOs with similar biochemical parameters, while AtHO‐2 lacks such an activity [10].

AtHO‐1 is a prototype of plant HOs and has been studied extensively. A long hypocotyl phenotype has been mapped to the HY1 locus, which harbors the AtHO‐1 gene and whose mutation causes a deficiency of phytochrome chromophore [11, 12, 13]. Mutation of the AtHO‐1 gene also generates a genomes uncoupled (gun) phenotype showing perturbed plastid‐to‐nucleus signaling [14, 15]. In addition, it is involved in drought tolerance by modulating the stomatal aperture and possibly acting as a negative regulator of drought stress signaling [16].

Until recently, most structural studies of canonical HOs were focused on mammalian and bacterial HOs [17]. The known structures include human HO‐1 and HO‐2 [18, 19, 20, 21, 22, 23], rat HO‐1 [24, 25, 26, 27], two cyanobacterial HOs from Synechocystis sp. PCC 6803, SynHO‐1 and SynHO‐2 [28, 29], and four pathogenic bacterial HOs, HemO from Neisseria meningitidis [30], Pseudomonas aeruginosa HO [31], HmuO from Corynebacterium diphtheriae [32, 33, 34], and Leptospira interrogans HO (LiHO) [35]. Mammalian HOs use NADPH‐cytochrome P450 reductase (CPR) as redox partner to obtain electrons from NADPH, and the electron transfer path has been revealed by the complex structures of rat CPR–HO‐1 [36, 37]. Bacterial and plant HOs prefer ferredoxin‐NADP+ reductase (FNR) and ferredoxin, but the interaction between these HOs and their redox partner(s) is unclear. The putative ferredoxin‐binding site of SynHO‐1/‐2 has been proposed based on molecular surface analysis [28, 29]. LiHO needs only FNR for the reaction with ferredoxin being dispensable, and a transient FNR–LiHO complex model has been proposed [35].

Very recently, crystal structure of soybean Glycine max HO‐1 (GmHO‐1) was reported and the interactions between GmHO‐1 and ferredoxin were characterized [38], offering the first structural insight into the catalytic mechanism of plant HOs. GmHO‐1 has a sequence similarity of 72% with AtHO‐1 [39], which implies that these two HOs are highly structurally conserved. Despite their similarity and the requirement of reduced ferredoxin as the primary redox partner, the structure–function relationship for plant HOs still await further characterization [40, 41]. In addition, the nonenzymatic heme degradation process to produce the biliverdin isomers, that is, a process referred to as coupled oxidation and commonly occurring in heme‐binding proteins such as myoglobin and cytochrome b 5 [42, 43], has not been tested for AtHO‐1. In this work, we characterize the heme‐binding and ‐degrading activities of AtHO‐1, differentiate the coupled oxidation and enzymatic processes, describe the heme–AtHO‐1 structure, and present a previously unreported feature of plant HOs.

Materials and methods

Protein expression and purification

2Gene sequence (The Arabidopsis Information Resource database identifier: At2g26670) encoding the mature AtHO‐1 (residues 55–282) was amplified by PCR and then inserted into the Novagen pET‐28a(+) vector between the Nco I and EcoR I restriction sites. The resulting construct encoded an N‐terminal His tagged AtHO‐1 and was transformed into Escherichia coli BL21(DE3) cells. Expression of the recombinant protein was induced by 200 μm of isopropyl β‐d‐thiogalactoside when the cell culture had an optical density of 0.8 at 600 nm. The culture was then grown at 16 °C for 16 h, pelleted, and resuspended in the lysis buffer (200 mm of NaCl and 20 mm of Tris–HCl, pH 7.5) with additional 5 mm of imidazole. Cell resuspension was lysed by sonication in an ice water bath and the debris was removed by centrifugation. The cleared lysate was incubated with nickel nitrilotriacetic acid agarose resin (QIAGEN, Shanghai, China) at 4 °C for 1 h, packed into a column, and washed with the lysis buffer supplemented with 50 mm of imidazole. Recombinant AtHO‐1 was eluted with 200 mm of imidazole in the lysis buffer. For further purification by size‐exclusion chromatography (SEC), the eluate was concentrated by ultrafiltration and then loaded onto a HiLoad 16/60 Superdex 200 column (GE Healthcare, Beijing, China) equilibrated and eluted with the lysis buffer. Peak fractions corresponding to recombinant AtHO‐1 were collected, pooled, and analyzed by SDS/PAGE. Recombinant Zea mays ferredoxin from and A. thaliana FNR were obtained as previously described [44].

To prepare the heme–AtHO‐1 complex, hemin (ferric chloride heme) was first dissolved in dimethyl sulfoxide to obtain a stock solution, and then hemin and the purified AtHO‐1 were mixed in a molar ratio of 2 : 1. The mixture was incubated at 4 °C for 1 h before being concentrated and applied onto a HiLoad 16/60 Superdex 200 column equilibrated and eluted with the lysis buffer. Peak fractions corresponding to the heme–AtHO‐1 complex were collected, pooled, and concentrated for coupled oxidation assay and crystallization.

Isothermal titration calorimetry

Isothermal titration calorimetry (ITC) experiment was performed on a MicroCal iTC200‐2 calorimeter (Malvern Panalytical, Westborough, MA, USA) at 25 °C. Hemin was dissolved to a concentration of 20 mm in 0.1 m NaOH, and then diluted ×20 with titration buffer (150 mm of NaCl and 100 mm of Tris–HCl, pH 7.5). The purified AtHO‐1 solution was changed to the titration buffer by ultrafiltration. The titration series consisted of 20 injections of hemin solution (first injection of 0.4 μL and subsequent injections of 2 μL) into the AtHO‐1 solution.

Coupled oxidation assay

The purified heme–AtHO‐1 complex was diluted to 10 μm for the assay, which was performed using the lysis buffer. Reagents were from Sigma‐Aldrich (St Louis, MO, USA) unless noted. The concentrations used were 3 μm, 0.2 mm, and 0.9 mm for catalase, ascorbate, and deferoxamine (DFO), respectively.

HO activity assay

Heme oxygenase activity was assayed following a previously described procedure [45]. Unless specified otherwise, the reaction mixture contained 10 μm of AtHO‐1, 2.5 μm of ferredoxin, 2.5 μm of FNR, 3 μm of catalase, 0.9 mm of DFO, 10 μm of hemin, and 400 μm of NADPH. Reaction was started by finally adding NADPH, and spectra were recorded from 350 to 800 nm every 4 min for 40 min.

Crystallization and structure determination

The purified heme–AtHO‐1 complex was concentrated to 10 mg·mL−1 for crystal screen. Crystals were grown with the sitting‐drop vapor‐diffusion method by mixing 1 μL of heme–AtHO‐1 complex with 1 μL of reservoir solution consisting of 0.17 m ammonium acetate, 85 mm sodium acetate trihydrate, pH 4.6, 25.5% (w/v) polyethylene glycol 4000, and 15% (v/v) glycerol at 16 °C. For data collection, crystals were transferred step by step into the reservoir solution containing 15%, 20%, and 25% (v/v) glycol, and then flash‐cooled in liquid nitrogen.

Diffraction data were collected at 100 K at the Shanghai Synchrotron Radiation Facility beamline BL17U1 and processed with the program hkl2000 [46]. The structure was determined by molecular replacement with the program phaser in phenix [47, 48]. The coordinates of C. diphtheriae heme oxygenase HmuO (PDB entry 1IW0) [32] were used as the search model. Two AtHO‐1 molecules were found in one asymmetric unit. The model was refined by iterative cycles of manual correction in coot [49] and automatic refinement in phenix [48]. The overall model quality was validated by molprobity [50]. Figures showing the structure were drawn with pymol (Schrödinger, LLC, New York, NY, USA).

Results

Purified AtHO‐1 is a mixture of apo‐ and heme‐bound forms

Recombinant AtHO‐1 purified by SEC has heme absorption maxima in the visible range (Fig. 1A). This indicates that the sample contains heme bound to the protein during its heterologous expression in E. coli. To estimate the ratio of unliganded (apo) to heme‐bound AtHO‐1, we titrated the sample to 10 μm heme (Fig. 1B). The Soret maximum has a red shift with gradual increment, confirming formation of the heme–AtHO‐1 complex. It is postulated that the initial rapid increase at 406 nm reflects the amount of heme–AtHO‐1 complex; after saturation of free heme, the increase at 406 nm reflects the heme–AtHO‐1 fraction of the added sample. By linear fitting, we calculated the ratio of apo to heme‐bound AtHO‐1, which was ca. 85–15%. The fact that partial AtHO‐1 binds heme during purification suggests that AtHO‐1 possesses a relative high affinity to heme. We then used ITC to quantify the affinity (Fig. 1C), and fitting of the titration curve yielded a submicromolar dissociation constant (K d = 0.26 ± 0.01 μm) between AtHO‐1 and heme. This K d value is ca. 6‐fold lower than reported values based on spectrophotometric titration [10, 40], and the discrepancy could be due to method difference and the fact that the sample used for ITC was a mixture of apo‐ and heme‐bound AtHO‐1.

Fig. 1

Characterization of purified AtHO‐1 protein. (A) The SEC profile, SDS/PAGE, and absorption spectrum for the purified recombinant AtHO‐1. The absorption spectrum (left inset) was recorded in a wavelength range from 250 to 800 nm. The peak fraction in the SEC profile was subjected to SDS/PAGE (right inset). (B) The absorption spectra of heme with increased concentration of AtHO‐1 (from red to purple). Heme spectrum in the absence of AtHO‐1 is in magenta. Inset: The absorbance difference at 406 nm with increased concentration of AtHO‐1. Error bars represent the standard deviation from three independent measurements. The fitted linear equations are in dashed lines. (C) Isothermal titration of AtHO‐1 with heme. Solid line in the lower panel shows the fit of the integrated heats to one single‐site binding model. [Colour figure can be viewed at wileyonlinelibrary.com]

Characterization of coupled oxidation

For AtHO‐1 activity characterization, we first tested the effect of ascorbate and catalase to separate coupled oxidation from enzymatic heme oxygenation (Fig. 2 and Fig. S1). The presence of 3 μm of catalase essentially inhibited the coupled oxidation; in the absence of catalase, the oxidation was enhanced by addition of 200 μm of ascorbate. The iron chelator DFO has been found to enhance iron release in human HO‐1 [51] and be required for the full activity of AtHO‐1 [40]. Our results confirmed that DFO is absolutely needed for the production of biliverdin. Therefore, catalase and DFO were included in the enzymatic assays (see below) for testing the reducing systems.

Fig. 2

Characterization of coupled oxidation. The absorbance difference at 406 nm of heme–AtHO‐1 (A) and at 680 nm of biliverdin (B). Additions of catalase, ascorbate, and ascorbate with DFO are shown in red, blue, and purple, respectively. Error bars represent the standard deviation from three independent measurements. [Colour figure can be viewed at wileyonlinelibrary.com]

Ferredoxin is redundant for FNR‐mediated AtHO‐1 activity

The assays were performed in the presence of 3 μm of catalase and 0.9 mm of DFO. The biological reducing system NADPH–FNR–ferredoxin was tested, and the necessity of each component was checked (Fig. 3 and Fig. S2). Absence of ferredoxin only perturbed the HO activity as reflected by changes of the four characteristic absorbance peaks, whereas the absence of NADPH or FNR abolished the activity. These results indicated that ferredoxin is dispensable for the NADPH‐dependent HO activity, suggesting that AtHO‐1 could directly interact with FNR to receive electrons from NADPH.

Fig. 3

FNR‐dependent AtHO‐1 activity assay. The absorbance difference at the Soret peak (A), 542 nm (B), 582 nm (C), and 680 nm (D). Removals of ferredoxin, FNR, and NADPH are shown in red, green, and purple, respectively. [Colour figure can be viewed at wileyonlinelibrary.com]

To define the role of FNR and ferredoxin for electron transfer, we first tested the concentration dependence of ferredoxin. When the FNR concentration was fixed at 0.05 μm, the activity of AtHO‐1 increased with the ferredoxin concentration from 0.05 to 2.5 μm (Fig. 4A and Fig. S3). As expected, in the absence of FNR, the activity was totally lost regardless of ferredoxin concentration change. We then tested the concentration dependence of FNR, and found that in the absence of ferredoxin, the activity increased with the FNR concentration from 0.05 to 2.5 μm (Fig. 4B and Fig. S4), indicating that FNR alone is capable of transferring electrons to heme. Interestingly, when ferredoxin was fixed at 2.5 μm, the activity decreased with the increasing FNR concentration (Fig. 4B), which suggests a possibility that FNR has low electron transfer efficiency compared with ferredoxin when the two are competing for AtHO‐1.

Fig. 4

Concentration relationship of FNR and ferredoxin for AtHO‐1 activity. (A) The AtHO‐1 activity assay under different concentrations of ferredoxin in the presence (filled symbols) or absence (open symbols) of FNR. (B) The AtHO‐1 activity assay under different concentrations of FNR in the presence (filled symbols) or absence (open symbols) of ferredoxin. Absorbance at 680 nm was shown and the conditions were labeled. [Colour figure can be viewed at wileyonlinelibrary.com]

AtHO‐1 structure

We then set to determine the structure of AtHO‐1, but the attempt to crystalize apo AtHO‐1 was unsuccessful. The heme–AtHO‐1 complex was crystallized and its structure was solved at 2.2‐Å resolution (Table 1). Like in all HOs, heme is located in the pocket between the N‐terminal and fifth helices of AtHO‐1 (Fig. 5A). The heme iron is coordinated by the imidazole group of His86 on the proximal side and a water molecule on the distal side, and the propionate groups are pointing out from the pocket. A conformational difference between AtHO‐1 and GmHO‐1 occurs with respect to the orientation of a propionate group (Fig. 5B). Within the pocket, a hydrogen‐bond network running from the distal water to the protein surface is formed by water molecules and residues including Tyr116, His206, Tyr230, Lys225‐Leu227, and Lys231 (Fig. 5C). On the proximal side, a hole at the backside runs from the α‐meso carbon to the heme pocket (Fig. 5D). Such a hole is also observed in GmHO‐1 structure [38] but not in human HO‐1 or SynHO‐1 (Fig. 6A), confirming the conservation of a specific feature for plant HOs.

Table 1

Data collection and structure refinement statistics of heme‐bound AtHO‐1.

PDB	7EQH
Diffraction data
Resolution (Å) a	50.00–2.20 (2.28–2.20)
Space group	P212121
Wavelength (Å)	0.979
Unit‐cell parameters
a, b, c (Å)	66.6, 84.8, 92.9
α, β, γ (°)	90, 90, 90
No. of measured reflections	197,033 (19,877)
No. of unique reflections	27,162 (2686)
Completeness (%)	99.8 (100)
Average redundancy	7.3 (7.4)
Mean I/σI	14.2 (2.3)
Wilson B‐factor (Å2)	31.49
R merge	0.131 (1.000)
R pim	0.058 (0.587)
CC1/2	0.965 (0.850)
Refinement
Resolution (Å)	34.75–2.20 (2.29–2.20)
No. of reflections used in refinement	25,120 (1647)
No. of reflections used for R free	1306 (82)
R work b (%)	20.8 (29.4)
R free c (%)	25.0 (35.8)
Number of atoms
Protein	3504
Ligand	86
Water	165
Average B‐factor (Å2)
Protein	35.13
Ligand	30.86
Water	37.40
R.m.s deviations
Bond length (Å)	0.004
Bond angles (°)	0.65
Ramachandran plot
Favored (%)	99.05
Allowed (%)	0.95

Values in parentheses are for highest resolution shell.

R work = ∑¦¦F o¦ − ¦F c¦¦/∑¦F o¦, where F o and F c are the observed and calculated structure factors, respectively.

R free is the cross‐validated R factor computed for a test set of 5% of the reflections, which were omitted during refinement.

Fig. 5

The heme–AtHO‐1 complex structure. (A) Overall structure of heme‐bound AtHO‐1. Protein backbone is in ribbon (colored in blue to red scheme from N‐ to C‐terminus) and the surface is in transparent blue. Heme (brown) is in ball and stick model. Inset shows the heme pocket. (B) Electron density at 1.0σ of heme (upper panel) and comparison with heme–GmHO‐1 (green, lower panel). Residues in AtHO‐1 are shown in blue, and polar interactions between heme and AtHO‐1 are in dashed cyan lines. Interactions between heme and GmHO‐1 are in purple. Water molecules are in spheres. (C) The hydrogen‐bond network in the distal side of heme. Electron density of heme and structured water molecules are shown. (D) The backside of heme pocket. Color scheme same as (A). [Colour figure can be viewed at wileyonlinelibrary.com]

Fig. 6

Comparison of HO structures, the FNR–ferredoxin structure, and an electron‐transfer model. (A) Surface potential representation of AtHO‐1, GmHO‐1, human HO‐1, and SynHO‐1. Heme is in stick model, and the position of AtHO‐1 Lys92 and GmHO‐1 Arg92 are indicated. (B) Surface potential of the ferredoxin–FNR complex. The left and right panels show the surface constituting the ferredoxin–FNR interface. (C) Electron‐transfer model. AtHO‐1 is proposed to interact with the FAD pocket to for direct electron transfer (dash line), compared with the canonical electron transfer pathway (solid lines). [Colour figure can be viewed at wileyonlinelibrary.com]

Discussion

In this study, we determined the crystal structure of heme–AtHO‐1 complex at a resolution of 2.2 Å. The surface potentials of AtHO‐1 and GmHO‐1 are highly similar despite differences in a heme propionate group and the exposed residues (Fig. 6A). A major difference is Lys92 in AtHO‐1 and Arg92 in GmHO‐1 (Fig. 6A), while the interactions of heme to HOs are identical (Fig. 5B). In mammalian HO‐1s, the charged surface around the heme pocket was proposed to facilitate the formation of complex with the electron donor NADPH–CPR [36, 37]. For cyanobacterial HOs, the surface is proposed to interact with the physiological oxidative partner, reduced ferredoxin [28, 29]. The [2Fe‐2S] cluster of ferredoxin directly receives electrons from FNR as shown in the FNR–ferredoxin complex structure (Fig. 6B) [52]. The reduced ferredoxin needs to dissociate from FNR to expose the [2Fe‐2S] side because this side is the HO‐interacting site as revealed by NMR titration [38].

The results presented here indicate that ferredoxin is dispensable for AtHO‐1 activity (Figs 3 and 4B). In the in vitro assay, reduced ferredoxin was obtained by supplying FNR and NADPH, and when ferredoxin was absent, the reaction still proceeded. Thus, FNR alone could transfer electrons to heme–AtHO‐1, although the process was slow. Addition of ferredoxin clearly enhanced HO reaction by promoting the efficient electron transfer to heme–AtHO‐1. This does not conflict with past assays in which the reduced ferredoxin was generated by light irradiation of isolated thylakoid membranes [40]. Because ferredoxin is abundant in chloroplast, the FNR–ferredoxin system should be the major in‐vivo electron donor for AtHO‐1. We conclude that electrons can also be directly transferred from FNR to AtHO‐1. This scenario has been observed in LiHO, which receives electrons from FNR via a possible pathway on the distal side [35]. Interestingly, the FNR in the aerobic spirochete L. interrogans is a close relative of plastid‐type FNR in eukaryotes. The hydrogen‐bond network observed in AtHO‐1 on the distal side allows a possibility for an electron pathway directly linking the flavin coenzyme to the substrate heme (Fig. 6C). The biological implication that AtHO‐1 catalysis can bypass ferredoxin and be solely dependent on FNR awaits to be tested in vivo.

Conflict of interest

The authors declare no conflict of interest.

Author contributions

JW, XL, and LL conceived and designed the project, JW, XL, J‐WC, TY, and YM acquired the data, XW and LL analyzed and interpreted the data, JW and LL wrote the paper.

Fig. S1. Spectra for characterization of coupled oxidation, related to Fig. 2. The assays were performed without additional reagents (A), with catalase (B), with ascorbate (C), and with ascorbate and DFO (D).

Click here for additional data file.

Fig. S2. Spectra for characterization of FNR‐dependent AtHO‐1 activity assay, related to Fig. 3. The assays were performed with NADPH, ferredoxin, and FNR (A), with NADPH and FNR (B), with NADPH and ferredoxin (C), and with ferredoxin and FNR (D).

Click here for additional data file.

Fig. S3. Spectra of the concentration dependence for ferredoxin, related to Fig. 4A. The concentration of ferredoxin ranges from 2.5 to 0.05 μM from top to bottom, (A) at fixed FNR concentration of 0.05 μM, (B) in absence of FNR.

Click here for additional data file.

Fig. S4. Spectra of the concentration dependence for FNR, related to Fig. 4B. The concentration of FNR ranges from 2.5 to 0.05 μM from top to bottom, (A) at fixed ferredoxin concentration of 0.05 μM, (B) in absence of ferredoxin.

Click here for additional data file.