Structure, function and evolution of the hemerythrin‐like domain superfamily

Abstract

Hemerythrin‐like proteins have generally been studied for their ability to reversibly bind oxygen through their binuclear nonheme iron centers. However, in recent years, it has become increasingly evident that some members of the hemerythrin‐like superfamily also participate in many other biological processes. For instance, the binuclear nonheme iron site of YtfE, a hemerythrin‐like protein involved in the repair of iron centers in Escherichia coli, catalyzes the reduction of nitric oxide to nitrous oxide, and the human F‐box/LRR‐repeat protein 5, which contains a hemerythrin‐like domain, is involved in intracellular iron homeostasis. Furthermore, structural data on hemerythrin‐like domains from two proteins of unknown function, PF0695 from Pyrococcus furiosus and NMB1532 from Neisseria meningitidis, show that the cation‐binding sites, typical of hemerythrin, can be absent or be occupied by metal ions other than iron. To systematically investigate this functional and structural diversity of the hemerythrin‐like superfamily, we have collected hemerythrin‐like sequences from a database comprising fully sequenced proteomes and generated a cluster map based on their all‐against‐all pairwise sequence similarity. Our results show that the hemerythrin‐like superfamily comprises a large number of protein families which can be classified into three broad groups on the basis of their cation‐coordinating residues: (a) signal‐transduction and oxygen‐carrier hemerythrins (H‐HxxxE‐HxxxH‐HxxxxD); (b) hemerythrin‐like (H‐HxxxE‐H‐HxxxE); and, (c) metazoan F‐box proteins (H‐HExxE‐H‐HxxxE). Interestingly, all but two hemerythrin‐like families exhibit internal sequence and structural symmetry, suggesting that a duplication event may have led to the origin of the hemerythrin domain.

Abbreviations

Fqo

F420H(2)‐dependent quinone reductase

HMM

hidden Markov model

IRP2

iron regulatory protein 2

LLM

luciferase‐like flavin monooxygenase

MCP

methyl‐accepting chemotaxis protein

PDB

protein data bank

PNPOx

pyridoxamine 5'‐phosphate oxidase

RIC

repair iron‐center

Introduction

Hemerythrin was initially described as a multimeric O₂ carrier‐protein with a binuclear nonheme iron center and with a distribution that, at first, appeared to be limited to three phyla of marine invertebrates, Brachiopoda, Priapulida, and Annelida.1 Over the past decade, however, hemerythrin‐like proteins have been identified in many taxonomically distant groups including humans,2, 3, 4 plants,5 bacteria,6, 7, 8, 9, 10 and archaea (PDB code: 2P0N). The function of these binuclear nonheme Fe‐containing proteins (e.g., hemerythrin, myohemerytrin, bacteriohemerythrin) is tightly related to O₂‐binding and activation,11, 12, 13 a trait indicative of the major evolutionary pressure exerted by atmospheric and oceanic oxygenation since the late Archaean Eon on the biosphere.14 It has been proposed that hemerythrin homologs that reversibly bind O₂ (hemerythrin, myohemerytrin,15 and bacteriohemerythrin16) belong to a monophyletic group that appeared during this period of increasingly oxidizing conditions.14

The basic structural fold of hemerythrin‐like proteins consists of an up‐and‐down four helix bundle with an overall right‐handed path.17, 18, 19 Most structures exhibit a binuclear cation‐binding site in which each helix contributes to the metal‐ion coordination with at least one residue. The first coordination sphere of the binuclear five‐coordinate/six‐coordinate nonheme Fe(II) site of O₂‐binding hemerythrins includes five histidines, two bridging carboxylates from a glutamic and an aspartic acid residues, and a µ‐oxo/hydroxo bridge.20 However, the cation‐binding residues are not conserved in all solved hemerythrin structures (Table 1). For instance, the catalytic site of hemerythrin in YtfE,24 a repair iron centers protein from Escherichia coli, has been shown to involve four instead of five histidine residues, and two carboxylate bridges from two glutamic residues instead of one glutamic and one aspartic acid residues. This arrangement results in a five‐coordinate/five‐coordinate binuclear Fe(II)‐binding site with nitric oxide‐reductase activity24 that is likely to be part of a defense mechanism against DNA damage associated with NO. A survey of the literature reveals that many other functionally characterized hemerythrin‐containing proteins have a cation‐binding motif that is different from the H‐HxxxE‐HxxxH‐HxxxxD motif of signal‐transduction and oxygen‐carrier hemerythrins. For instance, the primary structure of the Rv2633c hemerythrin‐like catalase from Mycobacterium tuberculosis (Uniprot ID: P9WL59),25 and hemerythrin homologs in Mycobacterium smegmatis (Uniprot IDs: A0QXI3, A0QV17, and A0R5J3);26 Anabaena sp. strain PCC7120 (Uniprot ID: Q8YS92);27 Aeromonas hydrophila (Uniprot ID: A0KMZ0);10 Acidothermus cellulolyticus (Uniprot ID: A0LQU2);28 Oryza sativa subsp. japonica (Uniprot IDs: V9G2Z0; and Q6AUD8) and Arabidopsis thaliana (Uniprot ID: Q8LPQ5),5 suggests that these proteins also have differences in the first coordination‐sphere of the iron centers, but there is no available confirmation from structural data.

Table 1.

Structural Characteristics of Cation‐coordination in Solved Hemerythrin‐like Domain Structures

Protein	Organism	Function	Coordination number	Ligation	Source [PDB code]
Hemerythrin (subunit A)	Themiste dyscritum	Reversible O2 binding	6C/5C	5H/1E/1D	[1HMD]20
Myohemerythrin	Themiste hennahi	Reversible O2 binding	6C/5C	5H/1E/1D	[2MHR]21
Bacteriohemerythrin (McHr)	Methylococcus capsulatus	Reversible O2 binding	6C/5C	5H/1E/1D	[4XPX]16
Hemerythrin‐like domain (DcrH)	Desulfovibrio vulgaris	O2 sensing	6C/5C	5H/1E/1D	[2AWC]22
Hemerythrin‐like domain (FBXL5)	Homo sapiens	Iron sensing	6C/5C	4H/3E	[3V5X]23
Hemerythrin‐like‐domain (YtfE)	Escherichia coli	2NO + 2H+ + 2e‐ → N2O + H2O	5C/5C	4H/2E	[5FNN]24
Hemerythrin domain (hypothetical protein NMB1532)	Neisseria meningitidis	Unknown	5C/5C	4H/2E	[2P0N]
Hemerythrin domain (uncharacterized protein PF0695)	Pyrococcus furiosus	Unknown	–	–	[3CAX]

Although the evolutionary relationships among a group of protein sequences closely related to signal‐transduction and oxygen‐carrier hemerythrins have been described elsewhere,14, 29, 30, 31 a comprehensive evolutionary analysis and classification of hemerythrin‐like proteins is lacking. In this study, we report the outcome of an extensive bioinformatic analysis of hemerythrin‐like proteins and present their classification into three major groups based on the conservation of cation‐coordinating residues.

Results and Discussion

Structural analysis

The structures of hemerythrin‐like homologs in the Protein Data Bank (PDB) were identified using the HHpred webserver.32 We found 27 structures corresponding to eight different proteins from different source organisms: YtfE (PDB IDs: 5FNN, 5FNP, 5FNY) from Escherichia coli, DcrH (PDB IDs: 2AVK, 2AWC, 2AWY, 3AGT, 3AGU, 3WAQ, 3WHN) from Desulfovibrio vulgaris; bacteriohemerythrin (PDB IDs: 4XPW, 4XPX, 4XPY) from Methylococcus capsulatus; NMB1532 (PDB ID: 2P0N) from Neisseria meningitidis; PF0695 (PDB ID: 3CAX) from Pyrococcus furiosus; myohemerythrin (PDB IDs: 1A7D, 1A7E, 2MHR) from Themiste hennahi; hemerythrin (PDB IDs: 1HMD, 2HMQ, 2HMZ, 1HMO) from Themiste dyscritum; and FBXL5 (PDB IDs: 3U9M, 3V5Y, 3U9J, 3V5X, 3V5Z) from Homo sapiens. To avoid redundancy, only one representative per protein was analyzed. Distinctive traits in the tertiary structure of these proteins were evaluated by pairwise comparisons of X‐ray crystal structures (Table 2). While seven of the analyzed eight structures showed a right‐handed four‐alpha‐helix bundle, characteristic of hemerythrin, a distinctive two‐helix swap was identified in the hemerythrin‐like structure of YtfE. This rearrangement preserves the up‐and‐down topology of the fold, but results in a left‐handed four‐helix bundle (Fig. 1). Because of this topological difference, the alignment of full protein structures produced suboptimal results (Supporting Information Table SI). We therefore performed further structural analysis of hemerythrin‐like proteins based on the comparison of their metal‐binding sites using MetalS2.33 The local similarity between pairs of proteins was evaluated by their MetalS2 score and by the percent identity of the superposition‐derived sequence alignment (Table 2). Scores lower than 2.25 indicate a high level of structural similarity.33

Table 2.

Metal S ² Score and Percent Identity of Pairwise Structural Alignments of Hemerythrin Homologs and the Uncharacterized Protein Q4MWP8

		1	2	3	4	5	6	7	8
Hemerythrin, subunit A (Uniprot ID: P02246)	1	0.000	0.637 (55%)	0.759 (50%)	0.797 (44%)	1.907 (14%)a	1.623 (31%)	1.611 (18%)a	1.814 (16%)a
Myohemerythrin (Uniprot ID: P02247)	2		0.000	0.832 (45%)	0.723 (42%)	1.937 (17%)a	1.514 (14%)a	1.514 (20%)a	1.748 (14%)a
Bacteriohemerythrin, McHr (Uniprot ID: Q60AX2)	3			0.000	0.740 (50%)	1.926 (20%)a	1.560 (21%)a	1.569 (18%)a	1.813 (18%)a
Hemerythrin‐like domain of DcrH (Uniprot ID: Q726F3)	4				0.000	1.853 (20%)a	1.559 (28%)a	1.870 (24%)a	1.946 (20%)a
Hemerythrin‐like domain of FBXL5 (Uniprot ID: Q9UKA1)	5					0.000	1.286 (25%)a	1.583 (29%)a	2.233b (26%)a
Hemerythrin‐like‐domain of YtfE (Uniprot ID: P69506)	6						0.000	1.061 (23%)a	1.489 (18%)a
Hemerythrin‐like domain of NMB1532 (Uniprot ID: Q9JYL1)	7							0.000	1.521 (18%)a
Uncharacterized protein Q4MWP8	8								0.000

The numbers in parenthesis refer to the percent identity of the superposition‐derived sequence alignment.

^aPercent identity <30%.

^bMetalS2 score >2.

Figure 1.

Ribbon diagram of distant hemerythrin homologues. In all structures, helices 1 and 2 are colored in dark green; helices 3 and 4 are colored in light green. Non‐homologous regions are shown in white. A disordered region in the tertiary structure of hemerythrin in FBXL5 is highlighted in purple. Gray arrows indicate the sequential arrangement of the helices in the hemerythrin fold.

The entire set of hemerythrin homologs aligned with MetalS2 scores of below 2. Hemerythrin, myohemerythrin, bacteriohemerythrin, and the hemerythrin‐like domain of DcrH from Desulfovibrio vulgaris are closely related (sequence identity >40%) and aligned with a MetalS2 score below 1. These hemerythrin homologs have a characteristically conserved 5H/1E/1D ligation of the binuclear iron site (Table 1).

The hemerythrin‐like domains of NMB1532 (PDB ID: 2P0N) and YtfE (PDB ID: 5FNN) are distantly related (sequence identity <25%), but they have a very similar tertiary structure, as shown by their structure alignment score (Table 2). The two‐helix swap in YtfE maintains the local structure of the binuclear site, in which helix α₁ and helix α₃ each donates one histidine residue, and both helix α₂ and helix α₄ donate a histidine and a glutamate. An important difference between these two structures is the presence of two manganese ions coordinated to the hemerythrin domain of NMB1532 from Neisseria meningitidis (PDB code: 2P0N). Because presently no functional data is available for NMB1532, it is unknown whether manganese ions are naturally present in hemerythrins.

The hemerythrin‐like domain of FBXL5, in which a disordered region substitutes part of helix α₃, exhibits the worst superimposition scores of the homologous set. The binuclear site of the hemerythrin domain in these proteins is asymmetrical: α₂ donates an additional glutamic acid, and the disordered region that replaces part of helix α₃ donates the third histidine residue of the cation‐coordination.

Cluster analysis

To map out sequence and evolutionary relationships between the members of the hemerythrin‐like superfamily in a comprehensive fashion, we searched for hemerythrin‐like sequences in 2580 fully sequenced genomes (Table 3) and identified a total of 6599 sequences, which were subsequently clustered in CLANS34 (Fig. 2) based on their all‐against‐all pairwise similarities as measured by BLAST P values. In addition to a large family of oxygen‐carrier and oxygen‐sensing hemerythrins, several independent clusters were identified. These clusters fall into three broadly defined groups based on the conservation of cation‐coordinating residues: signal‐transduction and oxygen‐carrier hemerythrins (H‐HxxxE‐HxxxH‐HxxxxD), hemerythrin‐like (H‐HxxxE‐H‐HxxxE), and metazoan F‐box (H‐HExxE‐H‐HxxxE) proteins. Their phylogenetic distribution indicates that the metazoan F‐box proteins set is the most recent one. Internal symmetry within most hemerythrin‐like and metazoan F‐box protein sequences (Fig. 3, Supporting Information Table SI) strongly suggests that the hemerythrin fold originated by duplication and fusion of an ancestral helix‐loop‐helix motif.

Table 3.

Presence/Absence of Hemerythrin Homologs

Phylogenetic group	Presence	Absence
Archaea (n = 171)	74	97
Nanoarchaeota (n = 2)	–	2
Euryarchaeota (n = 104)	41	63
Candidatus korarchaeota (n = 1)	1	–
Crenarchaeota (n = 56)	20	36
Thaumarchaeota (n = 9)	9	–
Unclassified archaea (n = 15)	3	12
Uncultured archaea (n = 4)	–	4
Bacteria (n = 3998)	2062	1936
Aquificae (n = 11)	9	2
Thermodesulfobacteria (n = 2)	2	–
Thermotogae (n = 12)	8	4
Caldiserica (n = 1)	1	–
Chrysiogenetes (n = 1)	1	–
Deferribacteres (n = 6)	6	–
Dictyoglomi (n = 1)	1	–
Chloroflexi (n = 34)	12	22
Actinobacteria (n = 561)	309	252
Planctomycetes (n = 26)	9	17
Chlamydiae (n = 11)	3	8
Verrucomicrobia (n = 13)	2	11
Lentisphaerae (n =1)	–	1
Candidatus omnitrophica (n = 3)	–	3
Nitrospirae (n = 12)	4	8
Acidobacteria (n = 10)	4	6
Synergistetes (n = 15)	–	15
Cyanobacteria/melainabacteria group (n = 98)	54	44
Firmicutes (n = 907)	431	476
Tenericutes (n = 71)	3	68
Fibrobacteres (n = 3)	2	1
Elusimicrobia (n = 3)	1	2
Bacteroidetes/chlorobi group (n = 334)	244	90
Candidatus cloacimonetes (n = 2)	1	1
Candidatus latescibacteria (n = 1)	1	–
Candidatus marinimicrobia (n = 1)	–	1
Deinococcus‐thermus (n = 17)	3	14
Armatimonadetes (n = 3)	1	2
Spirochaetes (n = 41)	30	11
Fusobacteria (n = 15)	5	10
Gemmatimonadetes (n = 7)	1	6
Proteobacteria (n = 1327)	857	470
Nitrospinae/tectomicrobia group (n = 5)	2	3
Unclassified bacteria (n = 442)	55	387
Uncultured bacterium (n = 1)	–	1
Eukaryota (n = 770)	392	378
Alveolata (n = 45)	3	42
Amoebozoa (n = 7)	4	3
Apusozoa (n = 1)	0	1
Cryptophyta (n = 1)	1	0
Euglenozoa (n = 19)	1	18
Diplomonadida (n = 3)	0	3
Haptophyceae (n = 2)	0	2
Heterolobosea (n = 1)	1	0
Choanoflagellida (n = 2)	1	1
Fungi (n = 373)	213	160
Metazoa (n = 221)	98	123
Fonticula (n = 1)	0	1
Ichthyosporea (n = 2)	2	0
Parabasalia (n = 1)	1	0
Rhizaria (n = 2)	2	0
Rhodophyta (n = 2)	2	0
Stramenopiles (n = 23)	7	16
Viridiplantae (n = 64)	56	8

Figure 2.

Cluster map of hemerythrin homologs. Cluster map of 6599 hemerythrin sequences in two‐dimensional space at a P value cutoff of 1e‐10. Dotted lines enclose three large groups formed at a P value cutoff of 1e‐13. Each dot represents a sequence; dots are colored by groups of sequences with known domain organization. (1) Q9PIQ3 and (2) Q0P932 from Campylobacter jejuni; (3) Q60AX2 from Methylococcus capsulatus, and Q726F3 from Desulfovibrio vulgaris; (4) Q9KSP0 from Vibrio cholerae; (5) Q9RJ01 from Streptomyces coelicolor; (6) Q92Z60 from Rhizobium meliloti; (7) Q8YS92 from Nostoc sp.; (8) P69506 from Escherichia coli; (9) Q7WX96 from Cupriavidus necator; (10) Q8U2Y3 from Pyrococcus furiosus; (11) A0KMZ0 from Aeromonas hydrophila; (12) Q9JYL1 from Neisseria meningitidis; (13) Q9UKA1 from Homo sapiens; (14) A0QXI3, A0R5J3, A0QV17 from Mycobacterium smegmatis; (15) Q8LPQ5 from Arabidopsis thaliana; (16) V9G2Z0 from Oryza sativa; and (17) Rv2633c from Mycobacterium tuberculosis.

Figure 3.

Repeat units of the hemerythrin‐like domain. Internal sequence symmetry was identified by HHpredID with a probability >90. Structural superimposition of the repeat units in (A) PF0695 from Pyrococcus furiosus (PDB ID: 3CAX), and in (B) YtfE from E. coli (PDB ID: 5FNN). Sequence alignments are based on the structural superimposition of the repeat units. Aligned residues are connected by a dot. Exact matches are connected by a colon. Amino acids that participate in cation‐coordination are marked with an asterisk.

O₂‐carrier hemerythrins and closely related sequences

Signal‐transduction and O₂‐carrier hemerythrins form a tightly connected group in the map containing 1424 hemerythrin sequences from bacterial, archaeal, and eukaryotic species, in which the motif H‐HxxxE‐HxxxH‐HxxxxD is conserved. These conserved positions, which represent a variation from the H‐HxxxE duplicate, contribute to the 5H/1E/1D ligation to iron (Table 1) in hemerythrin from Themiste dyscritum, myohemerythrin from Themiste hennahi, bacteriohemerythrin from M. capsulatus, and the hemerythrin‐like domain of DcrH from D. vulgaris. These hemerythrin homologs are all involved in O₂ transport and storage through their ability to bind O₂ reversibly,16, 20, 22, 23, 24 and O₂ sensing in signal‐transduction proteins by autoxidation of the binuclear iron site upon O₂ binding.35 Sequences with an H‐HxxxE‐HxxxH‐HxxxxD motif are highly conserved as shown by the low number of divergent projections in the cluster map, which may indicate a recent divergence of this group (Fig. 2, Supporting Information Table SI).

Hemerythrin‐like proteins containing signal transduction and chemotaxis domains were conspicuously identified in proteobacteria, where they also have been functionally characterized. For instance, the oxygen‐sensing protein DcrH (Uniprot ID: Q726F3) from D. vulgaris comprises an N‐terminal double sensory domain dCache_3 (Pfam accession: PF14827), followed by a domain of unknown function (Pfam accession: PF07889) and a methyl‐accepting chemotaxis‐protein signaling domain (Pfam accession: PF00015). Hemerythrin is located at the C‐terminal end of this protein. A different protein domain organization occurs in VC1216, a signal transduction protein from Vibrio cholerae in which hemerythrin is followed by a GGDEF diguanylate cyclase domain (Pfam accession: PF00990). Different arrangements of sensory and chemotaxis domains (such as TadZ_N, HPTransfase, MEKHLA, MCPsignal, TarH, and HAMP) were also present in sequences from Spirochaetes and Firmicutes species (Supporting Information Fig. S2), suggesting that in these cases, hemerythrin may be involved in a wide range of cellular responses to O₂ (Fig. 2, Table 4).

Table 4.

Putative Functions of the Most Frequently Identified Nonhomologous Domains in Hemerythrin‐like Proteins

Pfam ID (Accession)	Domain description	Putative functiona	Amino acid motif of the associated hemerythrin‐like domain
TadZ_N (PF16968)	N‐terminal domain of the pilus assembly protein TadZ	Signal‐transduction‐response receiver	H‐HxxxE‐HxxxH‐HxxxxD
HPTransfase (PF10090)	Histidine phosphotransferase	Signal transduction	H‐HxxxE‐HxxxH‐HxxxxD
MEKHLA (PF08670)	Shares similarity with the PAS domain	Signal sensor	H‐HxxxE‐HxxxH‐HxxxxD
GGDEF (PF00990)	GGDEF domain	Diguanylate cyclase, signal transduction	H‐HxxxE‐HxxxH‐HxxxxD
MCPsignal (PF00015)	Methyl‐accepting chemotaxis protein (MCP) signaling domain	Signal transduction	H‐HxxxE‐HxxxH‐HxxxxD
TarH (PF02203)	Tar (taxis towards aspartate and maltose, away from nickel and cobalt) ligand binding domain homolog	Signal transduction, chemotaxis	H‐HxxxE‐HxxxH‐HxxxxD
HAMP (PF00672)	HAMP domain (present in Histidine kinases, Adenyl cyclases, Methyl‐accepting proteins and Phosphatases)	Signal transduction	H‐HxxxE‐HxxxH‐HxxxxD
PDEase_II (PF02112)	cAMP phosphodiesterases class‐II	cAMP catabolic process	H‐HxxxE‐HxxxH‐HxxxxD
sCache_2 (PF17200)	Single cache (CAlcium channels and CHEmotaxis receptors) domain 2	Signal transduction, small‐molecule recognition	H‐HxxxE‐HxxxH‐HxxxxD
Popeye (PF04831)	Member of the conserved barrel domain of the cupin superfamily	Unknown	H‐HxxxE‐HxxxH‐HxxxxD
SprT‐like (PF10263)	Domain of unknown function	Unknown	H‐HxxxE‐HxxxH‐HxxxxD
DUF1858 (PF08984)	Domain of unknown function	Unknown	H‐HxxxE‐HxxxH‐HxxxxD
Prok‐RING_2 (PF14445)	Prokaryotic RING finger family 2	Associated with components of the ubiquitin‐based signaling and degradation system	H‐HxxxE‐H‐HxxxE
zf‐CHY (PF05495)	Zinc fingers motif	Unknown	H‐HxxxE‐H‐HxxxE
zf‐rbx1 (PF12678)	Zinc fingers motif	Unknown	H‐HxxxE‐H‐HxxxE
zinc_ribbon_6 (PF14599)	Zinc fingers motif	Unknown	H‐HxxxE‐H‐HxxxE
TAT_signal (PF10518)	TAT (twin‐arginine translocation) pathway signal sequence. Transport of folded proteins across energy‐transducing membranes	Transport across membranes	H‐HxxxE‐H‐HxxxE
E1‐E2_ATPase (PF00122)	Proton‐ATPase	Transport across membranes	H‐HxxxE‐H‐HxxxE
GST_N_3 (PF13417)	Glutathione S‐transferase, N‐terminal domain	Glutathione metabolism.	H‐HxxxE‐H‐HxxxE
Pyrid_ox_like (PF16242)	Pyridoxamine 5'‐phosphate oxidase like	Putative a defense mechanism against oxidative stress	H‐HxxxE‐H‐HxxxE
Hydrolase (PF00702)	Haloacid dehalogenase‐like hydrolase	Hydrolase activity	H‐HxxxE‐H‐HxxxE
UMPH‐1 (PF05822)	Pyrimidine 5'‐nucleotidase	Hydrolase activity	H‐HxxxE‐H‐HxxxE
Bac_luciferase (PF00296)	Luciferase‐like flavin monooxygenase (LLM) domain	Oxygenase activity	H‐HxxxE‐H‐HxxxE
MetRS‐N (PF09635)	N‐terminal domain of methionyl‐tRNA synthetase which adopts a glutathione S‐transferase (GST)‐like fold	Unknown	H‐HxxxE‐H‐HxxxE
ScdA_N (PF04405)	Repair of iron centers domain	Unknown	H‐HxxxE‐H‐HxxxE
DUF2249 (PF10006)	Domain of unknown function	Unknown	H‐HxxxE‐H‐HxxxE
F420H2_quin_red (PF04075)	F420H(2)‐dependent quinone (Fqo) reductase	Unknown	H‐HxxxE‐H‐HxxxE
DUF2481 (PF10654)	Domain of unknown function	Unknown	H‐HxxxE‐H‐HxxxE
DUF1858 (PF08984)	Domain of unknown function	Unknown	H‐HxxxE‐H‐HxxxE
PAS_10 (PF13596)	PAS: Per‐Arnt‐Sim domain (present in period circadian protein, aryl hydrocarbon receptor nuclear translocator protein, single‐minded protein)	Involved in oxidative stress protection	Not conserved
DUF438 (PF04282)	Domain of unknown function	Unknown	Not conserved
F‐box_4 (PF15966)	F‐box protein.	Protein‐protein interactions	H‐HExxE‐H‐HxxxE
LRR_9 (PF14580)	Leucine‐rich repeat.	Protein‐protein interactions	H‐HExxE‐H‐HxxxE

^aInformation about putative domain functions was gathered from InterPro.36

H‐HxxxE‐H‐HxxxE hemerythrins

At a P value cutoff of 1e‐13, a total of 4957 bacterial, archaeal, and eukaryotic sequences exhibiting a characteristic conservation of the H‐HxxxE repeat formed many distinct, but profusely connected clusters. These hemerythrin‐like sequences comprise more than a dozen groups, the majority of which are poorly studied. As shown in Table 4, some of the divergent sequences from this group could be annotated, as is the case for repair iron‐center (RIC) hemerythrins; PAS_10‐containing hemerythrins; hemerythrin‐like proteins with a C‐terminal DUF2249 domain (Pfam accession: PF10006); hemerythrin‐like ATPases; Bac_luciferase‐containing hemerythrins (Pfam accession: PF00296); F420H(2)‐dependent quinone reductases (Pfam accession: PF04075); pyridoxamine 5'‐phosphate oxidases (Pfam accession: PF16242); BRUTUS E3‐ligases; and hemerythrin‐containing glutathione S‐transferases (GST). Recently reported hemerythrin‐like cargo proteins detected by the encapsulin system37 were identified and are indicated on the cluster map (Fig. 2). This system is involved in iron mineralization and oxidative stress protection through encapsulation in Firmicutes.37

For the most part, linear combinations of domains in hemerythrin‐like proteins are both cluster‐ and phylum‐specific, with the clear exception of ScdA_N‐ and PAS_10‐containing hemerythrins, which were identified in taxonomically distant species.

Hemerythrin‐like sequences of repair iron‐center proteins form a large sub‐cluster (colored dark green in the map). These hemerythrin‐like proteins contain a domain of unknown function termed ScdA_N (Pfam accession: PF04405). Two characterized hemerythrin‐like RIC proteins, NorA from the denitrifier species Ralstonia eutropha and YtfE from Escherichia coli, have the ability to bind nitric oxide. Nitric oxide and reactive nitrogen species are deleterious products of denitrification and host immune system responses, particularly damaging to iron‐sulfur clusters38 and to DNA.39 The hemerythrin‐like domain of YtfE has been shown to catalyze the reduction of nitric oxide to nitrous oxide.24, 40

DUF438‐ and/or PAS‐containing hemerythrin‐like proteins (dark blue) often have an additional domain (DUF1858, Pfam accession: PF08984) similar to ScdA_N. This group is closely related to hemerythrin‐like RIC proteins. PF0695 is a PAS‐containing hemerythrin‐like protein from Pyrococcus furiosus that has been structurally determined by X‐ray crystallography (PDB code: 3CAX), but its biological function remains uncharacterized. Interestingly, multiple sequence alignments show that the first and the last iron‐coordinating histidine residues are not conserved in PF0695 and most DUF438‐ and/or PAS‐containing hemerythrin‐like proteins, suggesting that these may be naturally occurring metal‐free proteins. Internal structure and sequence symmetry was detected in the hemerythrin‐like domain of YtfE and PF0695 (HHrepID P values of 2.5e‐7 and 4.4e‐12, respectively). The N‐ and C‐ terminal halves of both hemerythrin‐like structures were superposed at RMSDs of less than 3Å (Fig. 3).

The group of hemerythrin‐like ATPase transporters includes two forms with slightly different domain organization; a C‐terminal hemerythrin‐like domain and an N‐terminal E1‐E2_ATPase domain (Pfam accession: PF00122) flank either a UMPH‐1 domain (Pfam accession: PF05822) or a hydrolase domain (Pfam accession: PF00702). This group includes two functionally characterized proteins from Acidothermus cellulolyticus 28 (Uniprot ID: A0LQU2) and from Sinorhizobium meliloti 41 (Uniprot id: Q92Z60). It has been suggested that both proteins participate in Fe²⁺ (or Ni²⁺) transport across membranes, and that the C‐terminal hemerythrin‐like domain could function as an iron sensor to avoid harmful intracellular iron overload.28, 41

Plant‐specific hemerythrin‐like sequences form two distinct groups: hemerythrin‐containing BRUTUS proteins and hemerythrin GSTs. BRUTUS proteins were identified in species of Chlorophyceae, Trebouxiophyceae, Mamiellophyceae, Klebsormidiophyceae, and Streptophytina. These sequences contain several zinc fingers motifs (zf‐CHY, zf‐rbx1, zinc_ribbon_6), and a Prok‐RING_2 domain (Pfam accession: PF14445), which suggests that they are E3‐ligases that participate in iron regulation,5, 42 a process equivalent to that present in animals (see below). Internal sequence symmetry found in most hemerythrin homologs was not detected in hemerythrin‐like sequences from BRUTUS proteins. Hemerythrin GST sequences were found exclusively in species of Streptophytina. While most hemerythrin GSTs have a GST_N_3 (Pfam accession: PF13417) N‐terminal domain, seven sequences in this group contain a structurally similar43 MetRS‐N fold (Pfam accession: PF09635). Hemerythrin GSTs may be involved in heavy metal‐detoxification processes in plants.44, 45

Metazoan iron‐sensing hemerythrins

The most distant group of homologues we have detected consists of a set of F‐box‐like iron‐sensing proteins possessing a conserved H‐HExxE‐H‐HxxxE motif. Leucine‐rich repeats are present in most of these proteins (Fig. 2). In the case of the human protein FBXL5, the N‐terminal hemerythrin domain undergoes conformational changes depending on oxygen and iron availability.3, 4 Both leucine‐rich repeats (Pfam accession: PF14580) and F‐box domains (Pfam accession: PF15966) mediate protein–protein interactions; the central F‐box domain interacts with the E3 ubiquitin–ligase complex, and leucine‐rich repeats have been proposed to have role in binding to the iron regulatory protein 2 (IRP2).3, 4

A case of local structure convergence

A similarity search using MetalS346 gathered structures with equivalent metal coordination sites and surrounding chemical species within 5 Å from the metal‐ion. A significant result (total score < 2) was obtained for Q4MWP8, an uncharacterized protein from Bacillus cereus G9241. Unlike hemerythrins, in which the cation coordination internally crosslinks the four helices of the fold, the binuclear nonheme coordination site in Q4MWP8 is located at the interface of four discontinuous fragments of sequence from bromodomain‐like folds (Fig. 4). Bromodomains are putative protein–protein interaction domains47 with no apparent phylogenetic relationship to the hemerytrin‐like domain superfamily, and the local structural similarity around the binuclear iron‐binding site apparently constitutes a case of local structural convergence (Fig. 4).

Figure 4.

Local structural similarity between hemerythrin and Q4MWP8 from Bacillus cereus G9241. Ribbon diagram of two evolutionarily unrelated protein structures, Q4MWP8 from Bacillus cereus G9241 (shown in gray, PDB ID: 3DBY) and hemerythrin‐like domain of DcrH (shown in green, PDB ID: 2AWC), aligned with MetalS2. The inset shows the superposition of the metal‐binding sites.

Conclusions

Here we present a bioinformatic analysis of hemerythrin homologs, which compose a diverse multifunctional protein domain superfamily. We have identified at least three broad groups within the hemerythrin‐like superfamily by well‐defined sequence and structure similarity criteria. These are characterized by having a set of conserved residues at putative cation‐coordinating sites. Sequences in the hemerythrin‐like group exhibit symmetrical traits both in sequence and structure, suggesting a possible origin of hemerythrin through a duplication and fusion event involving a primordial two‐up‐and‐down helix motif containing a single H‐HxxxE cation‐coordination site (Fig. 5). This duplication resulted in an increase of the functional properties of the metal site, as the contemporary role of characterized hemerythrins relies on the presence of both iron‐binding sites. Moreover, binuclear non‐heme Fe enzymes essentially perform O₂‐dependent reactions.13 This functional trait must have been incorporated to cellular metabolism as a response to free O₂ conditions. Both the sequence cluster topology and the specialized function of signal‐transduction and oxygen‐carrier hemerythrins (H‐HxxxE‐HxxxH‐HxxxxD) as well as F‐box proteins (H‐HExxE‐H‐HxxxE) suggest a recent divergence of these families from the core cluster. Incorporation of hemerythrin domains into proteins by domain shuffling events and lateral gene transfer appears to be a recurrent trait during the complex evolutionary history of this fold superfamily.

Figure 5.

Fold change in the evolution of the hemerythrin‐like domain superfamily. Schematic representation of the evolution of hemerythrins depicting possible scenarios for the origination of different handedness of the hemerythrin fold.

Materials and Methods

Homology search

To identify evolutionarily distant members of the hemerythrin‐like domain superfamily, a database of profile HMMs, comprising Protein Data Bank (PDB) entries clustered down to a pairwise sequence identity of 70% (PDB70), was searched using HHpred with default parameters,48 using as query the sequence of bacteriohemerythrin (Uniprot ID: Q60AX2) from Methylococcus capsulatus. Protein sequences having an aligned region with HHpred probability higher than 90% were retrieved. We constructed multiple sequence alignments of these aligned regions with HHblits49 using default parameters. The multiple sequence alignments were then converted to Hidden Markov Models (HMMs) with HMMER3.50 We used hmmsearch50 to identify statistically significant matches (E value cutoff of 1e‐3 or lower) to the generated HMMs in the sequence database Uniprot.51

Sequence analysis

To delineate the domain composition of proteins gathered by hmmsearch, profile HMMs were built for every sequence following the same methodology as for the initial homology search, and were subsequently compared to the PfamA 30.0 profile HMM database52 using HHsearch. Only protein domains identified with a probability of 70% or higher were considered.

Cluster analysis

Sequence regions comprising only hemerythrin‐like domains were clustered in CLANS34 based on their BLAST P values. Clustering was performed to equilibrium in two‐dimensional space at a P value cutoff of 1e‐10, using default settings. Sequence clusters formed at a cutoff value of 1e‐13 were aligned with MAFFT.53

Finally, in order to detect internal symmetry in hemerythrin‐like proteins, sequence clusters formed at a P value cutoff of 1e‐18 were aligned, and the resulting multiple sequence alignments were analyzed with HHrepID,32 using default parameters. Noncluster forming sequences were analyzed individually. To corroborate the presence of internal symmetry at the structure level, the repeat fragments were superposed in hemerythrin‐like proteins of known structure with TMalign.54

Local structural analysis of the metal ion sites

Three‐dimensional models of hemerythrin homologs revealed a helix swap on one of the hemerythrin‐like domain families, which made it difficult to obtain a correct structural alignment with most algorithms (Supporting Information Table SI). To evaluate the structural similarity between cation‐coordinating sites of hemerythrin homologs we used MetaS2, which is a local structure alignment strategy that starts by superimposing the metal ions and the donor atoms of a pair of structures. The MetaS2 algorithm scores alignments considering sequence similarity, fractional coverage of the smallest site and fragmentation.33 Lower scores indicate more similar environments of a pair of metal‐binding sites.33 For comparison, we also included the structure of an uncharacterized protein (Q4MWP8 from Bacillus cereus G9241, PDB code: 3DBY) with a cation‐coordination site structurally similar to that of hemerythrin homologs (Fig. 1). The Metals2 score was used as a measure of the local structural pairwise superposition of the metal‐binding sites compared.33

Supporting information

Supporting Information Figure S1

Supporting Information Figure S2

Supporting Information Figure S3

Supporting Information Figure S4

Supporting Information Table S1