A bioinformatics insight to rhizobial globins: gene identification and mapping, polypeptide sequence and phenetic analysis, and protein modeling.

Database search

Putative Glb sequences and Glb domains were identified in databases (Table S1) containing the genomes of rhizobial species and strains using the query sequences S. meliloti fHb; Vitreoscilla SDgb; Agrobacterium tumefaciens GCS; Methanosarcina acetivorans protoglobin; Methylacidiphilum infernorum sensor single domain globin; Mycobacterium tuberculosis tHb class 1; A. tumefaciens tHb class 2, and M. avium tHb class 3 (Genbank accession numbers AY328026, AAA75506, NP_354049, 2VEB_A, YP_001939425, NP_216058, WP_020813663 and BAN32501, respectively) and the SUPERFAMILY database (http://supfam.mrc-lmb.cam.ac.uk)²⁸. Resulting sequences were subjected to a FUGUE analysis (http://tardis.nibio.go.jp/fugue/prfsearch.html)²⁹ to determine the most similar Glb structure and presence of proximal H at the myoglobin-fold position F8. Putative Glbs had to satisfy the following criteria: length higher than or ~100 amino acids, a FUGUE Z score higher than 6 (which corresponds to 99% specificity²⁹) with known Glb structures, and the presence of proximal H at position F8.

Gene mapping and detection of promoter sequences

Scaffolds containing copies of the glb gene were used for mapping glbs. This included the detection of open reading frames (ORFs) ~5 kb up- and downstream to glbs and ORF length, transcription direction and localization in the +/- strand. Canonical (-10 and -35) and Fnr³⁰ promoter sequences and Shine-Dalgarno sequences were searched within 130 nucleotides upstream to the rhizobial glb genes either by using the search tool of MS Word^® or by pairwise sequence alignments using the ClustalX program (http://www.clustal.org/clustal2/)³¹.

Protein sequence alignments and phenetic analysis

Pairwise and multiple sequence alignments were performed using the ClustalX program³¹. Multiple sequence alignment was manually verified using the procedure described by Kapp et al.³² based on the myoglobin-fold³³. A phenogram was constructed from the aligned sequences using the UPGMA method from the ClustalX program. The resulting phenogram was edited using the iTOL program (http://itol.embl.de/)³⁴.

Modeling and analysis of the predicted proteins tertiary structure

The tertiary structure of rhizobial Glbs was modeled using the automated mode of the I-TASSER server (http://zhanglab.ccmb.med.umich.edu/I-TASSER/)^35–37, which also provided the best structural homologs to the query sequences. Models were edited using the VMD program (http://www.ks.uiuc.edu/Research/vmd/)³⁸ and Adobe Photoshop^® software. Distance and dihedral angles of amino acids at the heme prosthetic group were calculated using the distance and dihedral tools of the SwissPDBViewer program (http://spdbv.vital-it.ch/) as described by Gopalasubramaniam et al.³⁹ and Sáenz-Rivera et al.⁴⁰, respectively.

Bacterial growth, cell rupture and spectral analysis

Bradyrhizobium japonicum USDA38 and USDA58 were kindly provided by Drs. Donald Keister and Douglas Jones (United States Department of Agriculture, USA). All reagents were purchased from Sigma-Aldrich (St. Louis MO, USA). B. japonicum cells were grown in YM (Yeast Mannitol) broth (per 100 ml: KH₂PO₄, 50 mg; MgSO₄, 20 mg; NaCl, 10 mg; mannitol, 1 g; yeast extract, 50 mg, pH 7.0) for 3 to 5 days at 30°C with shaking at 200 rpm. Cells were harvested by centrifugation at 11,000 × g, pellets were resuspended in 50 mM Na-phosphate buffer (pH 7.2) containing 1 mM EDTA and 1 mM phenylmethylsulfonyl fluoride (PMSF). Cells were disrupted by sonication at maximum power (three cycles of 1 min each in ice) and incubation at 4°C overnight with gentle agitation after the addition of DNAse I (40 U/ml), RNAse A (3 U/ml) and lysozyme (2 mg/ml). The resulting solution was cleared by centrifugation at 22,000 × g for 40 min at 4°C, and the supernatant was fractionated with solid ammonium sulphate between 35 and 65% saturation. The resulting pellet was resuspended in 5 ml of 50 mM Na-phosphate buffer (pH 7.2) containing 1 mM EDTA and 1 mM PMSF and dialyzed for 18 h against the same buffer to remove the excess of salts. 0.5 to 1 ml aliquots of the dialyzed solution were used to obtain the dithionite reduced + CO minus dithionite reduced differential spectra in a Beckman DU6 spectrophotometer. Control spectra were obtained from commercial (Sigma-Aldrich) preparations of the sperm whale myoglobin and bovine blood hemoglobin.

Detection of Glb sequences in the genomes of α- and β-rhizobia

Recently, Vinogradov et al.⁷ reported that Glb sequences exist in the genomes of 96 rhizobia. However, this report did not provide the rhizobial Glb sequences or links to rhizobial scaffolds containing the Glb sequences. Hence, we searched in databases (see the Methods section and Table S1) in order to obtain rhizobial Glb sequences for analysis. We selected 62 out of the 96 rhizobial genomes reported by the above authors representing the major rhizobial genera, species and strains, which included α- and β-rhizobia (i.e. those classified within the α- and β-proteobacteria, respectively). A total of 197 glb sequences were detected in the 62 rhizobial genomes, corresponding to 7 fhbs, 47 sdgbs, 40 gcss and 103 thbs (4 thbs class 1, 56 thbs class 2 and 43 thbs class 3). Individual Glb nucleotide and polypeptide sequences and links to rhizobial scaffolds containing the Glb sequences are provided in Dataset 1 and Dataset 2, respectively. All the rhizobial genomes analyzed in this work contained glb sequences, thus indicating that glbs are widespread in rhizobia. However, protoglobin and sensor single domain globin sequences were not detected in the rhizobial genomes. This observation indicates that apparently only the fhb, sdgb, gcs and thb lineages evolved within rhizobia.

A distribution analysis showed that most (61) of the rhizobial genomes analyzed in this work contain thbs, either as single thbs (13) or in combination with fhbs, sdgbs and/or gcss (48). Furthermore, one rhizobial genome contained only a gcs and none contained only fhbs and sdgbs and the combinations fhbs + sdgbs, fhbs + gcss and sdgbs + gcss (Figure 1). These observations indicate that in the rhizobia analyzed in this work thbs predominate over other glbs and that in these bacteria fhbs, sdgbs and gcss mostly exist in combination with thbs. Also, analysis of the glb copy number showed that in the rhizobia analyzed in this work fhbs mostly exist as single copy (ranging from one to two copies), sdgbs mostly exist as two copies (ranging from one to four copies), gcss exist as either single or two copies (ranging from one to two copies) and thbs mostly exist as two copies (ranging from one to three copies) although quite a few thbs exist as single copy (Table 1). Thus, apparently rhizobial glbs mostly exist as either single or two copies.

Figure 1. Venn diagram illustrating the distribution of glb genes in the rhizobial bacteria analyzed in this work.

Numbers correspond to rhizobial genomes containing glbs.

Mapping of glb genes in the rhizobial genomes

The glb genes detected in this work were mapped within the rhizobial genomes in order to identify genes that flank nearby to and could coexpress with glbs. Mapping analysis showed that rhizobial glb copies are located in different scaffolds and that they are not tandemly arrayed. Figure S1A shows that either no ORFs or ORFs coding for hypothetical or non-identified proteins are located nearby most of the rhizobial fhb genes. However, genes coding for the transcriptional regulator NsrR, 2-nitropropane dioxygenase and NosR, Z, D, F, Y and X are located nearby cupnecN1fhb1, rhilegUPM1137fhb and sinmel1021fhb, respectively. Figure S1B shows that B. elkanii and B. japonicum sdgbs are mostly flanked by genes coding for proteins that function in nitrate/nitrite metabolism and sugar transport. Figure S1C shows that genes coding for proteins that function in chemotaxis are located nearby several rhizobial gcss, although genes coding for a peptide deformylase, sugar and nitrate transport proteins and NAD(P)H nitrate reductase are located nearby some other rhizobial gcss. Figure S1D shows that genes flanking the rhizobial thbs are rather variable. However, B. japonicum thbs are often flanked by genes coding for the transcriptional regulator Rieske Fe-S, shikimate kinase and alcohol dehydrogenase; mesorhizobia thbs are often flanked by genes coding for permeases and tRNA-Trp, and R. leguminosarum thbs are often flanked by genes coding for membrane proteins. Thus, if glb and flanking genes coexpress in rhizobia, and proteins coded by these genes function within the same metabolic pathways, the above observations suggest that rhizobial Glbs could play a variety of roles in rhizobial physiology, including nitrate/nitrite metabolism, transport processes, gene regulation and chemotaxis. Interestingly, with the exception of sinmel1021fhb which is flanked by nos and fix genes (Figure S1A)²⁶, nif and fix genes coding for proteins that function in N₂-fixation were not detected nearby the rhizobial glb genes. This observation suggests that rhizobial Glbs might not directly function in N₂-fixation.

Detection of promoter sequences upstream to the rhizobial glb genes

Identification of promoter sequences is crucial to an understanding of gene regulation and ultimately protein function within the cell's physiology. Hence, we searched for canonical (-10 and -35) promoters and the O₂- and NO-regulated Fnr promoter^30,41,42 within 130 nucleotides upstream to 44 selected rhizobial glb genes (i.e. those representative of major rhizobial Glb clades identified in this work (see Figure 2)). Also, we searched for Shine-Dalgarno sequences within the same region, which indicate that Glb transcripts could be translated into proteins. Results showed that, with the exception of burphySTM815thb1, burphySTM815thb2 and rhilupHPC(L)thb1, a -10 promoter is absent upstream of the selected rhizobial glbs. In contrast, with the exception of cupnecN1thb1 and rhilupHPC(L)thb2, a -35 promoter exists upstream of the selected rhizobial glbs. Searching for Fnr promoter sequences revealed that Fnr-like promoters exist upstream to 30 out of the 44 selected rhizobial glbs, including fhb, sdgb, gcs and thb genes. A Shine-Dalgarno sequence was detected upstream to most of the selected rhizobial glbs (Table 2). These observations suggest that the -35 promoter is a major canonical promoter that regulates most of the rhizobial glbs, that it is likely that several rhizobial glbs are regulated by levels of O₂ and NO throughout an FNR mechanism^41–44 and that rhizobial Glb transcripts are translated into proteins.

Sequence alignments and phenetic analysis of rhizobial Glbs

Pairwise sequence alignments showed that the rhizobial fHbs, SDgbs, GCSs and tHbs analyzed in this work are 34.6 to 85.4%, 6.7 to 100%, 10.9 to 100% and 3.5 to 100% identical, respectively. This indicates that variability among the rhizobial Glb sequences is high. Moreover, identity values for the fHbs globin and flavin domains were 39.1 to 93.7% and 26.5 to 81.1%, respectively, and identity values for the GCSs globin and transmitter domains were 17.5 to 100% and 5.9 to 100%, respectively. Thus, apparently in the rhizobial fHbs and GCSs analyzed in this work the globin domain is more conserved than the flavin and transmitter domains.

The average length and molecular mass for the rhizobial fHbs, SDgbs, GCSs and tHbs analyzed in this work are 400 amino acids and 44 kDa, 141 amino acids and 15 kDa, 510 amino acids and 55 kDa and 149 amino acids and 17 kDa, respectively. However, sequence analysis revealed that globin domain from BraelkUSDA76tHb1, BraelkUSDA94tHb1 and Braelk587tHb2 contains 119 to 237 extra amino acids at the N-terminal and 131 extra amino acids at the C-terminal, and that the globin domain from BrajapUSDA123tHb1, BrajapUSDA135tHb1, BraelkWSM1741SDgb2, RhietlCFN42GCS1, BrajapUSDA4tHb2 and BrajapWSM2793tHb3 contains 27 to 73 extra amino acids at the N-terminal. In contrast, a large deletion comprising helices A and B, CD loop and part of helix E was detected in the BraelkUSDA94SDgb2 sequence indicating that BraelkUSDA94SDgb2 is 89 amino acids in length (Figure S2).

Multiple sequence alignment showed that, with the exception of 21 GCSs, in the rhizobial Glbs analyzed in this work, the proximal (F8, located at position 322/323 in Figure S2) amino acid to the heme Fe is H. Apparently, in the above rhizobial GCSs, F8 is E. Amino acids other than H occupying the F8 position in bacterial Glbs were previously reported by Vinogradov et al.⁷. However, because H F8 is absolutely conserved in Glbs (i.e. from bacteria to mammals)^1,32,45–47, assigning E F8 to rhizobial (and other bacterial) GCSs should be taken with caution as this assignment might result from a sequence alignment artifact. Ideally, F8 from rhizobial GCSs should be identified by experimental methods, such as x-ray crystallography. Multiple sequence alignment also showed that in the rhizobial Glbs analyzed in this work, the distal (E7, located at position 285/289/290 in Figure S2) amino acid to the heme Fe is Q in fHbs, can be Q/R/K/M/L in SDgbs, Q in GCSs and can be H/F/L/V/R in tHbs. This indicates that distal Q is conserved in rhizobial fHbs and GCSs and that amino acids occupying the distal position in rhizobial SDgbs and tHbs are variable. The B10 and CD1 amino acids (located at positions 257 and 270/271/273 in Figure S2, respectively), which also participate in binding of ligands to the heme Fe^48–50, are Y and F in most of the rhizobial Glbs analyzed in this work followed by (in order of abundance) F, S and V and H, I, S and Y, respectively.

Figure 2. Phenetic relationships among Glbs detected in the genomes of rhizobial bacteria.

Phenogram was obtained from the Glbs sequence alignment shown in Figure S2. The fHb, SDgb, GCS, tHb class 1, tHb class 2 and tHb class 3 clusters are indicated with light blue, dark blue, red, light green, bright green and dark green, respectively. Stars indicate Glbs selected for the detection of promoter sequences upstream to the glb genes and Glb protein modeling.

A phenogram was constructed from the above multiple sequence alignment. Figure 2 shows that the rhizobial Glbs analyzed in this work segregate into two main lineages: one containing fHbs, SDgbs and GCSs, and the other containing tHbs (the fHb/SDgb/GCS and tHb lineages, respectively). This is consistent with the main evolutionary lineages identified in bacterial Glbs^1,51,52 thus indicating that major evolutionary patterns for rhizobial Glbs were identical to those for other bacterial Glbs. Rhizobial fHbs and GCSs cluster with rhizobial SDgbs within the fHb/SDgb/GCS lineage owing to the similarity between the fHb and GCS globin domains and SDgbs. This has been postulated to be the result of an early divergence from a common ancestor to the bacterial fHb and GCS globin domains and SDgbs^1,6. The tHb lineage segregates into rhizobial tHbs class 1, tHbs class 2 and tHbs class 3. Within this lineage the rhizobial tHbs class 3 segregate in ancestral position to the rhizobial tHbs class 1 and tHbs class 2. Also, the bradyrhizobial, azorhizobial, mesorhizobial, rhizobial and burkholderial tHbs class 3 segregate from each other; the segregation within rhizobial, sinorhizobial, mesorhizobial and β-rhizobial tHbs class 2 is rather conserved, and bradyrhizobial tHbs class 2 and class 3 segregate into the B. elkanii and B. japonicum tHb sublineages. These observations indicate that rhizobial tHbs evolved similarly to other bacterial tHbs^7,8,52 and that evolution of rhizobial tHb sublineages was rather conserved.

Modeling and analysis of the predicted rhizobial Glbs tertiary structure

Structure elucidation is essential to a full understand of a protein´s function within the cell´s physiology. The structure of a considerable number of bacterial and non-bacterial Glbs has been elucidated by x-ray crystallography. However, with the exception of a S. meliloti fHb whose tertiary structure was predicted using bioinformatics methods²⁶, the structure of rhizobial Glbs is not known. Hence, we used bioinformatics methods to predict and analyze the tertiary structure of 44 selected rhizobial Glbs (i.e. those representative of major rhizobial Glb clades identified in this work (see Figure 2 and Table S2)) using the best structural homologs as templates (Dataset 3).

Predicted structures for selected rhizobial SDgbs and fHbs and GCSs globin domain and tHbs fold into the 3/3- and 2/2-globin fold, respectively (Figure 3 to Figure 8). Figure 3 shows that structures among the predicted rhizobial fHbs are highly similar. Yet major differences were detected in the BurphySTM815fHb, CupnecHPC(L)fHb and RhilegUMP1137fHb flavin domains, which exhibited two additional helices. Dataset 3 shows that among globin domains from predicted rhizobial fHbs the distance of the proximal H and distal Q to the heme Fe is 1.44 to 2.47 Å and 6.71 to 15.35 Å, respectively. This observation suggests that the heme Fe in rhizobial fHbs is pentacoordinate.

Figure 3. Predicted structure of rhizobial fHbs (blue) overlapped to structural homologues (green).

Structural homologues are indicated in Dataset 3. Distal and proximal amino acids to the heme Fe and amino acids that interact with the FAD cofactor are shown in brown. Heme and FAD are shown in red and yellow, respectively. Helices within the globin domain are indicated with letters A to H. All structures are displayed in the same orientation.

Figure 4. Predicted structure of selected rhizobial SDgbs (blue) overlapped to structural homologues (green).

Structural homologues are indicated in Dataset 3. Distal and proximal amino acids to the heme Fe are shown in brown. Heme is shown in red. Helices are indicated with letters A to H. All structures are displayed in the same orientation.

Figure 5. Predicted structure of selected rhizobial GCSs globin domain (blue) overlapped to structural homologues (green).

Structural homologues are indicated in Dataset 3. Distal and proximal amino acids to the heme Fe are shown in brown. Heme is shown in red. Helices are indicated with letters A to H. All structures are displayed in the same orientation.

Figure 6. Predicted structure of class 1 CupnecN1tHb1 (blue) overlapped to the structural homologue Tetrahymena pyriformis tHb (PDB ID 3AQ5) (green).

Distal and proximal amino acids to the heme Fe are shown in brown; only potential distal E11 is shown in the CupnecN1tHb1 structure. Heme is shown in red. Helices are indicated with letters A to H.

Figure 7. Predicted structure of selected rhizobial tHbs class 2 (blue) overlapped to structural homologues (green).

Structural homologues are indicated in Dataset 3. Distal and proximal amino acids to the heme Fe are shown in brown; only potential distal E11 is shown in the tHbs structure. Heme is shown in red. Helices are indicated with letters A to H. Pre-helix F is indicated with the Greek letter φ. All structures are displayed in the same orientation.

Figure 8. Predicted structure of selected rhizobial tHbs class 3 (blue) overlapped to structural homologues (green).

Structural homologues are indicated in Dataset 3. Distal and proximal amino acids to the heme Fe are shown in brown; only potential distal E11 is shown in the tHbs structure. Heme is shown in red. Helices are indicated with letters A to H. All structures are displayed in the same orientation.

Figure 4 shows that 3/3-globin folding is highly conserved in the predicted structure of the rhizobial SDgbs AzodoeUFLA1-100SDgb, BraelkUSDA3254SDgb2, BraelkUSDA3259SDgb1 and BrajapUSDA38SDgb2. Major variations to 3/3-globin folding from predicted rhizobial SDgbs consisted of the existence of an unusually short helix E in BraelkUSDA94SDgb2, a long helix H in BraelkUSDA3254SDgb1 and BrajapUSDA124SDgb1, and the existence of a pre-helix A followed by a long loop at the N-terminal of BraelkWSM1741SDgb2. Dataset 3 shows that among the predicted rhizobial SDgbs the distance of proximal H and distal Q/R/K/M to the heme Fe is 2.11 to 4.44 Å and 5.08 to 6.63 Å, respectively. This observation suggests that the heme Fe in rhizobial SDgbs is either penta- or hexacoordinate.

Only the globin domain from bacterial GCSs has been crystalized and analyzed by x-ray crystallography^53,54 (Dataset 3). Crystal structure for the bacterial GCSs transmitter domain has not been elucidated. Hence, we only predicted and analyzed the tertiary structure of globin domains from the selected rhizobial GCSs. Figure 5 shows that the predicted rhizobial GCSs globin domain exhibits a 1.5- to 3-turn pre-helix A, that (with the exception of SinfreGR64GCS) no loop exists between helices A and B, and that helix H is unusually long in Rhietl8C3GCS, RhietlCIAT652GCS2 and RhilegGB30GCS2. Dataset 3 shows that among the predicted rhizobial GCSs globin domain distance of proximal H/E and distal Q to the heme Fe is 1.77 to 5.56 Å and 4.09 to 9.04 Å, respectively. This observation suggests that the heme Fe in the rhizobial GCSs globin domain is either penta- or hexacoordinate.

Figure 6 to Figure 8 show that 2/2-globin folding is highly conserved in the predicted rhizobial tHbs class 1, class 2 and class 3. Major variations to 2/2-globin folding from predicted rhizobial tHbs consisted of the existence of a 2.5-turn pre-helix A followed by a long loop at the N-terminal of (class 1) CupnecN1tHb1 (Figure 6); the existence of a one-turn pre-helix F (designated as φ in Figure 7⁸) in the rhizobial tHbs class 2; the existence of a long and extended C-terminal region in (class 2) BraelkUSDA94tHb1 (Figure 7), and the substitution of helix A by a long loop that connects to helix B through a 1- to 2.5-turn pre-helix B in (class 3) BraelkUSDA76tHb2, BrajapUSDA123tHb1, BurphySTM815tHb1, MeslotNZP2037tHb2 and Sinmel1021tHb2 (Figure 8). Dataset 3 shows that among the predicted rhizobial tHbs, the distance of proximal H and distal H/L/F to the heme Fe is 1.77 to 7.51 Å and 4.09 to 8.25 Å, respectively. This observation suggests that the heme Fe in the rhizobial tHbs is either penta- or hexacoordinate.

The above observations suggest that in spite of sequence variability (see the Sequence alignments and phenetic analysis of rhizobial Glbs subsection) the structure of rhizobial Glbs is similar to the canonical 3/3- or 2/2-globin folding of bacterial and non-bacterial Glbs. However, a number of predicted rhizobial Glbs exhibited variations at the N- and C-terminal regions suggesting that their structural properties could be different to those of canonical Glbs.

Data also shows that (with few exceptions) in addition to proximal and distal amino acids the distance of B10 and CD1 amino acids to the heme Fe and the orientation of proximal, distal, B10 and CD1 amino acids are similar within and among the predicted rhizobial SDgbs, fHbs and GCSs globin domain and tHbs. These amino acids participate in the binding of ligands to the heme Fe. Thus, these observations suggest that the mechanisms and chemistry for ligand binding are similar among the rhizobial Glbs.

Spectroscopic identification of putative Glbs in soluble extracts from Bradyrhizobium japonicum USDA38 and USDA58

The prerequisites for being able to infer a protein’s function are isolating and characterizing either native or recombinant proteins and detecting protein synthesis in vivo. No rhizobial Glb has been isolated and characterized thus far. However, spectroscopic evidence indicates that putative Glbs exist in soluble extracts from B. japonicum 505 (Wisconsin), R. leguminosarum bv. viciae, B. japonicum NPK63 and R. etli CE3 (see the Introduction section). In order to extend these analyses to other rhizobia, we analyzed soluble extracts from B. japonicum USDA38 and USDA58 by (dithionite reduced + CO minus dithionite reduced) differential spectroscopy using as controls the sperm whale myoglobin and bovine blood hemoglobin. Table 3 shows that absorption peaks and troughs in the Soret and Q regions for the B. japonicum USDA38 and USDA58, B. japonicum 505 (Wisconsin), R. leguminosarum bv. viciae, B. japonicum NPK63 and R. etli CE3 soluble extracts, Vitreoscilla VHb, E. coli K12 Hmp, sperm whale myoglobin and bovine blood hemoglobin are nearly identical. This preliminary evidence indicates that putative soluble Glbs are synthesized in B. japonicum USDA38 and USDA58. Interestingly, genes coding for SDgbs (brajapUSDA38SDgb1 and brajapUSDA38SDgb2) and tHbs (brajapUSDA38tHb1 and brajapUSDA38tHb2) were identified in the B. japonicum USDA38 genome (Dataset 1). Thus, it is likely that putative B. japonicum USDA38 Glbs corresponds to a combination of SDgbs and tHbs. Inferences from the preliminary results reported here should be confirmed by Glb detection, isolation and unequivocal identification after protein sequencing. This may open the possibility to carry out further experimental analyses on rhizobial Glbs.