ORIGINAL RESEARCH
published: 26 August 2016
doi: 10.3389/fmicb.2016.01343
Edited by:
Martin G. Klotz,
Queens College, City University
of New York, USA
Reviewed by:
James Hemp,
California Institute of Technology, USA
Eric Boyd,
Montana State University, USA
*Correspondence:
Dmitry A. Ravcheev
dmitry.ravcheev@uni.lu
Mikhail S. Gelfand
gelfand@iitp.ru
†
These authors have contributed
equally to this work.
Specialty section:
This article was submitted to
Evolutionary and Genomic
Microbiology,
a section of the journal
Frontiers in Microbiology
Received: 09 March 2016
Accepted: 15 August 2016
Published: 26 August 2016
Citation:
Tsoy OV, Ravcheev DA,
ˇ
Cuklina J
and Gelfand MS (2016) Nitrogen
Fixation and Molecular Oxygen:
Comparative Genomic
Reconstruction of Transcription
Regulation in Alphaproteobacteria.
Front. Microbiol. 7:1343.
doi: 10.3389/fmicb.2016.01343
Nitrogen Fixation and Molecular
Oxygen: Comparative Genomic
Reconstruction of Transcription
Regulation in Alphaproteobacteria
Olga V. Tsoy
1†
, Dmitry A. Ravcheev
2
*
†
, Jelena
ˇ
Cuklina
1,3
and Mikhail S. Gelfand
1,4,5,6
*
1
Research and Training Center on Bioinformatics, A.A. Kharkevich Institute for Information Transmission Problems, Russian
Academy of Sciences, Moscow, Russia,
2
Luxembourg Centre for Systems Biomedicine, University of Luxembourg,
Esch-sur-Alzette, Luxembourg,
3
Moscow Institute of Physics and Technology, Dolgoprudny, Russia,
4
Faculty of
Bioengineering and Bioinformatics, Moscow State University, Moscow, Russia,
5
Skolkovo Institute of Science and
Technology, Skolkovo, Russia,
6
Faculty of Computer Science, Higher School of Economics, Moscow, Russia
Biological nitrogen fixation plays a crucial role in the nitrogen cycle. An ability to fix
atmospheric nitrogen, reducing it to ammonium, was described for multiple species
of Bacteria and Archaea. The transcriptional regulatory network for nitrogen fixation
was extensively studied in several representatives of the class Alphaproteobacteria.
This regulatory network includes the activator of nitrogen fixation NifA, working
in tandem with the alternative sigma-factor RpoN as well as oxygen-responsive
regulatory systems, one-component regulators FnrN/FixK and two-component system
FixLJ. Here we used a comparative genomics approach for in silico study of the
transcriptional regulatory network in 50 genomes of Alphaproteobacteria. We extended
the known regulons and proposed the scenario for the evolution of the nitrogen fixation
transcriptional network. The reconstructed network substantially expands the existing
knowledge of transcriptional regulation in nitrogen-fixing microorganisms and can be
used for genetic experiments, metabolic reconstruction, and evolutionary analysis.
Keywords: bacteria, transcription factors, nitrogen fixation, comparative genomics, regulons
INTRODUCTION
Nitrogen is indispensable for all living species. The Earth atmosphere mainly consists of nitrogen,
but in the form of dinitrogen (N
2
) which is not available for living organisms. Only some Bacteria
and Archaea can convert N
2
into ammonium (NH
3
) to further synthesize organic compounds.
This process is known as biological nitrogen fixation. Biological nitrogen fixation plays a crucial
role in the nitrogen cycle as it returns the element from the geosphere and atmosphere to living
species (reviewed in
Dixon and Kahn, 2004).
Species capable of nitrogen fixation belong to Bacteria and Archaea, as no examples
of nitrogen fixation in Eukarya are known. Within Archaea, nitrogen fixation has been
observed in some methanogens (Methanobacteriales, Methanococcales, and Methanosarcinales).
Frontiers in Microbiology | www.frontiersin.org 1 August 2016 | Volume 7 | Article 1343
Tsoy et al. Nitrogen Fixation and Molecular Oxygen in Alphaproteobacteria
In Bacteria, nitrogen fixation is much more widely distributed
and has been characterized in phyla Actinobacteria, Bacteroidetes,
Cyanobacteria, Chlorobi, Chloroflexi, Firmicutes, and
Proteobacteria (Raymond et al., 2004; Dos Santos et al.,
2012
; Boyd and Peters, 2013).
Nitrogen fixation is often the limiting factor for crop and
natural ecosystem productivity making this process important for
agriculture (Dixon and Kahn, 2004). It is carried out that a large
number of nitrogen fixing bacteria in soils and in symbioses with
plants are Alphaproteobacteria.
The main enzyme catalyzing reduction of N
2
to ammonium
is the nitrogenase. All known nitrogenases require a FeS-cluster
and other metal-dependent cof actors for electron transduction.
The best studied and the most common is the molybdenum-
dependent nitrogenase which is encoded by t he nifHDK
genes (
Boyd and Peters, 2013). Alternative nitrogenases are
the vanadium- and iron-dependent nitrogenases encoded by
the vnfHDGK and anfHDGK genes, respectively (Rehder,
2000; Seefeldt et al., 2009; Hartmann and Barnum, 2010).
Since nitrogenases are irreversibly inhibited by dioxygen (O
2
)
(
Gallon, 1981; Boyd and Peters, 2013), bacteria developed
various strategies to avoid the latter, such as elimination of
O
2
through enzyme-catalyzed reactions, symbiotic nodules,
compartmentalization etc (Oelze, 2000; Boyd et al., 2015).
In spite of the importance of nitrogen fixation, the
regulation of this process has been well described in only
a few bacterial species. Diazotrophs have developed multiple
strategies in regulation of nitrogen fixation. Some regulate
nitrogen fixation post-translationally (
Huergo et al., 2012). For
example, in Rhodobacter capsulatus nitrogenase can be by
reversibly inactivated by the DraT protein (Masepohl et al.,
2002). Proteins NifL and NifI
1
I
2
, involved in post-translational
regulation in Klebsiella pneumoniae (Dixon and Kahn, 2004)
and Heliobacterium chlorum (
Enkh-Amgalan et al., 2006)
respectively, are absent in Alphaproteobacteria (
Boyd et al.,
2015). PII proteins, critical for Herbaspirillum seropedicae, have
been shown to be non-essential for nitrogen fixation in some
Alphaproteobacteria (Arcondeguy et al., 1997; Yurgel et al., 2010).
From early studies on nitrogen fixation in rhizobia it is known,
that nitrogen fixation cannot occur in nifA, fixJ, and fixK mutants.
All these genes encode transcription factors (
Hennecke, 1990). In
Proteobacteria, the nitrogenase genes are invariably activated by
NifA, while FixJ and FixK were shown to control nitrogen fixation
in response to microaerobic conditions in certain Rhizobiales
(Fischer, 1994). The regulation by NifA, FixJ, and FixK has been
well described in only a few model species. Whether functions
regulated by these f actors span beyond regulation of nitrogen
fixation, and what are other possible members of this regulatory
network, what are the relationships between these regulators, and
to what extent these networks are conserved in different species
is not clear.
In most nitrogen-fixing bacteria NifA is the master regulator
of nitrogen fixation. It works in association with the RNA-
polymerase sigma factor RpoN (sigma54; Sullivan et al., 2002;
Sciotti et al., 2003). RpoN re cognizes a −24/−12 promoter
sequence with the consensus TGGCACG-N
4
-TTGCW and
binds the NifA protein which in turn recognizes its own
binding site with the consensus sequence TGT-N
10
-ACA (
Barrios
et al., 1999
). As mentioned above, nitrogen fixation is very
sensitive to O
2
concentration, and NifA has an ability to
sense it through conserved cysteine residues. While NifA effects
nitrogen fixation itself, FixLJ and FnrN/FixK are responsible for
the cell adaptation to microaerobic conditions (
Sciotti et al.,
2003).
FixJ is a DNA-binding response regulator working in
tandem with sensor kinase FixL, constituting a two-component
regulatory system FixLJ (
Reyrat et al., 1993; Torres et al., 2011).
FixL directly senses O
2,
which binds the haem group in the sensor
domain and inactivates its kinase activity (Rey and Harwood,
2010; Sousa et al., 2015). FixJ is one of the major regulators
of nitrogen fixation, but the only known conserved member
of the FixJ regulon is fixK (Nellen-Anthamatten et al., 1998).
Several attempts have been made to determine its binding motif,
leading to varying results. Sequence analysis of the fixK promoter
region in Sinorhizobium. meliloti, Azorhizobium caulinodans
and Bradyrhizobium japonicum yielded a common CSNAATWT
motif at position −33 and an additional TAAG element
around position −64 (
Sciotti et al., 2003). This observation was
confirmed by DNase protection which showed that S. meliloti
FixJ binds to the fixK promoter in two different regions (−69,
−44) and (−57 and −31) relative to the transcription start
site (Galinier et al., 1994). The application of the SELEX
technique predicted two classes of FixJ-binding sites with the
consensus motifs GTAGTTTCCC and GTAMGTAG (
Ferrieres
and Kahn, 2002). The most recent studies analyzed FixJ-binding
sites in S. meliloti using a gel shift assay and the NMR
structure of the truncated C-terminal FixJ DNA-binding domain.
They determined the TAAGTAATTTCCCTTA sequence in the
upstream region of the fixK gene as the FixJ-binding site
(Kurashima-Ito et al., 2005). Therefore, the lack of a common
consensus complicates identification of new FixJ targets.
Some Rhizobiales, in particular, R. etli and R. leguminosarum,
lack FixJ, and thus use an alternative regulatory cascade
comprised of a hybrid histidine kinase hFixL and a response
regulator FxkR, analogous to FixJ and FixKf. A FxkR-binding
site wit h the consensus GTTACA-N
4
-GTTACA was identified
upstream of the fixKf gene by MEME and confirmed by the gel
shift assay (
Zamorano-Sanchez et al., 2012).
Other regulators of nitrogen fixation, FixK and FnrN, belong
to the Crp-Fnr superfamily of transcription factors (Matsui
et al., 2013; Rodgers et al., 2013). The main difference between
these proteins is the ability of FnrN to directly sense oxygen
with the iron–sulfur cluster. The cluster is formed by the
conserved N-terminal cysteine-rich domain and an additional
cysteine in the middle of the polypeptide chain (
Korner
et al., 2003). FixK and FnrN have been studied extensively in
Alphaproteobacteria and shown to control expression of oxidases,
hydrogen uptake, nitrogen metabolism, haem biosynthesis, and
nitrogen-fixation transcription f actors (Supplementary Table
S1 in the Supplementary Materials). The existence of two
closely related regulators, FnrN and FixK, is the main obstacle
for the genomic analysis of their regulatory interactions, as
both transcription factors recognize very similar binding sites
with the consensus motif TTGANCNNGATCAANG (
Cebolla
Frontiers in Microbiology | www.frontiersin.org 2 August 2016 | Volume 7 | Article 1343
Tsoy et al. Nitrogen Fixation and Molecular Oxygen in Alphaproteobacteria
and Palomares, 1994; Mes a et al., 2005, 2008; Bonnet et al.,
2013).
Hence, known regulatory networks for nitrogen fixation
in Alphaproteobacteria are rather complex, may vary between
species, and seem highly redundant due to the presence of
multiple transcription factors with very similar functions. The
experimental data about transcriptional regulation of nitrogen
fixation are difficult to obtain and thus are generated at a slow
rate. On the other hand, the growing number of complete
bacterial genomes and the development of comparative genomics
techniques allow us to analyze transcriptional regulatory
systems, addressing relevant biological questions. Here we
report the results of manual comparative analysis of the
regulatory network for nitrogen fixation in 50 complete genomes
belonging to five orders of Alphaproteobacteria. We predict
new members of the nitrogen fixation regulatory network and
reconstruct different types of regulatory cascades found in
Alphaproteobacteria. Based on these predictions, we propose a
model of the gradual growth of complexity for this network
and reconstruct possible evolutionary events that have formed
the network. We also have constructed t axon-specific profiles
for each of the analyzed regulators. The obtained dataset
including the information about the analyzed transcription
regulators, their binding sites and motifs, and re gulated genes
is deposited in the RegPrecise database (
Novichkov et al.,
2013).
MATERIALS AND METHODS
Comparative Genomics Approach for the
Reconstruction of Regulons
For the regulon reconstruction, we used a comparative genomics
approach (
Rodionov, 2007) implemented in the RegPredict
Web server (Novichkov et al., 2010). The workflow includes
inference of transcription regulator binding sites, construction
of positional weig ht matrices (profiles) for the binding-site
motifs, and further search for additional regulon members based
on predicted binding sites in gene promoter regions. To take
into account possible lineage-specific changes in binding site
motifs, we constructed individual profiles for genomes from each
considered order of Alphaproteobacteria.
To construct the profiles, the following steps were
implemented. An initial profile was constructed based on
known binding sites of a given regulator or on sequences of
binding sites predicted by phylogenetic footprinting, an approach
based on the analysis of conserved islands in multiple alignments
of DNA fragments (Shelton et al., 1997; McCue et al., 2001).
This profile was used for site prediction in genomes from a
given order. Two types of predicted sites were used to update the
taxon-specific profile: (1) sites predicted upstream of orthologs
of genes known to be regulated in at least one genome and (2)
sites upstream of operons with conser ved predicted regulation,
i.e., sites that were found upstream of orthologous genes in
at least than half of genomes from the analyzed order (the
consistency check approach;
Ravcheev et al., 2007; Rodionov,
2007). Operons were defined as groups of genes satisfying the
following criteria: same direction of transcription, intergenic
distance not exceeding 200 bp, absence of internal binding sites,
and conservation of the locus structure in a number of related
genomes. The obtained order-specific profile was further used
for the analysis of regulons in this order.
The initial profiles for the NifA and FnrN/FixK-like proteins
were constructed b ased on sequences of known binding sites
extracted from literature data. For the FixJ and FxkR proteins, the
initial profiles were constructed using phylogenetic footprinting.
The constructed profiles were further used to search for
additional regulon members using t he Run Profile tool in
RegPredict. The lowest score observed in the training set of
known and/or initially predicted binding sites was used as the
threshold for the genome-wide site search. To eliminate false
positives, the consistency check approach or functional relevance
of candid ate target operons were used.
Tools and Databases
Fifty studied genomes were downloaded from the
MicrobesOnline database (
Dehal et al., 2010). The initial
search for FnrN and FixK homologs was done using BLASTP
(Altschul et al., 1997) with known FnrN/FixK proteins (cutoff:
e-value = e-30, identity = 40%). The domain structure of
defined homologs was predicted with the Pfam database (Finn
et al., 2014) and used as an additional criterion: a protein
should the have N-terminal cyclic nucleotide-binding domain
(PF00027) and C-terminal Crp-like helix-turn-helix domain
(PF13545). Gene orthology was determined by the bidirectional
best hit criterion implemented in the GenomeExplorer software
(
Mironov et al., 2000) and valid ated by phylogenetic trees
from the MicrobesOnline database. Genes were considered
as orthologs if they: (1) formed a monophyletic branch in the
phylogenetic tree; and (2) demonstrated conserved chromosomal
gene context. Functional gene annotations were extracted from
the literature and uploaded from the SEED (Disz et al., 2010),
UniProt (Magrane and Consortium, 2011), MicrobesOnline, and
KEGG (Kanehisa et al., 2014) databases.
Multiple alignments for amino acid and nucleotide sequences
were constructed using a web version
1
of the MUSCLE tool
(Edgar, 2004) with default parameters. Phylogenetic trees were
constructed by the maximum-likelihood method (Felsenstein,
1996) and LG model for amino acid substitution (Le and Gascuel,
2008) implemented in −3.0 (Guindon et al., 2010) with default
parameters, and visualized with the Dendros cope (version 3.2.10,
built 19) program (Huson et al., 2007). Sequence logos for DNA
motifs were drawn with WebLogo (Crooks et al., 2004).
Data Availability
All predicted regulons including transcription regulators, their
binding sites, their regulated genes and operons, and functional
gene assignments are deposited in the RegPrecise database
(
Novichkov et al., 2013) and are freely available at http://
regprecise.lbl.gov/RegPrecise (Supplementary Table S2 in the
Supplementary Materials).
1
http://www.ebi.ac.uk/Tools/msa/muscle
Frontiers in Microbiology | www.frontiersin.org 3 August 2016 | Volume 7 | Article 1343
Tsoy et al. Nitrogen Fixation and Molecular Oxygen in Alphaproteobacteria
RESULTS
Distribution of Genes Encoding the
Nitrogenase and the Transcription
Regulators
To describe the transcriptional regulation for nitrogen fixation
systematically among Alphaprpteobacteria, we started with the
analysis of the distribution of genes encoding the nitrogenase
and the regulatory proteins in 50 genomes of Alphaproteobacteria
(Supplementary Table S3 in the Supplementary Materials).
nifHDK
The nitrogenase-encoding genes nifHDK were found in most
Rhizobiales and Rhodospirillales, and were not found in
Caulobacteriales. The nifHDK genes were also detected in one
Rhodobacterales genome, Rhodobacter sphaeroides 2.4.1, and one
Sphingomonadales genome, Zymomonas mobilis ZM4.
NifA
The orthologs of the transcription regulator NifA were found
in 10 Rhizobiales, 6 Rhodospirillales, 1 Sphingomonadales, and
1 Rhodobacterales genomes. No NifA orthologs were found
in Caulobacteriales. The NifA genes are always co-localized in
genome with the nifHDK genes, and in 10 studied genomes the
nifA genes are co-localized with the former (Supplementary Table
S3 in the Supplementary Materials).
FixLJ
Orthologs of the FixLJ two-component system were identified
in all four analyzed Caulobacteriales and in 15 of 20 analyzed
Rhizobiales. Previously, an alternative cascade called hFixL-FxkR-
FixKf has been identified in several Rhizobiales (
Zamorano-
Sanchez et al., 2012). Among genomes studied here, the
orthologs of these transcription factors were found in R. etli,
R. leguminosarum, and S. meliloti (Supplementary Table S3 in the
Supplementary Materials).
FnrN/FixK
Transcription factors related to FnrN and FixK that could also
be important for nitrogen fixation are distributed more widely
than genes for the nitrogenase itself (Supplementary Table S3 in
the Supplementary Materials). Some analyzed genomes contain
multiple copies of fnrN and/or fixK as usually more than one Crp-
Fnr superfamily protein per genome could be found (
Vollack
et al., 1999; Korner et al., 2003;Matsui et al., 2013)
FixK and FnrN are very closely related, so we performed a
detailed analysis of their orthologs, in order to be able to describe
the evolution of regulatory cascades involving these factors.
The FnrN/FixK orthologs were identified using the following
procedure: (1) homologs of previously known FnrN/FixK were
found in the analyzed genomes; (2) a phylogenetic tree for all
found homologs was constructed; (3) only proteins forming
monophyletic branches with known FnrN/FixK proteins were
retained for furt her analysis (Figure 1; Supplementary Figure S1
in the Supplementary Materials).
As the presence of cysteine residues required for formation
of oxygen-sensitive iron–sulfur clusters is a common feature
of the Fnr branch of the Crp-Fnr superfamily (
Korner et al.,
2003; Matsui et al., 2013), to distinguish between FnrN-like
and FixK-like regulators, we analyzed the detected proteins for
the presence of such cysteine residues. The proteins retaining
these cysteine residues formed a single monophyletic branch
on the tree (Figure 1) and, hence, were annotated as FnrN
orthologs. In addition to previously known FnrN proteins, this
branch also contains the AadR protein (locus tag: RPA4234)
from Rhodopseudomonas palustris (
Dispensa et al., 1992; Egland
and Harwood, 1999; Rey and Harwood, 2010), a transcription
regulator of anaerobic degradation of aromatic acids (see below).
Proteins lacking the cysteine residues form two branches. One
branch contains previously studied FixK proteins from various
Rhizobiales, whereas the second branch includes the FixKf (locus
tag: RHE_PF00508) protein of R. etli and is related to the FnrN
branch (Figure 1). Because of that, the latter group cannot
be considered as orthologs of FixK. On the other hand, the
absence of cysteine residues required for formation of the iron–
sulfur cluster (Supplementary Figure S1 in the Supplementary
Materials) shows that these proteins are not orthologs of FnrN. As
FnrN and FixKf branches are clearly separated from each other,
and both types of proteins are present in Rhizobium spp., we
propose that FnrN and FixKf are divergent paralogs.
As the FixKf branch is much closer to FnrN than to FixK,
we propose that FixKf and FixK independently have lost the
ability to form the iron–sulfur cluster for performing the same
functions. Thus, the FixKf proteins were defined as a separate
group of transcription regulators. Hence, among FnrN/FixK-like
proteins at least four groups of transcription regulators can be
distinguished: (1) FnrN (also known as FnrL in R. sphaeroides)
containing an iron-sulfur cluster, (2) regulator of anaerobic
degradation of aromatic acids AadR, also containing an iron-
sulfur cluster, (3) FixK proteins lacking an iron-sulfur cluster,
and (4) FixKf also lacking an iron-sulfur cluster, but much more
closely related to FnrN than to FixK.
Overall, FnrN/FixK-like proteins were found in 43 of 50
analyzed genomes. The distribution of these proteins is taxon-
specific. Caulobacteriales have only FixK proteins, whereas
Rhodobacterales, Rhodospirillales, and Sphingomonadales have
only FnrN proteins. In Rhizobiales, various combinations of
FixK, FnrN, and FixKf were observed – only FixK in three
genomes; only FnrN in two genomes; both FixK and FnrN in six
genomes; FixKf and FnrN in two genomes; and FixK and FixKf
in one genome (Supplementary Table S3 in the Supplementary
Materials). In all studied Alphaproteobacteria, except Z. mobilis,
FnrN/FixK-like proteins were found in all organisms having the
nitrogenase genes.
FixK/FixKf and FixLJ Co-occurrence
As the FixK proteins lack the iron-sulfur cluster, they are activated
in response to oxygen through the FixLJ-FixK regulatory cascade
(
Nellen-Anthamatten et al., 1998; Mesa et al., 2008; Reutimann
et al., 2010
). The fixK genes are present in all genomes having
the fixLJ operon, the only exception is R hizobium sp. NGR234,
lacking the fixLJ operon. For the fixKf genes such correlation was
not observed, as only in S. meliloti both fixKf and fixLJ genes were
found (Supplementary Table S3 in the Supplementary Materials).
Frontiers in Microbiology | www.frontiersin.org 4 August 2016 | Volume 7 | Article 1343
Tsoy et al. Nitrogen Fixation and Molecular Oxygen in Alphaproteobacteria
FIGURE 1 | Maximum-likelihood phylogenetic tree for FnrN/FixK-like proteins. Locus tags for proteins are shown. Previously analyzed proteins are shown in
bold. Solid branches correspond to proteins that have cysteine residues for the iron-sulfur cluster, dashed branches correspond to proteins lacking such cysteine
residues. Regulation by FixJ is shown by circles, regulation by FxkR, by triangles. The following proteins were used as the outgroups: the Fnr proteins from
Bordetella pertussis Tohama I (BP1197), Escherichia coli K-12 MG1655 (b1334), Neisseria gonorrhoeae FA 1090 (NGO1579), and Shewanella oneidensis MR-1
(SO2356), and the Anr protein from Pseudomonas aeruginosa PAO1 (PA1544).
Frontiers in Microbiology | www.frontiersin.org 5 August 2016 | Volume 7 | Article 1343