Genome-wide analysis of the response to nitric oxide n uropathogenic Escherichia coli CFT073
Abbreviations : ChIP-seq , chromatin immunoprecipitation and DNA sequencing ; DAVID , database for annotation visualization and integrated discovery ; EBSeq , empirical Bayes hierarchical model for inference in RNA-seq experiments ; iNOS , inducible nitric oxide synthase ; NONOate , compound with formula R1R2N - ( NO2 ) - N5O ; MACS2 , model-based analysis for ChIP-seq ; MEME , multiple Em for motif elicitation ; PSWM , position-specific weight matrix ; RNA-seq , whole transcriptome shotgun sequencing ; RSEM , RNA-Seq by expectation-maximization ; RT-PCR , reverse transcriptase PCR ; TCA , tricarboxylic acid ; TPM , transcripts per million ; UPEC , uropathogenic Escherichia coli ; UTI , urinary tract infection . 
Data Summary
1 . 
RNA-seq data have been deposited in the GEO database ; accession number : GSE69830 ( url -- http://www.ncbi.nlm . 
nih.gov/geo/query/acc.cgi ? 
acc5GSE69830 ) 
2 . 
ChIP-seq data have been deposited in the GEO database ; accession number : GSE69829 ( url -- http://www.ncbi.nlm . 
nih.gov/geo/query/acc.cgi ? 
acc5GSE69829 ) 
Introduction
Extraintestinal Escherichia coli are a group of bacteria that can survive as harmless human intestinal inhabitants but are serious pathogens when they enter the appropriate environment ( Welch et al. , 2002 ) . 
Uropathogenic Escheri-chia coli ( UPEC ) strain CFT073 is one such pathogen that is a causative agent of urinary tract infections ( UTIs ) and was isolated from the blood of a woman suffering from acute pyelonephritis ( Mobley et al. , 1990 ) . 
UPEC is responsible for 80 % of all symptomatic and asympto-matic UTIs ( Roos et al. , 2006 ) . 
During UTI , the mucosal inflammatory response is activated , which causes neutro-phils to infiltrate and migrate through the tissues and into the urine ( Godaly et al. , 2001 ) . 
Thus , UPEC is exposed to the defence mechanisms of the innate immune system . 
The antimicrobial properties of nitric oxide ( NO ) are exploited by cells of the innate immune system . 
In response to stimulation by proinflammatory cytokines and lipopolysaccharides of microbial pathogens , phagocy-tic cells express an inducible nitric oxide synthase ( iNOS ) , which oxidizes arginine to produce NO ( Fang , 1997 ; Fang & Vazquez-Torres , 2002 ; Mowat et al. , 2010 ) . 
iNOS is also expressed in epithelial cells of the urinary tract ( Poljakovic et al. , 2001 ) . 
NO ( and related N radicals that are derived from NO ) target proteins containing iron -- sulfur clusters , haem and thiols ( Fang , 1997 ; Fang & Vazquez-Torres , 2002 ; Kim et al. , 1995 ; Ren et al. , 2008 ) . 
Because of the cytotoxic properties of NO and its conge-ners , bacteria that proliferate within the host often employ strategies to convert NO into a non-toxic product . 
E. coli has three known enzymes that detoxify NO . 
Flavohaemoglobin ( Hmp ) is an NO denitrosylase that oxidizes NO to nitrate , and may reduce NO to N2O in the absence of oxygen ( Gardner & Gardner , 2002 ; Hausladen et al. , 2001 ) . 
Hmp has a role in protecting UPEC from NO stress in the host environment ( Svensson et al. , 2010 ) . 
Flavorubredoxin ( FlRd or NorV ) together with its NADH-linked reductase NorW reduces NO to N O ( Gomes 2 et al. , 2000 ) . 
FlRd is sensitive to oxygen in vitro , and so the enzyme has been described as an ` anaerobic ' NO reductase ( da Costa et al. , 2003 ; Gardner et al. , 2002 ) . 
The periplasmic nitrite reductase Nrf can catalyse the reduction of NO to ammonia under anaerobic conditions , a reaction that might contribute to defence against NO ( Poock et al. , 2002 ) . 
Recently , it has been suggested that the hybrid cluster protein Hcp is a key player in the response to NO under anaerobic conditions , when E. coli is lacking all previously known NO scavenging enzymes ( Cole , 2012 ) , although the enzymic activity of Hcp remains enigmatic . 
The response to NO in E. coli involves several transcription factors , including NsrR , FNR , SoxR , OxyR , NorR and Fur ( Bodenmiller & Spiro , 2006 ; Cruz-Ramos et al. , 2002 ; D'Autre ́aux et al. , 2002 ; Gardner et al. , 2002 ; Hausladen et al. , 1996 ; Pomposiello & Demple , 2001 ) . 
Most of these regulators primarily sense other signals such as oxygen , hydrogen peroxide and iron ( Bodenmiller & Spiro , 2006 ) , but NorR and NsrR serve as dedicated sensors of NO ( Tucker et al. , 2010 ) . 
The mononuclear non-haem iron centre of NorR directly senses NO , in response to which NorR activates transcription of norVW ( D'Autre ́aux et al. , 2005 ; Tucker et al. , 2010 ) . 
NsrR is an [ Fe -- S ] protein that is an NO-sensitive repressor of its target genes ( Bod-enmiller & Spiro , 2006 ; Filenko et al. , 2007 ; Pullan et al. , 2007 ) . 
The hmp gene is subject to complex regulation by multiple regulators including NsrR and FNR ( Spiro , 2007 ) . 
Apart from hmp , the NsrR regulon contains various genes implicated in the NO stress response , such as ytfE , hcp and the nrf operon ( Tucker et al. , 2010 ) . 
In addition to recognizing an 11 -- 1 -- 11 bp inverted repeat sequence in its target promoters , it has been suggested that NsrR can also bind to a single copy of the 11 bp motif ( Partridge et al. , 2009 ) . 
In E. coli , the only known genes activated by NorR are norVW , but binding sites for NorR have been found upstream of the hmp gene in Pseudomonas aer-uginosa , Pseudomonas putida and Vibrio cholerae and the gene encoding a respiratory NO reductase in the denitrifier Ralstonia eutropha ( Rodionov et al. , 2005 ; Stern et al. , 2012 ; Tucker et al. , 2004 ) . 
Low concentrations of NO are generated endogenously by E. coli as a by-product of respiratory nitrate and nitrite reduction ( Corker & Poole , 2003 ; Ji & Hollocher , 1988 ) . 
Nitrate and nitrite are sensed directly by the NarXL and NarQP two-component regulatory systems ( Gunsalus , 1992 ; Stewart , 1993 ) . 
Thus , during nitrate or nitrite respiration , complex changes occur in the transcriptome that are mediated by NarXL / NarQP in addition to the above-mentioned NO-responsive regulators ( Constantinidou et al. , 2006 ) . 
Global gene expression analysis has been used to study UPEC during UTI and demonstrates that UPEC is directly exposed to NO in the host environment ( Hagan et al. , 2010 ; Haugen et al. , 2007 ; Roos & Klemm , 2006 ; Snyder et al. , 2004 ) as well as to nitrate in urine ( Green et al. , 1981 ; Radomski et al. , 1978 ) . 
UPEC is more resistant to the stress imposed by acidified nitrite than K-12 strains of E. coli ( Bower & Mulvey , 2006 ) and may also be more resistant to a prolonged exposure to NO ( Svensson et al. , 2006 ) , in which case toxicity might be due to N radicals derived from NO . 
We have shown that CFT073 recovers from an exposure to NO no better than a K12 strain and that recovery is partly , although not entirely , dependent on Hmp ( Spiro , 2011 ) . 
Thus , we were motivated to undertake a deeper analysis of the determinants of NO resistance in CFT073 . 
In this paper , we show that apart from Hmp , FlRd is a major contributor to aerobic NO detoxification in UPEC . 
We also show that CFT073 possesses at least one novel anaerobic NO scavenging mechanism in addition to Hmp and FlRd . 
We use expression analysis to examine the response of CFT073 to a physiological source of NO , and map NsrR binding sites in the CFT073 genome . 
Methods
Bacterial strains and growth conditions . 
The strains used in this work are listed in Table S1 ( available in the online Supplementary Material ) . 
The rich medium was L broth ( per litre : 10 g tryptone , 5 g yeast extract , 5 g NaCl ) , supplemented with 0.5 % glucose for anaerobic cultures . 
To treat cultures with NO , 50 mM spermine NONOate ( which releases two equivalents of NO with a half-life of 39 min at 37 uC ; Cayman Chemicals ) was added to cultures during the early exponential phase ( OD650 of 0.15 -- 0.3 ) . 
[ NONOate has the chemical formula R R N - ( NO ) - N5O . ] 
Anaerobic cultures were 1 2 2 grown in filled bottles supplemented with 20 mM nitrate where indicated . 
Cultures for RNA isolation were grown anaerobically in MOPS minimal medium ( Neidhardt et al. , 1974 ) supplemented with 0.05 % Casamino acids , 0.5 % glucose ( and 20 mM nitrate as indicated ) , and 5 mg vitamin B1 ml . 
Gene deletions were made using the 21 method of Datsenko & Wanner ( 2000 ) . 
Oxygen and nitric oxide consumption assays . 
For oxygen consumption assays , 30 ml cultures grown aerobically in L broth ( with and without 1 h of 50 mM spermine NONOate treatment ) were harvested , washed and resuspended in 50 mM HEPES ( pH 7.4 ) , 100 mM NaCl , 5 mM KCl , 1 mM MgCl2 , 1 mM NaH2PO4 , 1 mM CaCl2 and 1 mM glucose ( Stevanin et al. , 2000 ) . 
All samples were resuspended at equal cell densities . 
Oxygen consumption was measured using a Clark-type oxygen electrode ( Hansatech Instruments ) . 
A 100 ml aliquot of cells was added to 500 ml buffer in a capped , waterjacketed chamber at 37 uC . 
The NO sensitivity of oxygen uptake was measured by addition of 20 mM proli NONOate ( which releases two equivalents of NO with a half-life of 1.8 s at 37 uC ; Cayman Chemicals ) when the oxygen concentration was 60 , 120 or 180 mM . 
For NO consumption assays , 30 ml of cultures grown anaerobically in L broth ( treated with 50 mM spermine NONOate for 2 h or grown with 20 mM nitrate ) were harvested and resuspended in the same HEPES buffer ( without glucose ) . 
The water-jacketed chamber housing an amperometric NO-specific electrode ( ISO NOP electrode ; WPI Instruments ) was maintained at 37 uC . 
Cell suspension ( 0.5 ml ) and buffer ( 1.5 ml ) were added to the chamber , and oxygen was removed with 5 ml of 1 M glucose , 5 ml of 30 mg glucose oxidase ml and 5 ml of 7 mg catalase ml . 
When oxygen was 21 21 undetectable , 20 mM proli NONOate was added and the rate of NO consumption was measured . 
The same procedure was used to measure NO consumption by cell fractions . 
Cell fractionation . 
Cells were fractionated using a modification of a previously described procedure ( Alefounder & Ferguson , 1980 ) . 
Cultures were grown anaerobically in 300 ml L broth supplemented with 0.5 % glucose , and in some cases were treated with 50 mM spermine NONOate . 
Cultures were harvested by centrifugation and washed twice in 10 mM potassium phosphate buffer ( pH 7.6 ) . 
Cell pellets were resuspended in 5 ml spheroplasting buffer ( 0.5 M sucrose , 3 mM sodium EDTA and 0.1 M Tris/HCl , pH 8.0 ) , 0.2 mg lysozyme ml was added and the suspension was 21 incubated at 30 uC for 30 min . 
The suspension was centrifuged at 12 200 g for 15 min at 4 uC , and the supernatant ( periplasmic fraction ) was kept on ice . 
The pellet was resuspended in 1 ml of 0.1 M Tris/HCl ( pH 8.0 ) and slowly added drop-wise into 4.5 ml water with constant stirring at 4 uC . 
After the mixture became homogeneous , it was centrifuged at 47 800 g for 1 h at 4 uC . 
The supernatant ( cytoplasmic fraction ) was stored on ice and the pellet ( membrane fraction ) was resuspended in 0.1 M Tris/HCl ( pH 8.0 ) and kept on ice . 
Malate dehydrogenase was used as a marker to check the integrity of cell fractions . 
Malate dehydrogenase activity ( Sutherland & McAlister-Henn , 1985 ) was detected only in cytoplasmic fractions . 
RNA sequencing . 
Cultures were grown in triplicate as described above , and total RNA was isolated using the Qiagen RNeasy Protect Bacteria Mini kit . 
For rRNA depletion , samples were treated using the MICROBExpress Bacterial mRNA Enrichment kit ( Life Technologies ) according to the manufacturer 's instruct-ions . 
Samples were cleaned with the Zymo RNA Clean and Concentrator kit ( Zymo Research ) and then subjected to a second cycle of rRNA depletion . 
RNA was recovered by ethanol precipitation . 
Library preparation and whole transcriptome shotgun sequencing ( RNA-seq ) was performed at the University of Texas Southwestern Medical Center Genomics and Microarray Core Facility . 
The program Bowtie ( Langmead et al. , 2009 ) was used to align RNA-seq reads to the genome of E. coli CFT073 ( GenBank accession no . 
AE014075 .1 ) with default par-ameters . 
To estimate transcript abundances , transcripts per million ( TPM ) values were calculated using RNA-Seq by expectation-maximization ( RSEM ; Li & Dewey , 2011 ) . 
Gene annotations were obtained from the European Nucleotide Archive ( accession no . 
AE014075 .1 ) . 
Differential expression between conditions with and without NO treatment was analysed using EBSeq ( an empirical Bayes hierarchical model for inference in RNA-seq experiments ; Leng et al. , 2013 ) . 
A change greater than twofold and a false discovery rate cut-off of 0.05 were used to determine significant differential expression . 
To identify functional categories of differentially expressed genes and to identify enriched pathways , we used the DAVID ( database for annotation , visualization and integrated discovery ) gene functional classification tool with default statistical parameters and Benjamini correction ( Huang et al. , 2007 ) with an adjusted P-value cut-off of 0.05 . 
RNA-seq data have been deposited in the GEO database , accession number GSE69830 ( Data Citation 1 ) . 
Reverse transcriptase PCR ( RT-PCR ) . 
Wild-type CFT073 and 3X mutant cells were grown in triplicate as described for RNA-seq . 
Total RNA was isolated using the Qiagen RNeasy Protect Bacteria Mini kit , and 2 mg of total RNA was used to make cDNA using Ambion 's RETROscript Reverse Transcription kit . 
Then , 1 ml of cDNA was used as template to amplify genes that were chosen for validation . 
PCR was performed using Thermo Scientific DreamTaq PCR Master Mix and 10 ml of PCR product was run on a 2 % agarose gel . 
Chromatin immunoprecipitation followed by highthroughput sequencing . 
Chromatin immunoprecipitation ( ChIP ) was performed as described previously on cultures grown aerobically in L broth to mid-exponential phase ( Efromovich et al. , 2008 ) . 
Immunoprecipitated and purified DNAs ( 10 ng ) from three cultures of CFT073 were collected for sequencing , along with 10 ng of the input DNA as a control . 
Samples were sheared by sonication to within a size range of 200 -- 600 bp . 
DNA fragments were treated using an Epicentre End-It DNA End Repair kit and 39 A overhangs were added with DNA polymerase I ( Klenow fragment ) . 
Adapters from the IlluminaTruSeq DNA sample preparation kit were ligated using LigaFast ( Promega ) and DNAs were amplified by PCR using primers provided in the IlluminaTruSeq DNA sample preparation kit and Phusion DNA polymerase ( NEB ) . 
Products of the ligation reaction and PCR amplification in the range 300 -- 400 bp were purified by 2 % agarose gel electrophoresis . 
DNA concentrations were measured using Qubit dsDNA HS Assay kits ( Invitrogen ) . 
DNA sequencing was done on the Miseq ( Illumina ) platform following the manufacturer 's instructions . 
For one replicate , a single-end reads , 60 bp run was performed . 
For the other two replicates , a paired-end reads , 100 bp run was performed . 
Sequence reads were aligned with the published E. coli CFT073 genome ( AE014075 .1 ) using the software package Bowtie with the parameters ` bowtie - k 1 - X 500 - m 1 ' ( Langmead et al. , 2009 ) . 
Peaks were identified using the peak finding algorithm of MACS2 ( Zhang et al. , 2008 ) , with default parameters . 
For motif analysis , multiple Em for motif elicitation ( MEME ) was used to identify over-represented sequences ( Bailey & Elkan , 1994 ) . 
PatSer was used to search the genome for the presence of the NsrR position-specific weight matrix ( PSWM ) ( Hertz & Stormo , 1999 ) . 
A precision -- recall curve was constructed to determine the optimal threshold for predicting high-quality NsrR binding sites . 
Precision was defined as the ratio of true positives ( locations with an NsrR ChIP-seq peak and a predicted NsrR binding site ) to true positives plus false positives ( locations with a predicted NsrR binding site but no NsrR ChIP-seq peak ) . 
Recall was defined as the ratio of true positives divided by true positives plus false negatives ( locations with an NsrR ChIP-seq peak but no NsrR predicted binding site ; Myers et al. , 2013 ) . 
ChIP-seq data have been deposited in the GEO database , accession number GSE69829 ( Data Citation 2 ) . 
Results
FlRd contributes to aerobic NO detoxification
The norV gene is absent from current annotations of the E. coli CFT073 genome ( Welch et al. , 2002 ) . 
However , we have resequenced the norV region and found a sequen-cing error in the published sequence that leads to a frame-shift mutation . 
The corrected norV sequence encodes a protein product that is 99 % identical to FlRd ( NorV ) of E. coli K-12 ( 476/479 residues identical ) . 
Thus , we performed experiments to determine whether E. coli CFT073 expresses an active FlRd , and to assess its contribution to NO detoxification . 
The E. coli K-12 FlRd functions as an ` anaerobic ' NO reductase , although it is capable of reducing NO in vivo in microaerobic cultures growing under an atmosphere containing * 5 mM oxygen ( Gardner et al. , 2002 ) . 
FlRd can also function as an oxygen reductase , albeit with a rather low affinity for oxygen ( Gomes et al. , 2002 ) . 
Enzymes from the flavo-diiron family in other organisms are inactivated by oxygen in vitro ( Silaghi-Dumitrescu et al. , 2003 , 2005 ) , so a view has emerged that FlRd functions in NO detoxification only in cultures growing anaerobically . 
However , the norVW genes can be induced by sources of NO in aerobic cultures , so a physiological role for FlRd under aerobic conditions can not be excluded ( Hutchings et al. , 2002 ; Mukhopadhyay et al. , 2004 ) . 
Respiration in E. coli is sensitive to NO ( Yu et al. , 1997 ) and the activity of an NO scavenging enzyme can be studied by observing its ability to protect aerobic respir-ation from NO inhibition ( Stevanin et al. , 2000 ) . 
Washed cell suspensions respiring oxygen were exposed to NO in a Clark-type oxygen electrode , and the inhibitory effect of NO was observed as a transient decrease in the rate of oxygen consumption . 
In these experiments , oxygen uptake returns to normal after an interval that depends upon the ability of the strain to scavenge NO ( Fig. 1 ) . 
Wild-type cells that were exposed to NO during growth exhibited NO-resistant oxygen uptake , while the respir-ation of cells that were not pre-exposed to NO was sensitive to NO ( duration of inhibition 1.9 +0.04 min , Fig. 1e ) . 
In an hmp mutant , respiration became more resistant to NO if the culture was pre-induced with NO ( 4.3 +0.7 versus 1.4 +0.4 min inhibition for uninduced and induced cells , respectively ; Fig. 1a ) . 
Respiration of the norVW mutant was completely NO-resistant if cells were induced with NO , but NO-sensitive ( 1.7 +0.6 min inhibition ) if cells were not induced ( Fig. 1b ) . 
Thus , both hmp and norVW mutants show evidence of an NO-induci-ble scavenging activity , which is , presumably , FlRd and Hmp , respectively . 
Accordingly , in an hmp norVW double mutant ( Fig. 1c ) , oxygen consumption was equally sensitive to NO whether or not the cultures were exposed to NO ( 6 +0.1 and 6.3 +0.3 min inhibition for uninduced and induced cells , respectively ) . 
Under the growth and assay conditions used for these experiments , NO-inducible NO scavenging by CFT073 can be entirely accounted for by the combined activities of Hmp and FlRd . 
By measuring NO inhibition of respiration of the hmp mutant at different oxygen concentrations , we concluded that FlRd can scavenge NO in vivo in the presence of as much as 180 mM oxygen ( Fig. 1g -- i ) . 
Complementing the hmp norV double mutant with norV on a plasmid expressed from an inducible promoter showed that FlRdmediated protection of respiration was restored ( data not shown ) . 
Interestingly , complementation failed if norV was expressed on a plasmid from its own promoter , as we have observed before ( Hutchings et al. , 2002 ) , and successful complementation required expression from a heter-ologous promoter . 
A novel inducible anaerobic NO scavenging activity
We generated a triple mutant of E. coli CFT073 lacking the three known NO detoxification systems , Hmp , FlRd and NrfA ( this strain is designated UTD692 , and will be referred to as ` 3X ' ) . 
Washed cells of the 3X strain grown anaerobically with nitrate ( nitrate provides a source of endogenously generated NO under anaerobic conditions ; Ji & Hollocher , 1988 ) , or induced anaerobically with NO showed a higher rate of NO consumption compared with an uninduced strain ( Fig. 2 ) . 
Thus , in cells grown and assayed anaerobically , there is evidence of a novel NO-inducible activity . 
A wild-type strain grown and assayed under similar conditions showed only moderately increased rates of NO consumption compared with the 3X mutant ( Fig. 2 ) . 
The dominant activity of Hmp requires molecular oxygen , so Hmp is not expected to contribute to NO consumption under these assay conditions . 
The measured activity is therefore a combination of FlRd and the novel activity , the latter seeming to be a major contri-butor in cells grown and assayed anaerobically . 
Cell fractionation experiments revealed this activity to be associated with the cell membrane ( Fig. 2c ) . 
Interestingly , NO uptake by membrane fractions of the triple mutant required neither an exogenous reductant nor an oxidizing agent . 
Possible candidates for the source of this activity include NirB ( Vine & Cole , 2011b ) and the hybrid cluster protein , Hcp ( Cole , 2012 ; Vine & Cole , 2011a ) , although in both cases NO reduction would be dependent on NADH . 
Introduction of nirB and hcp-hcr mutations into the 3X mutant , either individually or in combination , had no effect on the NO scavenging activity , so NirB and Hcp can be excluded as the source of the activity we observe . 
Introduction of a mobA mutation ( which eliminates nitrate reductase activity ) into the 3X mutant eliminated the response to nitrate , confirming that nitrate reduction and therefore endogenous NO generation is probably required for nitrate-mediated induction of the NO scavenging activity . 
In a 3X nsrR mutant , NO scavenging could be induced by nitrate but not by NO in anaerobic cultures ( Fig. 3 ) . 
The same outcome was seen in a CFT073 nsrR mutant ( data not shown ) , which further demonstrates that under these conditions , Hmp , which is de-repressed in an nsrR mutant ( Filenko et al. , 2007 ) , is not functional . 
In the 3X strain with an fnr mutation , the NO scavenging activity became constitutive ( Fig. 3 ) , implying that FNR acts negatively on the expression of the gene ( s ) encoding the activity . 
In a 3X narL narP mutant , the activity could be induced by nitrate but not NO in anaerobic cultures ( data not shown ) . 
Individual 3X narL and 3X narP mutants behaved like the 3X parent strain . 
In further attempts to identify the source of the NO scavenging activity , we tested candidate genes by introdu-cing the corresponding deletion mutations into the 3X mutant . 
Candidates were identified on the basis of one or more of the following criteria : ( 1 ) a primary structure suggesting a possible role in NO metabolism ; ( 2 ) a previously described expression pattern matching our observations described above ; and ( 3 ) an expression pattern in our RNA-seq data ( see below ) similar to the behaviour of the novel activity . 
Some genes were also tested that might be indirectly involved in expression of this activity ( e.g. mobA , required for the activity of enzymes containing the molybdopterin guanine dinucleotide cofactor ) . 
In this way , we showed that 26 genes or operons are not required for expression of the novel NO scavenging activity ( aegA , betA , cydAB , cyoAB , fdhE , hcp-hcr , mobA , ndh , nirB , nrdA , poxB , putA , rnr , sdhA , tehAB , yceJI , ydcX , ydhXV , yeaR-yoaG , yebE , yedY , yeiH , ygbA , yhaM , yibIH , ytfE ) . 
Further efforts to identify the source of the novel NO scavenging activity described in this paper have so far proved unsuccessful . 
Vine & Cole ( 2011b ) have reported that a norVW nrf nirB hmp quadruple mutant of E. coli K-12 consumes NO at rates comparable to those of the wild-type strain . 
Our data suggest that Hmp and FlRd are the major NO scavenging activities of E. coli CFT073 under aerobic conditions , and the novel activity we describe makes a significant contribution to NO consumption only under anaerobic growth conditions . 
Transcriptional profiling
We used transcriptomics ( RNA-seq ) to explore the response of E. coli CFT073 to a source of NO . 
In part , this experiment was motivated by a desire to identify the gene ( s ) encoding the novel NO scavenging activity . 
Therefore , we used a strain and growth conditions identical to those used for the initial detection of this activity , as described above . 
That is , the transcriptome of the triple mutant was analysed in anaerobic cultures grown in the presence and absence of nitrate . 
By differential gene expression analysis of the RNA-seq data , we identified 525 upregulated and 649 downregulated genes in the nitrate-treated cultures ( genes showing greater than two-fold change with a 0.05 false discovery rate cut-off ) . 
Among the most highly upregulated genes ( Table 1 ) were some that are known to be regulated by NsrR and to respond to NO , including yeaR , ytfE , hcp-hcr and ygbA ( Filenko et al. , 2007 ; Pullan et al. , 2007 ) . 
Previous transcriptomics experiments with E. coli K-12 have also shown that members of the NsrR regulon are de-repressed in cultures grown anaerobically with nitrate ( Constanti-nidou et al. , 2006 ) . 
RNA was extracted from anaerobically grown cultures , so it was surprising that the most highly upregulated genes included some involved in oxidative phosphorylation ( cyoABCDE , sdhABCD and nuoEF ) and the tricarboxylic acid ( TCA ) cycle ( acnB , icdA , sucD , lpdA , sdhABCD , fumA and mdh ) . 
Several genes encoding ABC transporters ( dpp operon , proVW , gltK , kpsMT ) were also upregulated . 
Other genes showing increased expression in nitrate-grown cells were those involved in nitrogen metabolism ( nitrate respiration ) , DNA repair and the SOS response , bacterial motility and chemotaxis . 
The most highly downregulated genes included some encoding enzymes involved in anaerobic metabolism , including hydrogenase ( hya operon ) and formate dehydrogenase ( fdhF ) . 
At least some of these regulatory effects may reflect inactivation of FNR by NO ( Cruz-Ramos et al. , 2002 ; Justino et al. , 2005 ; Pullan et al. , 2007 ) , or regulation by the nitratesensing two-component systems NarXL and NarQP ( Constantinidou et al. , 2006 ) . 
Genes involved in glycolysis , gluconeogenesis , pyruvate metabolism [ eno , pykF , pgi , fba , gapA , glk , pgk , ldhA , maeA ( sfcA ) ] , the pentose phosphate pathway ( talA , tktb , pgl ) and the metabolism of sugars ( fructose , sucrose and mannose ) showed decreased expression . 
Iron transport genes ( feoAB ) and some stress response genes ( clpB , dnaJ , dnaK , dps ) also showed reduced expression . 
Fig . 
S1 ( a ) provides an overview of the differentially expressed genes based on their occurrence in pathways , and Fig . 
S1 ( b ) provides an overview based on functional categories . 
It is known that the citric acid cycle is repressed in E. coli grown anaerobically in nitrate with glucose as the carbon source ( Prohl et al. , 1998 ) . 
Under anaerobic conditions with glucose and nitrate , ArcA represses operons encoding a-ketoglutarate dehydrogenase and succinate dehydrogen-ase , thus preventing complete oxidation of glucose to carbon dioxide ( Prohl et al. , 1998 ) . 
However , our RNA-seq data for the 3X mutant show increases in the transcript levels of the genes encoding members of both these enzyme complexes as well as various other genes involved in the intact aerobic TCA cycle . 
This suggests that nitrate ( and possibly NO generated by nitrate respiration ) impacts the TCA cycle by allowing complete oxidation of acetyl-CoA and this may have a role to play in energy generation in the presence of NO . 
Our data are consistent with a recent finding that the TCA cycle is necessary for UPEC fitness in vivo ( Alteri et al. , 2009 ) , conditions in which UPEC is known to be exposed to NO ( Lundberg et al. , 1996 ; Mysorekar et al. , 2002 ; Poljakovic et al. , 2001 ) as well as nitrate ( Green et al. , 1981 ; Radomski et al. , 1978 ) . 
Numerous studies indicate that iNOS expression levels increase and NO is released in the urinary tract during bacterial infection ( Kaboré et al. , 1999 ; Lundberg et al. , 1996 ; Smith et al. , 1996 ; Wheeler et al. , 1997 ) . 
Also , dietary nitrate is excreted into urine where UPEC can potentially use it for respiration , generating NO as a byproduct ( Corker & Poole , 2003 ; Ji & Hollocher , 1988 ; Lidder & Webb , 2013 ) . 
Although the role of NO is presumably anti-microbial , the pathogen may use host NO ( as well as NO derived from nitrate respiration ) as a signal to induce the expression of virulence genes . 
We make this suggestion based on the observation that virulence-related genes are upregulated in UPEC cultures exposed to nitrate and/or to NO produced by nitrate respiration ( Fig . 
S1c ) . 
The secreted autotransporter toxin ( Sat ) is a serine protease that causes cytoplasmic vacuolation and histological damage in urinary-tract-derived epithelial cells ( Guyer et al. , 2002 ) . 
The flagellar gene fliC contributes to fitness of UPEC and enhances its pathogenesis ( Lane et al. , 2005 ) , and other genes involved in flagellar assembly ( fliD , fliS , fliR , motAB and flgM ) were upregulated in nitrate-grown cells . 
The fimbrial site-specific recombinases fimB and fimE can be associated with virulence , as they regulate type 1 fimbrial gene expression . 
Type 1 fimbriae are known to enhance E. coli virulence in the urinary tract ( Bryan et al. , 2006 ; Connell et al. , 1996 ) . 
Haemolysin ( encoded by hly genes ) is a cytotoxin for renal proximal tubular epithelial cells and a haemolysin-deficient CFT073 mutant demonstrates significantly reduced cytotoxicity ( Mobley et al. , 1990 ) . 
The sialic acid capsule proteins ( encoded by kps genes ) , also known as K antigens , encapsulate bacteria so that they can evade unspecific host responses ( Jann & Jann , 1992 ; Rowe et al. , 2000 ) . 
BipA is identified as a virulence regulator in enteropathogenic E. coli that regulates several processes such as flagellamediated motility , resistance to host defence peptides and group 2 capsule gene clusters ( Farris et al. , 1998 ; Rowe et al. , 2000 ) . 
The dipeptide binding protein DppA delivers dipeptides to its cognate ABC-type transporter proteins . 
As sugar sources such as glucose , maltose and lactose are rare in the urinary tract , it is suggested that dipeptides and certain amino acids such as D-serine are important sources of nutrients for UPEC ( Haugen et al. , 2007 ) . 
Upregulation of dppA is observed during UTI ( Subashchandrabose et al. , 2014 ) . 
A periplasmic osmoprotectant , ProV , is upregulated during UTI ( Subashchandrabose et al. , 2014 ) and we observe upregulation during our growth conditions as well . 
Very often , pathogens experience increased osmotic pressure at the site of infection and hence acquire osmoprotectants from the environment ( Lewis et al. , 2012 ) . 
KpsMT , DppA and ProV are ABC-type transporters that , under appropriate conditions , become important for viability , virulence and pathogenicity ( Davidson et al. , 2008 ) . 
Upregulation of all these viru-lence-related genes suggests that , while NO provides a defence mechanism for the host and nitrate provides an electron acceptor for pathogens , either or both could also provide a useful signalling mechanism for pathogenic bacteria to induce virulence . 
Several genes were selected from the RNA-seq dataset for validation using RT-PCR . 
Expression of these genes was measured in both the wild-type strain and the 3X mutant grown anaerobically with nitrate ( the same conditions used for RNA-seq ) . 
Of seven genes upregulated by nitrate according to RNA-seq data , six ( sdhA , kpsM , bipA , cyoA , dppA and ytfE ) were also upregulated in RT-PCR data , in both strains ( Fig. 4 ) . 
The seventh ( hlyA ) was upregulated only in the 3X mutant . 
Three genes that were downregulated by nitrate in RNA-seq data ( asr , hycA and fdhF ) were also downregulated according to RT-PCR ( Fig. 4 ) . 
On the basis of this selection of genes , we conclude that most changes observed in the transcriptome of the 3X mutant in response to nitrate are likely also to occur in the wild-type parent . 
The NsrR regulon of E. coli CFT073
As the E. coli CFT073 genome is * 0.6 Mb larger than that of E. coli K-12 , it is of interest to determine the extent to which regulatory networks of the two organisms differ . 
Thus , we used chromatin immunoprecipitation and DNA sequencing ( ChIP-seq ) to identify NsrR binding sites in the E. coli CFT073 genome . 
Cultures expressing 3Xflagtagged NsrR were grown aerobically . 
After ChIP , libraries constructed from precipitated DNAs were sequenced using the Illumina Miseq platform . 
The peak finding algor-ithm MACS2 was used to identify putative NsrR binding sites , with a false discovery rate of 0.01 . 
Ninety-four significant peaks [ 2log10 ( P-value ) w10 with fold enrichment greater than 2 ] were identified in at least two of the three biological replicates . 
In total , 52 % of the binding sites ( 49 of 94 ) in E. coli CFT073 were located in putative promoter regions ( within 350 bp of the start codon ) and the remaining 48 % were found either within coding regions or between the coding regions of convergent genes . 
These potentially functional 49 NsrR binding sites are shown in Table 2 . 
The presence of promoter-associated NsrR binding sites identifies target genes that potentially belong to the NsrR regulon . 
Of these promoters bound by NsrR in vivo ( Table 2 ) , 19 ( grxD , hypA , ytfE , ygiG/folB , hmp , ybjW/hcp , feoA , ybeM , yihF , yccM , yibD/waaH , yieI , yohK , ygiF , trxB , yggS/yggT , dgoK , rfe , yfhB/pgpC ) were identified in a previous ChIP-chip analysis of NsrR binding sites in E. coli K-12 ( Partridge et al. , 2009 ) . 
Twenty of the remaining sites are associated with genes ( hycB , phnC , arnA , livJ , wzxE , tufB , yfbT , glpA , yjbN , deoB , murB , recJ , dkgB , c3139 , rhaB , arcB , c3976/nanK , fimC , c3934/h flB and malF ) that have homologues in E. coli K-12 , and 10 ( c0118 , c0233 , c2471 , c5065 , c0813 , c2514 , c0650 , c4580 , c4214 and c5205 ) are specific to E. coli CFT073 . 
In E. coli K-12 , the nrfA promoter is bound by NsrR ( Partridge et al. , 2009 ) and is repressed by NsrR according to microarray and reporter fusion data ( Filenko et al. , 2007 ) . 
In our ChIP-seq data , NsrR binding was also detected upstream of the transcription unit that includes nrfA . 
In strain CFT073 , an additional gene upstream of nrfA ( c5065 ) is predicted to be co-expressed with nrfABCD . 
The c5065 gene encodes a small protein of 65 aa residues . 
We have confirmed the sequence of this reading frame in the CFT073 genome . 
The genome location and expression pattern of the c5065 gene suggest that its product may have a role in the response to NO stress in CFT073 . 
Nineteen of the 49 potential NsrR targets show differential expression with a fold change greater than 1.5 for the CFT073 3X mutant strain in the presence of a physiological source of NO ( Table 3 ) . 
The hycB , c0118 , feoA , ybeM , c5065 , c0813 , c2514 , glpA , deoB , hypAB , yohJK and yfhB / pgpC genes were downregulated , among which hycB , c0118 , c5065 , c0813 , c2514 , glpA and deoB are newly detected potential NsrR targets in CFT073 . 
The grxD , folB , ybjW/hcp , ytfE , yjbN , trxB and c5205 genes were upregulated and c5205 and yjbN are potential NsrR targets newly detected in CFT073 . 
The glpA gene , which encodes anaerobic glycerol-3-phosphate dehydrogenase subunit A , is downregulated in UPEC strain UTI89 exposed to acidi-fied nitrite ( Bower et al. , 2009 ) . 
The livJ gene , which encodes a periplasmic Leu/Ile/Val-binding protein , is upre-gulated during in vitro growth in human urine ( Snyder et al. , 2004 ) . 
The fimC gene encoding type 1 fimbriae is upregulated in vivo during UTI ( Snyder et al. , 2004 ) . 
The rfe gene was upregulated in vivo compared with growth in human urine in vitro ( Hagan et al. , 2010 ) . 
The E. coli 
K-12 homologue of yfbT is upregulated in the presence of a source of NO ( Hyduke et al. , 2007 ) . 
In E. coli K-12 , the expression of arcB and malF was increased and decreased , respectively , after treatment with NO ( Hyduke et al. , 2007 ) , and phnC was upregulated by treatment with 1 mM S-nitrosoglutathione or acidified nitrite ( Mukho-padhyay et al. , 2004 ) . 
Computational analysis of NsrR binding sites in CFT073
The 49 peaks located in putative regulatory regions were used to construct a PSWM for NsrR binding sites in the CFT073 genome . 
Two hundred base pairs centred on the nucleotide with the largest tag density within each of the peaks was analysed ( Myers et al. , 2013 ) . 
The sequence of NsrR in CFT073 is identical to that in E. coli K-12 , and evidence from previous studies suggests that NsrR binding sites have two copies of an 11 bp motif arranged as an inverted repeat with 1 bp spacing ( Partridge et al. , 2009 ) . 
So , we first used MEME to identify over-represented palindromic sequences with the parameters ` - mod zoops-nmotifs 1 - minw 23 - maxw 23 - revcomp -- pal ' to see if the same motif could be retrieved . 
Motifs matching the search criteria could be found in 20 of the 49 peak regions . 
As expected , the predicted NsrR binding site in CFT073 is similar to that for E. coli K-12 ( Fig. 5 ) . 
A precision -- recall curve ( see Methods ) was constructed using the NsrR PSWM with two inverted repeats and searching throughout the genome of CFT073 to determine the optimal threshold for predicting high-quality NsrR binding sites . 
Using an ln ( P-value ) of 214.28 as the cut-off , where we had both relatively high precision and recall , there were 27 predicted NsrR binding sites with the 11 -- 1 -- 11 inverted repeat ( palindrome ) motif in the CFT073 genome ( Table 4 ) . 
Four of these predicted targets were not detected by the ChIP-seq data ( tehA , yeaR , yhiX and ygbA ) . 
Among them , yeaR and ygbA are known to be regulated by NsrR ( Bodenmiller & Spiro , 2006 ; Lin et al. , 2007 ) , and the ygbA promoter was reported to be bound by NsrR in E. coli K-12 according to previous ChIP-chip data ( Partridge et al. , 2009 ) . 
Likewise , tehA was implicated as an NsrR target in E. coli K-12 by the same ChIP-chip data and by repressor titration ( Bodenmiller & Spiro , 2006 ) , and it was shown to be upregulated in the urinary tract in an asymptomatic bacteriuria strain of E. coli ( Roos & Klemm , 2006 ) . 
By contrast , reporter fusion data suggest that tehA is not regulated by NsrR ( Bodenmiller & Spiro , 2006 ) ; conflicting reports may reflect differences in growth conditions or genetic background . 
Minimally , we can conclude that yeaR and ygbA are probably false negatives in our ChIP-seq data . 
The gadX ( yhiX ) gene was reported to be induced by NO through an indirect NsrR-dependent mechanism in E. coli O157 : H7 ( Branchu et al. , 2014 ) , but the presence of an NsrR binding site upstream of gadX may indicate a direct regulatory mechanism . 
There is evidence that a single 11 bp motif can function as an NsrR-binding site in E. coli K-12 ( Partridge et al. , 2009 ) . 
So we combined the two halves of the 11 -- 1 -- 11 palindro-mic motif , and reconstructed a PSWM of 11 bp . 
The new 11 bp PSWM was used to scan the 49 200 bp sequences flanking all the peak regions using the P-value 26 cut-off of 10 . 
In this analysis , 38 of 49 peaks had at least one single motif , and the updated sequence logo for the 11 bp motif is shown in Fig. 5 ( b ) . 
Discussion
Flavohaemoglobin ( Hmp ) , flavorubredoxin ( FlRd ) and respiratory nitrite reductase ( Nrf ) have been extensively studied to understand their role in combating NO in E. coli and Salmonella enterica ( Clarke et al. , 2008 ; Gardner , 2005 ; Gardner & Gardner , 2002 ; Gardner et al. , 2002 ; Gomes et al. , 2002 ; Mills et al. , 2008 ; Poock et al. , 2002 ; van Wonderen et al. , 2008 ) . 
In S. enterica , it has been suggested that additional NO detoxification mechanisms are expressed in the absence of Hmp , and that the availability of different NO detoxification mechanisms under different environmental conditions is an important contri-butor to virulence ( Mills et al. , 2008 ) . 
Work in E. coli K-12 lacking the known NO detoxification mechanisms has also suggested the existence of an additional major pathway of NO metabolism ( Vine & Cole , 2011b ) , which may be the same activity that we have observed in CFT073 . 
In UPEC strains , the only system known to protect the pathogen from NO is Hmp ( Svensson et al. , 2006 , 2010 ) . 
Competitive infection of UTI mouse models with wild-type and hmp-deleted UPEC strains showed a decreased ability of the mutant to infect ( Svensson et al. , 2010 ) . 
The roles of NrfA and FlRd have not been studied in the pathogen , but in this paper we show that FlRd is a major contributor to NO metabolism in UPEC , and that there is an additional NO-inducible activity yet to be identified . 
Our data suggest that the respiratory nitrite reductase Nrf makes only a minor contribution to NO metabolism . 
Previous transcriptomic studies have suggested that UPEC experiences iron and oxygen limitation in the urinary tract ( Hagan et al. , 2010 ; Snyder et al. , 2004 ) . 
It has also been proposed that the ability of UPEC to adapt to low oxygen may be critical for successful bladder colonization during UTI ( Subashchandrabose et al. , 2014 ) . 
As UPEC is potentially exposed to NO ( host derived and/or endogenously generated from nitrate respiration ) , nitrate and low oxygen in vivo , our choice of growth conditions for transcriptomics is relevant to the host environment . 
Our data ( Fig. 4 ) suggest that responses to nitrate in the 3X mutant also occur in the wild-type parent . 
Nevertheless , the use of the 3X mutant may exacerbate responses to nitrate and NO compared with the parent strain . 
Our results have highlighted a group of interesting genes ( including some that are virulence-associated ) that we believe are good candidates for further investigation , including in vivo approaches . 
A disadvantage of the use of nitrate as a source of NO is that we can not necessarily disentangle the effects of endogenously generated NO from direct effects due to nitrate . 
Thus , the changes observed in the transcriptome are likely to be mediated by nitrate sensing systems ( NarXL/NarQP ) in addition to those responsive to NO ( principally NorR and NsrR ) . 
At least for those genes shared with E. coli K-12 , we can use prior knowledge to infer some of the regulatory consequences of nitrate exposure . 
For example , upregulation of members of the NsrR regulon is strong evidence for the generation of physiologically significant concentrations 
In this work , upregulation of genes involved in respiration and electron transport , along with genes associated with the TCA cycle , suggests that the pathogen uses these mechanisms to maximize energy generation during NO stress . 
Decreases in the levels of transcripts involved in glucose metabolism ( glycolysis and gluconeogenesis ) and upregulation of genes involved in dipeptide transport suggests that during NO stress , glucose may not be the energy source used by the pathogen . 
Differential expression of virulence-associated genes and genes on pathogenicity islands as a consequence of nitrate exposure suggest a role for nitrate and/or NO in pathogenesis . 
These experiments were performed with a mutant strain compromised in its ability to metabolize NO . 
By ChIP-seq we identified NsrR binding sites in the CFT073 genome . 
Of 49 NsrR binding sites in promoter regions , 19 are associated with genes that were nitrate-responsive in the RNA-seq data . 
This discrepancy may reflect differences in the strains used , or the growth conditions used for the two experiments ( aerobic growth for ChIP-seq , anaerobic growth for RNA-seq ) , although there is no published evidence to suggest that NsrR binding to DNA is sensitive to oxygen in vivo . 
Another possible explanation is that at some binding sites NsrR exerts weak or no regulation , as we have observed previously for E. coli K-12 . 
As was the case for E. coli K-12 ( Partridge et al. , 2009 ) around half of mapped sites were within coding regions or between convergently transcribed genes . 
Similar results have been obtained with other regulatory proteins , for example Fur ( Seo et al. , 2014 ) , and this is not surprising behaviour for a DNA-binding protein with a relaxed sequence specificity . 
We assume that most sites in this category have no biological function , although some may regulate the activity of promoters driving expression of small or anti-sense RNAs . 
We found strong NsrR binding signals upstream of some hypothetical proteins of unknown function , some of them specific to CFT073 ( meaning not present in E. coli K-12 ) . 
Examples are c0118 and c0233 , which are homol-ogues of each other . 
Both c0118 and c0233 have two copies of a conserved helix -- turn -- helix domain that is often found in transposases and is likely to bind DNA . 
Both proteins are implicated as transposases or derivatives in the clusters of orthologous groups of proteins ( COGs ) database . 
Transposase genes are frequently associated with pathogenicity islands , and NsrR has been implicated in regulating pathogenicity island genes in E. coli O157 : H7 ( Branchu et al. , 2014 ) . 
Therefore , it would be interesting to study the function of c0118 and c0233 to see if they are related to the pathogenicity of CFT073 , and to determine if NsrR is involved in the regulation of pathogenicity island genes . 
Of the genes implicated as possible NsrR targets by ChIP-seq that were also differentially regulated in response to NO , two-thirds were downregulated in the presence of a source of NO . 
This behaviour is consistent with positive regulation by NsrR , as has been reported previously ( Branchu et al. , 2014 ) , or with indirect effects of NsrR . 
Some genes associated with NsrR binding sites were not differentially regulated in the RNA-seq experiment , which may indicate that these genes are subject to multiple regulatory mechanisms , such that regulation by NsrR is revealed only under specific growth conditions . 
An additional possibility is that there is a category of promoter that is bound by , but not regulated by , NsrR . 
In conclusion , the response of UPEC strain CFT073 to NO overlaps substantially with that of E. coli K-12 . 
In both cases , Hmp and FlRd provide the principal NO detoxification mechanisms , although there is evidence of additional activities yet to be identified . 
Anaerobic growth in the pre-sence of nitrate ( and therefore low concentrations of endogenously generated NO ) causes a major reprogramming of the transcriptome . 
Major players in regulating differential gene expression under these conditions are likely to be NarXL , NarQP , FNR and NsrR . 
The NsrR regulon of CFT073 overlaps significantly with that of E. coli K-12 , but our data also suggest that NsrR ( and therefore NO ) may regulate several CFT073 genes that do not have homologues in E. coli K-12 . 
Acknowledgements
We are grateful to Harry Mobley and Barry Wanner for generously providing strains and plasmids . 
We thank Zhenyu Xuan for help with data analysis and Yunfei Wang for help with sequencing . 
M. Q. Z. acknowledges financial support from the National Institutes of Health ( award HG001696 ) and the Cecil H. and Ida Green Endowment .