All chemicals were purchased from Sigma–Aldrich (Munich, Germany) if not stated otherwise. Bacterial strains, oligonucleotides, plasmids, and phages are listed in Tables S1 and S2. E. coli strains were cultivated at 37°C in Luria-Bertani (LB) broth or on LB-agar supplemented with 100 μg/ml ampicillin or 100 μg/ml hygromycin, as required.

Mycobacterium smegmatis mc² 155 and transformants were grown in LB media or on Middlebrook phage (MBP) agar (Supplementary Methods) supplemented with 100 μg/ml hygromycin if necessary. Liquid cultures were incubated at 37°C in a shaking incubator at 100 rpm. MAP DSM44135 wildtype (wt), MAPΔfurA, and the complemented strain MAPΔfurA^C were grown in Difco^TM Middlebrook 7H9 broth or on solid Middlebrook 7H10 agar (Beckton Dickinson, Franklin Lakes, NJ, USA) supplemented with mycobactin J (2 mg/l, Allied Monitor, Fayette, MO, USA), 2.5% glycerol and 10% OADC (0.06% oleic acid, 5% albumin, 2% dextrose, 0.085% NaCl, 0.003% catalase; MB-complete broth) and 50 μg/ml hygromycin and/or 50 μg/ml kanamycin, as required. Agar plates were sealed in plastic bags after inoculation and incubated at 37°C for 8–10 weeks. Liquid cultures were grown in Duran® laboratory glass bottles on a magnetic stirrer at 100 rpm and 37°C until they reached an OD₆₀₀ of 1.0 and used for further experiments.

For growth kinetics, a starter culture was grown to an OD₆₀₀ of 1.0 in MB-complete broth. Bacteria were harvested by centrifugation at 3,800 ×g for 15 min at 4°C, resuspended in MB-complete broth, singularized with glass beads (Ø 3 mm) by vortexing and inoculated into fresh medium to obtain an initial OD₆₀₀ of 0.2. The OD₆₀₀ of cultures was measured twice a week and bacteria were grown until they entered the stationary phase.

The response of MAPwt and MAPΔfurA to oxidative stress was examined upon exposure of bacteria to hydrogen peroxide (H₂O₂). Cells were grown in MB-complete broth and harvested at an OD₆₀₀ of 1.0 as described above. The bacterial pellet was washed twice with PBS and singularized cells were transferred to 50 ml Middlebrook 7H9 broth supplemented with 2 mg/l mycobactin, 2.5% glycerol, and 10% ADC lacking catalase (5% albumin, 2% dextrose, 0.085% NaCl; MB-cat broth) to obtain an OD₆₀₀ of 1.0. H₂O₂ was added to a final concentration of 10 mM and cultures were exposed for 2 h at 37°C with gentle shaking (100 rpm) in Erlenmeyer flasks. Following, bacteria were harvested and immediately used for RNA-extraction.

Construction of a furA (map1669c) deletion mutant of MAPwt was performed by specialized transduction according to Park et al. (2008) with minor modifications. A detailed protocol is given in the Supplementary Methods file.

For complementation of the mutant strain the furA gene, including 88 bp of the upstream UTR, was amplified by PCR using primers oMAP-furA-K1/oMAP-furA-K2. The PCR product was digested with XbaI and HindIII and ligated into the corresponding sites of the mycobacterial integrative vector pMV306 (Stover et al., 1991), carrying a kanamycin resistance. The resulting plasmid pMAP-furA1101 was transformed into electro-competent MAPΔfurA cells and plated on MB-complete agar with 50 μg/ml hygromycin and kanamycin. Successful deletion and integration of furA was confirmed by PCR.

The intracellular survival of MAPwt, MAPΔfurA, and MAPΔfurA^C was determined in the mouse macrophage cell line J774A.1 (Ralph et al., 1975). Macrophage cells were seeded in tissue culture dishes and were maintained in Dulbecco’s modified Eagle medium (DMEM) supplemented with 10% fetal calf serum (FCS), 1% glutamine, 100 units/ml penicillin, 100 μg/ml streptomycin at 37°C and 8% CO₂. Twenty-four hours prior to infection, medium was changed to antibiotic-free DMEM. For infection experiments, MAP cells were grown to an OD₆₀₀ of 1.0, harvested, singularized and stored in MB-complete broth containing 10% glycerol at -80°C until further usage. For infection, MAP cells were thawed on ice, diluted in antibiotic-free DMEM to an OD₆₀₀ of 0.15 and incubated with the macrophages (2.0 × 10⁶ cells per cell culture dish) as described earlier (Kuehnel et al., 2001), resulting in a MOI of 10. The mean values of the initial inocula (five experiments) of all strains (MAPwt, MAPΔfurA, MAPΔfurA^C) are shown in the Supplementary data files. At indicated time points, infected macrophages were washed twice with PBS and scraped off the plates in 1 ml PBS containing 0.1% SDS. J774A.1 cells were disrupted by passage through a 24-gauge needle. The homogenates as well as the infection inocula were 10-fold serial diluted in PBS and 25 μl of the 10^-4 and 10^-5 dilutions were plated on supplemented MB-complete agar plates in duplicate. Colony forming units (CFU) were determined after incubation for up to 10 weeks at 37°C and normalized to the initial inoculum.

Induction of ROS production by macrophages upon infection with MAPwt, MAPΔfurA, and MAPΔfurA^C was determined by use of CellRox® Deep Red (Life Technologies), according to the manufacturer’s protocol. In brief, macrophages were cultured and infected as described above and incubated for 2 h at 37°C/8% CO₂ or treated with the chemical menadione as a positive control at a final concentration of 100 μM (Malorni et al., 1993). CellRox® Deep Red was added to a final concentration of 2.5 μM 30 min before harvesting and detection. Signals were measured at 660 nm by flow cytometry (Guava® easyCyte 8HT, Millipore, Billerica, MA, USA) and shown as mean fluorescence intensity (MFI) of four independent experiments.

Macrophages were seeded on coverslips and infected as described above. After 30 min and 2 h of infection, cells were fixed with 3% formaldehyde in PBS for 10 min, washed twice with PBS and treated with blocking buffer (PBS/1% BSA/10% FCS) for 20 min at room temperature. For staining of extracellular mycobacteria, coverslips were incubated with a 1:100 dilution of a polyclonal rabbit anti-MAP-HBHA serum (Supplementary Methods) in blocking buffer at room temperature in a humid chamber for 45 min. After washing with PBS, coverslips were incubated with goat anti-rabbit IgG coupled to Alexa Fluor® 488 (Life Technologies, Darmstadt, Germany) for 30 min and subsequently washed with PBS. To label intracellular mycobacteria, cells were permeabilized with Triton X-100 (0.1%) for 5 min at room temperature, washed in PBS, followed by incubation with a 1:100 dilution of polyclonal rabbit anti-MAP-HBHA serum in blocking buffer for 45 min at room temperature. After washing, coverslips were treated with goat anti-rabbit IgG coupled to Alexa Fluor® 568 (Life Technologies, Darmstadt, Germany) for 30 min. Coverslips were washed three times in PBS and, after one brief washing step with ddH₂O, samples were mounted using ProLong® Gold with DAPI (Life Technologies, Darmstadt, Germany). Mounted samples were examined using a TCS SP5 confocal laser scanning microscope equipped with a 63×/1.4–0.6 NA HCX PL APO objective (Leica, Wetzlar, Germany). Image stacks were acquired using 1 Airy unit pinhole diameter in sequential imaging mode to avoid bleed through. Image stacks were deconvolved using Huygens® Essential (Scientific Volume Imaging, Hilversum, The Netherlands), maximum intensity projections were calculated for display purposes and adjusted for brightness and contrast using ImageJ/Fiji (Schindelin et al., 2012). After this labeling procedure, extracellular mycobacteria appear yellow, whereas intracellular mycobacteria are stained red.

Mouse infection experiments were approved by the Lower Saxony Federal State Office for Consumer Protection and Food Safety, Germany (reference number 08/1504). In each group, 9 female C57BL/6 mice aged 8 weeks (Charles River, Erkrath, Germany) were challenged intraperitoneally with an infection dose of 1 × 10⁸ bacteria of MAPwt and MAPΔfurA in 200 μl Dulbecco’s Phosphate-Buffered Saline (DPBS, Life Technologies GmbH, Darmstadt, Germany). DPBS was used as negative control. Mice were sacrificed after 4 weeks, and liver and spleen were weighed. Bacteria were quantified by plating of homogenized tissue on MB-complete agar plates. Histology was performed in the Mouse Pathology Platform at HZI Braunschweig as recently described (Meißner et al., 2014; Suwandi et al., 2014).

Genomic DNA was prepared as previously described (Stratmann et al., 2004). Plasmids were prepared using NucleoBond® AX kit (Macherey Nagel GmbH, Düren, Germany) according to the manufacturer’s protocol. Total RNA from bacteria was isolated according to Rustad et al. (2009) with minor modifications. In brief, bacterial pellets were resuspended in TRIzol® reagent, mechanically disrupted in a FastPrep® instrument (ThermoSavant, Carlsbad, CA, USA) and RNA was separated by chloroform, chloroform-isoamylalcohol (49:1 v/v) extraction, and precipitation with 2-propanol. Total RNA was treated twice with 50 U DNase I (Roche, Mannheim, Germany) according to the manufacturer’s protocol. Integrity and quality of RNA was determined by gel electrophoresis and spectrophotometric analysis using an Epoch instrument (Biotek, Bad Friedrichshall, Germany) at 260 nm.

For cDNA synthesis, 4 μg of DNA-depleted RNA were diluted with RNase free water up to a volume of 20 μl and incubated for 10 min at 70°C with 0.4 μg random primers (Promega, Madison, WI, USA). After 5 min cooling on ice, samples were split in two and mixed with 5x RT buffer, 10 mM dNTP’s, and either 200 U MMLV-superscript transcriptase (Promega) or buffer without transcriptase as a negative control. The mix was incubated for 1 h at 42°C, following incubation for 5 min at 85°C. Samples were diluted with 90 μl ddH₂O and stored at -20°C for further analysis. 2.5 μl of each cDNA sample was mixed with 0.4 μM primer and 10 μl SYBR-Green Mix (Qiagen, Düsseldorf, Germany) in a total volume of 20 μl. Quantitative real-time PCR (qRT-PCR) experiments were performed using a Mx3005P qPCR system (Agilent Technologies, Santa Clara, CA, USA) with a thermal cycling profile as follows: segment 1, 20 min at 95°C, 1 cycle; segment 2, 45 s at 95°C, 1 min at 58°C, 1 min at 72°C, 45 cycles; segment 3, 1 min 95°C, 30 s 55°C, 30 s 95°C, 1 cycle. Each sample was analyzed in duplicate. Ct values were shown either as absolute data or normalized to the housekeeping gene gap and expressed as fold change to the untreated control (ΔΔct). All oligonucleotide primer pairs (Table S1) were tested for efficacy with a serial dilution of genomic DNA.

RNA deep sequencing was used to determine the regulon of FurA in MAP. Sequencing libraries of MAPwt and MAPΔfurA RNA, of three independent samples each, were prepared and sequenced using 36 bp single-ends sequencing on a Genome Analyzer IIx (Illumina, San Diego, CA, USA) or 50 bp single-ends sequencing on a HiSeq2500 instrument (Illumina), respectively. In brief, libraries of 300 bp were prepared according the manufacturer’s instructions “Preparing Samples for Sequencing of mRNA” (Illumina) and “Protocol for ScriptSeq^TM v2 RNA-Seq Library Preparation” (Epicentre Biotechnologies, Madison, WI, USA). Quality of the libraries was validated using an Agilent Bioanalyzer (Agilent Technologies) following the manufacturer’s instruction. Cluster generation was performed using the Illumina cluster station; sequencing on the Genome Analyzer IIx or HiSeq2500 followed a standard protocol. The fluorescent images were processed to sequences and transformed to FastQ format using the Genome Analyzer Pipeline Analysis software 1.8.2 (Illumina). The sequence output was controlled for general quality features using the fastq-mcf tool of ea-utils (Aronesty, 2013) and was mapped against the genome sequence of the reference strain MAP-K10 (AE016958) using BWA version 0.7.5 (Li and Durbin, 2009) and SAMtools (Li et al., 2009) for storing nucleotide sequence alignments. Strain MAP-K10 was used as a reference strain as the genome of MAP DSM44135 has not been sequenced yet. Raw data sets are available in the European Nucleotide Archive Repository ¹. For data analyses, reads of the 50 bp sequencing were clipped to 36 bp reads. Following, all sequences were computed with Rockhopper tool (McClure et al., 2013). Genes with a q-value ≤0.01 were considered as significantly differentially expressed and genes with raw counts in all replicates were included for further analysis (Table S3). The fold change was calculated as the expression value of MAPwt/MAPΔfurA.

Differentially expressed genes identified by Rockhopper analysis were further processed with Blast2Go tool (Conesa et al., 2005) and NCBI blastx ² as well as tuberculosis database ³ to investigate functions of putative proteins. Additionally, the putative proteins were clustered into orthologous groups (COG), based on the genome of MAP-K10 ⁴. Data are expressed as mean ± SEM, and statistical analyses were performed using GraphPad Prism 5.03 (GraphPad, San Diego, CA, USA). Depending on the experiment, either the non-parametric t-test (Mann–Whitney) or one-way ANOVA test (Kruskal–Wallis) were used. A p-value of <0.05 between samples and controls was considered as statistically significant.

Prediction of putative FurA binding sites in MAP (NCBI accession No. NC_002944) was performed by FIMO analysis (Bailey et al., 2009). Initially, we generated a consensus sequence for FurA binding by comparison of sequences located -70/+30 bp up-/downstream of predicted FurA translation start sites in Mtb (NC_000962, Rv1909c), M. bovis BCG pasteur (NC_008769, BCG_1948c), M. avium ssp. avium 104 (NC_008595, MAV_2752), MAP (NC_002944, MAP_1669c), and M. smegmatis (NC_008596, MSMEG_6383) with MEME Suite (Bailey et al., 2009). The resulting binding motif T[CT]TTGACT[CG][AG]TTCCA[GA]A[AT]AA[GT][GT][GC][ACG][GC][TG]CAT[AT] was subsequently submitted to FIMO analysis. Nucleotides in brackets are variable, single nucleotides are conserved.

Ferric uptake regulator A (FurA) has been described as a global regulator in E. coli and was proposed to be involved in oxidative stress response and regulation of iron homeostasis (Zheng et al., 1999). In mycobacteria, iron homeostasis is maintained by IdeR, and FurA is assumed to take part in general stress response due to its genomic co-localization and its (partly) co-transcription with katG. However, the regulatory role of FurA has not yet been fully understood.

To gain insight into the role of FurA in MAP, we generated a furA (map1669c) deletion mutant MAPΔfurA by specialized transduction. In addition, a complemented strain was constructed by introducing the integrative plasmid pFurA-MAP1101, harboring the furA gene under control of its own promoter, into the ΔfurA mutant strain, resulting in strain MAPΔfurA^C. Deletion of the furA gene in MAPΔfurA and complementation of MAPΔfurA^C were confirmed by PCR (Figure 1A). Expression of furA and the adjacent gene katG were analyzed by qRT-PCR. furA transcripts were detected in MAP wildtype (MAPwt) and MAPΔfurA^C but not in MAPΔfurA. The expression levels of katG were similar in all strains (Figure 1B), indicating that the genetic manipulation had no polar effect on the expression of downstream located genes. Next we compared growth of MAPwt, MAPΔfurA, and MAPΔfurA^C in MB-complete broth. As shown in Figure 1C, all strains exhibited similar growth kinetics. Growth of MAPΔfurA and MAPΔfurA^C was slightly reduced in comparison to MAPwt, and both strains reached stationary growth phase at slightly lower OD₆₀₀ than MAPwt. These effects might be due to the antibiotic supplements in the growth media.

To analyze the regulatory role of FurA in MAP, we prepared RNA from MAPwt and MAPΔfurA grown to an OD₆₀₀ of 1.0 in MB-complete broth and performed RNA deep sequencing. Gene expression profiles of three independent replicates of each strain were analyzed with the Rockhopper analysis tool (McClure et al., 2013), and genes with a q-value of ≤0.01 were considered as significantly differentially expressed. Differences in gene expression were calculated as fold change of expression values calculated by Rockhopper analysis (MAPΔfurA vs. MAPwt).

In total, we identified 48 genes, which were differentially expressed in the mutant strain. Compared to the wt strain, 13 genes were higher (Table 1) and 35 lower (Table 2) expressed in MAPΔfurA. Clustering of putative proteins encoded by the differentially expressed genes into orthologous groups (COG) revealed that almost 40% of the differentially expressed genes were assigned to metabolism, 20.75% to cellular processes and signaling, 9.43% belong to the group of poorly characterized proteins, and 30.19% were not assigned to any COG (Tables 1 and 2). No genes related to iron homeostasis, such as siderophores (mycobactin) and siderophore uptake systems (Neilands, 1992; Escolar et al., 1998; McHugh et al., 2003) or iron storage proteins bfr, fntA (Abdul-Tehrani et al., 1999; Wildermann et al., 2004), were affected by the furA deletion in MAP.

Among the 13 higher expressed genes, 11 were organized in 4 putative operons. The majority of these genes encodes for proteins predicted to be involved in stress response or redox processes. Within this group, the gene cluster map1589c–1587c, encoding for two alkyl hydroperoxide reductases involved in oxidative stress and drug resistance as well as intracellular survival in Mtb (Sherman et al., 1999; Master et al., 2002), and a NAD(P)H nitroreductase (map1743c), which has been shown to be associated with detoxification of nitroaromatic components in Mtb (Purkayastha et al., 2002), exhibited the most significant expression changes (Table 1).

The group of lower expressed genes is comprised of 35 genes. Aside from furA, this group consists of two putative transporters, five genes predicted to be involved in metabolism, three stress response-associated genes and 13 genes were of unknown function. The remaining 11 genes can be associated to virulence which is represented by two putative invasion proteins (map1203/1204), two putative antigens (map0047c/map2168c), one Pro-Pro-Glu protein (PPE; map1003c), one PE-PGRS glycin-rich protein (map3868), one nlpc p60 protein (map1272c), and one mammalian cell entry (mce) family protein (map2114c) as well as three other enzymes (map1706, map1967c, map3421c) for cell entry and persistence. Altogether, map0047c, map0847 (conserved protein), and map4206c (putative efflux ABC transporter) showed the most significant differences in expression compared to the wt (Table 2). In addition, the 5′ UTRs of map1589c (ahpC), map0081 (putative oxidoreductase), map0847, and map3421c (putative diguanylate cyclase phosphodiesterase) featured a predicted FurA binding site (Tables 1 and 2). Overall, these analyses indicated a considerable involvement of MAP FurA in the adaptation of MAP to the host cell environment and general stress response.

To confirm the results of our RNA deep sequencing analyses and to further specify the involvement of FurA in regulation, we analyzed the transcription of ahpC and ahpD (map1589c/1588c, higher expressed in MAPΔfurA), and of map0847 and map0047c (lower expressed in MAPΔfurA) by qRT-PCR. MAPwt, MAPΔfurA, and the complemented strain MAPΔfurA^C were grown in MB-complete broth to an OD₆₀₀ of 1.0, RNA was extracted and analyzed by qRT-PCR. As a control, we included the mbtB gene (map2177c) which has been shown to be regulated iron-dependently by IdeR in Mtb (Gold et al., 2001) and MAP (Janagama et al., 2009). As expected, the expression of mbtB was similar in all strains. In agreement with our RNA sequencing analyses, ahpC, ahpD, map0847, and map0047c were up- and down-regulated in the mutant, respectively. In the complemented mutant MAPΔfurA^C expression levels were mostly restored to wt levels (Figure 2). The differences to the wt strain might be due to the slightly lower expression level of furA in MAPΔfurA^C compared to MAPwt (Figure 1B). Overall, these results confirm the regulatory influence of FurA on expression of ahpC, ahpD, map0847, and map0047c.

AhpC and AhpD are metal dependent alkyl hydroperoxide reductases and constitute an antioxidant system important in reduction and neutralization of a wide range of organic hydroperoxides and peroxynitrite (Wood et al., 2003). Genes encoding for AhpC/D have been shown to be induced by peroxide stress mediated by OxyR in other mycobacteria (Dhandayuthapani et al., 1997). However, in MAP these genes seem to be regulated by FurA which resembles PerR mediated activation in Bacillus and Campylobacter (van Vliet et al., 1999; Helmann et al., 2003). Thus, we exposed MAPwt and MAPΔfurA cultures without catalase supplement to oxidative stress by addition of 10 mM H₂O₂ for 2 h and analyzed ahpD expression by qRT-PCR. As a negative control we used the FurB dependent but FurA independent regulated mptA (map3736c) gene (Eckelt et al., 2014).

As shown in Figure 3, expression of ahpD in MAPΔfurA was increased compared to the wt and supports RNA-sequencing data. Upon peroxide treatment, both strains responded with an induction of ahpD expression. However, the relative fold change in MAPΔfurA was considerably higher than in MAPwt compared to the untreated control.

In contrast mptA expression was similar in all strains at both conditions (standard/oxidative stress).

These data on the one hand side indicate that repression of ahpC/D by FurA can be relieved by peroxide. On the other hand they demonstrate that an additional peroxide sensing regulatory activity, most probably by OxyR or a OxyR-like regulator, co-regulates ahpC/D expression in MAP.

It has been shown that ahpC/D are necessary for defense against ROS and RNS in the mycobacterial phagosome (Sherman et al., 1995; Wilson et al., 1998; Master et al., 2002). By this pathogenic mycobacteria are able to survive within macrophages, enabling metabolic adaptation to the phagosomal compartment which is necessary for persistence. The above data showed that FurA is involved in the regulation of MAP response to oxidative stress, as indicated by the regulation of enzymes involved in neutralization of oxygen and nitrogen radicals [ahpC/D, NAD(P)H nitroreductase]. Therefore, we aimed to address the role of FurA for intracellular survival of MAPwt, MAPΔfurA, and MAPΔfurA^C in macrophages. First we confirmed that the MAP strains are exposed to ROS after being taken up by the macrophages. For this, J774A.1 macrophages were infected with MAPwt, MAPΔfurA, and MAPΔfurA^C. After 1.5 h of infection, cells were treated with the reagent CellROX Deep Red for 30 min. This dye fluoresces upon oxidation and allows the detection of ROS. As positive control we included macrophages treated (+) with menadione (Figure 4B) which is known to induce the generation of ROS in eukaryotic cells (Malorni et al., 1993). As shown in Figures 4A,B, the MFI measured by FACS analysis in macrophages infected with MAPwt, MAPΔfurA and the complemented mutant strain was higher than in untreated (-) macrophages (Figure 4B), indicating the generation of ROS induced by all MAP strains. Next, cells were infected with MAPwt, MAPΔfurA, and MAPΔfurA^C as described above and bacterial survival was determined after 2 h, 2 and 7 days by CFU-counting after serial dilution plating. After 2 h and 2 days infection we found significantly higher numbers of viable MAPΔfurA cells compared to MAPwt (Figure 4C). Complementation of MAPΔfurA almost completely restored the phenotype to wt level. Confocal analyses of macrophages infected with the different MAP strains for 30 min or 2 h after inside-outside staining of bacteria showed no differences in mycobacterial uptake by and adhesion to the macrophages (Figure 4D). These findings clearly exclude a better phagocytosis of MAPΔfurA which could account for the initial higher survival of MAPΔfurA in the macrophages. Interestingly, the percentage of viable MAPΔfurA cells decreased over time from ∼60% (of the initial inoculum) at 2 h to ∼40% at day 2 and to ∼18% at day 7. In contrast, no change in survival was visible in the wt and complement strain, indicating that the mutant was also being killed (Figure 4C). Calculating the survival rates at day 2 and 7 relative to the number of bacteria taken up by the macrophages after the 2 h infection period revealed that MAPΔfurA even though being more resistant to initial killing, is much more efficiently killed during the time course of macrophage infection (data not shown). Overall, these data indicate that MAPΔfurA can better resist the initial killing by the macrophage most probably caused by oxidative stress but is susceptible to macrophage defense mechanisms present at later time points of infection or less adapted to the phagosomal life style.

Hence, we analyzed the role of FurA for survival in the host. For this, C57BL/6 mice were infected intraperitoneally with 1 × 10⁸ bacteria of exponentially grown cultures (OD₆₀₀ of 1.0) and were sacrificed 4 weeks post infection. As shown in Figure 5, no significant differences in weight of liver and spleen, as well as number of bacteria or granuloma in the liver upon infection with MAPwt and MAPΔfurA were detected. However, as a trend we observed lower CFU numbers in the livers of MAPΔfurA infected cells. Overall, the lack of FurA in MAP did not alter survival rates compared to the wt and suggests that FurA might be dispensable for MAP-infection in the mouse model.