The Nocardia cyriacigeorgica GUH-2 genome shows ongoing adaptation of an environmental Actinobacteria to a pathogen’s lifestyle

Virulence status of N. cyriacigeorgicaGUH-2

Virulence of N. cyriacigeorgica GUH-2 strain was confirmed by intravenous injection of approximately 10⁷ CFU in the tail of BALB/c mice. The death rate rose to 60% 7 days after infection. Autopsy indicated death to be due to septicemia with formation of nodules in several organs: kidneys, liver, brain, spleen, lungs and heart (Figure 1). Microscopic analysis of nodules showed high numbers of poly- and mono-nucleated inflammatory cells, and N. cyriacigeorgica GUH-2 cells. Some mice were injected a lower number (3.5 × 10⁵ CFU) of N. cyriacigeorgica GUH-2 cells and developed abnormal behavior: rigidity, stooped posture, hemiparesis and vertical yes-yes head shaking probably due to a brain infection.

General features of the Nocardia cyriacigeorgicagenome

GUH-2 chromosome harbors eleven insertion sequences (see IS section), and 15 CDS encoding putative phage proteins. Putative virulence genes are scattered along the chromosome without distinguishable pathogenicity island (see Additional file 1 for a complete listing). Several syntons of 5 CDS or more were found conserved between N. cyriacigeorgica, N. farcinica, Rhodococcus jostii and Mycobacterium tuberculosis (Figure 2). Non-conserved regions frequently showed distinct G + C % biases. Regions encoding the largest putative proteins from the genome were found to have features of non ribosomal peptide synthetases (NRPS) and showed a high G + C content. The only exception is NOCYR_4710 CDS which has a G + C content slightly lower than the averaged one, and encodes a putative Dipeptidyl carboxypeptidase Dcp located near a transposase-related CDS (Figure 2).

To better understand the events that had led to the present-day N. cyriacigeorgica GUH-2 genomic structure, its core genome was delimited by identifying CDS conserved in a panel of closely related Actinobacteria (Figure 3). 15% (805 of 5477) of the CDS was found in a putative common ancestor to all genomes except Frankia sp. CcI3 (Figure 4). Of these CDS, 80% could be assigned a function and, as expected, a large proportion was inferred to play part in basic bacterial functions such as synthesis of proteins, nucleosides and nucleotides, amino acids, co-factor prosthetic groups and carriers, and of the cell envelope (see Additional file 2 for the full listing). Interestingly, the only CDS exclusively shared between N. cyriacigeorgica and M. tuberculosis were those of IS987, an insertion sequence, suggesting a likely transfer of this IS between these species. The N. cyriacigeorgica and N. farcinica genomes were found to share 74.5% of their CDS, delimiting a Nocardia pangenome of about 4.5 Mb. 1398 CDS (25.5%) of N. cyriacigeorgica were not found in the N. farcinica genome. Most of these CDS are of unknown function (78%) but some are likely involved in phosphonate transport and metabolism (phytase), synthesis of fatty acids, glutamate metabolism, nitrite/nitrate transport, and RuBisCO production. Conversely, 2253 CDS of N. farcinica were not found in the N. cyriacigeorgica genome (data not shown). Again, a high proportion of these CDS (69%) were “unk” CDS (unknown function) and a few could be related to particular activities like synthesis of thiocyanate (toxic compound) (4 CDS), catabolism of urea (6 CDS), of auxins (6 CDS) and lignin, heavy metal resistance and virulence (9 CDS). These differences were in line with biochemical tests such as measurement of nitrate reductase and urease activities, performed to differentiate these two species. 11.6% CDS (633 out of 5477) of N. farcinica and N. cyriacigeorgica genomes were not recorded in other actinobacterial sequenced genomes (Figure 3). These two Nocardia strains have an equivalent genome size which is longer than the ones of primary actinobacterial pathogens (Table 1). 16.6% (777 kb) of N. cyriacigeorgica CDS content was not recorded in any other organisms referenced in the databases. These CDS were sometimes related to mobile and extrachromosomal elements but most of them were of “unknown function”. A search for amplified CDS among N. cyriacigeorgica genome revealed 161 occurrences, of which 132 are duplicated, 21 are triplicated and 8 are quadruplicated. More than 29% of these CDS could not be given a particular function, and 79% were also found in N. farcinica. IS (1 CDS), mce (4 CDS), transcriptional regulators (12 CDS) and nitrate reductase (6 CDS) were part of these amplified regions. More than 23% of these CDS were found among regions of genomic plasticity (RGP) (see Additional file 3 for more details).

Out of the 5477 CDS present in N. cyriacigeorgica genome, 4016 CDS (i. e. 73%) could be assigned to a COG [see Additional file 4]. The proportion of these COGs among Actinobacteria was similar, with a slightly higher occurrence of CDS among the “transcription” (K) and “signal transduction” (T) COGs of the Nocardia genomes. Correspondence analysis of the number of CDS per COG among a set of actinobacterial species was performed, to identify a possible bias related to the pathogenic nature of the species. This analysis showed that the number of CDS per COG could differentiate primary pathogens from non-pathogens. This “pathogen pattern” was more significant than the COG organization bias inferred from species belonging to a same genus. On the other hand, all non-pathogens had closer COG patterns even though some were part of different genus or part of a genus showing pathogenic species. COG patterns of the Nocardia genomes were found similar to those of non-pathogens (Figure 5, but also see the Additional file 4). A correspondence analysis on the functional domains inferred from the CDS of the “transcription” COG was performed (Additional file 5 and Additional file 6). This analysis did not segregate the dataset according to the pathogenic nature of the species regardless of their genus. Instead, the numbers of CDS per sub-division were found similar between the Nocardia genomes and similar to those observed among M. tuberculosis.

Regions of genomic plasticity

A lineplot graph between N. farcinica and N. cyriacigeorgica genomes, representing synteny results of series of 5 CDS or more, was performed in order to visualize the distribution of variable and conserved regions (Figure 6). This analysis showed the number of conserved CDS between these chromosomes to increase towards their respective origin of replication. The lowest concentration of these CDS was observed around the chromosomal terminus of replication. Overall, the organization of these variations was quite similar between the two halves of the circular chromosome creating a mirror-like effect indicative of increasing evolutionary constraints from the terminus towards the origin of replication (on both strand). The chromosomal terminus is partially visible on the circular map of N. cyriacigeorgica genome, and shows a large segment with a distinct G + C content according to the GC plot (Figure 2). The Artemis Comparison Tool (ACT) was used to refine these analyses and identify DNA segments >4.5 kb or containing more than 5 CDS. These DNA regions were not showing all the features of genomic islands such as a tRNA gene at one end, an integrase CDS or a G + C bias distinct from the one of the genome. Twenty-two RGP could be detected using this approach (named RGP-Cy#) (Table 2). Twenty-one of these RGP were also detected with the RGPfinder tool of the MaGeplateform. RGP-Cy6 was not detected in this latter analysis because of its length of 4.9 kb. Of these RGP, four could be considered genomic islets (< 10 kb), and the largest RGP was of about 80 kb. These RGP represented a total of 704 kb i.e. 11.4% of the genome and encoded 622 CDS. The average G + C% content of these RGP is of 65.5%, with values ranging from 60.1% to 68.6%. RGP boundaries were analyzed. tRNA or tmRNA genes were detected at the extremity of ten of these. Direct DNA repeats were observed for RGP-Cy8 from positions 1098658 to 1098672 and 1109676 to 1109690, at its left and right ends, respectively. Four RGP showed IS sequences. Five RGP contain CDS implicated in DNA modification processes such as integrases, recombinases, endonucleases and excisionases. These CDS were probably involved in the acquisition of these RGP. Ten RGP did not show any of the above features. Most CDS on these selected RGP encode putative proteins and transcriptional regulators of unknown function (74%). Some CDS, likely representing a benefit for N. cyriacigeorgica GUH-2, could be identified and predicted to encode a catalase (RGP-Cy3), a limonene monooxygenase (RGP-Cy16), and a sulfonate ABC transporter (RGP-Cy21). Furthermore, CDS involved in nitrate metabolism were identified on the RGP-Cy14. A cobalamin-independent methionine synthase (metE) and three CDS involved in citrate metabolism were found on RGP-Cy15 while pglY and pglZ involved in phage defense were found on Cy10 (Table 2).

PCR screenings were designed to investigate the distribution of these RGP among 83 N. cyriacigeorgica strains (Additional file 7). Prevalence of these RGP was quite variable, with some not being detected in other strains, and some being found among up to 69% of the strains. RGP-Cy4 and RGP-Cy8 were only found in the N. cyriacigeorgica GUH-2 genome and showed all the features of mobile elements. A cladogram was built using the RGP distribution patterns based on the positive and negative PCR results (Figure 7). Dataset of inner RGP and RGP-ends DNA targets were analyzed separately or together, and gave similar cladograms. All strains harboring 5 or more RGP were grouped in the “GUH-2 complex”. Other strains belonged to a “type strain complex”. In fact, RGP patterns of 30 strains were found to be in the “GUH-2 complex”, while patterns of 56 strains were allocated to the “type strain complex”. Among the “GUH-2 complex”, some of the selected RGP were highly prevalent: RGP-Cy1, RGP-Cy6, RGP-Cy11, RGP-Cy15, RGP-Cy16, RGP-Cy18 and RGP-Cy21 were found in 23, 18, 25, 29, 23, 20, and 26 strains respectively. Strain N7 harbored the highest number of RGP found in the N. cyriacigeorgica GUH-2 genome (15/22 positive PCR products). Among the type strain complex, RGP- Cy2, -Cy5, -Cy6, -Cy7, -Cy9, -Cy10, -Cy16, -Cy18, -Cy19, -Cy20, and -Cy21 were not recorded. No RGP seemed to be representative of this latter complex, and RGP-Cy3, RGP-Cy15 and RGP-Cy17 were the most prevalent. RGP-Cy3 was the most broadly distributed (69 positive strains) among the N. cyriacigeorgica species. PCR screenings targeting the left and right ends of RGP were defined to estimate the level of conservation of their respective proximal DNA region. The left (L) and right (R) ends of RGP-Cy9, -Cy15, and -Cy16 were broadly detected whereas only one end could be detected for RGP-Cy2 (R), -Cy3 (L), -Cy4 (R), -Cy7 (R), -Cy11 (R), -Cy17 (R), -Cy18 (L), -Cy19 (R), -Cy20 (L), -Cy21 (L), and -Cy22 (R) among all the strains tested. The L and R ends of the RGP-Cy5, Cy6, Cy10, and Cy11 were detected among the “GUH-2 complex”, while only RGP-Cy1 (L), -Cy3 (R), -Cy7 (L), -Cy11 (L), -Cy13 (L), -Cy20 (R), -Cy21 (R) could be detected among the “GUH-2 complex” (data not shown). It is noteworthy that an attempt was made at comparing this RGP classification with phylogenetic relationships inferred from 16S rDNA sequences. However, significant sub-groups among N. cyriacigeorgica could not be resolved with this marker (data not shown).

RGP represent gains in CDS likely to confer novel properties but CDS losses can also represent adaptations to particular habitats. 193 N. farcinica CDS were found mISing in the N. cyriacigeorgica GUH-2 genome. Among these, 35% encode putative proteins of unknown function but four clusters of CDS involved in glutamine metabolism, in phenyl acid acetic degradation, in thiocyanate degradation and nitrogen metabolism, and in urease synthesis were found mISing (see Additional file 3 for further details).

The particular case of insertion sequences (IS)

On the N. cyriacigeorgica genome, eleven IS belonging to five different families were detected (IS3, IS21, IS200, IS256 and ISNCY). A re-evaluation of the N. farcinica IS genome content was performed, and led to the identification of fifteen IS belonging to five families (IS3, IS5, IS200, IS481, and IS630) (Additional file 8). Only ISNfa14 and ISNcy8 of the IS3 family were found located in the same DNA site of these two actinobacteria. These IS share 86% DNA identity. N cyriacigeorgica genome showed DNA signatures of two Tn3 transposons but one appeared to be truncated. N. farcinica genome also showed the presence of a Tn3 in which the transposase CDS shares 85% DNA identity with the one of N. cyriacigeorgica transposon 2. Inverted repeats (IR) and/or direct repeats (DR) were identified for five N. cyriacigeorgica IS and for three N. farcinica IS. ISNcy2 copies (with DNA identities going from 82 to 100%) in N. cyriacigeorgica genome were found related to IS987 (75% DNA identity) which was only previously detected in the M. tuberculosis and M. bovis genomes. The N. cyriacigeorgica GUH-2 genome harbors the ISNcy1 in four copies with one being partial and three being identical. Three IS were found in several copies in N. farcinica. ISNfa1 of the IS481 family was found in two copies with 99% DNA identity; ISNfa2 of the IS5 family in eight copies with identities going from 72% to 99%; and ISNfap1, a partial element of the Tn3 family, having 85% DNA identity with a copy found on N. farcinica larger plasmid.

The distribution of six of these IS/Tn elements was investigated among a sub-set (n = 18) of the N. cyriacigeorgica strains screened in the RGP section, a set (n = 11) of N. farcinica strains, and a panel of Nocardia species type strains (n = 12) (Additional file 9). Only ISNcy4 (Tn3) and ISNfa2 (IS427 of the IS5 family) were detected among both N. cyriacigeorgica and N. farcinica but their prevalence was quite different from one species to another. In fact, both of these elements were more prevalent among the panel of strains of N. farcinica than N. cyriacigeorgica strains selected for this study. ISNcy2 (IS51 of the IS3 family) was the most prevalent among N. cyriacigeorgica but was absent from the N. farcinica strains tested. This ISNcy2 was also found among the N. otitidiscaviarum type strain. ISNcy3 was not detected in N. farcinica. Its prevalence was higher among the GUH-2 complex of N. cyriacigeorgica (which was defined according to the distribution of RGP in the section above). A similar situation was observed for ISNcy4. ISNcy2 and ISNcy3 were not detected among the Nocardia type strains that were selected for this analysis, suggesting a distribution restricted to N. cyriacigeorgica. ISNfa1 and ISNfa5 were not detected in N. cyriacigeorgica and appeared to be restricted to N. farcinica strain 10152. ISNcy4 (Tn3) and ISNfa1 (IS481) share 89% identity, and were detected among several strains of N. farcinica and N. cyriacigeorgica. A cladogram was built from this IS distribution pattern analysis and was found to match the RGP distribution patterns (data not shown).

Virulence-related functions

Several genes were previously found involved in virulence among Actinobacteria. Blast and keyword-based searches allowed identification of some of these CDS among the N. cyriacigeorgica genome. Six complete mce (mammalian cell entry) loci containing yrbE and mas (mce associated) CDS were found. Analysis of CDS encoding cell wall components involved in virulence revealed 85-kDa antigens family proteins (4 CDS), lipoproteins (19 CDS) and PE_PGRS/PPE family proteins (5 CDS). Two superoxide dismutase CDS (sod) and three catalase ones were also identified. Nitrate reductase CDS (narBGHIJKY, and nirBD), trehalose 6,6’-dimycolate transferase (one CDS) and RuBisCO (2 CDS) were detected and found to be clustered.

Extracellular enzymes were identified by searching for particular domains including a peptide leader and low number of transmembrane domains. The inferred secretome of N. cyriacigeorgica GUH-2 was compared with the ones of other Actinobacteria, and showed a majority (70%) of putative proteins of unknown function, several proteases, lipases as well as a transcriptional regulator and members of the mce genes (Additional file 10 shows these exported CDS in more detail). Comparison of the putative proteins of the N. cyriacigeorgica and N. farcinica secretomes showed high identities (between 81 to 86%), much higher than those observed with other Actinobacteria (60%).

The N. cyriacigeorgica GUH-2 genome shows a good potential for the synthesis of a number of metabolites that could be antimicrobials or proteasome inhibitors. Seven CDS were predicted to encode polyketides synthases (PKS), and 17 CDS were predicted to encode NRPS (non-ribosomal peptide synthetases). Among N. farcinica, 4 PKS and 15 NRPSs were found including the cluster of CDS previously described as producing a lipid-soluble iron-binding nocobactin. A cluster of nine CDSs similar to the coelibactine synthesis cluster in Streptomyces coelicolor was also identified in N. cyriacigeorgica GUH-2’s genome. NOCYR_4800 was found to be the longest CDS of the genome (43689 pb) and likely to encode a NRPS with twelve modules. Two other NRPS operons containing two and three CDS could synthesize molecules containing thirteen and eleven building blocks, respectively. A 2-amino-9,10 epoxi-8-oxodecanoic acid was predicted for NOCYR_0751. This putative acid shows a structure similar to the epoxomicin proteasome inhibitor (epoxy group) but chemical assays will be required to confirm its synthesis. Other softwares did not find this structure.

It is noteworthy that a correlation was observed between the isoelectric point and the molecular weight of the inferred proteins of the N. cyriacigeorgica GUH-2 genome. The largest proteins of the genome, constituted mainly by NRPS and PKS, showed similar isoelectric points, suggesting a compartmentalization inside the cytosol that could allow a coordinated functioning of these enzymes (Figure 8).

Genetic potentials and metabolic profiling

Phenotypic microarray datasets including antibiotic resistances were compared with the functional predictions made from the annotated N. cyriacigeorgica GUH-2 genome. Phenotypic profiling was performed by testing N. cyriacigeorgica ability to grow with various carbon and nitrogen sources. KEGG pathways were used to find the CDS involved in these pathways but also as a reference for explaining an absence of growth under certain conditions. (Additional file 11). Only seven amino acids (L-asparagine, L-aspartic acid, L-cysteine, L-glutamic acid, L-glutamine, L-histidine, L-lyxose, L-tyrosine) could be used as carbon and/or nitrogen sources. The absence of some CDS previously shown in other bacteria to be involved in the substrate catabolism or transport could explain several negative results. However, the lack of growth with L-aspartic acid, L-lysine and L-proline could not be explained by the absence of particular CDS.

The osmoadaptation capacity N. cyriacigeorgica GUH-2 was also tested and found to be high under high salt concentrations. This property could involve CDS encoding transport proteins and/or osmoprotectants like ectoine and betaine (Additional file 11). Growth tests were also performed under various pH. N. cyriacigeorgica GUH-2 was shown to grow at pH 7 and 8 but not pH 5.2. This dataset was completed by an analysis of the antibiotics resistance pattern of N. cyriacigeorgica GUH-2 using plate assays. N. cyriacigeorgica GUH-2 was confirmed to have a type VI pattern with resistance against several sulfonamides, aminoglycosides, tetracyclines and penicillins. It was found sensitive to cefamandole, cefotaxime, amikacin, and imipenem. Many CDS likely playing part in these antibiotic resistances were found in the N. cyriacigeorgica genome. A mutation in the gyrase A gene that can prevent ciprofloxacin binding, was detected. Several CDS were inferred as encoding β-lactamases in the GUH-2 genome, and could be responsible for the observed ampicillin, carbenicillin, oxacillin and penicillin resistances. The paromomycin and tobramycin resistances observed can be due to the expression of one of the three aminoglycoside phosphotransferases detected by the annotation process. Macrolide resistances could involve one of the six ribosomal RNA methyltransferases encoded by the GUH-2 genome. These may cause a decrease in the affinity of macrolides for the 50S ribosomal unit. The fourteen drug or multidrug efflux transporters encoded by this genome could be involved in tetracycline, penimepicycline, polymyxin B, paromomycin, D,L-serine hydroxamate, sisomicin, sulfamethazine, novobiocin and/or sulfadiazine resistance (data not shown) by extruding these molecules.

Virulence determinants

The N. cyriacigeorgica GUH-2 genome analyses revealed several genetic determinants related to virulence. Some of these are part of RGP but some were also part of the Nocardia or Actinobacteria pangenomes inferred from comparisons with the full genome sequence of N. farcinica and other Actinobacteria. The genome of N. cyriacigeorgica harbors virulence-related CDS such as Mce coding genes described as important virulence factors of M. tuberculosis[30]. Mce can act as transmembrane transporters favoring macrophage invasion [31]. These CDSs are organized in operons containing two yrbE CDS followed by six mce CDS and sometimes two mce associated (mas) CDSs. There are six copies of the complete mce operon in N. cyriacigeorgica and N. farcinica whereas four copies are identified in M. tuberculosis genome. Although, the importance of these CDSs in M. tuberculosis virulence has been shown [32], their detection in multicopies among M. smegmatis and R. jostii (six and four clusters, respectively) suggests a function not limited to cell entry [33]. Having access to the N. cyriacigeorgica mce CDS will allow transcriptomics experiments to identify the mce CDSs which are turned on during host cell colonization.

Other N. cyriacigeorgica CDSs besides mce have been described as important in the infection and cell invasion processes. In particular, the expression of superoxide dismutase and catalase CDS were observed during N. cyriacigeorgica macrophage invasion and were suggested to be involved in resistance towards oxidative stresses [34]. Two sod and three kat CDS were found in the N. cyriacigeorgica genome. The katA CDS is harbored by the RGP-Cy3 but the other two kat CDSs are found on conserved genomic regions. Other CDSs can also have a complementary action during Nocardia growth in macrophages. The trehalose 6,6’-dimycolate transferase CDS was found related to the 85-kDa antigen family protein [35] that can promote M. tuberculosis survival in macrophages by decreasing both phagosomal acidification and phagolysosomal fusion [36, 37]. An encoded N. cyriacigeorgica hemolysin which can disrupt the phagolysosome membrane was also annotated and could also favor survival in macrophages [38, 39]. The isocitrate lyase CDS observed on RGP-Cy15 could also be part of the N. cyriacigeorgica macrophage invasion process by preventing host cell apoptosis as observed in M. tuberculosis[40]. An intracellular pathogen not only requires defense mechanisms against macrophage antibacterial processes but also needs to survive under low oxygen pressure such as the one observed in poorly irrigated tissue of the mammal body [41]. CDS involved in such processes can thus also be defined as virulence-related determinants. N. cyriacigeorgica contains denitrification CDSs (narBGHIJKY, and nirBD) of which five are harbored by RGP-Cy14. N. cyriacigeorgica was initially described as an obligate aerobe but presence denitrifying CDS suggests an ability to grow under anaerobic conditions. However, the conditions allowing growth under low oxygen pressure remain to be defined.

Several metabolites, proteins, enzymes and lipids which are not directly involved in colonization but are at the frontline during host infection can also play a role in virulence such as cell wall constituents and some extracellular enzymes/metabolites. The cell wall is a protection for bacteria but can also be a target for the immune system. For example, PE/PPE serine α/β hydrolases membrane proteins are important in Mycobacterium pathogenic species to avoid detection or killing during their intracellular life in a variety of host cells [42, 43]. However, their low number in N. cyriacigeorgica and N. farcinica could be an effect of their opportunistic status (environmental cycle) and could explain their poor ability to escape the immune system of a healthy host. Nocardia can also secrete several enzymes that may interfere with the host cell metabolism [25], in particular SODs [44], lipases [45] and proteases [46]. CDSs encoding such enzymes were observed in the N. cyriacigeorgica GUH-2 genome. Eight CDSs were found encoding extracellular lipases and 10 encoding extracellular proteases. Nevertheless, much more CDSs encoding putative secreted enzymes have been reported in Mycobacterium such as the ESAT (early secreted antigenic target) proteins [47]. N. cyriacigeorgica GUH-2 genome harbored three ESAT CDSs.

Another major group of secreted molecules that can play a role in virulence are siderophores. Nocardia strains can produce several siderophores like formobactin [48], amamistatin [49], brasilibactin [50], asterobactin [51] and nocobactin [52]. Nocobactin synthesis was previously found encoded by two genetic clusters, with cluster I positioned 195 kb from cluster II. The N. cyriacigeorgica GUH-2 genome showed a cluster I organization different from the one observed in the N. farcinica genome. The gene nbtH is absent in the N. cyriacigeorgica GUH-2 genome and replaced by a putative formyltransferase CDS, which could play a role in preventing the transfer of an acyl chain to the ϵ-amino group of lysine. The N. cyriacigeorgica GUH-2 cluster II shows high identities with the one of N. farcinica but additional CDS encoding NRPS and an exported protein of unknown function were recorded. The NRPS CDSs could play a role in the synthesis of this siderophore and change some of its properties. A coelibactin-like siderophore is also likely to be produced by N. cyriacigeorgica GUH-2. A conserved synteny was observed with S. coelicolor CDS involved in its synthesis except for a supplementary CDS encoding a cytochrome P450 protein.

These Actinobacteria are also known to synthesize other extracellular metabolites through NRPS and PKS-related processes. The N. cyriacigeorgica GUH-2 genome was found to encode 12 PKS-related CDSs and 22 NRPS-related ones. It is considered that the substance produced by N. cyriacigeorgica GUH-2 which can cause brain damages by inducing apoptosis and a dopamine depletion would be encoded by a NRPS or PKS [25, 26]. Here, a putative metabolite produced by an operon of three NRPSs could be predicted to have a structure similar to the one of epoxomicin [53]. Such molecules have epoxy groups that could inhibit the functioning of proteasomes. NRPS and PKS are also involved in the synthesis of antibiotics, and could have been involved in the synthesis of transvalencin Z [54], DA-7218 [55], and nocardithiocin [56]. However, the molecules produced by these synthases are often difficult to obtain in large quantities and are hard to purify from culture filtrates without genetic manipulations.

Phenotypic and genomic plasticities

Metabolic profilings showed N. cyriacigeorgica GUH-2 abilities at growing on a variety of substrates. In most cases, the CDSs involved in transport and catabolism of these substrates were found in the genome. However, some metabolic activities inferred from the genome could not be confirmed by growth tests e. g. the growth on L-glycine, L-lysine, and L-proline. Interestingly, CDSs involved in L-lysine and L-proline catabolism were detected outside RGP but showed a codon adaption index (CAI) below the average (data not shown). This codon bias could be related to the low expression of these CDSs, and might be the consequence of a recent acquisition. An overexpression of these genes by genetic manipulations of N. cyriacigeorgica GUH-2 would be needed to test this hypothesis. Low CAI was also observed for CDSs identified when in vivo activity was assessed. In this case the presence of multiple CDS (with both high and low CAI values) assigned to this function probably allowed a sufficient gene expression level to observe the expected phenotype.

A high turnover of RGP appears to occur among N. cyriacigeorgica. N. cyriacigeorgica GUH-2 closest RGP profile among a collection of 83 strains showed the absence of 7 RGP and the conservation of 15 ones. RGP profilings thus suggest a good N. cyriacigeorgica competence towards DNA acquisition but the mechanisms involved remain to be determined. RGP-Cy8 appeared to have been recently acquired by N. cyriacigeorgica GUH-2, and was found to harbor all the trademarks of mobile genomic islands. This RGP could be a good candidate for the design of a DNA cloning vector [57]. No RGP was found related to prophage-like elements. This is different from the situation observed in corynebacterial and mycobacterial pathogenic genomes, where such prophage-like DNA contains virulence genes [28].

A high turnover of IS elements was also observed among Nocardia. However, while several IS were found among the N. cyriacigeorgica (n = 16) and N. farcinica (n = 26) genomes in this work, previous studies had only identified two IS among the Nocardia: ISNfa1 (N. farcinica) [41] and IS204 (N. mexicana) [58]. These low numbers suggested Nocardia strains to have a poor propensity at acquiring exogenous DNA which are often acting as IS shuttles. Here, we clearly demonstrate the opposite. The Nocardia genomes were found to be rich in IS elements and diversity. IS of eight families were recorded suggesting a high genomic tolerance towards these elements and a frequent acquisition by these Actinobacteria. However, about 45% of the observed IS did not show IR and DR. A lack of such sequences could be indicative of a loss of transposition autonomy or of selective pressures leading to their fixation at a particular site because of functional benefits. This would need to be further investigated. IS were found involved in the emergence of pathogenic clones by reducing genome size [59]. There are also some reports showing a good match between the presence of a particular IS and infra-specific diversifications e.g. [60]. Distribution analysis also showed species-specific IS elements among the Nocardia e.g. ISNcy2 and ISNfa2 being restricted to N. cyriacigeorgica and N. farcinica, respectively. Furthermore, ISNcy3 was restricted to the N. cyriacigeorgica GUH-2 complex, and ISNfa5 of two particular clones of N. farcinica. These data support the hypothesis of a good match between IS acquisitions and infra-specific diversifications. Interestingly, distribution analysis indicated similar RGP/IS repartition profiles among N. cyriacigeorgica strains. These profiles divided N. cyriacigeorgica into two clusters: (1) those with patterns similar to strain GUH-2, and (2) those with patterns similar to the type strain. These similarities in the evolutionary patterns of these elements suggest a strong association. RGP could have been the genetic shuttles for some of these IS elements. In fact, even though most IS elements were found distributed over the genome without particular insertion site preferences, some e.g. ISNcy5, ISNcy2-b, ISNCy4, & ISNcy1-a4, were found harbored by RGP. For example, ISNcy2-b could have been acquired with RGP-Cy13. This IS is significantly different from the other ISNcy2-related copies found in the N. cyriacigeorgica GUH-2 genome, and has DNA signatures only detected on the copy found on this RGP. Interestingly, a division of N. cyriacigeorgica into two phylogenetic clusters was inferred by McTaggart et al.[13]. However, the strain collection of this latter study was different from the one of this work. It would be interesting to apply McTaggart et al. [13] approach to see if the IS/GI sub-groups would match such phylogenetic clusters. These analyses would add further support for the role of IS/GI elements in bacterial diversification and speciation. No relation between the habitat nor the geographical origin of the strains used in our study and the observed RGP/IS patterns could be inferred.

It is noteworthy that ISNcy2 was found similar to IS987 of M. tuberculosis and M. bovis, and was not detected in other sequenced bacterial genomes. In order to relate IS divergences to the evolution of Actinobacteria, the two orthologous IS alleles named ISNfa14 and ISNcy8, found in N. farcinica and N. cyriacigeorgica, respectively, were used as molecular clocks. These IS have 86% identity. Using this value, one can consider ISNcy2 and IS987 diversification (75% identity) to be more ancient. PCR screenings showed ISNcy2 to be broadly distributed among N. cyriacigeorgica whereas it was totally absent from N. farcinica and other Nocardia type strains. This would suggest an acquisition by N. cyriacigeorgica at the moment of its differentiation.

Outdoor-related functional ISues

Although a significant number of virulence genes were identified, with some being harbored by RGP, the main drivers of N. cyriacigeorgica evolution appear to be related to its environmental cycle outside mammalian hosts. Nocardioses are without a doubt of environmental origin, and infections are mainly the consequence of exposure to soil or water sources of these opportunistic pathogens. Indeed, several characteristics of the N. cyriacigeorgica GUH-2 genome reflect the environmental origin of this species. For instance, N. cyriacigeorgica genome size is in the range of what is expected for saprophytic bacteria (from 6 Mb to 11 Mb). This higher size range is related to the selection of CDS likely improving metabolic potentials and regulatory processes, and allowing growth under a wider range of environmental constraints [61]. These observations were further supported by COG analyses which showed Nocardia genomes to share a configuration similar to the ones observed among non-pathogenic Actinobacteria. Interestingly, both Nocardia genomes have a higher proportion of K and T COG CDS. These COGs encode DNA and proteins with regulatory functions, suggesting an ongoing evolution towards regulatory fine tunings of their genetic potentials that likely lead to expression patterns favoring growth under more diverse conditions.

Mice experimentations

Female BALB/c mice (pathogen-free) of 18-20 g (approximately 8 weeks old) were maintained by the “Institut Claude Bourgelat” (VetAgroSup, Marcy l’Etoile, France) in accordance with protocols approved by the board of ethics for animal experimentations.

N. cyriacigeorgica GUH-2 strain was grown in brain-heart infusion broth (BHI-P) to mid-log phase at 37°C with mild rotational agitation (150 rpm). The broth was centrifuged at low speed (55 g) to pellet bacterial clumps, and cell concentration was adjusted at approximately 3.5 × 10⁶ CFU/ml. A 0.1 ml of this suspension was injected intravenously (IV) through the lateral tail vein into each mouse, as described in details by Kohbata and Beaman [18]. Each mouse received approximately 3.5 × 10⁵ CFU.

Genome sequencing and assembly

The genome sequence of N. cyriacigeorgica GUH-2 is publicly available at http://www.genoscope.cns.fr/agc/mage[63]. Sequences and annotations data have been deposited at the EMBL database (http://www.ebi.ac.uk/ena/) and given the accession number FO082843.1.

Genome annotation and analysis

Global comparative study

We compared genomes of A. mediterranei (CP002896.1), C. diphtheriae (BX248353), C. glutamicum (BA000036), M. tuberculosis (AL123456), M. smegmatis (CP000480), N. farcinica (AP006618), N. cyriacigeorgica (FO082843), R. equi (ADNW00000000) and R. jostii (CP000431) using various graphic tools implemented on the MaGe Microscope platform and ACT. The phylogenetic tree of life of the Actinobacteria was built from a MLSA data set on the basis of [13]. We used Clusters of Orthologous Groups (COGs) automatic annotation and correspondence analysis (CA) with R software (http://www.R-project.org) [69] to graphically infer global trends between the above genomes. Nocardia IS families were identified using BLAST analyses against an IS database at https://www-is.biotoul.fr/.

Deleted CDS from the N. cyriacigeorgica GUH-2 genome were identified by searching CDS in common with N. farcinica, R. equi and R. jostii and looking for those absent from N. cyriacigeorgica genome using the Phyloprofiles exploration tool of MaGe. Only CDS presenting more than 40% amino-acid identities over 80% of the length of the shortest sequence were considered. Duplicated genes were detected in the N. cyriacigeorgica GUH-2 genome with the same tool, N. cyriacigeorgica genome was compared against itself and CDS presenting more than 70% predicted amino-acid identities were selected.

Distribution of IS and selected RGP among a set of Nocardiastrains

Rapid DNA extractions were performed on the suspension of 10 to 20 colonies of Nocardia cells in 100 μL of sterile water. The mixture was heated to 55°C for 15 min, and 5 units of achromopeptidase (Wako chemicals, Richmond VA) were added before incubating the suspension at 70°C for 15 min. Cells were centrifuged and the supernatant containing DNA was kept for further analyses. PCR screenings were designed to investigate the distribution of all selected RGP reported in this work, and for a set of selected IS elements. Primers were defined using primerselect (DNASTAR), and are indicated in the Additional file 12. All PCR reactions were performed in a final volume of 25 μL, containing 2.5 μL of each primers at 10 μM, 2.5 μL of 10X PCR buffer, 0.75 μL of MgCl₂ 50 mM, 0.25 μL of 10 mM DNTPs, 5% DMSO w.vol^-1, and 1 μL of the extracted DNA solution. PCR cycle is: 95°C 300 sec, 95°C 30 sec, annealing temperature depending on primers used, 72°C from 30 sec to 90 sec (× 35) and 72°C 300 sec. PCR products were visualized by electrophoresis using 1% agarose gels, and staining with ethidium bromide.

Identification of the secretome

Protein secretion in Gram-positive bacteria occurs mainly through general secretory (Sec) and twin arginine translocation (Tat) pathways, and to a lesser extent by ABC (ATP-binding cassette) type transporters. There are also minor pathways such as the Early Secreted Antigen Target (ESAT-6) machinery described in Mycobacterium[70, 71]. The secretome was analyzed for pathogenic and non pathogenic Actinobacteria using SignalP and TMhmm [72] as well as PSORTb [73]. These bioinformatics tools can detect trans-membrane domains [74]. None of these tools is error-free but their combined use yields a set of proteins that were previously shown, in most cases, to be secreted proteins [72, 75].

Metabolic profiling

N. cyriacigeorgica GUH-2 strain was grown for three days at 37°C on Middlebrook 7H10 (supplemented with 0.5% glycerol and 1% Middlebrook OADC enrichment) agar plates. N. cyriacigeorgica GUH-2 was inoculated to 20 mL of IF-0a GN/GP (Biolog Inc, Hayward CA, USA) and homogenized to obtain a 81% transmittance solution, free of bacterial clumps. This suspension (880 μL) was added to 10 mL of IF-0a GN/GP (Biolog Inc) supplemented with 1 mL of solution specific of each Omnilog plate. Dye mix (120 μL) F or H (depending on the Omnilog plate use) was added. This mix (100 μL) was added to each well of the selected Omnilog plates, and the plates were incubated. Cellular respiration was measured by monitoring formation of dark blue tetrazolium crystals over a 72 h time period.