The ins and outs of metal homeostasis by the root nodule actinobacterium Frankia

Background Frankia are actinobacteria that form a symbiotic nitrogen-fixing association with actinorhizal plants, and play a significant role in actinorhizal plant colonization of metal contaminated areas. Many Frankia strains are known to be resistant to several toxic metals and metalloids including Pb2+, Al+3, SeO2, Cu2+, AsO4, and Zn2+. With the availability of eight Frankia genome databases, comparative genomics approaches employing phylogeny, amino acid composition analysis, and synteny were used to identify metal homeostasis mechanisms in eight Frankia strains. Characterized genes from the literature and a meta-analysis of 18 heavy metal gene microarray studies were used for comparison. Results Unlike most bacteria, Frankia utilize all of the essential trace elements (Ni, Co, Cu, Se, Mo, B, Zn, Fe, and Mn) and have a comparatively high percentage of metalloproteins, particularly in the more metal resistant strains. Cation diffusion facilitators, being one of the few known metal resistance mechanisms found in the Frankia genomes, were strong candidates for general divalent metal resistance in all of the Frankia strains. Gene duplication and amino acid substitutions that enhanced the metal affinity of CopA and CopCD proteins may be responsible for the copper resistance found in some Frankia strains. CopA and a new potential metal transporter, DUF347, may be involved in the particularly high lead tolerance in Frankia. Selenite resistance involved an alternate sulfur importer (CysPUWA) that prevents sulfur starvation, and reductases to produce elemental selenium. The pattern of arsenate, but not arsenite, resistance was achieved by Frankia using the novel arsenite exporter (AqpS) previously identified in the nitrogen-fixing plant symbiont Sinorhizobium meliloti. Based on the presence of multiple tellurite resistance factors, a new metal resistance (tellurite) was identified and confirmed in Frankia. Conclusions Each strain had a unique combination of metal import, binding, modification, and export genes that explain differences in patterns of metal resistance between strains. Frankia has achieved similar levels of metal and metalloid resistance as bacteria from highly metal-contaminated sites. From a bioremediation standpoint, it is important to understand mechanisms that allow the endosymbiont to survive and infect actinorhizal plants in metal contaminated soils. Electronic supplementary material The online version of this article (doi:10.1186/1471-2164-15-1092) contains supplementary material, which is available to authorized users.


Background
Frankia is a soil dwelling diazotrophic actinobacteria that forms a symbiosis with a variety of woody dicots, primarily of the nitrogen-fixing clade of Eurosids [1,2]. Actinorhizal plants are found worldwide in a broad range of ecological conditions [3]. The symbiosis with Frankia allows these actinorhizal host plants to colonize harsh environmental terrains including highly contaminated or nutrient-poor soils [4]. Frankia cultures exhibit elevated levels of tolerance to various heavy metals including Pb 2+ , Al 3+ , SeO 2 3− , Cu 2+ , AsO 4 , and Zn 2+ [5]. The levels of tolerance to several heavy metals by some Frankia strains are even greater than those of Cupriavidus metallidurans, a wellcharacterized metal-resistant β-proteobacterium isolated from metal-contaminated sediment [6]. As a potential consequence of their association with plants that frequently grow in poor soils with low buffering capacity, Frankia may be exposed to high levels of metals. Under these low-buffering-soil conditions, metals may be freely solubilized from the soil substrate through the action of organic acids, phenolics, and protons produced from both plant and microbial communities [7].
Nearly a quarter of all proteins require a metal cofactor, many of which are toxic at elevated levels [8]. Frankia are highly versatile as a saprophyte, plant symbiont, diazotroph, and producer of secondary metabolites, and therefore require many essential metals for growth [9]. The effects of metals on Frankia physiology have been investigated in culture [10,11], in planta [12,13], and in the field [14][15][16]. However, little is known about the molecular details and mechanisms of metal resistance and homeostasis by these actinobacteria.
Generally, bacteria have a combination of low and high affinity permeases and ATP-driven metal importers for each of the essential metals [17]. Toxic metals (mercury, lead, cadmium, arsenic, etc.) are often unintentionally imported with the beneficial ones due to similar atomic and aqueous divalent ion radii [18,19]. This ion mimicry adds a layer of complexity as cells must not deplete reserves of beneficial metals while exporting their toxic counterparts. Metal export occurs primarily through four mechanisms: (1) cation diffusion facilitators (CDF), (2) resistance-nodulation-cell division (RND) complexes, (3) P-type ATPases, and (4) metal-specific efflux permeases [20]. Many multi-gene transporters are found in a single transcriptional unit (operon), or a separately transcribed functional unit (cassette). An intricate system of metal import, binding, storage, metal cofactor handling, efflux, and cellular redox maintenance mechanisms is needed to maintain metal homeostasis [18,21]. This complex system must be responsive to external growth conditions, metabolic state, and various cellular activities (e.g. nitrogen fixation), making metal homeostasis one of the most complex biological processes.
For this study, the genomes from the cultured Frankia strains CcI3, ACN14a, QA3, EAN1pec, EuI1c, EUN1f, DC12 and CN3 were used as the databases. These strains represent three of the four Frankia lineages, have large genome size ranges (5.43 -9.97 Mbp), and have diverse levels of association with host plants. In this study, we used bioinformatics and comparative genomics approaches to identify metal homeostasis and toxic metal resistance mechanisms in Frankia.

Part 1: Beneficial metals and metal homeostasis
In the absence of genetic tools to assess gene function in Frankia, comparative genomics provides a means of exploring Frankia metal homeostasis and resistance capabilities. A combination of bioinformatics techniques were used to identify metal homeostasis mechanisms in Frankia ( Figure 1) including: meta-analysis of 16 gene array studies to identify conserved mechanisms (Additional file 1), phylogenetic grouping with characterized metal homeostasis mechanisms (Additional files 2, 3, 4, 5, 6, 7) gene neighborhood synteny comparison (Additional file 8), and the presence of conserved metal binding residues in protein alignments (Additional file 9). From these analyses, several groups of potential metal homeostasis mechanisms were identified (Additional file 10). Relevance of metal resistance mechanisms was correlated to previously established metal and metalloid tolerance levels [5]. Figure 1 shows a composite diagram of the Frankia resistance/tolerance mechanisms, while the profiles for the individual Frankia genomes are found in the supplemental materials (Additional files 11,12,13,14,15,16,17,18). Discussed below are the unique combinations of import, chaperone, modification, and export mechanisms involved in metal homeostasis and toxic metal resistance in eight sequenced Frankia strains.

Iron homeostasis
Iron, zinc, and manganese are essential to most, if not all, organisms [45][46][47]. Analysis of the Frankia genomes revealed hundreds of genes that encode proteins requiring iron (Table 1). Interestingly, the more metal-resistant Frankia strains and C. metallidurans had significantly higher proportion of iron metalloproteins than other bacteria (Table 1). Three types of iron importers were identified in the Frankia genomes: (1) a high-affinity permease, (2) iron-siderophore ABC cassettes, and (3) Fe 3+ ABC cassettes. A closely related (33% identity) homolog of the ferrous iron transporter (FTR1) from Saccharomyces cerevisiae was found present in all of the Frankia genomes ( Table 2). The Frankia FTR1 homologs were in a possible operon with an iron-transport lipoprotein and a Dyp-type haem peroxidase. These proteins may function collectively for iron assimilation.
All of the Frankia strains also have at least one ABC cassette, either FhuBCD or FepBCDG that functions to transports iron-siderophore complexes ( Table 2). Except for strain CcI3, the fepBCDG operons in the Frankia genomes contained an enterobactin-type permease (COG4997) and a hydroxamate-type permease (COG0609) (Additional file 2). This heterodimeric iron permease complex may allow the transporter to accept a wider variety of iron chelating substances, such as xenosiderophores of the rhizosphere community, providing a competitive advantage [45,48]. The presence of an additional iron importing cassette, either citrate or haem-type, was found in several Frankia strains (Additional file 10). Frankia produces both catecholate and hydroxamate siderophores, making the presence of multiple Fe-siderophore importing ABC cassettes physiologically relevant [49,50].
During symbiosis, an increased iron acquisition is required by Frankia for both nitrogenase activity and hemoglobin production [51]. Only the symbiotic Frankia strains contain a third type of iron-ABC cassette (FbpABC), which is used for ferric iron import (Additional file 10). The supplementary high affinity iron transporter may help Frankia procure additional iron needed for nitrogenase or for metalloproteins to deal with host-derived phenolic compounds produced during symbiosis [52]. There is a further connection between iron and nitrogen fixation. Nitrogen fixation by Frankia and cyanobacteria requires boron [53], and boron facilitates ferric iron Figure 1 Overview of metal homeostasis mechanisms in Frankia. Schematic diagram of known and putative metal homeostasis systems in Frankia. Symbols for proteins involved in metal ion uptake transporters, chaperones, modification enzymes, efflux transporters, and surface binding protein and efflux systems are shown (left to right) with arrows to indicate the flow of metals through the cell. Evidence for displayed mechanisms can be found in Additional file 10. Gene counts for displayed mechanisms are shown in Table 2. transport via FbpABC [54]. Increased oxidative stress occurs from the import of the more toxic ferric-iron and from the host immunity during symbiosis [52,55]. The nitrogenase enzyme needs iron. but is permanently inactivated once oxidized During nitrogen fixation, the iron cofactored superoxide dismutase (SodF) is up-regulated in Frankia [56]. This link between iron and nitrogen-fixation explains why SodF is only found in the N 2 -fixing Frankia strains (Additional file 10). The iron chaperone, IscA, is involved in the production of iron-sulfur clusters and was found in all of the Frankia genomes ( Table 2). The nitrogen-fixing strains have a second copy of the IscA, which is likely used for the Fe-S containing nitrogenase complex [57]. Once inside the cell, iron is removed from siderophores and reduced to ferrous iron by the reductase FhuF [45]. Orthologs of FhuF are 3X more abundant in the nitrogen fixing strains (Table 2), and possibly work in concert with the ferric iron ABC transporters found only in these strains. The ferrous iron is incorporated into proteins by ferrochelatase or placed into bacterioferritin (Bfr) or Dps storage complexes [48]. The ferrous iron efflux protein (FieF) is in the family of cation diffusion facilitators (CDF) which may export both divalent zinc and iron [58]. FieF may help alleviate iron stress in Frankia by exporting surplus iron.

Zinc and manganese homeostasis
All of the Frankia genomes have hundreds of zinc and manganese metalloproteins, nearly as many as for iron ( Table 1). The necessity of zinc is well established, both as a structural component of macromolecules and as a cofactor in enzymatic reactions, with up to 10% of proteins in a genome requiring a zinc cofactor [59,60]. Manganese, as an antioxidant in the cytoplasm, helps detoxify oxygen radicals as a cofactor for SodN, and replaces the more toxic iron when cells are under oxidative stress [46,61]. Zinc and manganese are structural analogs and share transport mechanisms [46]. While the natural resistance-associated macrophage protein (NRAMP)-type manganese permease (MntH) is found in closely related actinomycetes including Geodermatophilus and Mycobacterium, the mntH genes were conspicuously absent in Frankia genomes. However, the Frankia genomes contained low affinity magnesium permeases, CorA and MgtE (Table 2), which transport several types of divalent cations including zinc [62]. Additionally, Frankia sp. strain CcI3 and F. alni strain ACN14a contained the zinc-iron permease (ZIP).
All of the Frankia genomes possessed high-affinity Mnt/ Znu-type ABC transporter. Under low nutrient conditions, this high affinity Mnt/Znu -type ABC cassette functions for both Mn 2+ and Zn 2+ uptake, as seen with M. tuberculosis [63]. There is little understanding of the appropriate placement of zinc into metalloproteins once it enters the cytoplasm. Certain GTPases are metallochaperones, some of which deliver nickel or cobalt (UreJ/HypB and CobW, respectively), while other proteins (YciC) are suspected to deliver zinc to metalloproteins [64]. The Frankia homologs of these GTPases (COG0523) fall phylogenetically between CobW/YjiA and YciC, both of which are zinc chaperones (Additional file 3).
Some Frankia strains are resistant up to 8 mM zinc (data not shown), which is a level comparable to the metal resistant C. metallidurans isolated from a zinc contaminated site [6]. The absence of zinc-type ATPases (EC:3.6.3.3 and 3.6.3.5) and heavy metal RND (HME-RND) exporters in the Frankia genomes indicates their reliance on other resistance mechanisms. In addition to FieF, two more cation diffusion facilitators (CzcD and  Import permeases ZnT-9 Counts indicate the total number of instances of the mechanism based on protein identifiers listed in Additional file 10. Symbols were assigned either by the closest ortholog in well characterized organisms or most commonly used symbol from the literature. ZnT-9) were present in all of the Frankia genomes and probably contribute to zinc resistance ( Figure 2). CzcD is found up-regulated in both gram-positive and gramnegative bacteria for a variety of metal cations (e.g. Zn 2+ , Cd 2+ , Cu 2+ ) and, being under control of ArsR, it is also up-regulated by arsenic oxyanions (Additional file 1). The CzcD orthologs found in the Frankia genomes are highly similar to those found in the array studies ( Figure 2). As one of the few orthologs to known metal export mechanisms, CzcD may contribute significantly to Frankia metal resistance. Unlike CzcD, the exact function of the other CDF transporter (ZnT-9) remains elusive [65]. Due to its ionic mimicry of zinc, cadmium likely follows the same route into and out of the cell, being imported through general ion permeases (e.g. CorA) and exported via cation diffusion facilitators [59]. CadD, which is involved in low-level cadmium resistance [66], was present in three Frankia genomes, but its contribution to metal resistance is not clear based on metal resistance patterns alone ( Table 2). Manganese has low toxicity and there are few known bacterial manganese-specific exporters [46]. Because of its similarity to zinc and iron, Frankia probably exports manganese via cation diffusion facilitators, as was demonstrated for Arabidopsis and Streptococcus pneumoniae [67,68].

Nickel and cobalt homeostasis
Several other trace elements act as cofactors for hundreds of metalloproteins, including: nickel [69], cobalt [70], copper [71], molybdenum [72], and selenium [73,74]. Although these trace elements are used by most organisms, only 17% of them use all five as cofactors [47]. All of the Frankia strains appear to utilize all five trace elements (Table 1), with each strain having at least three different metalloenzymes per element (data not shown). Nickel import occurs through active transport via the nickel ABC cassette (NikABCDER) or by high (HoxN/NhlF) and low affinity (CorA and MgtE) permeases [61]. All of the Frankia genomes, except strain EUN1f, have a high affinity nickel permease (NhlF) ( Table 2). Additionally, the Frankia genomes possessed a wide range (4 to 23) of the binding component (nikA) genes and multiple combinations of the remaining subunits (nikBCDE) (Additional file 10). Nickel and oligopeptide ABC transporter subunits are highly homologous and multiple nickel/oligopeptide cassettes are frequently present in a genome [69]. Protein sequence alignments showed that oligopeptide (DppA) and nickel transporters of Escherichia coli and Bacillus subtilis are more closely related to each other than any of the Frankia homologs (data not shown), making the separation of these functions in Frankia impossible without experimentation. NikA/ DppA homologs are also heme binding, creating a relationship between iron and nickel transport, possibly due to NiFe cofactored enzymes (e.g. hydrogenase) [75]. The nickel-iron link may explain both the large number of NikA homologs, which were directly proportional to the number of bacterioferritins in their respective Frankia strains (Additional file 10), and the large number of iron and nickel co-factored proteins in Frankia (Tables 1). Once inside the cell, chaperones, such as UreG and HypAB, bind and insert nickel into the appropriate protein (e.g. urease) [76]. Both of these chaperones were found in all of the Frankia genomes (Additional file 3). At elevated levels, nickel becomes toxic due to displacement of cognate metal cofactors in enzymes that facilitate electron transfers [77]. With B. subtilis, the cation diffusion facilitator (CzcD) was more highly expressed with nickel exposure over that of any other cation (Additional file 1). Frankia strains, which are equally resistant to nickel as E. coli and B. subtilis [5], have orthologs to CzcD Bsub ( Figure 2) and so may also use CzcD to export surplus nickel. Frankia sp. strain EAN1pec, EUN1f, and CN3 genomes contained a nickel permease that phylogenetically clusters with the nickel and cobalt efflux permease (RcnA) from E. coli rather than NhlF (Additional file 4). This transporter provides only a low-level of nickel resistance [78], but may have another function since all of the Frankia strains are similarly resistant to Ni 2+ .
Cobalt plays an important role as a cofactor for corrinoids, such as cobalamin (vitamin B12), and several other metalloenzymes including methionine synthase and bromoperoxidase [70]. The cobalt import cassette (CbiMNOQ) is an energy coupled factor (ECF)-type cobalt transporter involved in cobalamin formation [79]. Only Frankia EUN1f and EAN1pec genomes possessed CbiMNOQ. The other Frankia strains may be obtaining sufficient cobalt via the CbtA, CorA, or MgtE permeases ( Table 2). These permeases allow passage of Co 2+ and other divalent cations [61].
Cobalt toxicity is due to generation of reactive oxygen species and mis-incorporation into iron metalloproteins [70]. To guide correct insertion of cobalt into proteins, all of the Frankia genomes possessed both aerobic (CobN) and anaerobic (CbiX) cobalt chaperones. While Frankia is not an anaerobe, it can lower cytoplasmic oxygen to protect the nitrogenase complex inside nitrogen-fixing vesicles surrounded by multiple layers of hopanoid lipids [80]. At least two copies of the cobalt exporter (corC) gene were found in all of the Frankia genomes (Additional file 10). In several different bacteria, this transporter was up-regulated with a variety of di-and tri-valent metals (Additional file 1), and may function as the exporter counterpart to the general metal importer CorA.

Copper homeostasis
Bacterial cuproenzymes are typically involved in cellular redox cycling reactions, such as oxidation of carbon compounds (e.g. laccase, tyrosinase), electron transport (e.g. cyanins, cytochrome c), and reduction of inorganic oxides (e.g. nitrite reductase, superoxide dismutase) [81]. In Gram-negative bacteria, the use of copper is limited to either the periplasm or cytoplasmic membrane, and so there are few known copper-specific importers [71]. In Gram-positive bacteria, the absence of a periplasm for handling copper necessitated cytoplasmic involvement. This likely drove the evolution of a copper influx permease (YcnJ) and the more copper-resistant cellular thiols (e.g. mycothiol) [82,83]. The proposed process of copper homeostasis in Frankia is as follows: copper import via YcnJ [84], detoxification by metallothioneins (MrnA) [85], chaperoning of copper to cuproenzymes or exporters by CopZ [86], and exported by transporters such as CopA and CopCD [20,81]. Furthermore, three of Frankia strains (DC12, EuI1c, and CN3) are highly resistant to copper at MIC values of 2, 5, and 20 mM, respectively [5]. Studies on copper resistance are complicated due to the presence of multiple resistance mechanisms (CopZA, CopABCD operon, and the CusCBA) [85]. With Frankia, mechanisms for copper resistance were not obvious. Frankia genomes did not possess a CusCBA system, and all of them have at least one Cu 2+ -ATPase (copA) and two copCD genes ( Table 2). Amino acid substitutions in copies of CopA and CopD are likely contributors to the high levels of copper resistance in the resistant Frankia strains (discussed below).
Bacterial P-type ATPases fall into two types: one exports copper and silver (EC:3.6.3.4), the other generally exports lead, zinc, and cadmium (EC:3.6.3.5) [20]. Protein sequence comparison with experimentally verified P-type ATPases indicated that 14 out of 15 Frankia ATPases were of the Cu/Ag-type (CopA) (Figure 3). CopA proteins are distinguished by the amino acid motif, CPxALGLATP, found in trans-membrane segment 6, compared to the xPCALVxxTx motif of Pb/Cd/Zn-type ATPases (Figure 3). Several Frankia genomes contained duplications of copA, but all copA genes in copper sensitive strains were on a distinct phylogenetic branch (Additional file 5). Many CopA proteins function as a low rate copper efflux system for cytochrome c oxidase, rather than as a resistance mechanism [87]. Both heme and cytochrome c biosynthesis operons are located in the copA gene neighborhood of Frankia genomes (Additional file 8). It is possible that most Frankia CopA proteins function to provide copper to electron transport chain components, or to maintain low cytoplasmic copper levels during heme synthesis [84].
Two duplicate CopA proteins (FrCN3draft_2755 and FraDC12draft_3523) contained several amino acid substitutions with negatively charged residues (Asp and Glu) and had a His rich N-terminus (Figure 3). These proteins also grouped phylogenetically with well-characterized copper-resistance CopA transporters rather than other Frankia Cu 2+ -ATPases (Additional file 5). This evidence would suggest that these Cu 2+ -ATPases may function as copper resistance mechanisms for CN3 and DC12.
YcnJ, was found in each Frankia genome (Table 2), and consists of a fused CopC/CopD polypeptide that may function as a copper importer under copper limiting conditions [84]. Another copper transport system is the copABCD operon (copA here being a multi-copper oxidase not a Cu 2+ -ATPase), which has been implicated in both copper import and export [81,88,89]. CopC and CopD are considered a functional unit, with CopC binding copper, possibly for delivery to the copper permease CopD [81]. All Frankia strains have both copper importing YcnJ and copper importing/exporting CopCD systems (Additional file 6). In Frankia, CopD is a polypeptide with an attached cytochrome C oxidase (Caa3) domain. CopC and SenC may work in tandem with this CopD-Caa3 protein to delivery copper to membrane bound cytochromes [90].
Both YcnJ and CopCD transporters tend to contain more charged and metal-binding amino acids in the resistant Frankia strains (not shown) and they form separate groups phylogenetically (Additional file 6). Frankia sp. strain EuI1c also has a third CopD (FraEuI1c_1869) which lacks the Caa3 domain, has increased metal binding amino acids, and groups phylogenetically with the two CopD proteins from the highly copper resistant C. metallidurans (Additional file 6). These two copD genes (Rmet_5668 and Rmet_6115) are up-regulated either with copper or with copper and other divalent cations (respectively) (Additional file 1).
The transcriptional response to copper was examined in Frankia sp. strain EuI1c [91]. All of the predicted copper transporters (CopD and YcnJ) and copper binding proteins (CopC and CopZ) are up-regulated in the presence of elevate copper levels [91]. In summary, duplication and modification of amino acids in copper transporters, CopD for strain EuI1c and CopA for strains DC12 and CN3, may explain the high resistance in these Frankia strains. Interestingly, these three copper-resistant strains are also atypical (non-nitrogen fixing) Frankia strains, although there is no clear link between nitrogen fixation and copper sensitivity.

Molybdenum homeostasis
Molybdenum is found as part of an iron-sulfur cofactor in nitrogenase or combined with a tricyclic pterin forming the molybdenum cofactor (MoCo) [72]. MoCo oxidoreductases are nearly ubiquitous in life, and include many important enzymes classes including several dehydrogenases and ferredoxin oxidoreductases [92]. Molybdenum enters the cell through the high affinity ABC cassette (ModABC) or through low affinity sulfate permeases, SulT and SulP, due to its chemical similarity to sulfate [73]. All of the Frankia genomes have the ModABC cassette and several copies of SulP for high and low-affinity molybdate transport ( Table 2). The molybdate chaperone (ModG) was only found in the genomes of nitrogen-fixing Frankia strains ( Table 2), suggesting ModG is involved in the synthesis of the iron-molybdate nitrogenase cofactor [72]. The non-nitrogen-fixing Frankia genomes CN3, DC12, and EuI1c contain a sulfate export protein (TauZ). Since molybdate will enter through general sulfate importers, it is possible that TauZ could function to export surplus molybdate. All nitrogen-fixing Frankia strains lack TauZ, possibly to retain their molybdenum reserves.

Selenium homeostasis
Selenium, in the form of selenocysteine or selenomethionine, is found in several stress proteins including glutathione peroxidase, alkyl hydroperoxidase, and multiple disulfide reductases [93]. The deprotonated electrons of selenium cofactors make the selenoproteins reduction-oxidation reactive, explaining why many identified selenoproteins are involved in thiol and oxidative stress resistance [93]. Since selenite generates these stresses in the cell, the stress-related selenoproteins may function in both detoxification and removal of free selenite ions from the cytoplasm [94]. Only 20% of sequenced bacteria are known to contain selenoproteins [47]. All of the Frankia genomes contained selenoproteins ( Table 1).
As a sulfur analogue, selenium oxyanions enter cells through sulfate transporters such as SulP and CysPUWA, but selenite competitively inhibits the sulfate uptake through SulP [73]. In this study, four Frankia strains (DC12, EUN1f, EuI1c, and CN3) have elevated levels of selenite resistance (up to 3.5 mM) compared to E. coli K12 (0.5 mM) [5]. Only the genomes from these selenite-resistant Frankia strains contained both the SulP and CysPUWA sulfate permeases genes (Table 2), which are likely to prevent sulfur starvation in the presence of selenite.
Once inside the cell, selenite is reduced to inert insoluble elemental selenium, which has red color [95]. A red precipitate is formed by the selenite resistant Frankia strains CN3, EuI1c, EUN1f and DC12 implying selenite reduction [5]. Recently, selenite reduction and selenium nanosphere production was verified in Frankia strain EuI1c [96]. Small thiol-containing molecules, like glutathione, reduce selenite to seleno-diglutathione, which is further reduced to selenopersulfide by glutathione reductase [95]. Only Frankia EuI1c, EAN, and CN3 genomes contained glutathione synthase (GSH) ( Table 2). However, all of the Frankia genomes contained synthase proteins for small thiols like mycothiol (MSH) ( Table 2), which may substitute for glutathione for metal resistance [97]. Selenium is also reduced enzymatically either through thioredoxin and its reductase (TrxA and B) or nitrate/nitrite reductases [98]. These reductases are found in all Frankia genomes, likely to supply the cell with selenopersulfide. The selenopersulfide is either integrated into proteins via selenocysteine or spontaneously dismutates into elemental selenium [95]. The Frankia CN3 genome has a second type of nitrate reductase (NasC) that was associated with nitrite reductases (NirBD), and may contribute to the greater levels of selenite resistance in Frankia strain CN3 (Additional file 10).
Selenium-resistant strains also contained sulfite oxidase genes (FraEuI1c_6626, FrCN3DRAFT_7646, FrEUN1f_7579). Sulfite oxidase (SorA) converts selenite to the less toxic selenate [98,99]. In summary, Frankia selenite resistance is likely due to alternate sulfate transporters (CysPUWA) that prevents sulfur starvation. Additional selenite resistance may result from oxidation of selenite to the less toxic selenate using SorA, or from selenite reduction. The selenite reduction observed in resistant strains could occur through several mechanism including mycothiol, TrxAB, YedY, or NasC/NirBD.

Part 2: Toxic metals Arsenate resistance
Being structurally analogous to phosphate, arsenate enters the cells through phosphate uptake transporters and acts as a metabolic poison [100]. Arsenate resistance involved the reduction of arsenate to arsenite, which is transported out of the cell [101]. Arsenate reduction to arsenite occurs by either the glutaredoxin-dependent ArsC (KO:K00537) or a thioredoxin-dependent phosphotyrosine-phosphatase (PTPase) termed ArsC2 (KO:K03741) [102]. These reductases are typically found in an ars operon with an arsR regulator and one of several types of arsenite transporters (arsAB, arsB, acr3, or aqpS) [100]. Arsenate reduction by Frankia likely involves ArsC2 and mycothiol, which substitutes for glutathione in the arsenate reduction process [102]. Glyoxalase I isomerizes hemithioacetal adducts (glutathione recycling), is responsive to various heavy metals, and is often found associated with ars operons in bacteria [103][104][105]. A cadmium-inducible glyoxyalase I (CadI) was also found within all of the potential Frankia arsenate resistance operons (data not shown). This evidence suggests CadI may assist in reducing glutathione/mycothiol-arsenical complexes.
Several Frankia strains are significantly more resistant to arsenate than E. coli (>50 mM vs. 10 mM), but not to arsenite (0.1 mM vs. 3 mM) [5]. A similar pattern of arsenate resistance and arsenite sensitivity was observed with Sinorhizobium meliloti, which has a novel pathway for arsenate resistance [106]. In S. meliloti, ArsC reduces arsenate to arsenite which is exported out of the cell via an aquaglyceroporin (AqpS) [106]. Only one (SMc02648) of the three AqpS homologs in S. meliloti is involved in arsenite export [106]. Most Frankia genomes contained at least two aqpS genes ( Table 2). Phylogenetic analysis placed Frankia AqpS proteins on two of three distinct branches of aquaglyceroporins (Additional file 7). One branch of Frankia Aqps proteins was orthologous to SMc02648. The gene neighborhood of these Frankia aqpS genes contained the arsR, arsC2 and cadI genes (not shown). Frankia EuI1c genome contained 2 copies of aqpS genes, but was 10-fold more sensitive to arsenate than other Frankia strain. The duplicated aqpS gene (FraEuI1c_6606) was missing the last two transmembrane domains, and had more positively-charged arginine residues (40 vs 9) on the long cytoplasmic c-terminal stretch. Possible explanations for this strain increased arsenate sensitivity is that this protein may form only a dysfunctional channel, or retain arsenic oxyanions on the positively charged cytoplasmic domains.
A second branch of the aquaporin (AqpZ) proteins contains two members that were up-regulated with heavy metals (Additional file 7). These AqpZ proteins were nearly identical to AqpS except they lacked the long cytoplasmic c-terminus (data not shown). The third branch of this phylogenetic tree contained homologs that were related to the glycerol uptake facilitators (GlpF), which are also known for arsenite-importing in both eukaryotes and prokaryotes [100]. No glpF genes were found in any of the Frankia genomes. Another arsenite transporter (ACR3), an arsenite sodium symporter, was identified in a few Frankia genomes (Table 2), but its presence did not correlate to the levels of arsenate or arsenite resistance.

Chromate resistance
Chromate enters the cell through sulfate permeases, and is a potent inducer of oxidative stress and DNA repair mechanisms [107]. Chromate resistance involves export by the chromate permease (ChrA) followed by reduction either by ChrB or by a number of other proteins (nitrite reductase, quinone oxidoreductases, cytochromes, glutathione, etc.) [107,108]. Most Frankia strains are more resistant to chromate than E. coli, B. subtilis, and C. metallidurans [5]. The presence or absence of ChrA in the Frankia genomes did not correlate to their levels of resistance. Other known chromate-resistance determinants [108][109][110] were either not present in the Frankia genomes or did not correlate to chromate resistance levels.
By looking at genes present only in more chromate resistant (MIC >1 mM) Frankia strains, 2 candidate chromate resistance genes were identified. One was a metallo-β-lactamase, which is unlikely to be involved in chromate resistance. The other gene coded for a possible metallothionein, which contained 3 or 4 CXXC motifs interspersed with multiple histidines (Additional file 9). Metallothioneins are involved in chromate binding and reduction through thioester bond formation [111], and are found in association with the ChrA transporter in related actinobacteria [112]. Therefore, it is possible that this metallothionein contributes to chromate resistance in Frankia.

Lead resistance
Lead enters cells through Fe 2+ and Ca 2+ transporters, and exerts its toxicity by displacing these cations at their binding sites in metalloproteins [113]. The most well-characterized lead resistance mechanism occurs in C. metallidurans. This mechanism involves lead export through the Zn 2+ -type ATPase (PbrA), followed by precipitation on the cytoplasmic membrane through the combined action of a PAP-2 phosphatase and signal peptidase (PbrB and PbrC, respectively) [114]. Lead easily forms stable lead-phosphate precipitates. Several gram-positive and gram-negative bacteria exploit this property to detoxify lead as intra-or extracellular precipitates [115].
All of the Frankia strains are highly lead resistant with up to 8-fold higher levels compared to C. metallidurans and at least 2-fold greater than S. viridochromogenes levels [5]. Frankia was also able to bind significant amounts of lead as a precipitate [116]. In a genome-wide array study of C. metallidurans under lead stress, the most lead-responsive gene was the copA-like gene (Rmet_3524) [6]. The lead resistance genes, pbrABCDT, are up-regulated primarily with zinc and cadmium stress with only pbrA being up-regulated with lead stress [6]. All of the Frankia genomes contained copA genes and these gene products were significantly divergent from the other characterized CopA proteins (Additional file 5). One possibility is that the Frankia Cu 2+ -ATPases also export other metals like Pb 2+ . Phosphate utilization (e.g. polyphosphate phosphatase) and cell surface modification proteins (e.g. UDP-glucose 4-epimerase) were located within the highly conserved gene neighborhood of the Frankia copA genes (Additional file 8). These proteins are known to play a role in the surface binding of Pb 2+ [117].
In a Bacillus isolate which lacked the pbrA and pbrT genes, lead was exported by alternate transporters, but is still bound extracellularly [118]. Among the genes that are up-regulated with lead stress, there are 3 paralogous operons containing the genes for undecaprenyl phosphatase (UppP) and DUF347 protein (Additional file 1). Although UppP does not necessarily cause metal resistance, it may help to stop the futile metal import-export cycle by precipitating metals on the cell surface [119]. The DUF347 protein is 6 trans-membrane-segment protein that may work in conjunction with the co-expressed UppP as another metal export-precipitation system. In Thalassiosira pseudonana, DUF347 is the most highly up-regulated gene that was divalent metal-specific [120] supporting this hypothesis. All of the Frankia genomes contained at least one copy of DUF347 and multiple copies of UppP (Table 2). In contrast, only 3 of the 18 non-Frankia bacterial genomes contained a DUF347 or multiple UppP orthologs (Additional file 10). The presence of DUF347 or multiple UppP orthologs could explain the high lead resistance levels for these Frankia strains.

Tellurite resistance
Tellurite is considered a biologically useless metalloid. Due to its induction of oxidative stress in bacteria, tellurite is toxic at micromolar levels [121]. Tellurite enters bacterial cells through either the phosphate transporter (PitA) or nitrate transporters [122,123]. Similar to selenite, tellurite is detoxified through a reduction mechanism by nitrate/nitrate reductases and small cellular thiols like glutathione, leaving an insoluble black precipitate [121,124]. Additionally, there are specific tellurite resistance genes including telAB, terABCDEZ, and tehAB, but little information exists on their exact functions [121]. While tellurite resistance levels have not been established for Frankia, there were several factors that suggested potential tellurite resistance. Many Frankia strains were also resistant to other metal oxyanions like selenite and arsenate. Secondly, all of the Frankia genomes had a full set of terABCDEZ genes and nitrite reductases (NirBD). Bacteria having terABCDEZ operons are frequently resistant up to 4 mM tellurite [125]. Frankia CcI3 and CN3 genomes also had nitrate reductase (NasC) gene, which could be used for selenite or tellurite reduction [121]. To test our hypothesis and to distinguish between tellurite and selenite resistance, Frankia strain CcI3, which is selenite-sensitive, was grown on tellurite agar. Frankia strain CcI3 was able to grow in the presence of tellurite and produced black elemental tellurium in medium containing up to 3 mM TeO 3 (Additional file 19).

Metal regulation
The MerR family contains several types of self-regulating transcriptional activators of genes involved in copper/silver resistance (CueR), gold (GolS), mercury resistance (MerR), zinc (ZntR), cadmium (CadR), lead (PbrR), cobalt (CoaR), oxidative stress (SoxR), nitrogen metabolism (GlnR), biofilm formation (MlrA), and heat shock (DnaK) [126][127][128]. There are between 9 and 20 MerR proteins in Frankia genomes (Additional file 10), which phylogenically cluster into ten groups (Additional file 20). Four of these nine groups could be metal responsive based on a conserved amino acid motif (Additional file 10) derived from sequence alignments of metal induced MerR genes from compiled gene arrays (Additional file 1). None of the metal-related merR genes are located in operons with metal resistance proteins, so are likely regulating acting in trans.
The second largest family of metal regulators in Frankia include members of ArsR/SmtB-family. Originally characterized as a repressor of the arsenic responsive ars operon, members of this family can respond to a variety of metals (nickel, zinc, copper, cadmium, lead), causing de-repression of regulatory targets [129]. In Frankia, ArsR proteins phylogenetically cluster into 5 groups. One groups is part of the potential arsenic resistance operon (Additional file 21). There is also a large group of ArsR-like proteins in a potential operon with activator of HSP90 ATPases (AHSA1), though a direct link between ArsR and AHSA1 expression has not been elucidated. A third group shows sequence similarity to the nickel sensor from M. tuberculosis (NmtR), though retain no synteny or proximity to metal transporters. A small group of ArsR proteins are orthologous with the cadmium and lead responsive CmtR of M. tuberculosis and nickel responsive SrnR of Streptomyces griseus [130,131]. Members of this group of Frankia ArsR proteins are in a possible operon with the metal exporter CzcD (Additional file 21). The last group shows no homology to any characterized ArsR family proteins and are not proximal to other metal related genes.
The other Frankia strains contain orthologs to RicR Mtub (Additional file 22), which is responsive to copper, iron, and zinc [134]. A branch of Frankia CsoR-like proteins (Franean1_3360, FrEUN1f_0873, FrCN3DRAFT_5236) cluster with an uncharacterized regulator (Rv1766) and to a lesser extent the copper and nickel responsive The iron transport repressor (Fur) binds a variety of essential divalent metals including iron (Fur), zinc (Zur), nickel (Nur), manganese (Mur), or cobalt (Co-Fur) the combination of which determines the level of Fur transcription [137]. Bacteria contain an additional layer of Fur-family proteins (eg. PerR) which also bind metals and regulate the transcription of Fur [137]. Orthologs of Fur and Nur have been previously examined in Frankia which retain a high level of protein sequence conservation to characterized proteins [138,139]. In Frankia, Fur is located in an operon with either KatA or KatG catalases (Additional file 23), which rely on iron (heme) to function [140,141]. There are two phylogenetic groups of Frankia Nur proteins (Additional file 23). For cluster I Frankia nur is located adjacent to a metallopeptidase (gcp) and the nickel chaperone ureG. In cluster III and IV Frankia genomes, nur is upstream of an ironsulphur cluster repair protein (YgfZ). The characterized S. coelicolor Nur ortholog is also co-located with YgfZ and has been demonstrated to regulate nickel import genes (nikABCDE) and oxidative stress (sodF) [139].
Another Fur paralog (Zur) regulates zinc importers, and is itself regulated by ArsR as demonstrated in M. tuberculosis [142]. ArsR is zinc responsive and may work with Zur to regulate intracellular zinc through repression and derepression of the cation transporter CzcD [143]. While all Frankia genomes contained a Zur regional to the zinc importer ZntABC, Frankia sp. strain DC12 has a second Zur in an operon with ArsR (Additional file 23). The remaining Fur protein in Frankia (PerR) is in operon with rubrerythrin, which binds iron and zinc and is involved in oxidative stress response [144]. The Frankia genomes contain a second type of iron regulatory protein, diptheria toxin repressor (DtxR) (Additional file 10), which regulate several iron related proteins including siderophores and iron storage proteins [145]. The Frankia DtxR proteins are distinguished from the closely related manganese regulator (MntR) by the conservation of four amino acid residues previously identified as iron binding [146].

Conclusions
Frankia has successfully adapted to a challenging habitat: an acidic rhizosphere full of secondary metabolites, plant phenolics, and solubilized metals. To accomplish this feat, Frankia has a full complement of proteins that use all major beneficial metals. Frankia metal resistance patterns appear to be unique to each strain. The high levels of copper resistance appears to be due to duplicated copD genes in Frankia sp. strain EuI1c and unusual copA genes in strains CN3 and DC12. Selenite resistance involves an alternate sulfur importer (CysPUWA) working with reductases to produce elemental selenium. Nickel, cobalt, zinc, and cadmium resistance are likely the result of a combination of CDF proteins. Lead appears to be exported via Cu 2+ -ATPases or DUF347, and precipitated as a bound metal phosphate similar to other bacterial species. Frankia arsenate resistance appears to be achieved through a novel mechanism utilizing an aquaporin (AqpS) that was discovered in nitrogen-fixing S. meliloti. The terABCDEZ genes and various reductases provide Frankia CcI3 (and possibly other Frankia strains) with tellurite resistance. Frankia has achieved similar levels of metal and metalloid resistance as bacteria from highly metal-contaminated sites. The potential for use of Frankia-actinorhizal symbiosis in bioremediation efforts makes further research into expression of these genes worthwhile.

Gene identification in Frankia
An initial list of genes involved in Frankia metal homeostasis was created using a combination of comparative analyses, motif searches, PERL algorithms, and microarray data (Additional file 24). Comparative analysis of Frankia and other genomes was performed primarily using the Integrated Microbial Genome (IMG) database [147]. Metal homeostasis genes identified in the literature that lacked locus tags or gene accession numbers were identified through the UniProt database [148]. Orthologs of genes from functionally characterized proteins of other microorganisms were identified in Frankia using the embedded BLASTp software on IMG. Proteins with a minimum of 30% identity and e-values <10 −20 were considered orthologs.

Low scoring and hypothetical proteins
The large phylogenetic distance between Frankia and model organisms, such as E. coli and B. subtilis, meant many of the closest metal transport homologs in Frankia had less than 30% identity and e-values of >10 −20 . These low-scoring genes, as well as all hypothetical proteins were: 1. examined for conservation of domains from other databases including protein families (Pfam) [149], gene ontology [150], and cluster of orthologous groups (COG) [151], 2. individually analyzed for conserved metal binding residues as compared to characterized metal resistance genes, 3. visually checked for neighborhood synteny and clustering with other metal homeostasis genes using native IMG tools [147].

Gene array meta-analysis
Genes from 16 different published whole genome array studies involving bacterial response to heavy metals were analyzed as follows. Any potential metal homeostasis proteins that were significantly (based on originating paper cut-off ) up-regulated were included for analysis. Genes were assigned an IMG identifier by downloading the genome data from IMG for each organism and matching genes from the arrays to the IMG data using Excel tools. The genes were uploaded to IMG and the various information about the genes (symbol, COG numbers, protein domains…etc) were obtained and exported. The expression data were matched to the downloaded IMG gene information and this collective data was combined and analyzed in Excel (Additional file 1).
To normalize gene expression between studies, the genes of each study were ranked by dividing the fold change of each gene by the maximum fold change in that study. Thus, the gene with the maximum response has a value of 1 and the remaining genes are a fraction of the maximal response (Additional file 1). Up-regulated genes in multiple different species/array studies were selected for analysis. BLASTp searches were done using IMG to identify orthologs of these genes in the Frankia genomes.

Assessment of metal binding potential
Using the regular expression search function on IMG and in-house PERL algorithms, the Frankia genome were searched for proteins containing either conserved metal binding motifs obtained from the literature or high metalbinding potential as previously described [112]. Since each genome has a unique amino acid distribution, all proteins in each genome were scaled by % of each individual amino acid. Proteins in the top 1% highest cysteine or histidine were considered potentially metal binding and examined for other metal homeostasis factors, including: neighboring metal-related genes, presence/absence in metal resistant strains, orthology to genes up-regulated with metal in array studies.

Phylogenic tests of relatedness
To distinguish highly similar mechanisms or metal specificity, phylogenetic trees were constructed with proteins from Frankia and organisms that were either: 1. from metal homeostasis literature, 2. from past studies involving Frankia metal resistance, and/or 3. from the compiled gene array studies. General comparative organisms for gene identification can be found in Additional file 10 and include either organisms used in the Richards et. al. 2002 analysis of Frankia metal resistance or organisms with model metal homeostasis systems in 4 major categories: actinobacteria, Gram positive and negative bacteria, and eukaryotes (to demonstrate genetic distance). Alignments were created using ClustalW in BioEdit [152], and neighbor-joining trees with 1000 bootstraps were constructed using MEGA software [153]. Functionality or metal specificity was tentatively assigned based on grouping with experimentally characterized proteins from the literature.

Ethics
This study did not involve humans, animals or plants in any manner.
arrows to indicate the flow of metals through the cell. Information at the