The predominant route for microbial utilization of phosphonates has been thought to be the “C–P lyase” pathway (Ternan et al., 1998) and its presence and activity in marine microorganisms has recently been demonstrated (e.g., Dyhrman et al., 2006). However the broad-specificity C–P lyase multienzyme complex has not yet been fully characterized (Hove-Jensen et al., 2011) although a major technical obstacle to its elucidation has very recently been overcome with the first report (Kamat et al., 2011) of the in vitro reconstitution of the pathway in Escherichia coli and identification of the reaction intermediates. In E. coli the C–P lyase complex is comprised of 14 gene products which together catalyze the difficult dephosphonation reaction in a range of structurally diverse phosphonates which may contain either activated or unactivated C–P bonds (Wackett et al., 1987; Yakovleva et al., 1998). In each case the product is the corresponding hydrocarbon and P_i (Figure 1B). Of these 14 genes, phnCDE are thought to encode an ATP-binding cassette transporter, and phnF a repressor protein, while the products of phnNOP are thought to have regulatory or other accessory roles (Hove-Jensen et al., 2010, 2011). Seven gene products (PhnGHIJKLM) constitute the “core” components of the C–P lyase reaction pathway, with PhnJ now known to catalyze its central reaction, the S-adenosyl-l-methionine-dependent radical cleavage of 5-phosphoribosyl-1-phosphonate to produce 5-phosphoribosyl-1,2-cyclic phosphate and the corresponding alkane (Hove-Jensen et al., 2011; Jochimsen et al., 2011; Kamat et al., 2011). Variant forms of the C–P lyase enzyme complex with different substrate ranges are also known to exist (e.g., Kertesz et al., 1991; Sviridov et al., 2012), and other genes of as-yet unknown function have frequently been found associated with the C–P lyase operon (see Occurrence of the phnJ Component of the C-P Lyase Operon in Sequenced Bacterial Genomes). In addition the role of PhnK remains unknown (Kamat et al., 2011).

In all bacteria studied to date the genes that encode the C–P lyase pathway(s) have been found to be arranged in tandem and transcribed from a single promoter in an operon that is a member of the Pho regulon – a suite of genes whose products include proteins required for Pi-scavenging, and for the catabolism of alternative P sources. Expression of the genes of the Pho regulon is induced under conditions of Pi-starvation but is Pi-repressible (Quinn et al., 2007). Thus in T. erythraeum the genes of the C–P lyase pathway were expressed in situ at Pi concentrations of ∼6 nmol l⁻¹, but not under Pi-replete conditions (Dyhrman et al., 2006). Similarly the aerobic production of methane from methylphosphonate by mixed marine microbial communities (likely to be mediated by C–P lyase) was demonstrated under Pi-stressed conditions (Karl et al., 2008).

All C–P bond cleavage enzymes had originally been thought to be similarly subject to Pho regulon control; however, it seems likely that in this event phosphonates could be utilized by microorganisms only as a source of P, since any Pi released during their catabolism in excess of that required for balanced cellular growth would repress any further degradation (McGrath et al., 1997; Quinn et al., 2007). Indeed, further phosphonate uptake by cells would also be repressed because the relevant high-affinity transporters are also Pho-regulated. However field and laboratory observations (Benitez-Nelson et al., 2004; Beversdorf et al., 2010) suggest that phosphonate degradation in the oceans continues under conditions of Pi sufficiency. And indeed three inducible enzymes capable of C–P bond cleavage have been identified and characterized (Quinn et al., 2007). Unlike C–P lyase, which is capable of the cleavage of both activated and unactivated C–P bonds, each of these phosphonohydrolases is uniquely specific to a natural-product phosphonate that contains an electron-withdrawing carbonyl group in the beta position to the P atom. Expression of all three enzymes has been shown to be Pi-insensitive and hence independent of Pho regulation – thus they are likely to play a role in the marine environment that is complementary to that of C–P lyase.

The phnX gene product phosphonoacetaldehyde hydrolase (EC 3.11.1.1; phosphonatase) cleaves the C–P bond of its unique substrate to yield acetaldehyde and Pi (LaNauze et al., 1977; Dumora et al., 1989); it is encoded within a degradative operon for the mineralization of 2-aminoethylphosphonic acid (2-AEP, ciliatine). 2-AEP replaces its structural analog ethanolamine phosphate as a component of the phosphonoglycolipids common in the cell membrane of many invertebrates and is considered – along with its N-methylated derivatives – to be the most abundant and ubiquitous marine C–P compound (Hilderbrand, 1983; Hori et al., 1984; Quin and Quin, 2001). The mechanism of C–P bond cleavage by phosphonatase, a member of the halo acid dehalogenase superfamily, involves the formation of a Schiff base intermediate between a lysine residue at the active site of the enzyme and the phosphonoacetaldehyde carbonyl group; this activates the phosphonate group for attack by an active site nucleophile (Baker et al., 1998; Morais et al., 2004). The phosphonatase pathway of 2-AEP degradation (Figure 1C) also includes a 2-AEP:pyruvate aminotransferase (EC 2.6.1.37), encoded by phnW, whose products are phosphonoacetaldehyde and alanine (Kim et al., 2002). The Pi-starvation-inducible expression of this pathway as a part of the Pho regulon has been described in Enterobacter aerogenes and Salmonella typhimurium, neither of which is able to utilize 2-AEP as sole carbon and energy source (Jiang et al., 1995). However in other bacterial strains, e.g., Pseudomonas putida NG2 and Pseudomonas aeruginosa PAO1 that are able to mineralize 2-AEP as sole C, N, and P source, expression of the operon is substrate-inducible, mediated by the product of an adjacent gene that encodes a LysR-like transcriptional activator (LTTR; Ternan and Quinn, 1998a; Quinn et al., 2007).

The phnA gene product, phosphonoacetate hydrolase, was identified in an activated sludge isolate of Pseudomonas fluorescens that was capable of utilizing phosphonoacetate as sole C and P source (McMullan et al., 1992). The enzyme is a member of the alkaline phosphatase superfamily and forms acetate and Pi from its substrate following metal-assisted hydrolytic cleavage of the C–P bond (McGrath et al., 1995; Agarwal et al., 2011; Kim et al., 2011; Figure 1D). In its original host phosphonoacetate hydrolase, together with an associated phosphonate transporter, was found to be encoded within an operon whose expression was regulated by a phosphonoacetate-responsive LTTR (Kulakova et al., 2001, 2009). More recently, however, the phnA gene has been found in other bacterial strains within an operon that also encodes 2-AEP transaminase (PhnW) and a novel NAD⁺-dependent phosphonoacetaldehyde dehydrogenase (PhnY) whose product is phosphonoacetate (Borisova et al., 2011; Kim et al., 2011; Figure 1E). The enzyme sequence encoded by this operon thus represents an alternative route of 2-AEP mineralization to that of the phosphonatase pathway (see Phosphonoacetaldehyde Hydrolase), and is likely to be a major route of biogenic phosphonoacetate production (Cooley et al., 2011).

This enzyme was identified by Ternan and Quinn (1998b) in cells of a Burkholderia strain that utilized phosphonoalanine (2-amino-3-phosphonopropionic acid) as sole C, N, and P source. Phosphonoalanine, the phosphonate analog of aspartic acid, is formed by the transamination of phosphonopyruvate (Horigane et al., 1989). It is a constituent of membrane phosphonolipids and is widely distributed amongst marine invertebrates (Kittredge and Hughes, 1964); its degradative pathway involves an initial transamination to phosphonopyruvate, followed by cleavage of the C–P bond by a metal-dependent phosphonohydrolase to yield pyruvate and Pi (Figure 1F). Phosphonopyruvate hydrolase (EC 3.11.1.3) is a member of the phosphoenolpyruvate phosphomutase/isocitrate lyase superfamily and was shown in Variovorax sp. Pal2 to be encoded within an operon that also contains phosphonoalanine transaminase and phosphonate transporter genes; its expression is regulated by an adjacent phosphonoalanine- and phosphonopyruvate-responsive LTTR (Kulakova et al., 2003, 2009). Phosphonopyruvate hydrolase shares 41% sequence identity with the C–P bond-forming enzyme, phosphoenolpyruvate phosphomutase (see Biosynthesis of Phosphonates above); a single amino acid substitution at the active site is likely to be responsible for the functional divergence of the two enzymes (Chen et al., 2006).

There is also evidence for the existence of a number of less well-characterized C–P bond cleavage activities that are likely to be of significance in marine environments:

129 sequenced strains (almost 10% of the genomes contained in the NCBI GenBank database) contained the ppd gene indicative of the PEP phosphomutase/phosphonopyruvate decarboxylase pathway of phosphonate biosynthesis; seven of these were of marine origin. Of the 129 strains, 48 also possessed a pathway for phosphonate degradation, and 10 the associated phnCDE operon encoding a phosphonate transporter linked to phosphonate catabolism (Huang et al., 2005). The phosphonate biosynthetic pathway was distributed amongst distant bacterial groups, but predominantly within the members of the proteobacteria, actinobacteria, and clostridia (Figure 2). Analysis of their ppm/ppd operon structures failed to reveal any clearly conserved gene composition, which may reflect the wide range of phosphonate metabolites that are produced by microorganisms (Circello et al., 2010; Johannes et al., 2010; Lee et al., 2010). Nevertheless, two enzymes frequently found to be encoded within the operon (in addition to PEP phosphomutase and phosphonopyruvate decarboxylase) were 2-AEP aminotransferase (PhnW) – observed in 60 strains – which plays a key role in the production of 2-AEP, and an acetaldehyde dehydrogenase (found in 16 strains), which shares strong homology (44% identity, 60% similarity) with the phosphonoacetaldehyde dehydrogenase (PhnY) known to be involved in the production of phosphonoacetate from phosphonoacetaldehyde and found in the phosphonoacetate hydrolase pathway of 2-AEP degradation (Borisova et al., 2011; see Phosphonoacetate Hydrolase). In addition nine Gram-negative bacterial strains possessed a copy of the phosphonate biosynthetic operon that was flanked by genes encoding enzymes involved in exopolysaccharide biosynthesis; it is known that the capsular polysaccharide B complex associated with virulence in Bacteroides fragilis is characterized by its high content of 2-AEP.

Analysis of the distribution of phnJ, which encodes a catalytic component of the C–P lyase multienzyme system, indicates that C–P lyase is (after “phosphonatase,” see Occurrence in Sequenced Bacterial Genomes of the phnX Gene Encoding Phosphonoacetaldehyde Hydrolase) the second most abundant pathway for phosphonate degradation amongst sequenced bacteria, and that it is widely distributed amongst diverse bacterial groups (Figure 3). Its 223 homologs, including 44 from marine strains, occur mainly in the alpha-, beta-, and gamma-proteobacteria. (Interestingly, however, a copy of the entire operon, and a second copy of its phnCDE genes, were found in the genome of the archaeon Haloquadratum walsbyi; further analysis suggests that these copies have separate proteobacterial origins).

Great diversity was found in the composition and architecture of the C–P lyase operon (including insertions, substitutions, and deletions of components); this confirms a similar observation by Huang et al. (2005) and suggests that the operon has been subject to multiple horizontal gene transfers. All these genetic variations could represent potential variations in the specificity or functionality of the reaction sequence encoded, and thus provide possible evolutionary advantages to their hosts. For example domain of unknown function (DUF) 1045, which is not conventionally designated as one of the 14 genes of the C–P lyase operon, was in fact present in 78 of the 223 operons identified (almost all from the alpha- and beta-proteobacteria, and including 26 of the 44 marine strains in which the operon was found). In 86% of instances DUF1045 flanked or lay close to phnN which is thought to encode a “feeder” activity for the C–P bond cleavage pathway (Hove-Jensen et al., 2011). Expression of both DUF1045 and phnN has been found to be highly induced in the presence of methylphosphonate or the absence of Pi in Acidithiobacillus ferrooxidans (Vera et al., 2008).

Genomic data thus appear to provide support for existing biochemical evidence (see C–P Lyase above) that the C–P lyases may be considered as a family of closely related multienzyme systems that nevertheless vary in, for example, their substrate specificities. The high incidence of “optional” genes suggests that these may provide additional functional features. In terms of their regulation, ∼90% of C–P lyase operons were apparently controlled by flanking GntR regulators, confirming numerous findings that this pathway is inducible only under conditions of Pi-starvation (Wackett et al., 1987). Surprisingly, however, genes predicted to encode alternative regulators (including those from the LysR, IclR, DeoR, LuxR, MarR, ArsR, TetR, and TonB families) were found adjacent to C–P lyase operons in some 6% of cases, while in a small number of others no putative regulatory genes could be identified.

Genes encoding the phosphonatase pathway of 2-AEP degradation, which comprises 2-AEP-pyruvate transaminase (PhnW) and phosphonoacetaldehyde hydrolase (PhnX; see Phosphonoacetaldehyde Hydrolase), were found in 235 bacterial genomes of which 50 were of marine origin, mostly from within the alpha- and gamma-proteobacteria. It is thus the most abundant phosphonate degradation pathway amongst the 1384 bacteria whose sequences are currently available; this may reflect the fact that 2-AEP is thought to be the most widely distributed biogenic phosphonate in the environment (White and Metcalf, 2007). The distribution of genes encoding the phosphonatase operon is shown in Figure 4; some 44% belonged to the gamma-proteobacteria, of whom the majority were potential pathogens (Salmonella, Vibrio, etc.) that have been implicated in gastrointestinal diseases. Predicted phnX-containing strains belonging to the bacillales, bacteroidetes, and clostridia are also associated with similar pathologies. To date, there is no direct evidence to link phosphonoacetaldehyde hydrolase activity to pathogenesis; however it has been reported that low Pi conditions can enhance the virulence of pathogenic species such as P. aeruginosa and Salmonella typhimurimum (Beier and Gross, 2006; Lamarche et al., 2008). Additionally, mutations in Pho regulon genes have been thought to be responsible for the decrease of colonization rates by intestinal strains of E. coli and Vibrio cholerae (von Krüger et al., 1999, 2006). Since phnX in most gamma-proteobacteria was associated with GntR regulators and hence likely to be expressed under Pi limiting conditions (Gebhard and Cook, 2008), it is conceivable that phosphonoacetaldehyde hydrolase may play a role in the infection process.

The phosphonatase operons in Bacillaceae and clostridial strains were observed to be flanked by deoR, and in bacteroidetes by araC regulatory sequences. Only in alpha-proteobacteria (mostly marine isolates), and in beta-, and gamma-proteobacteria (all pseudomonads) were the phosphonatase pathways associated with LysR type regulators. No such strains have been implicated in gastrointestinal infections.

Several features of the phosphonatase operon structures provide strong evidence that PhnX is encoded as part of a genomic island which can be easily transferred amongst bacteria. “Pick-and-mix” horizontal gene transfer is suggested not only by the broad range of regulatory systems apparently involved in the control of the operon expression (including GntR, LysR, DeoR, AraC, and TetR), but also by the insertion of genes such as that encoding 4-aminobutyrate transferase, or DUF6 homologs (the latter is found in some marine strains; DUF6 can also occur within some C–P lyase operons). Furthermore the various rearrangements of the operon components (phnW,X and the phosphonate transporter genes phnC,D,E) that were observed is a feature that is almost entirely absent in pathways of apparently shorter evolutionary history, e.g., that involving phnA (see Occurrence in Sequenced Bacterial Genomes of the phnA Gene Encoding Phosphonoacetate Hydrolase).

The phnA gene was much less widely distributed in sequenced bacterial genomes than those encoding the C–P lyase or phosphonoacetaldehyde hydrolase pathways. Figure 5 shows the affiliation of the 69 bacterial strains that encoded the 89 phnA homologs detected during the bioinformatic analysis (these were predominantly Burkholderia spp. within the beta-proteobacteria). Only six strains were identified as marine isolates; however five of these are members of the Roseobacter clade, which is known to be the predominant bacterial group in the near-surface waters of the oceans.

The limited distribution of phnA within other phyla suggests that phosphonate degradation via phosphonoacetate hydrolase is a pathway of recent development or that it lacks the transferability observed in the cases of the C–P lyase and phosphonoacetaldehyde hydrolase pathways. Nevertheless, the phnA gene was frequently duplicated within the genomes of the species that possessed it (e.g., in 20 different strains of beta-proteobacteria). Only three sequenced strains possessed duplications of phnX, and five strains a duplication of the C–P lyase operon. Mechanisms of duplication have been suggested to consolidate the existence of pathways in the host genome, as has been reported for phthalate metabolism in Rhodococcus sp. strain DK17 (Choi et al., 2007).

In contrast with the C–P lyase and phosphonatase pathways with their apparently longer evolutionary histories, analysis of those operons that contained phnA homologs indicated that these were almost exclusively controlled by only two kinds of regulators: GntR or LysR. Of these, GntR regulators were exclusively encoded in gamma-proteobacteria, with LysR present in all alpha- and some beta-proteobacterial strains (including all six marine strains identified). Rearrangements in the structure of the operons that contained phnA were very limited; most (71 out of 89) contained the phnY gene that encodes phosphonoacetaldehyde dehydrogenase and phnW that encodes 2-AEP aminotransferase, suggesting that phosphonoacetate hydrolase (phnA) is more commonly a component of a degradative sequence for the catabolism of 2-AEP (see Phosphonoacetate Hydrolase), than directly inducible by phosphonoacetate as is the case in the operon originally characterized in P. fluorescens 23F by Kulakova et al. (2001), in which neither phnW or phnY is present.

In contrast to the wide distribution exhibited by the phosphonate degradative sequences analyzed above, pph was found to be present in only three sequenced bacterial strains, all Burkholderia spp. In each instance the associated operon was controlled by a LysR regulator and contained a copy of the phnW gene coding for 2-AEP-aminotransferase. This suggests that the operon associated with phosphonopyruvate hydrolase activity may be of relatively recent origin and that it has not been subject to extensive gene transfer.

In addition to the pathways responsible for the breakdown of phosphonates, the distribution of genes that encode other activities associated with their metabolism was analyzed.

The gene that encodes the PhnD periplasmic binding protein of the phosphonate PhnCDE transporter was very widely distributed across bacteria (Figure 6), even more so than those encoding the C–P lyase catalytic protein, PhnJ, or phosphonoacetaldehyde hydrolase. Numbers of predicted homologs were particularly high amongst marine bacteria; of the 294 strains in the NCBI database that were confirmed to possess a phnD homolog, 67 were of marine origin. In 177 of the strains identified, phnD was juxtaposed with a phosphonate degradative gene (thus 60% of strains capable of transporting phosphonates into the cell have an apparent ability to degrade them). On the other hand, 117 other strains that contained phnD did not encode any known pathway of phosphonate degradation, suggesting that not all phosphonate degradation pathways have yet been discovered (as has been proposed by Palenik et al., 2003). It is also possible that phosphonates are not always used as a P source, but after transport into the cell may be incorporated into phosphonolipids, phosphonoproteins, or other macromolecules by anabolic routes. And conversely, 289 strains that possessed genes associated with a known phosphonate degradation pathway showed no evidence of possessing the PhnCDE route of phosphonate acquisition – suggesting the existence of as-yet unknown uptake routes.

An apparent example of a recent pathway of phosphonate metabolism is encoded by the operon designated as PhnY*/PhnZ and discovered by Martinez et al. (2010) in a clone from a fosmid library from the Sargasso Sea. The enzymes of the pathway have not yet been characterized, but it was experimentally proven that they conferred the capacity to consume phosphonates as a P source on an E. coli phn mutant strain (Martinez et al., 2010). PhnY*is thought to be a 2-oxoglutarate dioxygenase, and PhnZ a protein of the HD phosphohydrolase family (Martinez et al., 2010). Bioinformatic analysis indicated that only five sequenced bacterial strains possessed a copy of this operon. Interestingly, four of these were isolated from the marine environment, suggesting that this pathway may have evolved in the oceans and may now play a significant role in P acquisition in marine microorganisms. Three of the strains (Plesiocystis pacifica SIR-1, and Prochlorococcus marinus str. MIT9301 and MIT9303) possess a copy of the phosphonate transporter operon phnCDE but had been considered to lack any recognized phosphonate degradation pathway. As is the case with the phosphonopyruvate hydrolase pathway of C–P bond cleavage, the PhnY*/PhnZ pathway appears to have undergone no extensive evolutionary rearrangement. In this paper the phnY gene described by Martinez et al. (2010) is designated as phnY*.