Methylphosphonic Acid Biosynthesis and Catabolism in Pelagic Archaea and Bacteria

Article abstract of DOI:10.1016/bs.mie.2018.01.039

Inorganic phosphate is essential for all life forms, yet microbes in marine environments are in near constant deprivation of this important nutrient. Organophosphonic acids can serve as an alternative source of inorganic phosphate if microbes possess the appropriate biochemical pathways that allow cleavage of the stable carbon–phosphorus bond that defines this class of molecule. One prominent source of inorganic phosphate is methylphosphonic acid, which is found as a constituent of marine-dissolved organic matter. The cycle of biosynthesis and catabolism of methylphosphonic acid by marine microbes is the likely source of supersaturating levels of methane in shallow ocean waters. This review provides an overview of the rich biochemistry that has evolved to synthesize methylphosphonic acid and catabolize this molecule into Pi and methane, with an emphasis on the reactions catalyzed by methylphosphonic acid synthase MpnS and the carbon–phosphorus lyase system. The protocols and experiments that are described for MpnS and carbon–phosphorus lyase provide a foundation for studying the structures and mechanisms of these and related enzymes.

Full text of DOI:10.1016/bs.mie.2018.01.039

CHAPTER THIRTEEN
Methylphosphonic Acid
Biosynthesis and Catabolism
in Pelagic Archaea and Bacteria
†
,1
‡,1
Emily C. Ulrich*, Siddhesh S. Kamat , Bjarne Hove-Jensen ,
§
,1
David L. Zechel
*
Carl R. Woese Institute for Genomic Biology, University of Illinois at Urbana—Champaign,
Urbana, IL, United States
Indian Institute of Science Education and Research (IISER), Pune, Maharashtra, India
†
‡
Aarhus University, Aarhus, Denmark
Queen’s University, Kingston, ON, Canada
§
1
Corresponding authors: e-mail address: siddhesh@iiserpune.ac.in; hove@mbg.au.dk;
dlzechel@chem.queensu.ca
Contents
1
. Introduction
352
352
355
362
377
377
377
379
380
381
381
389
397
400
406
409
413
418
419
419
1
1
1
1
.1 The Methane Paradox
.2 Biosynthesis of Methylphosphonic Acid
.3 Catabolism of Methylphosphonic Acid by Carbon–Phosphorus Lyase
.4 Rationale for the Chosen Protocols
2
. Methylphosphonic Acid Synthase
2.1 Heterologous Expression and Purification of MpnS
2.2 Assaying MpnS Activity With 2-Hydroxyethylphosphonic Acid
2.3 Kinetic Characterization of MpnS
3
. Carbon–Phosphorus Lyase Pathway
3
3
3
3
3
3
3
.1 Analysis of CdP Lyase Pathway Reaction Intermediates
.2 Phn(GHIJ) K and Phn(GHIJ) Protein Complexes
.3 PhnI, Nucleosidase
.4 PhnI (PhnG, PhnH, PhnL), Glycosyltransferase
.5 PhnM, Diphosphohydrolase
.6 PhnJ, Carbon–Phosphorus Lyase
.7 PhnP, 5-Phospho-α-D-Ribosyl 1,2-Cyclic Phosphodiesterase
2
2
4
. Summary
Acknowledgments
References
Methods in Enzymology, Volume 605
ISSN 0076-6879
https://doi.org/10.1016/bs.mie.2018.01.039
# 2018 Elsevier Inc.
All rights reserved.
3
51

352
Emily C. Ulrich et al.
Abstract
Inorganic phosphate is essential for all life forms, yet microbes in marine environments
are in near constant deprivation of this important nutrient. Organophosphonic acids
can serve as an alternative source of inorganic phosphate if microbes possess the appro-
priate biochemical pathways that allow cleavage of the stable carbon–phosphorus
bond that defines this class of molecule. One prominent source of inorganic phosphate
is methylphosphonic acid, which is found as a constituent of marine-dissolved organic
matter. The cycle of biosynthesis and catabolism of methylphosphonic acid by marine
microbes is the likely source of supersaturating levels of methane in shallow ocean
waters. This review provides an overview of the rich biochemistry that has evolved
to synthesize methylphosphonic acid and catabolize this molecule into Pi and methane,
with an emphasis on the reactions catalyzed by methylphosphonic acid synthase MpnS
and the carbon–phosphorus lyase system. The protocols and experiments that are
described for MpnS and carbon–phosphorus lyase provide a foundation for studying
the structures and mechanisms of these and related enzymes.
1. INTRODUCTION
1
.1 The Methane Paradox
Methane is a potent greenhouse gas with numerous anthropogenic, biotic,
and abiotic sources. It is also consumed globally though combustion and
microbial oxidation. Because methane plays such a prominent role in atmo-
spheric chemistry, it is necessary to estimate the net flux of this molecule by
accounting for sources and sinks in a global methane budget. Current esti-
mates place the world’s oceans and freshwater lakes as a minor contributor
to global methane emissions, estimated to be 2% of the global total of 0.5
megatones per year (reviewed in depth by Reeburgh) (Reeburgh, 2007).
The distribution of methane in the oceanic water columns has long pres-
ented a scientific mystery. The ocean can be viewed primarily as a vessel
for microbial oxidation of methane, which results in low nanomolar con-
centrations throughout the water column that are either at equilibrium or
undersaturated with respect to the atmosphere. The exception is a spike
in methane concentration to 5nM at a depth (100–150m) that lies just above
the thermocline that separates warmer surface-mixed waters from deeper
and colder waters (Fig. 1; Scranton & Brewer, 1977). This spike, referred
to as the mixed methane layer maximum, has perplexed scientists since
the 1970s, as 5nM methane is supersaturating with respect to the atmo-
sphere, and thus implies net generation of methane in shallow ocean waters.
While methanogenesis by anaerobic bacteria is well known, this pathway

Methylphosphonic Acid Biosynthesis and Catabolism
353
2
0
3
24
25
26
27
s
n moles CH4
l
1.0
2.0
3.0
CH4
O
O
1
2
00
00
HO P CH3
OH
HO P OH
CH4
OH
Pi
1
300
AII-86
Station 2186
s
4
00
Measured [CH4]
Predicted [CH4]
5
00
2
4
6
8
´
10–5 mL STP CH4
I
Fig. 1 Variation in methane concentration (open circles) with ocean depth at a station
approximately 500km East of Puerto Rico (Scranton & Brewer, 1977). Figure adapted with
permission from Elsevier.
requires anoxic conditions that are only available in the sediments of the
ocean floor. The paradox presented by saturating levels of methane in shal-
low, oxygen-rich ocean waters has spurred a decade’s long hunt for a less
oxygen-sensitive pathway for methane production. Only recently does evi-
dence suggest that the “ocean methane paradox” is the product of microor-
ganisms living in the mixed surface waters, which degrade methylphosphonic
acid 1 into methane and inorganic phosphate (Pi) (Karl et al., 2008; Repeta
et al., 2016).
The formation of methane from 1 arises from an elaborate enzymatic
reaction that has evolved to produce Pi from alkylphosphonic acids (Pn)
by cleaving a chemically inert carbon–phosphorus (CdP) bond. While
Pi is a central ion that underpins the biochemistry of all life forms, a shadow
phosphorus biochemistry has evolved in microorganisms that is based on the

354
Emily C. Ulrich et al.
biosynthesis and catabolism of Pn, of which 1 is the simplest example.
While Pn are better known to the public as synthetic compounds, with the
herbicide glyphosate (RoundUp) being the best example, a large number
of Pn also occur as natural products, serving structural and antibiotic roles
for the organisms that produce them. Some well-known examples include
2
-aminoethylphosphonic acid 2, the most common naturally occurring Pn,
fosfomycin 3, a widely used antibiotic, and the herbicide phosphinothricin
(Fig. 2). Within the Pn natural product family 1 is part of a small subset
4
of structurally simple Pn that are known to either serve as building blocks
to create more complex natural products, or to serve as sources of Pi
when this nutrient is limiting (reviewed in Horsman & Zechel, 2017).
Some of these, notably 1 (Metcalf et al., 2012; Sosa et al., 2017), 2,
2
2
-amino-1-hydroxyethylphosphonic acid (Watanabe et al., 2001), and
-hydroxyethylphosphonic acid 8 (Kaya et al., 2006; Sosa et al., 2017; Yu,
Price, Evans, & Metcalf, 2014), are observed to appear as Pn esters with lipids,
polysaccharides, and proteins. Incorporation of Pn into other biomolecules
can occur to surprisingly high levels: up to 60% of the phospholipids in the
cilia of the rumen protozoan Tetrahymena pyriformis have 2 in place of pho-
sphocholine as a head group (Kennedy & Thompson, 1970). Such biomole-
cules comprise the dissolved organic matter of marine environments that
31
sustains microbial growth. Recently, P NMR spectroscopy was used to
directly observe 1, 8, and hydroxymethylphosphonic acid 17 as the most
abundant Pn associated with the polysaccharide fraction of dissolved organic
matter that was isolated from the ALOHA station near Hawaii (Repeta et al.,
3
1
2
016). Similarly, P NMR spectroscopy has shown that the cyanobacterium
Trichodesmium erythraeum IMS101 can synthesize Pn, most likely 1 and 2, and
incorporate these compounds into lipids, proteins, and glycans such that 10%
of the total biomolecule-associated phosphorus is represented by Pn
(Dyhrman, Benitez-Nelson, Orchard, Haley, & Pellechia, 2009). This strat-
egy appears to be widespread in cyanobacteria, which have been shown to use
Pn suchas 2, a near perfect surrogate for phosphoethanolamine, as head groups
for lipids under Pi limiting conditions (Van Mooy et al., 2009). This suggests
Fig. 2 Examples of Pn natural products. Methylphosphonic acid 1,
2-aminoethylphosphonic acid 2, fosfomycin 3, and phosphinothricin 4.

Methylphosphonic Acid Biosynthesis and Catabolism
355
that in addition to servingas a Pi source and a motif forantibiotic biosynthesis,
Pn can serve a third role in reducing the demand for Pi in biosynthesis. It has
been estimated that using Pn to synthesize phosphonolipids can reduce the
total cellular phosphorus demand by 10–40% (Van Mooy et al., 2009).
Therefore, in marine microbes, we find an intriguing biochemical cycle of
Pn synthesis, storage, and catabolism, sometimes all within a single species.
1.2 Biosynthesis of Methylphosphonic Acid
The simple structure of 1 belies a rich biochemistry, largely shared with
other Pn natural products, that is dedicated to its biosynthesis and catabo-
lism. The biosynthesis of 1, like almost all known Pn natural products, begins
with the formation of the CdP bond (Scheme 1). This occurs through
the reversible conversion of phosphoenolpyruvate 5 to phosphonopyruvate
6
by the enzyme phosphoenolpyruvate mutase (PepM) (Seidel, Freeman,
Seto, & Knowles, 1988). Because this reaction is common to almost all
of the discovered natural product Pn, the gene encoding PepM can be
used as a marker to identify Pn biosynthetic gene clusters in genomes,
and even phylogenetically predict Pn structural diversity (Ju et al., 2015;
Ju, Doroghazi, & Metcalf, 2014; Yu et al., 2013). The equilibrium between
5 and 6 is unfavorable due to the greater strength of the OdP bond relative to
the CdP bond (Bowman, McQueney, Barry, & Dunaway-Mariano, 1988);
therefore the next reaction is irreversible, involving the decarboxylation of
6
by the thiamine pyrophosphate-dependent enzyme phosphonopyruvate
decarboxylase (Ppd). The product, phosphonoacetaldehyde 7, is reduced
+
by a NAD /NADH-dependent dehydrogenase (Pdh) to form 2-
hydroxyethylphosphonic acid 8, a key intermediate in the biosynthesis of a
number of Pn (Shao et al., 2008), and the penultimate precursor to 1.
1.2.1 Methylphosphonic Acid Synthase
The CdC bond cleaving step that converts 8 to 1 is a fascinating example
of how evolution can repurpose the reactivity of a common enzyme scaffold.
The enzyme catalyzing this step, methylphosphonic acid synthase (MpnS),
was discovered by scanning the genome of Nitrosopumilus maritimus, a com-
mon marine archaeon, for Pn biosynthetic gene clusters using pepM as a
probe (Metcalf et al., 2012). The gene encoding MpnS was found in asso-
ciation with three genes (pepM, ppd, pdh) necessary to encode the synthesis of
8
. MpnS belongs to a cupin subfamily of mononuclear nonheme Fe(II)-
dependent enzymes that include hydroxypropylphosphonic acid epoxidase
HppE) (Kuzuyama, Seki, Kobayashi, Hidaka, & Seto, 1999; Liu et al.,
(

Scheme 1 Overview of the microbial biosynthesis of methylphosphonic acid 1 followed by its catabolism into phosphate containing metab-
olites and methane by the carbon–phosphorus lyase pathway.

Methylphosphonic Acid Biosynthesis and Catabolism
357
2
001) and 2-hydroxyethylphosphonate (8) dioxygenase (HEPD) (Cicchillo
et al., 2009), which are involved in the biosynthesis of 3 and 4, respectively.
Unlike the better-known α-ketoglutarate-dependent family of nonheme Fe
(II)-dependent dioxygenases, MpnS and HEPD function as oxygenases
without a requirement for α-ketoglutarate as a cosubstrate (Scheme 2A
and B). Both MpnS and HEPD cleave the CdC bond of 8, but the fate
of molecular oxygen differs in each enzyme.
The N. maritimus MpnS reaction is novel in that the CdC bond of 8 is
cleaved to form 1 and CO using only molecular oxygen and Fe(II). As the
2
input of external electrons is not required for CdC bond cleavage, all four
electrons that are used to reduce O to H O are derived from 8. A detailed
2
2
mechanism for this unusual reaction has been proposed based on extensive
studies with substrate analogs, isotopic labeling, site-directed mutagenesis,
and X-ray crystallography (Born et al., 2017; Cooke, Peck, Evans, & van
der Donk, 2012; Metcalf et al., 2012; Peck, Chekan, Ulrich, Nair, & van
der Donk, 2015; Peck et al., 2017; Zhu, Peck, Bonnot, van der Donk, &
Klinman, 2015). The X-ray crystal structure of N. maritimus MpnS revealed
that the Fe(II) ion is coordinated by two histidine residues and one gluta-
mine (Scheme 3A, I) (Born et al., 2017). Such Fe(II) coordination had
Scheme 2 Oxidative reactions catalyzed by methylphosphonic acid synthase (MpnS)
(
A) and the homologous Fe(II)-dependent enzymes 2-hydroxylethylphosphonic acid
dioxygenase (HEPD) (B), and 2-hydroxypropylphosphonic acid epoxidase (HppE) (C).

358
Emily C. Ulrich et al.
Scheme 3 Proposed mechanisms for (A) MpnS and HEPD contrasted with (B) HppE.

Methylphosphonic Acid Biosynthesis and Catabolism
359
not been observed previously: mononuclear nonheme Fe(II) enzymes,
including HEPD (Cicchillo et al., 2009) and HppE (Higgins, Yan, Liu,
Liu, & Drennan, 2005), typically bind Fe(II) with two histidine residues
and either an aspartate or a glutamate residue, called a 2-histidine/
1
-carboxylate facial triad. The substrate 8 is observed to bind to the Fe(II)
nucleus in a bidentate fashion using a phosphoryl oxygen and the
-hydroxyl (Born et al., 2017). Molecular oxygen is proposed to bind the
2
Fe(II) center and be reduced to superoxide (Scheme 3A, I to II) (Cooke
et al., 2012). Abstraction of the pro-S hydrogen of C2 forms a ketyl radical
(II to III), which in turn reduces the Fe(III) ion to Fe(II) and forms phos-
phonoacetaldehyde 7 (III to IV). Attack of the bound hydroperoxide on
the aldehyde with concomitant formation of a ferryl species is proposed
(
IV to V), which could also arise directly from homolytic scission of the per-
oxide bond to hydroxylate the ketyl radical (III directly to V) (Peck et al.,
017). The Fe(IV) oxo species then abstracts the hydroxyl hydrogen to form
an alkoxy radical (V to VI), thereby promoting β-scission of the CdC bond
a common reaction for alkoxy radicals), and forming a methylphosphonate
2
(
radical and formate (VI to VII). Hydrogen atom transfer from formate to the
methylphosphonate radical forms 1, while reduction of Fe(III) by the CO2
radical anion restores the Fe(II) state and produces CO (VII to VIII to I).
2
Strong evidence suggests the methylphosphonate radical formed in VII
(Scheme 3A) is a key mechanistic branch point that distinguishes the reac-
tivity of N. maritimus MpnS and HEPD. While MpnS quenches the
methylphosphonate radical by hydrogen transfer from formate, HEPD
directs the transfer of the Fe(III)-bound hydroxyl to form hydro-
xymethylphosphonic acid 17 and formate as final products (Hirao &
Morokuma, 2010, 2011; Peck et al., 2011; Whitteck et al., 2011). The var-
iation of one of the facial-triad Fe(III) ligands, from glutamate to glutamine,
plays a key role in this divergent reactivity. In the MpnS active site, the
amide NH of the Fe(III) coordinating ligand Gln152 is ideally located to
2
provide a hydrogen bond to formate that is produced upon CdC bond
cleavage (Born et al., 2017). It was proposed that by anchoring formate
in proximity to the methylphosphonate radical, hydrogen transfer from for-
mate to the radical is facilitated and 1 is produced (Scheme 4A). In the case of
Streptomyces viridochromogenes HEPD, the corresponding Glu176 side chain
bound to the Fe(III) ion cannot simultaneously anchor the formate with
a hydrogen bond; thus this product is either released or rotated while bound
to Fe(III), placing the formate hydrogen out of reach. Consequently, the
methylphosphonate radical reacts with the Fe(III)-bound hydroxide

360
Emily C. Ulrich et al.
Scheme 4 The role of Fe(III) coordinating and second sphere amino acids side chains in
directing the fate of a methylphosphonate radical in (A) N. maritimus MpnS,
(
B) S. viridochromogenes HEPD, and (C) S. albus HEPD.
producing 17 (Scheme 4B). It should be noted that a mechanistic variation
of Scheme 4A and B involving transfer of hydride from formate, or hydrox-
ide from Fe(II), to a methylphosphonate carbocation would also satisfy the
current experimental data on MpnS and HEPD (Peck et al., 2017).
The presence of glutamine as a Fe(III) ligand is not sufficient in itself to
confer MpnS activity. This is observed in a homologous enzyme from Strep-
tomyces albus that, despite the presence of Gln153 as a Fe(III) ligand, has
HEPD activity (Born et al., 2017). The structure of S. albus HEPD revealed
an important role for a pair of second-sphere active site residues, Gly184 and
Tyr163, with the former creating a space for Gln153 to release and rotate
away from the Fe(III) ion, and the latter stabilizing this alternate conforma-
tion with a hydrogen bond (Scheme 4C). The mobility of Gln153 was pro-
posed to allow formate to move away from the methylphosphonate radical
and prevent hydrogen atom transfer. In the case of N. maritimus MpnS, the
glycine is replaced by a bulky isoleucine side chain (Ile184), which sterically
prevents Gln152 from rotating away from the Fe(III) ion (Scheme 4A).
Gratifying support for the mobility hypothesis was obtained from the

Methylphosphonic Acid Biosynthesis and Catabolism
361
conversion of S. albus HEPD to an equally active MpnS, with no residual
HEPD activity, by substitution of the two second sphere residues to the
N. maritimus MpnS counterparts (Y163F/G184I) (Born et al., 2017).
These experiments have thus identified a minimal sequence motif, glu-
tamine as a facial-triad ligand for Fe(III), and isoleucine to anchor the glu-
tamine ligand in place, that can accurately identify MpnS gene products
from genome sequences. This led to the discovery and functional annotation
of an MpnS from Pelagibacter ubique (Born et al., 2017). Unlike sequences
encoding HEPD and HppE, which are found in marine and terrestrial bac-
teria, mpnS has only been detected so far in marine microorganisms. Surpris-
ingly, mpnS sequences also appear in the eukaryotic genus Symbiodinium, a
marine dinoflagellate. From Global Ocean Survey data, the mpnS sequence
has been detected in 0.6% of marine microorganisms (Metcalf et al., 2012).
While this may appear to be a low frequency, it is important to note that
some of the genre captured in this value, such as Pelagibacter and
Nitrosopumilus, are among the most commonly found organisms in the
sea. This implies that the biosynthesis of 1 creates a substantial latent pool
of methane.
The HppE reaction is profoundly different than that of HEPD or MpnS,
despite a very similar active site architecture (Higgins et al., 2005). Unlike
these enzymes, HppE does not incorporate molecular oxygen into products
and instead functions as a dehydrogenase. Early studies proposed that HppE
used molecular oxygen to extract two hydrogens and two electrons from
S-2-hydroxypropylphosphonic acid 18 to form the epoxide ring in 3
(Higgins et al., 2005; Liu et al., 2001, 2003). The proposed catalytic cycle
required an external source of two additional electrons to reduce O to
2
water (Yan, Munos, Liu, & Liu, 2006). However, more recently it was dis-
covered that HppE can efficiently use hydrogen peroxide as an oxidant
(
Scheme 2C) (Wang et al., 2013). The proposed mechanism involves the
oxidation of the Fe(II) nucleus by peroxide to form a ferryl species
Scheme 3B, I to II), which proceeds to remove the pro-R, C1 hydrogen
(
and one electron from (S)-18 (Hammerschmidt, 1991; Hammerschmidt &
Kaehlig, 1991), thus restoring the iron nucleus to the Fe(II) state without the
input of external electrons. The reactivity patterns of substrate analogs sup-
port the formation of a carbocation intermediate (IV) (Chang, Dey, et al.,
2013; Chang, Mansoorabadi, & Liu, 2013; Woschek, Wuggenig, Peti, &
Hammerschmidt, 2002). This carbocation is quenched by stereospecific
attack of the adjacent C2 hydroxyl, forming the epoxide ring in 3. When
compared to MpnS and HEPD, HppE is notable for its apparent selectivity

362
Emily C. Ulrich et al.
for H O over O as an oxidant in order to access this reaction manifold. It is
2
2
2
still not fully understood how this selectivity is achieved, or how H O2
2
would be produced in the cell for HppE activity. The reader is encouraged
to see the following reviews by Nair and van der Donk (2011) and Peck and
van der Donk (2017) that include in depth discussions on the structures and
mechanisms of MpnS, HEPD, and HppE.
1
.3 Catabolism of Methylphosphonic Acid by
Carbon–Phosphorus Lyase
1.3.1 Strategies for Cleaving CdP Bonds
Marine microorganisms are chronically starved for Pi. A common strategy
used by microorganisms to survive under low Pi conditions is to derive this
nutrient through catabolism of Pn, hypophosphite, and phosphite. As noted
above, 1 and 2 occur frequently in the membranes, proteins, and glycans of
marine microorganisms, thus providing a source of Pn. However, in order to
extract Pi from Pn, specialized enzyme pathways are required that can cleave
the CdP bond, which is notably resistant to cleavage by thermal, photo-
chemical, and typical hydrolytic enzymes (e.g., phosphatases). Currently
three enzyme mechanisms are known to cleave CdP bonds of Pn: (1)
hydrolytic, as exemplified by the phosphonoacetaldehyde and phospho-
noacetate hydrolases, (2) oxidative, by the nonheme Fe(II)-dependent
oxygenases PhnY* and PhnZ, and (3) radical, by CdP lyase. These path-
ways were recently reviewed (Horsman & Zechel, 2017; Hove-Jensen,
Zechel, & Jochimsen, 2014; Kamat & Raushel, 2013b). The CdP lyase
reaction is notable for two reasons. First, while the hydrolytic and oxidative
enzymes are substrate specific, CdP lyase is distinguished by its ability to
release Pi from a wide structural range of Pn (Frost, Loo, Cordeiro, & Li,
1
987; Parker, Higgins, Hawkes, & Robson, 1999; Wackett, Shames,
Venditti, & Walsh, 1987), as well as organophosphinates (requiring cleavage
of two CdP bonds) (Wackett, Shames, et al., 1987), and phosphite (requir-
ing HdP bond cleavage) (Metcalf & Wanner, 1991). It is also the only
enzyme system known to catabolize 1. Second, CdP bond cleavage does
not occur directly on 1, but instead follows conjugation of 1 to a sugar as
part of a multistep process called the CdP lyase pathway.
1.3.2 The phn Operon
In E. coli, the CdP lyase pathway is encoded by the phn operon and consists
of 14 genes phnCDEFGHIJKLMNOP (Chen, Ye, Zhu, Wanner, & Walsh,
1
990; Metcalf & Wanner, 1991, 1993b, 1993a). The phn operon is found

Methylphosphonic Acid Biosynthesis and Catabolism
363
widely in bacterial genomes, including marine bacteria (Martinez,
Ventouras, Wilson, Karl, & DeLong, 2013). Huang and colleagues have
reviewed the considerable variation that is observed across species in the
structure of the phn operon, which suggests that this sequence has undergone
multiple horizontal gene transfer events, thereby highlighting the utility of
this operon to bacterial survival (Huang, Su, & Xu, 2005). Using the strictly
conserved phnJ sequence as a marker, Quinn and colleagues have identified
23% of sequenced marine bacterial genomes as containing phn operons
(Villarreal-Chiu & Quinn, 2012). Accordingly, such bacteria have the
potential to catabolize 1 into Pi and methane. Analysis of metagenomic
DNA samples acquired by the Global Ocean Survey reveals a similar story,
with 30% of the 83 sampled ocean sites testing positive for the phn operon
(Villarreal-Chiu & Quinn, 2012). The phnJ sequence occurs in the equiv-
alent of 5% of all marine genomes sampled by the Global Ocean Survey. Karl
and colleagues observed a similar frequency of phn genes in metagenomic
samples isolated from the Sargasso Sea (Karl et al., 2008). Such data point
to a widespread microbial potential to produce methane from 1. Indeed,
in a milestone paper, Karl and colleagues demonstrated that seawater deprived
of Pi but supplemented with 1 would produce methane (Karl et al., 2008).
More recently, Repeta and colleagues estimated that daily CdP lyase-
mediated turnover of only 0.2–0.3% of the marine reservoir of 1 would
be sufficient to account for the methane paradox (Repeta et al., 2016).
DeLong and colleagues have exploited the ability of marine bacteria to
survive Pi deprivation to select for genes encoding catabolism of Pn and
other P sources. This has led to the discovery of the oxidative pathway
encoded by phnY*phnZ (Martinez, Tyson, & DeLong, 2010) and the obser-
vation of widespread use of genes encoding enzymes for the acquisition and
oxidation of phosphite (Martinez, Osburne, Sharma, DeLong, & Chisholm,
2012; Polyviou, Hitchcock, Baylay, Moore, & Bibby, 2015). The essential
role of CdP lyase in catabolism of 1 by marine bacteria was demonstrated
when 1 as a Pi source efficiently selected for bacterial species that possessed
phn operons (Martinez et al., 2013). Similarly, when an Escherichia coli phn
deletion strain was used to screen a fosmid library of marine metagenomic
DNA for utilization of 1 as a Pi source, only fosmids with phn operons were
selected (Martinez et al., 2013). However, in marine environments, micro-
organisms are more likely to encounter 1 and other Pn not as free molecules,
but as esters with dissolved organic matter, such as polysaccharides (Repeta
et al., 2016). Addressing this issue, Sosa and colleagues enriched bacteria
from the Sulfitobacter and Oceanospirillaceae families by amending

364
Emily C. Ulrich et al.
seawater with polysaccharide-Pn esters obtained from dissolved organic
matter (Sosa et al., 2017). In such polysaccharides, 20% of the total
P content is represented by Pn (Repeta et al., 2016). The bacterial isolates
produced methane from the polysaccharides, consistent with the action of
CdP lyase on 1, and phn operons were found in 20% of the sequenced
genomes (Sosa et al., 2017). Surprisingly, these strains also produced ethyl-
ene from marine polysaccharide-Pn and free 8, which is at odds with the
radical mechanism of CdP lyase discussed below (Sosa et al., 2017). It is
observed that phn operons are found throughout the water column at the
ALOHA station, but the frequency peaks (1/200 genomes) where methane
reaches supersaturation (125–200m) (Sosa et al., 2017). Because CdP lyase
cannot process esters of Pn, 1 and other Pn must be released from dissolved
organic matter by an enzyme with phosphonate monoesterase activity, and
indeed genes encoding such an enzyme appear to be present in the genomes
of the Sulfitobacter isolates (Sosa et al., 2017). It is also noteworthy that the
phnO sequence, encoding an aminoalkylphosphonate N-acetyltransferase
(Errey & Blanchard, 2006), is absent from marine phn operons that have
been selected this way. PhnO is used in CdP lyase pathways that catabolize
aminoalkylphosphonates such as 2 (Hove-Jensen, McSorley, & Zechel,
2
012); thus its absence is consistent with N-acetylation being a superfluous
reaction for an organism that subsists on 1 as a Pi source.
The phn operon is a member of the phosphate (Pho) regulon and thus its
expression is induced in response to Pi limitation (Wackett, Wanner,
Venditti, & Walsh, 1987) (for reviews on the Pho regulon see Hsieh &
Wanner, 2010; McGrath, Chin, & Quinn, 2013). The detection of environ-
mental Pi levels is performed by seven membrane-associated proteins. Four
of these proteins form the phosphate-specific transporter (Pst), which is the
primary means for transporting Pi into the cell. Pst is an ABC-type mem-
brane transport system consisting of a Pi-specific periplasmic binding protein
PstS, a transmembrane component comprised of PstA and PstC, and an
ATPase component PstB. Pst communicates with a two-component mem-
brane system consisting of the membrane protein PhoR and the response
regulator PhoB that controls expression of genes of the Pho regulon. Com-
munication between Pst and PhoR/PhoB is mediated by an inhibitory pro-
tein PhoU: when Pi levels are low, PhoU ceases to inhibit PhoR/PhoB,
leading to expression of the Pho regulon (Gardner, Johns, Tanner, &
McCleary, 2014). A second level of control of the phn operon may be
exerted by the phnF gene product, which encodes a transcriptional regula-
tor. Although phnF is not strictly conserved in phn operons (Huang et al.,

Methylphosphonic Acid Biosynthesis and Catabolism
365
2
005), this sequence is observed in marine phn operons that encode the
catabolism of 1 (Martinez et al., 2013). The Mycobacterium smegmatis PhnF
binds to a DNA sequence between phnF and phnD and represses the tran-
scription of the Pi transport operon, phnDCE when Pi levels are high
(Gebhard et al., 2014; Gebhard & Cook, 2008). However, the ligand that
releases PhnF from its repressor site is unknown.
In E. coli induction of expression of phn occurs when Pi concentrations
fall below approximately 4μM (Hsieh & Wanner, 2010; Metcalf & Wanner,
1991; Wanner, 1993), a value that exceeds the maximal value found in most
marine environments (Karl, 2014). Using measurements from the ALOHA
station as an example (Karl, 2014), the Pi concentration varies from 50 to
200nM in the eutrophic zone (0–200m depth), then increases to 3μM
for depths greater than 1000m. For this reason, the phn operon is often
actively expressed in marine bacteria, such as the cyanobacterium genus
Trichodesmium (Dyhrman et al., 2006), which is widely found in tropical
and subtropical oceans. In E. coli, the expression of the phn operon can be arti-
ficially rendered constitutive by deleting pstS (Hove-Jensen, Rosenkrantz,
Zechel, & Willemo e¨ s, 2010). This has proven very useful for studying Pn
catabolism by CdP lyase in vivo as Pi does not need to be omitted from
the culture medium (Hove-Jensen et al., 2012; Hove-Jensen, McSorley, &
Zechel, 2011).
1
.3.3 Transport of Methylphosphonic Acid by PhnCDE
The beginning of the CdP lyase pathway for 1 involves transport into the
bacterial cell. This is performed by PhnCDE, which comprises an ABC
transporter where PhnD is a Pn periplasmic binding protein, PhnE is a trans-
membrane protein, and PhnC is an ATPase component (Scheme 1). In
addition to association with the phn operon, the phnCDE sequence has been
observed in microbial genomes either alone (Gebhard, Tran, & Cook, 2006;
Martinez et al., 2012) or in association with gene clusters encoding different
mechanisms of Pi utilization (Martinez et al., 2012). Differences in ligand
specificities are observed between PhnD homologs from different bacterial
species, likely reflecting recruitment of this transporter for different roles.
This is demonstrated by Feingersch and colleagues who used ITC measure-
ments to compare PhnD homologs (Feingersch et al., 2012). E. coli PhnD
can effectively bind to 1 (K ¼18.4μM), 2 (0.1μM), ethylphosphonic acid
D
(
1.4μM), aminomethylphosphonic acid (16.6μM), but not to Pi or phos-
phite. In contrast, the marine bacterial species Prochlorococcus MIT9301 has
two transporters encoded by two separate loci, ptxABCD and phnDCE

366
Emily C. Ulrich et al.
(Martinez et al., 2012). Feingersch demonstrated that Prochlorococcus PtxB
can bind to phosphite (K ¼2.0μM), 1 (0.8μM), and ethylphosphonic acid
D
(2.8μM) but not 2. In contrast, Prochlorococcus PhnD can bind tightly to phos-
phite (0.12μM), 1 (39μM) and Pi (54.9μM), but not ethylphosphonic acid
or 2 (Feingersch et al., 2012).
1.3.4 Ribosylation of Methylphosphonic Acid by PhnI(GHL)
The first committed step in the catabolism of 1 by CdP lyase involves a
glycosyl transfer reaction by PhnI. Evidence of this reaction had been
inferred by Avila, Draths, and Frost (1991) and by Hove-Jensen et al.
(2011) from the observation of α-D-ribosyl-1-alkylphosphonates in cultures
of E. coli grown on Pn, but the phn gene product that was responsible for this
transformation was not known. The breakthrough occurred when Kamat
and Raushel recognized that E. coli PhnI, a protein that had previously been
shown by Jochimsen and colleagues as part of a PhnGHIJK protein complex
(
Jochimsen et al., 2011), had weak sequence similarity to nucleosidases
Kamat, Williams, & Raushel, 2011). Based on this clue, Kamat and Raushel
(
were able to show that PhnI alone could catalyze the hydrolysis of GTP and
ATP to form D-ribose-5-triphosphate 19 and the corresponding base
(Scheme 5) (Kamat, Williams, et al., 2011). Hydrolysis was observed even
in the presence of 1. However, in combination with PhnG, PhnH, and
PhnL, PhnI was shown to catalyze the transfer of ribose to 1 using GTP
or ATP as glycosyl donors, forming 9 with inversion of anomeric stereo-
chemical configuration. Therefore, a minimal set of proteins comprising
Scheme 5 Reaction of the carbon–phosphorus lyase component PhnI with nucleoside
triphosphates either alone or in combination with PhnG, PhnH, and PhnL. The
oxacarbenium ion transition state mimic 21 is a potent inhibitor of PhnI.

Methylphosphonic Acid Biosynthesis and Catabolism
367
PhnG, PhnH, PhnI, and PhnL is essential to dictate the glycosyl acceptor
specificity (H O or 1) of PhnI. The PhnGHIL catalyzed reaction also accepts
2
dATP as a donor for transfer to 1, thereby forming the 2-deoxy analog 20
(Kamat, Williams, Dangott, Chakrabarti, & Raushel, 2013). The
oxocarbenium ion transition state that is common to glycosyltransferase
reactions is also stabilized by PhnI, as shown by the potent inhibition of PhnI
by the transition state analog 21 (K ¼20nM) (Kamat, Burgos, & Raushel,
I
2013) (Scheme 5). It is important to note that the PhnG, PhnH, PhnI, and
PhnL polypeptides are unstable when reconstituted in vitro, which presented
a unique challenge to reconstituting the PhnI reaction. In the experiment
described above, the individual proteins require glutathione-S-transferase
(GST) tags to enhance solubility. However, assembly and catalysis only
occur upon removal of the GST tag in situ by addition of a protease, and
even then, precipitation was observed during the course of the reaction
(Kamat, Williams, et al., 2011).
1
.3.5 Triphosphate Hydrolysis by PhnM
PhnM belongs to the amidohydrolase superfamily (Seibert & Raushel,
005). Unlike the PhnG, PhnH, PhnI, and PhnL polypeptides described
2
in the previous section, PhnM could be produced and purified from
E. coli without a solubility tag. Kamat and Raushel demonstrated that PhnM
catalyzes the hydrolysis of the α–β bond of the 5-triphosphate group of 9,
producing the 5-phosphate derivative 10 (Scheme 1) (Kamat, Williams,
1
2
8
18
et al., 2011). Running the reaction in H O produced 10 with O incor-
porated into the 5-phosphate and not diphosphate; thus PhnM directs the
hydrolytic water molecule toward the α-phosphate of 9. In vivo, the
5-phosphoryl groups of 9 and 10 are sensitive to nonspecific phosphatase
activity. If these intermediates accumulate in cells, one can observe the pro-
duction of the shunt product α-D-ribosyl-1-methylphosphonate. Frost and
colleagues were the first to indirectly observe the PhnI and PhnM reaction
products by isolating α-D-ribosyl-1-ethylphosphonate from an E. coli mutant
3
2
subsisting on P-labeled ethylphosphonic acid (Avila et al., 1991). Likewise,
Hove-Jensen and colleagues observed α-D-ribosyl-1-phosphite and several
31
α-D-ribosyl-1-alkylphosphonates by P NMR spectroscopy in cultures of
E. coli phnP or phnJ deletion mutants, depending on the Pn source used in
the experiment (Hove-Jensen et al., 2011, 2012). These shunt products
31
are readily observed in culture supernatants by P NMR spectroscopy
because the signals appear far downfield (δ 13–30ppm) in the spectra relative
to more common biological phosphate esters and diesters.

368
Emily C. Ulrich et al.
1.3.6 CdP Bond Cleavage by PhnJ
A first clue into the identity of the intermediate that underwent CdP bond
cleavage came from mutagenesis studies in E. coli. Hove-Jensen and col-
31
leagues observed by P NMR spectroscopy that an E. coli phnJ deletion
strain accumulated α-D-ribosyl-1-methylphosphonate when fed 1 as a Pi
source, whereas the phnP deletion strain accumulated this compound plus
α-D-ribosyl-1,2-cyclic phosphate (Hove-Jensen et al., 2011). These com-
pounds correspond to the 5-phosphorylated intermediates 10 and 11 shown
in Scheme 1. Intermediate 10 was proposed to be the substrate for CdP
lyase, with CdP bond cleavage generating 11. The latter was shown to
be the substrate for the phosphodiesterase PhnP, discussed in the next sec-
tion. Kamat and Raushel hypothesized that PhnJ was the enzyme that
cleaved the CdP bond of 10 to form 11 and methane (Kamat, Williams,
et al., 2011). Underpinning this hypothesis was the presence of four con-
served cysteine residues in the PhnJ sequence, which suggested PhnJ might
bind an iron–sulfur ([4Fe–4S]) cofactor. Such enzymes are well known to
perform radical reactions (Broderick, Duffus, Duschene, & Shepard,
2
014; Frey, Hegeman, & Ruzicka, 2008; Wang et al., 2014) and would
explain the radical basis of the CdP lyase reaction that had been inferred
over 20 years earlier by feeding Pn based radical probes to E. coli (Frost
et al., 1987; Shames, Wackett, LaBarge, Kuczkowski, & Walsh, 1987).
Iron–sulfur cluster-dependent enzymes are highly sensitive to oxidation,
which also proved to be the case with PhnJ. Kamat, Raushel, and colleagues
were successful in producing and purifying GST tagged PhnJ from E. coli
under aerobic conditions, yielding a black protein solution that contained
2
mol Fe ion per mol PhnJ (Kamat, Williams, et al., 2011). This was con-
sistent with the presence of an oxidized and nonfunctional [4Fe–4S] cluster.
After equilibrating the enzyme with Fe(II) and Na S under anaerobic con-
2
ditions the PhnJ solution turned brown-red and had an absorbance spectrum
2+
consistent with a cluster in a [4Fe–4S] oxidation state. Upon reduction
with dithionite, PhnJ had absorbance and EPR spectra that reflected a
1
+
[
4Fe–4S] oxidation state (Kamat, Williams, et al., 2013), which is the
active form of the cofactor in radical S-adenosyl-L-methionine (SAM)-
dependent enzymes. Kamat, Raushel, and colleagues demonstrated that
PhnJ converts 10 to 11 in the presence of SAM 22, with coproduction of
0
methane, 5 -deoxyadenosine 24, and L-methionine (Scheme 6) (Kamat,
Williams, et al., 2011). The reaction proceeded only under anaerobic con-
ditions and with the addition of dithionite. This is consistent with PhnJ using

Methylphosphonic Acid Biosynthesis and Catabolism
369
Scheme 6 Proposed mechanism for CdP bond cleavage by the PhnJ component of
0
carbon–phosphorus lyase. Ado, adenosine; dAdo, 5 -deoxyadenosine.
1
+
0
a [4Fe–4S] cluster as a one electron donor to reductively cleave the 5 -C-S
0
bond of SAM, yielding a 5 -deoxyadenosyl radical 23 (Scheme 6, I to II).
Notably, PhnJ was inactive as the GST fusion; therefore this reaction, like
that of PhnI supplemented with PhnG, PhnH, and PhnL described above,
was initiated by removing the GST tag in situ with a protease. Precipitation
of PhnJ and limited turnover suggests that this activity is unstable outside of
the context of the PhnGHIJK protein complex described below. Neverthe-
less, reconstitution of the activity of PhnJ in vitro sets the stage for probing the
mechanism of CdP bond cleavage.

370
Emily C. Ulrich et al.
A surprising aspect of PhnJ is that this enzyme represents a convergent evo-
lutionary approach to forming a [4Fe–4S] cluster. Radical SAM-dependent
enzymes typically use three conserved cysteine side chains to bind the
[4Fe–4S] cluster (Shisler & Broderick, 2012). In contrast, PhnJ contains four
conserved cysteine residues within the sequence CX CX CX C (where X is
2
21
5
any amino acid) that does not match the distinct CX CX C sequence found
2
3
in the radical SAM superfamily of enzymes (Frey et al., 2008). Substitution of
the four cysteines (Cys241, Cys244, Cys266, and Cys272) in PhnJ for alanine
produced inactive PhnJ variants (Kamat, Williams, et al., 2013). However,
PhnJ Cys272Ala could still form the [4Fe–4S] cofactor. Therefore, Cys272
appears to play a key role in catalysis, while the other three cysteines can
be assigned roles in binding the [4Fe–4S] cluster.
A key question in the mechanism of PhnJ is how generation of the
0
5
-deoxyadenosyl radical 23 by the [4Fe–4S] cluster could initiate CdP
bond cleavage. This was probed by determining the origin of the new
0
hydrogen atoms in the reaction products 5 -deoxyadenosine 24 and meth-
ane. Deuterium was not incorporated into 24 when the PhnJ reaction was
performed in D O, indicating that the hydrogen abstracted by 23 to form
2
0
the 5 -methyl group of 24 does not originate from a solvent exchangeable
position, either from within the PhnJ active site or from a substrate mole-
cule. However, radical SAM enzymes are also known to use 23 to generate
active site glycyl radicals, which are less reactive due to captodative stabili-
zation of the radical by the neighboring α-amino and carbonyl groups
(Broderick et al., 2014; Frey et al., 2008). To test this possibility, PhnJ
was produced in E. coli grown in minimal medium supplemented with deu-
2
terated glycine ([2,2- H ]-glycine), thereby replacing all the glycine residues
2
2
with [2,2- H ]-glycine. The reaction with [2,2- H ]-glycine labeled PhnJ
2
2
2
in H O produced 24 with a mass consistent with incorporation of a single
2
0
deuterium atom in the 5 -methyl group. A single deuterium atom was also
2
incorporated into methane by [2,2- H ]-glycine labeled PhnJ, but only with
2
a single turnover in H O; multiple turnovers led to unlabeled methane.
2
Conversely, the reaction of unlabeled PhnJ in D O produced CDH . Over-
2
3
all this indicates that the new hydrogen in methane originates from a PhnJ
glycine residue, but over multiple turnovers the glycine α-carbon hydrogens
are acquired from the bulk solvent. Finally, by incorporating R- and
S-deuterium-labeled glycine into PhnJ, it was determined that under single
turnover conditions the pro-R hydrogen of a glycine residue is abstracted by
2
3, and by inference, the pro-S hydrogen is incorporated into methane.
These experiments demonstrated that PhnJ is a glycyl radical SAM
enzyme. Because E. coli PhnJ has eight glycine residues, the catalytic glycine

Methylphosphonic Acid Biosynthesis and Catabolism
371
was first identified by running the PhnJ reaction for multiple turnovers in
D O, which according to the previous experiments, must exchange the cat-
2
alytic glycine hydrogens with deuterium (Kamat, Williams, et al., 2013).
A single deuterium-labeled peptide was identified by LC–MS/MS analysis
of protease-digested PhnJ, which contained the conserved glycine residue,
Gly32. The corresponding PhnJ Gly32Ala variant was unable to cleave the
CdP bond of 10 to produce methane, but still produced 24. Moreover,
2
running the reaction with PhnJ Gly32Ala labeled with L-[2,3,3,3- H ]-
4
alanine produced deuterium labeled 24, demonstrating that Ala32 in the
PhnJ Gly32Ala variant could still transfer a hydrogen atom to the radical
2
3, but not transfer a hydrogen during CdP bond cleavage of 10 to produce
methane. Overall, these experiments identified Gly32 as the site of glycyl
radical formation in PhnJ.
With a sequence of hydrogen transfer events mapped within the active
site of PhnJ, a mechanism for CdP bond cleavage can now be proposed
(
[
Scheme 6) (Kamat & Raushel, 2015). The PhnJ iron–sulfur cluster in the
4Fe–4S] oxidation state reductively cleaves SAM to form the 5 -
1
+
0
deoxyadenosyl radical 23 and L-methionine (I to II). The catalytic cycle
of PhnJ is initiated when 23 abstracts the pro-R hydrogen of Gly32 to
form a glycyl radical and 24 (II to III). The glycyl radical is then proposed
to abstract a hydrogen from Cys272 to form a thiyl radical (III to IV).
Addition of the thiyl radical to the phosphorus center of 10 is thought
to produce a phosphoranyl radical that promotes β-scission of the CdP
bond (IV to V). The resulting methyl radical is then quenched by abs-
traction of the pro-S hydrogen of Gly32, forming methane (V to VI).
It is important to note that early in vivo studies by Frost and colleagues
revealed this step to be remarkably efficient (Frost et al., 1987); using
cyclopropylmethylphosphonic acid as a “radical clock” substrate, abstrac-
tion of the PhnJ Gly32 pro-S hydrogen by the corresponding cyclo-
propylmethyl radical can be estimated to occur with rate constant of
9
ꢁ1
6
.5ꢀ10 s . This high rate constant would be helpful in preventing inac-
tivation of PhnJ by nonspecific hydrogen abstraction or addition reactions
by the highly reactive methyl radical.
The CdP bond cleavage step (V to VI) predicts the formation of a cova-
lent intermediate VI involving a phosphothioester linkage to Cys272. Turn-
over of this linkage is affected by intramolecular attack of the ribose
2
-hydroxyl on the thioester, leading to α-D-ribosyl-1,2-cyclic phosphate
1
1 (VI to III). This step resets PhnJ for another round of catalysis. Preliminary
evidence for covalent attachment of a phosphothioester intermediate VI to
Cys272 has been observed by trapping PhnJ with a 2-deoxy analog of 10,

372
Emily C. Ulrich et al.
which is unable to turn over due to the missing 2-hydroxyl (Kamat, Williams,
3
1
et al., 2013). A P NMR spectroscopic signal consistent with a pho-
sphothioester linkage (δ 25ppm) was observed on a proteolytically generated
peptide from this sample, but the linkage was too labile to observe as an
adduct by mass spectrometry.
1
.3.7 Rescue of a Dead-End Intermediate by the Phosphodiesterases
PhnP and PhnPP
An interesting feature of the PhnJ reaction is the production of
-phospho-α-D-ribosyl-1,2-cyclic phosphate 11 as a product of CdP bond
5
cleavage. Rather than evolving an ability to activate a water molecule to
hydrolyze the phosphothioester intermediate, thereby directly forming met-
abolically useful α-D-ribosyl-1,5-bisphosphate 12 (Scheme 1), Nature opted
to exploit the greater effective molarity of the adjacent ribose 2-hydroxyl.
While this makes sense in terms of chemical reactivity, it does create a prob-
lem in that 11 is not immediately useful to the cell, and if it is not hydrolyzed
to form 12 the cells will not survive. For this reason, hydrolysis of the 1,2-
cyclic phosphate 11 is performed by a dedicated phosphodiesterase, either
PhnP or PhnPP.
Metcalf and Wanner first identifiedphnP as the last essential gene of the phn
operon through mutagenesis in E. coli (Metcalf & Wanner, 1993a). Later,
32
31
Hove-Jensen and colleagues observed by P radiolabeling and P NMR
spectroscopy that E. coli phnP deletion strains accumulated 10 and 11, indicat-
ing that PhnP played a pathway specific role (Hove-Jensen et al., 2011, 2012).
The phnP sequence encodes a hydrolase from the metal ion-dependent
β-lactamase superfamily. Biochemical studies and X-ray crystallography
revealed that PhnP uses a pair of manganese ions to activate water and the
phosphoryl group to perform hydrolysis of a variety of phosphate diesters,
but not monoesters (He et al., 2011; Podzelinska et al., 2009). In vitro
PhnP will hydrolyze substituted methylphenylphosphates, bis(p-nitrophenyl)
0
0
phosphate, and 2 ,3 -cyclic nucleotides. Despite this broad substrate specific-
ity, PhnP is highly specific toward the regioselective ring opening of the
1
,2-cyclic phosphate of 11 to form 12 (Hove-Jensen et al., 2011). This reac-
tion represented an important link to another CdP lyase component,
PhnN, which had been previously shown by Hove-Jensen and colleagues
to phosphorylate 12 to form 5-α-D-phosphoribosyl-1-pyrophosphate
(
PRPP) 16 (Hove-Jensen, Rosenkrantz, Haldimann, & Wanner, 2003).
In some phn operons, including those of marine bacteria (Martinez et al.,
2
013), the phnP sequence is replaced with a gene that encodes the cyclic

Methylphosphonic Acid Biosynthesis and Catabolism
373
phosphate dihydrolase PhnPP. This enzyme is a member of polymerase
and histidinol phosphatase subgroup of the amidohydrolase superfamily,
and is variously annotated in databases and the literature as DUF1045
(Huang et al., 2005), rcsF (Martinez et al., 2013), or cog0613 (Ghodge,
Cummings, Williams, & Raushel, 2013). Ghodge, Raushel, and colleagues
demonstrated that PhnPP from Eggerthella lenta performed two successive
hydrolysis reactions on 11, first opening the 1,2-cyclic phosphate to form
D-ribose-2,5-bisphosphate 13 (Scheme 1), followed by hydrolysis of the
2
-phosphate to form D-ribose-5-phosphate 14 and Pi (Ghodge et al.,
013). The reaction of PhnPP with 11 is unusual in that it has the opposite
2
regiospecificity relative to PhnP, and thus leads to another metabolic dead
end in the form of 13. The overall reaction is additionally novel in that PhnPP
solves the problem posed by 13 with a specific phosphate monoesterase activ-
ity. Therefore, PhnPP is a rare example of an enzyme that has specific phos-
phate monoesterase and diesterase activities. In addition to releasing Pi
for metabolic use, the product 14 can either enter the pentose phosphate
pathway or be converted to 12 via isomerization by phosphopentose mutase
(DeoB) followed by phosphorylation of the 5-hydroxyl of ribose. The
latter reaction, however, has been ascertained only in archaeal species of
the Thermococcales order (Aono, Sato, Imanaka, & Atomi, 2015).
1.3.8 Synthesis of PRPP by PhnN
The serendipitous discovery of the function of PhnN was the first unambig-
uous clue that CdP lyase was a pathway that fed into general metabolism.
PhnN is an ATP-dependent kinase that converts the PhnP product 12 to
PRPP 16 (Scheme 1). 16 is required by bacterial cells for the synthesis of
+
pyrimidine and purine nucleotides, histidine, tryptophan, and NAD
(
reviewed previously by Hove-Jensen et al., 2017; Jensen, Dandanell,
Hove-Jensen, & Willemoes, 2008). Normally 16 is synthesized by the prs
gene product, PRPP synthase. Hove-Jensen and colleagues attempted to
select for mutations in an E. coli prs deletion strain that would relieve the
+
requirement for supplemental NAD (Hove-Jensen et al., 2003). Mutations
+
were acquired within the pst-phoU operon that suppressed the NAD
requirement. It was previously known that mutations within this sequence,
which control the Pho regulon, can lead to constitutive expression of the phn
operon. This led to the determination that phnN could suppress the require-
+
ment for NAD by E. coli prs, and thus play a redundant role in the cell. This
was the first enzyme activity to be assigned to a phn gene product. As well,
the structural similarity between 16 and α-D-ribosyl-1-ethylphosphonate,

374
Emily C. Ulrich et al.
discovered in E. coli over a decade earlier by Frost and colleagues (Avila et al.,
991), suggested CdP lyase performed catabolism of Pn through a catabolic
pathway involving ribose.
1
1
.3.9 Unknown Functions and the Phn(GHIJ) K Protein Complex
2
The functions of the CdP lyase components PhnG, PhnH, PhnK, and
PhnL are either unknown or opaque. While PhnK and PhnL have sequences
that resemble ATP binding proteins, reactions with this molecule or any
nucleotide have yet to be demonstrated with these individual proteins. As
noted previously, PhnL plays an important role in modifying the ribosyl
acceptor specificity (H O or Pn) of the nucleosidase PhnI together with
2
PhnG, PhnH, and PhnL (Kamat, Williams, et al., 2011). Complicating this
picture is the observation that PhnG, PhnH, and PhnK are components of a
larger, 240kDa multiprotein complex with the composition Phn(GHIJ) K
2
(Jochimsen et al., 2011; Ren et al., 2015; Seweryn et al., 2015). Jochimsen
and colleagues demonstrated that this complex can be synthesized in sol-
uble form in E. coli and purified either through multiple steps of ammonium
sulfate precipitation, ion exchange chromatography, and size exclusion
chromatography, or more conveniently in one step by Ni-NTA chroma-
tography when PhnK has a C-terminal His tag (Jochimsen et al., 2011).
6
Smaller complexes comprised of PhnGI and PhnGHIJ can also be generated
and purified by Ni-NTA chromatography (Jochimsen et al., 2011; Ren
et al., 2015). Notably, expression of phnL, phnM, or any other phn gene
is not required for assembly of Phn(GHIJ) K, nor do other polypeptides
2
copurify with this complex. The Phn(GHIJ) K complex can also be
2
observed in the gas phase when the complex is ionized under native condi-
tions in ESI–MS (Ren et al., 2015). By raising the ionization and collisional
energy of the MS instrument it was possible to observe the loss of each com-
ponent of the Phn(GHIJ) K complex in the order PhnH–PhnI–PhnH–
2
PhnI–PhnG–PhnI. PhnGI was the smallest fragment observed, suggesting
that the overall complex is successively assembled by PhnGI and PhnHJ
heterodimers (Scheme 7).
An X-ray crystal structure of a Phn(GHIJ) protein complex, also des-
2
˚
ignated the CdP lyase core complex, has been determined to 1.7A reso-
lution (Scheme 7) (Seweryn et al., 2015). Overall the complex forms a
two-fold symmetric octamer. The core of the complex is formed by homo-
dimer of PhnI. Each PhnI in turn interacts with a copy of PhnG and PhnJ.
The interaction between PhnI and PhnG is extensive, achieved by a
β-hairpin and a C-terminal helix in PhnG that clamps onto a grove in PhnI.

Methylphosphonic Acid Biosynthesis and Catabolism
375
2
Scheme 7 Overview of the Phn(GHIJ) core-complex (PDB ID: 4XB6) and its proposed
assembly by the heterodimers PhnGI and PhnHJ. PhnH can also form a homodimer in
2+
solution. Putative active sites in PhnI and PhnJ are occupied by Zn (yellow spheres). The
4Fe–FS] coordinating cysteine residues in the PhnJ active site are shown as gray sticks
[
2+
and the catalytic Cys272 is shown as red sticks. Note that a Zn is observed in place of an
intact [4Fe–FS] cluster. The catalytic Gly32 residue in PhnJ is shown as red spheres. Pro-
tein structures were rendered in PyMol (DeLano, 2002).

376
Emily C. Ulrich et al.
PhnI also has long N- and C-terminal helices that are used to bind a copy of
PhnJ. A copy of PhnJ forms a reciprocal interaction with both copies of
PhnI using an α-helical domain that has been labeled the central insertion
domain. Finally, PhnH binds to the complex by forming a heterodimer with
PhnJ, an interaction that is mediated by conserved α-helices in each protein.
Notably, despite sharing little sequence similarity, the closest structural homo-
log to PhnH is PhnJ. The PhnHJ heterodimer is all the more intriguing when
it is considered that PhnH alone will form a stable homodimer in solution
utilizing the same α-helical interaction (Adams et al., 2008).
The location of the PhnI and PhnJ active sites has been tentatively iden-
tified in the Phn(GHIJ) structure (Seweryn et al., 2015). A total of four sites
2
are present, one for each copy of PhnI and PhnJ, each marked by a bound
zinc ion. Because the [4Fe–4S] cluster of PhnJ was not reconstituted anaer-
obically prior to crystallization, a zinc ion is instead observed bound in each
copy of PhnJ by Cys241, Cys244, Cys266, and Cys272. Three of these res-
idues, Cys241, Cys244, and Cys266, are involved in binding the [4Fe–4S]
cluster, while Cys272 is proposed form a covalent reaction intermediate
during CdP bond cleavage (Scheme 6) (Kamat & Raushel, 2015).
Surprisingly, Gly32, which transiently forms a glycyl radical and is pro-
posed to abstract the hydrogen from the thiol side chain of Cys272, is
˚
found over 30A away from this residue. It is clear that a significant con-
formational change must occur to bring Gly32 and Cys272 together
˚
(
within approximately 3A) to allow hydrogen atom transfer to occur. It
is likely that a reactive conformation is achieved only when a functional
4Fe–4S] cluster is formed in the active site, or that the complex is capable
of large conformational changes during catalysis.
[
The proposed active site of the glycosyltransferase PhnI is unusual in that
it is partially formed by PhnJ. The active site appears at the end of a tunnel
that leads to the surface of the Phn(GHIJ) complex, and substrate access may
2
be controlled by the central insertion domain of PhnJ. A zinc ion is observed
bound in this cavity by His328 and His333 of PhnI while a third ligand,
2+
His108 is donated by PhnJ. Because Zn typically prefers a tetrahedral
coordination state, one ligand site is available to form a Lewis acid interac-
2+
tion with a substrate molecule. Substitution of any of the Zn coordinating
histidine side chains for alanine creates an inactive CdP lyase pathway in
vivo, thereby confirming the importance of this site to overall pathway func-
tion (Seweryn et al., 2015). However, what must be reconciled is the obser-
vation that in vitro PhnI requires PhnG, H, and L for catalysis (Scheme 5),
but not PhnJ. Additionally, while PhnI is observed to interact directly with
PhnG in the Phn(GHIJ) protein complex (and can be copurified as a PhnGI
2

Methylphosphonic Acid Biosynthesis and Catabolism
377
complex), no direct interaction is observed with PhnH. Finally, PhnL must
only transiently visit PhnI during catalysis as this protein does not copurify as
part of any complex.
The symmetry of the Phn(GHIJ) complex is disrupted by PhnK. While
2
PhnK could not be crystallized with PhnGHIJ despite copurifying with this
complex, a single copy of PhnK has been observed bound to Phn(GHIJ)2
using negative stain electron microscopy (Seweryn et al., 2015) and single
particle cryoelectron microscopy (Yang, Ren, Raushel, & Zhang, 2016).
PhnK binds to the central insertion domain of PhnJ. While two identical
PhnK binding sites are available on the Phn(GHIJ) complex, only a single
2
copy of PhnK can bind due to steric interference with the second binding
site (Yang et al., 2016). This raises the possibility that PhnK exerts an allosteric
effect and may induce the Phn(GHIJ) protein complex to have half-site reac-
2
tivity (Bernhard & MacQuarrie, 1973), where only one PhnJ and PhnI active
site is available for catalysis in a given cycle. Adding significance and intrigue
to the allosteric hypothesis is the exposure of the PhnJ active site residue
Gly32 to solvent when bound to PhnK (Yang et al., 2016). In the unbound
copy of PhnJ, Gly32 remains buried, and presumably unavailable for catalysis.
1.4 Rationale for the Chosen Protocols
MpnS and CdP lyase are the focus of the following protocols as they rep-
resent unique and key transformations that lead to and from 1. MpnS and
CdP lyase both catalyze novel bond cleavage reactions (CdC and
CdP) that offer opportunities to study the activation of molecular oxygen
by iron, the reactivity of radical intermediates, and the role of multiprotein
complexes in enzyme catalysis. Accordingly, these protocols offer a starting
point to the investigator who is interested in studying these and related
enzymes, and provide greater detail on nuanced, but critical, steps that
are often abbreviated in the literature and assume prior expertise.
2. METHYLPHOSPHONIC ACID SYNTHASE
2
.1 Heterologous Expression and Purification of MpnS
The mpnS gene from N. maritimus has been expressed in E. coli using several
different methods. A typical expression involves induction with 0.5mM
isopropyl-β-D-thiogalactopyranoside (IPTG) and incubation for 5h at
30°C, followed by cooling the cells for 10min before harvesting (Metcalf
et al., 2012). The method outlined below was described previously (Born
et al., 2017).

378
Emily C. Ulrich et al.
2
•
•
•
•
•
•
•
•
•
•
•
.1.1 Equipment
Sterilization facilities
Incubator with thermostat and shaking device
Spectrophotometer for optical density (OD) measurement at 600nm
French pressure cell (Thermo Electron Corporation)
Nickel-nitrilotriacetic acid (Ni-NTA) agarose resin (Qiagen)
Amicon Ultra centrifugal filters (EMD Millipore)
NanoDrop spectrophotometer (Thermo Scientific)
PD10 columns (GE Healthcare)
€
Akta FPLC system (GE Healthcare)
Superdex 200 preparative grade resin (GE Healthcare)
SDS-PAGE apparatus
2.1.2 Materials and Reagents
•
E. coli Rosetta 2 (DE3) pLysS cells for expression (other BL21 strains may
work as well, but these have not been tested)
•
•
•
•
•
•
•
IPTG
LB
Ampicillin
Chloramphenicol
DNase
Lysozyme
50mM HEPES pH 7.5, 200mM KCl, 20mM imidazole, 10% (v/v)
glycerol
•
•
•
50mM HEPES pH 7.5, 200mM KCl, 50mM imidazole, 10% (v/v)
glycerol
50mM HEPES pH 7.5, 200mM KCl, 250mM imidazole, 10% (v/v)
glycerol
50mM HEPES pH 7.5, 200mM KCl, 10% (v/v) glycerol
2.1.3 Protocol
1
. Grow E. coli Rosetta 2 (DE3) pLysS cells (Novagen) harboring the pET-
1
5b-mpnS plasmid in LB medium supplemented with 100 μg/mL ampi-
cillin and 12.5 μg/mL chloramphenicol while shaking (200rpm) at 37°C
until an OD600 of 0.6–0.8 is reached.
. Place the culture flasks in an ice water bath for 10min before addition of
2
IPTG to a final concentration of 0.1mM. Incubate the cultures for
1
0–12h while shaking (200rpm) at 18°C.

Methylphosphonic Acid Biosynthesis and Catabolism
379
3. Harvest the cells by centrifugation, and wash the pellets with 25mM
HEPES pH 7.5 before resuspension in lysis buffer (50mM HEPES
pH 7.5, 200mM KCl, 20mM imidazole, 10% (v/v) glycerol).
4
. Incubate the resuspended cells with lysozyme (1mg/mL) and DNase
(0.01mg/mL) at 4°C for 30min before two passages through a chilled
French pressure cell (Thermo Electron Corporation). Remove the cell
debris by centrifugation at 35,000ꢀg for 1h before protein purification
by Ni-NTA chromatography at 4°C.
5
. Incubate the supernatant with Ni-NTA resin equilibrated with 10 col-
umn volumes of lysis buffer (2.5mL of resin per L of expression culture)
for 30min before collecting the flow-through. Wash the resin twice
with 10 column volumes of wash buffer (50mM HEPES pH 7.5,
200mM KCl, 50mM imidazole, 10% (v/v) glycerol). Elute the protein
with elution buffer (50mM HEPES pH 7.5, 200mM KCl, 250mM
imidazole, 10% (v/v) glycerol).
6
. Pool one column volume fractions based on absorbance at 280nm
measured by a NanoDrop spectrophotometer (Thermo Scientific).
Concentrate the protein to approximately 2mL using an Amicon Ultra
centrifugal filter with a 30kDa molecular weight cut off. Finally, desalt
the protein solution using a PD10 column equilibrated with storage
buffer (50mM HEPES pH 7.5, 200mM KCl, 10% (v/v) glycerol).
. Assess the purity by SDS-PAGE, and measure the concentration by
absorbance at 280nm using a NanoDrop spectrophotometer with a the-
7
ꢁ
1
ꢁ1
oretical extinction coefficient of 64,540M cm calculated using the
SIB ExPASy Bioinformatics Resources Portal (Artimo et al., 2012; Born
et al., 2017). The typical yield is about 18mg of protein per L of E. coli
culture.
8
9
. Store the final N-terminal His fusion protein at ꢁ80°C in aliquots.
6
. Additionally, the enzyme can be purified further by size exclusion chro-
matography using Superdex 200 preparative grade resin (GE Healthcare)
equilibrated with at least two column volumes of storage buffer on an
€
Akta FPLC.
2
.2 Assaying MpnS Activity With 2-Hydroxyethylphosphonic Acid
31
The conversion of 8 to 1 and CO by MpnS can be readily monitored by P
2
NMR spectroscopy. The activity assay for MpnS has been described previ-
ously (Cooke et al., 2012).

380
Emily C. Ulrich et al.
2
•
.2.1 Equipment
NMR spectrometer with a P probe
Anaerobic chamber with an atmosphere of 97% N /3% H (Coy
31
•
2
2
Laboratory Products)
2
•
•
•
•
•
•
•
•
.2.2 Materials and Reagents
Reaction buffer: 50mM HEPES pH 7.5
ammonium iron(II) sulfate hexahydrate
8, synthesized as previously described (Metcalf et al., 2012)
MpnS stock solution
1 as a synthetic standard (Sigma-Aldrich)
®
Chelex 100 sodium form (Sigma-Aldrich)
Amicon Ultra centrifugal filters (EMD Millipore)
D O
2
2.2.3 Protocol
1
. Reconstitute MpnS (100 μM final concentration) in an anaerobic
chamber (Coy Laboratory Products) with an atmosphere of 97%/3%
N /H with 1M equivalent of Fe(NH ) (SO ) in 50mM HEPES
2
2
4 2
4 2
pH 7.5 with incubation for 10min on ice.
2
3
4
. Set up an assay (500 μL) aerobically, which typically contains 10 μM
enzyme and 2mM 8 in 50mM HEPES pH 7.5. Incubate for 2h at room
temperature.
®
. Incubate the reaction mixture with Chelex for 20min while shaking,
and remove the protein using a Millipore centrifuge filter with a 30kDa
molecular weight cut off.
3
1
. Add D O (100 μL, 20% final v/v) before analysis by P NMR spectros-
2
1
copy with H decoupling. The observed chemical shifts of 8 and 1 are
approximately δ 18 and 21–24ppm, respectively. Because Pn chemical
shifts are sensitive to pH around their pK values, product identity is con-
a
firmed by spiking with a synthetic standard of 1. Under these conditions,
the reaction is observed to have run to completion.
2
The rate of oxidation of 8 to 1 and CO by MpnS can be followed by
.3 Kinetic Characterization of MpnS
2
monitoring the consumption of molecular oxygen. Experiments to measure
oxygen consumption are performed on a Clark-type oxygen electrode.
This assay method has been described previously by Cooke and colleagues
(Cooke et al., 2012).