Towards a characterization of the structural determinants of specificity in the macrocyclizing thioesterase for deoxyerythronolide B biosynthesis

Article abstract of DOI:10.1016/j.bbagen.2015.11.007

Type I polyketide synthases (PKSs) are giant multidomain proteins that synthesize many therapeutics and other natural products. The synthesis proceeds by a thiotemplate mechanism whereby intermediates are covalently attached to the PKS. The release of the final polyketide is catalyzed by the terminal thioesterase (TE) domain through hydrolysis, transesterification, or macrocyclization. The PKS 6-deoxyerythronolide B synthase (DEBS) produces the 14-membered macrolide core of the clinically important antibiotic erythromycin. The TE domain of DEBS (DEBS TE) has well-established, empirically-defined specificities for hydrolysis or macrocyclization of native and modified substrates. We present efforts towards understanding the structural basis for the specificity of the thioesterase reaction in DEBS TE using a set of novel diphenyl alkylphosphonates, which mimic substrates that are specifically cyclized or hydrolyzed by DEBS TE. We have determined structures of a new construct of DEBS TE alone at 1.7 ?, and DEBS TE bound with a simple allylphosphonate at 2.1 ? resolution. Other, more complex diphenyl alkylphosphonates inhibit DEBS TE, but we were unable to visualize these faithful cyclization analogs in complex with DEBS TE. This work represents a first step towards using DEBS TE complexed with sophisticated substrate analogs to decipher the specificity determinants in this important reaction.

Full text of DOI:10.1016/j.bbagen.2015.11.007

Towards a characterization of the structural determinants of specificity in the
macrocyclizing thioesterase for deoxyerythronolide B biosynthesis
Panos Argyropoulos, Fabien Bergeret, Christophe Pardin, Janice M.
Reimer, Atahualpa Pinto, Christopher N. Boddy, T. Martin Schmeing
PII:
DOI:
Reference:
S0304-4165(15)00324-4
doi: 10.1016/j.bbagen.2015.11.007
BBAGEN 28330
To appear in:
BBA - General Subjects
Received date:
Revised date:
Accepted date:
21 August 2015
13 November 2015
20 November 2015
Please cite this article as: Panos Argyropoulos, Fabien Bergeret, Christophe Pardin, Jan-
ice M. Reimer, Atahualpa Pinto, Christopher N. Boddy, T. Martin Schmeing, Towards
a characterization of the structural determinants of specificity in the macrocyclizing
thioesterase for deoxyerythronolide B biosynthesis, BBA - General Subjects (2015), doi:
10.1016/j.bbagen.2015.11.007
This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.

ACCEPTED MANUSCRIPT
Towards a characterization of the structural determinants of specificity
in the macrocyclizing thioesterase for deoxyerythronolide B
biosynthesis
Panos Argyropoulos,a† Fabien Bergeret,b† Christophe Pardin,a Janice M. Reimer,b
Atahualpa Pinto,a Christopher N. Boddya* and T. Martin Schmeing,b**
aDepartment of Chemistry and Biomolecular Sciences, Centre for Catalysis Research and
Innovation, University of Ottawa, Ottawa, ON K1N 6N5, Canada
bDepartment of Biochemistry, McGill University, Montréal, QC H3G 0B1, Canada
Groupe de Recherche Axé sur la Structure des Protéines (GRASP), McGill University,
Montréal, QC H3G 0B1, Canada
†These authors contributed equally to the work.
* Corresponding author. Tel.: 613-762-8900; fax: 613-562-5170.
** Corresponding author. Tel.: 514-398-2331; fax: 514-398-2983.
Email addresses: cboddy@uottawa.ca (C.N. B.) martin.schmeing@mcgill.ca (T. M. S.)
1

ACCEPTED MANUSCRIPT
Key Words
Thioesterase, polyketide synthase, non-hydrolyzable acyl-enzyme intermediates,
inhibition, high-resolution structure
Highlights
 1.7 Å resolution structure of a new construct of the erythromycin thioesterase
 diphenyl phosphonate inhibitors of the erythromycin thioesterases were
synthesized and assayed
 2.1 Å resolution structure of allylphosphonate adduct of the erythromycin
thioesterase
 Slow maturation of initial phosphonate-enzyme adduct limits use of diphenyl
phosponates esters
Abstract
Type I polyketide synthases (PKSs) are giant multidomain proteins that synthesize many
therapeutics and other natural products. The synthesis proceeds by a thiotemplate
mechanism whereby intermediates are covalently attached to the PKS. Release of the
final polyketide is catalyzed by the terminal thioesterase (TE) domain through hydrolysis,
transesterification, or macrocyclization. The PKS 6-deoxyerythronolide B synthase
(DEBS) produces the 14-membered macrolide core of the clinically important antibiotic
erythromycin. The TE domain of DEBS (DEBS TE) has well-established, empirically-
defined specificities for hydrolysis or macrocyclization of native and modified substrates.
We present efforts towards understanding the structural basis for the specificity of the
thioesterase reaction in DEBS TE using a set of novel diphenyl alkylphosphonates, which
mimic substrates that are specifically cyclized or hydrolyzed by DEBS TE. We have
2

ACCEPTED MANUSCRIPT
determined structures of a new construct of DEBS TE alone at 1.7 Å, and DEBS TE
bound with a simple allylphosphonate at 2.1 Å resolution. Other, more complex diphenyl
alkylphosphonates inhibit DEBS TE, but we were unable to visualize these faithful
cyclization analogs in complex with DEBS TE. This work represents a first step towards
using DEBS TE complexed with sophisticated substrate analogues to decipher the
specificity determinants in this important reaction.
3

ACCEPTED MANUSCRIPT
Introduction
Type I polyketide synthases (PKSs) are giant multimodular proteins found in bacteria and
fungi that synthesize secondary metabolites.[1,2] These polyketides play a crucial role in
the fitness of microorganisms in their environment, impacting growth, colonization,[3]
survival, and virulence.[4,5] Because of their highly selective and potent binding to
specific biological targets, polyketides have been exploited as pharmaceuticals, including
clinical antibiotics (erythromycin) anticancer agents (ixabepilone) and
immunosuppressants (rapamycin).[2,6]
Type I PKSs synthesize polyketides from acyl-CoA building block substrates, using thio-
templated logic, where intermediates are covalently shuttled between active sites attached
to the phosphopantetheine arm of an acyl carrier protein (ACP) domain. Synthesis occurs
in three stages, a loading stage, an elongation-modification stage, and an off-loading
stage.[2] The most common off-loading mechanism in type I PKS biosynthesis is
catalyzed by the thioesterase (TE) domain.[2,7,8] The TE domain uses a two-step
mechanism (Figure 1). In the first step, the polyketide chain, which is linked by a
thioester to the ACP, is transferred to the active site Ser of the catalytic Ser-His-Asp
triad[8,9] of the TE through nucleophilic attack. The second step cleaves the acyl-TE
intermediate, either by an intermolecular attack of an exogenous nucleophile (hydrolysis
or transesterification) or an intramolecular attack by an internal nucleophile
(macrolactamization, macrolactonization, or Claisen-like condensation), to release the
completed polyketide from the PKS.
4

ACCEPTED MANUSCRIPT
Figure 1. Thioesterases, which are serine hydrolases and possess the typical Ser-His-Asp
serine hydrolase catalytic triad, play a key role in off-loading completed polyketide
chains from the acyl carrier proteins of their biosynthetic pathways. This is catalyzed via
a two-step mechanism. In the first step the completed polyketide chain is transferred to
the active site Ser to make covalent complex between the TE and the polyketide chain. In
the second step, nucleophilic attack on the activated ester bond of the acyl-enzyme
intermediate leads to release of the polyketide. This can occur through either
intramolecular nucleophilic attack, giving a macrocycle (arrow in red) or attack by water
(arrow in blue) releasing the free acid. The mechanism controlling partitioning between
macrocyclization and hydrolysis is uncharacterized.
Although the loading step typically has low substrate selectivity,[10] the release step is
often highly selective. The best-characterized TE is from the 6-deoxyerythronolide B
synthase (DEBS TE) (Figure S1). DEBS TE shows high stereo-, regio-, and chemo-
selectivity in the release step. In vivo, it is able to cyclize a number of truncated or
elongated erythronolide analogs forming 6, 8, 12, 14 or 16-membered rings.[11–14] In
vitro, both the absolute stereochemistry of the nucleophilic alcohol and the presence of a
carbonyl in the polyketide backbone are essential for cyclization.[15–18]
Structures of DEBS TE in the absence of ligands[19,20] gave high-resolution views of
this important enzyme. Accompanying docking studies were used to suggest that H-
bonding played a key role in controlling selectivity.[19] However, experimental testing of
5

ACCEPTED MANUSCRIPT
this model called the conclusion into doubt.[21] The related TE domain from the
pikromycin biosynthetic pathway (Pik-TE) has been determined in complex with non-
hydrolyzable acyl analogs derived from diphenyl phosphonate esters.[22,23] The
structures suggested that hydrophobic packing was a key mediator of enzyme-substrate
interactions. However, the non-hydrolyzable acyl-enzyme intermediates used in the study
had limited similarity to substrates known to undergo macrocyclization, thus tempering
the use of these models for identifying discrete enzyme-substrate interactions that control
the off-loading step. Ultimately, it is not yet possible to predict whether a particular
substrate will undergo macrocyclization or hydrolysis.[10,17] Predicting and controlling
the mode of release would be very useful for bioengineering experiments.
Here we present efforts towards characterizing the structural determinants of specificity
of DEBS TE. We have designed a set of substituted diphenyl phosphonates that mimic
substrates known to be cyclized or hydrolyzed by DEBS TE. These phosphonates
covalently modify the active site Ser of DEBS TE and inhibit its function. We have
designed and produced a modified construct of DEBS TE that is fully active and
susceptible to inhibition by diphenyl phosphonates. We determined the structure of this
novel DEBS TE construct at higher resolution (1.7 Å) than those previously available and
also present a high-resolution structure of DEBS TE covalently modified with an
allylphosphonate. Finally, we describe the challenges associated with producing non-
hydrolyzable acyl-enzyme intermediates from complex phosphonates for use in
understanding the selectivity of DEBS TE for macrocyclization versus hydrolysis
activity.
6

ACCEPTED MANUSCRIPT
Results and Discussion
Acyl- and peptidylphosphonates are a class of inhibitors that have been widely used to
form a stable tetrahedral intermediate of serine proteases and TEs.[22–25] The successful
use of diphenyl phosphonates in obtaining high-resolution acyl-enzyme intermediate
structural data with the Pik TE,[22,23] suggested that these reagents would be applicable
to the characterization of the enzyme-substrate interactions that govern substrate
selectivity in DEBS TE. It should be possible to decipher the molecular basis for
intramolecular cyclization by DEBS TE using phosphonate molecules that are highly
analogous to DEBS TE substrates known to undergo either cyclization or
macrocyclization.[17]
A new construct of the TE domain of DEBS
Our initial goal was thus to obtain a higher resolution structure of DEBS TE than
previously published.[19,20] The Khosla and Stroud groups obtained two crystal
structures for the DEBS TE, herein designated by their protein data base (PDB) codes,
TE1KEZ at 2.8 Å (PDB ID: 1KEZ) and TE1MO2 at 3.0 Å resolution (PDB ID: 1MO2). The
N- and C-termini of these two structures were disordered, potentially disturbing the
crystal packing and limiting the quality of diffraction pattern. Based on this observation,
we truncated the N- and C-termini of DEBS TE, to generate a new construct containing
residues S15 to S283 (TE15-283) (Figure S2). This TE construct was overexpressed in
Escherichia coli BL21(DE3) and purified to homogeneity by liquid chromatography with
a high yield of 50 mg protein per liter of culture. This construct is robust and hydrolyzes
7

ACCEPTED MANUSCRIPT
the N-acetyl cysteamine thioester of 3-hydroxyheptanoate with a rate (V/[E]t) of 0.25
min-1 (Figure 2.), maintaining the same activity as the existing DEBS TE construct, which
has turnover numbers between 0.1 - 0.8 min-1 for a wide variety of substrates [15, 17, 21].
In addition, the enzyme could be stored at 37 C for 2 months without becoming inactive
(data not shown).
Figure 2. Inhibitors, model substrate and activity of DEBS TE. Panel A displays the
chemical structures of the diphenyl phosphonates (1-3) used in this study. Panel B shows
8

ACCEPTED MANUSCRIPT
the activity assay used to test the TE domains. Substrate (4) is hydrolyzed by DEBS TE
releasing the free acid and N-acetylcysteamine (HSNAC). Panel C presents the
normalized velocity of TE mediated hydrolysis of 4 under the listed conditions.
We attempted to reproduce the crystallization reported by the Khosla and Stroud
groups,[19,20] but the crystals we obtained in these conditions with TE15-283 did not
diffract well. Therefore, we performed sparse matrix crystallization screening using
sitting-drop vapour diffusion and obtained several new growth conditions producing large
monocrystals. The best conditions were optimized to a precipitant solution of 0.1M Na-
cacodylate pH=6.5, 35-38% PEG-300 and 0.18-0.24M Ca-acetate. These crystals
belonged to space group P 3 2 1, and were phased using molecular replacement with
TE1KEZ as the search model. There is one molecule per asymmetric unit, all residues are
visible in the electron density map, and the model has been refined to an Rfactor of 15.6%
and an Rfree of 18.6% at 1.7 Å (Table 1) (PDB ID: 5D3K). These crystals are robust and
reproducible, and the increased resolution obtained made them ideal for structural
analysis of DEBS TE in complex with substrate analogues.
Overview of the structure of TE
DEBS TE15-283 adopts the classic TE  hydrolase fold and has the catalytic triad of Ser-
142, Asp-169 and His-259 (Figure 3) located in the middle of a long substrate channel, as
previously reported.[20] Although TE15-283 is overall quite similar to TE1KEZ and
TE1MO2(C-RMSD of 1.2 Å and 2.4 Å, respectively), there are some conformational
differences between the three structures in two neighbouring, functionally important
areas.
9

ACCEPTED MANUSCRIPT
Figure 3. The structure of TE15-283 in comparison to TE1KEZ and TE1MO2. Panel A shows
TE15-283 (brown) in cartoon representation with the catalytic residues in sticks. Panel B
shows that the overall conformation is very similar to TE1KEZ (orange) and TE1MO2
(green). Panel C shows that there is a substantial difference in conformation of the lid
element. Panel D shows that at the active site Tyr171 hydrogen bonds to the carbonyl of
Trp242, rather than to the amide of Lys243.
The lid region is a mobile area known to interact with TE domain substrates, and is
usually designated as α-helices 6 and 7 in TE domains. In DEBS TE the lid is between
residues 176 to 193, and is a single -helix between residues 178 and 190 (α-helix 6/7).
The orientation of α-helix 6/7 differs between the three structures, and in TE15-283 it is
longer (residues 176-193) and kinked by ~65 degrees (Figure 3C). pH has been proposed
to cause this difference in orientation, as repulsive interactions between residues that
10

ACCEPTED MANUSCRIPT
become charged would result in changes to the size and the shape of the substrate
channel.[20] The lid in TE15-283 structure (determined at pH 6.5) forms a long and closed
substrate channel, similar to TE1KEZ (pH 7.2), while at more basic pH, the substrate
channel is short and open (TE1MO2, pH 8.5).
The active site His259 is in a loop (residues 256-260) N-terminal to helix 9, and nearby
to the long loop (residues 230-251) between -strands 8 and 9. A hydrogen bond between
the hydroxyl of Tyr171 and the backbone amide of Lys243, reported with TE1KEZ, was
suggested to be important for forming the active site and placing His259 in the correct
orientation for catalysis.[21] At higher pH (TE1MO2, pH 8.5), or in the absence of this
hydroxyl (Y172F mutant), the interaction is lost and His259 is no longer positioned
correctly for catalysis. Accordingly, in TE1MO2, the conformation of residues 232-245 is
completely altered and displaced by the lid, helix 9 unravels, and the loop with His259
is splayed away from the active site. TE15-283 shows an intact active site similar to TE1KEZ
The interaction we observe, however, is the donation of a hydrogen bond from the
hydroxyl of Tyr171 to the backbone carbonyl of Trp242, rather than the acceptance of one
from the backbone amide of Lys243 (Figure 3C). It is unclear whether the difference from
1KEZ is because the new maps allow better placement of residues 242 and 243, or if this
represents a novel conformation.
.
Novel simple phosphonates inhibit DEBS TE activity
To validate the use of diphenyl phosphonates in producing a non-hydrolyzable acyl-
enzyme intermediate with DEBS TE, we synthesized a number of simple diphenyl
11

ACCEPTED MANUSCRIPT
phosphonates. Five different phosphonates were constructed using two different synthetic
approaches: a Grignard or alkyl lithium type addition strategy and an SN2 type strategy
(Figure 4). All phosphonates were obtained at high purity and acceptable yields.
Figure 4. Synthesis of simple diphenyl phosphonates. Panel A shows the general scheme
for the synthesis of simple diphenyl phosphonates via nucleophilic addition to diphenyl
phosphoryl chloride. The four substrates accessed via this approach (3, 5-7) are shown. a)
diphenyl phosphoryl chloride, M-R, ether -78 ºC, 3h, 8-24 %. Panel B shows the
synthesis of the simple phosphonate 10 via nucleophilic attack by diphenyl phosphite
onto an alkyl bromide. b) DBU, MeCN, 0 ºC to 20 ºC, overnight, 26%. Panel C shows
the normalized activity of DEBS TE for hydrolysis of 4 following an 18 h pre-treatment
with diphenyl phosphonates.
Amino phosphonate 10 proved challenging to access. While phosphonates 3, 5-7 were
easily accessed via addition of commercially available Grignard and alkyl-lithium to
12

ACCEPTED MANUSCRIPT
diphenyl phosphoryl chloride, accessing 10 via nucleophilic addition to diphenyl
phosphoryl chloride was unsuccessful. We thus focused on an alkylation route involving
addition of the cesium salt of diphenylphosphite to alkyl halides, as described by
Salvatore et al. [26] While this proved successful and provided sufficient quantities of 10
for completion of the study, the reaction suffered from competing O-alkylation.
As the phosphonates are expected to modify the active site Ser142, treatment of DEBS TE
with the phosphonates should inhibit enzymatic activity. DEBS TE was treated with
phosphonates or a vehicle control for 18 h and subsequently evaluated as to its ability to
hydrolyze the N-acetyl cysteamine thioester of 3-hydroxyheptanoate, 4. The normalized
initial velocities relative to the DMSO control (Figure 4C) show that 18-hour treatment
with most phosphonates inhibit DEBS TE activity. The diphenyl allylphosphonate 3 was
particularly noteworthy, and inhibited DEBS TE by approximately 87 %.
Structure in complex with an diphenyl allylphosphonate
Treatment of TE15-282 with diphenyl allylphosphonate 3, similarly led to a loss of activity
((V/[E]t) =0.03 min-1 , 88 % inhibition) demonstrating that DEBS15-282 TE is also
inhibited by diphenyl phosphonates (Figure 2). Based on this result, soaking experiments
with 3 and TE15-282 were performed and the resulting structure determined at a resolution
of 2.1 Å (Table 1) (PDB ID: 5D3Z). The electron density map unambiguously indicates
that the active site Ser of the TE was modified by 3 (Figure 5). The density indicates that
both phenyl groups are absent, and the allyl chain becomes disordered after the first
carbon. One of the phosphonate oxygens is within hydrogen bonding distance of His259 of
13

ACCEPTED MANUSCRIPT
the catalytic triad, and the other is making water-mediated hydrogen bonds to the
hydroxyl side chain of Thr76 and the backbone amides of Thr76 and Ala77. Otherwise,
there are no other contacts between the allylphosphonate and the enzyme, and no
structural differences between the unliganded and allylphosphonate-bound TE15-282. The
lack of structural rearrangement upon ligand binding is consistent with similar
observations with Pik-TE and the phosphonate biosynthetic pathway,[22,23] but is
seemingly at odds with the induced-fit mechanism for substrate recognition originally
proposed for these enzymes.[20]
Figure 5. Election density shows the allylphosphonate covalently bound to catalytic
Ser142 at the active site of TE15-283
.
Synthesis of complex phosphonates
The success of generating high-resolution structural data from soaking of TE15-283 crystals
with 3 encouraged us to synthesize more complex phosphonate analogues. Pinto et al.
previously identified a set of diastereomeric substrates where the syn diastereomer
14

ACCEPTED MANUSCRIPT
(Figure S3, S1) was macrocyclized by DEBS TE and the anti diasteromer (Figure S3, S3)
was hydrolysed and not macrocyclized by DEBS TE (Figure S3, S4).[17] We thus
expected that the diphenyl phosphonate analogs of these substrates, 1 and 2, would
provide us with the ideal set of compounds to identify the molecular basis for intra-
molecular cyclization.
Phosphonates 1 and 2 were synthesized using a convergent strategy that relied on
coupling of diphenyl 6-aminohexylphosphonate (16) to both the syn and anti
diastereomers of a protected 3,5-dihydroxy carboxylic acid (Figure 6). The two
carboxylic acids were synthesized following similar routes, both involving sequential
asymmetric Brown allylations to set the required stereogenic centers. Both syntheses
began with an in-situ asymmetric Brown allylation of phenylacetaldehyde, 11.[27] The
homoallylic alcohol was then protected as a silyl ether (12) and the olefin was oxidatively
cleaved to give aldehyde 13.[28] A second Brown allylation followed by silylation was
used to install the syn configured 1,3-diol functionality in 14. While 14 was produced
with in-situ preparation of the (–)ipc2allyl borane, the anti substrate (anti-14) was
produced using the commercially obtained (+)ipc2allyl borane followed by silylation.[29]
In our hands the in-situ preparation of (+)ipc2allyl borane led to lower stereoselectivity
with this substrate. The protected diols 14 and anti-14 were subjected to oxidative
cleavage of the olefin followed by Pinnick oxidation [30] to give the two acids 15 and
anti-15.
15

ACCEPTED MANUSCRIPT
Figure 6. The synthesis of phosphonate 1 is shown. Phosphonate 2 was synthesized from
13 using following similar chemistry. a) 1. (-)-(Ipc)2BOMe, allylMgBr, Et2O, 0 ºC, 1h;
then phenylacetaldehyde, Et2O, -100 ºC, 2h; 2. Imidazole, TBSCl, DMF, r.t., overnight,
55% b) K2CO3, K3Fe(CN)6, K2OsO7•(H2O)2, 1:1 t-BuOH:H2O, rt, overnight; then
Pb(OAc)4, DCM, 0.5h 54% c) 1.(-)-(Ipc)2BOMe, allylMgBr, Et2O, 0 ºC, 1h; then
phenylacetaldehyde, Et2O, -100 ºC, 2h; 2. Imidazole, TBSCl, DMF, r.t., overnight, 40%
d) 1. K2CO3, K3Fe(CN)6, K2OsO7•(H2O)2, 1:1 t-BuOH:H2O, rt, 12h; then Pb(OAc)4,
DCM, 0.5h; 2. NaClO2, NaH2PO4, H2O, 2-methyl-2-butene, t-BuOH, r.t., overnight, 63%
e) HCl•Dioxane, r.t., 1h, f) 1. TBTU, DIEA, DMF, r.t., overnight 2. HF, pyridine, MeCN,
r.t., 5h, 28% over 2 steps.
Completion of the synthesis required coupling the carboxylic acids (15, anti-15) with
deprotected amino phosphonate. Deprotection of 7 was effected with HCl·dioxane and
the free amine, 16, was coupled to both 15 and anti-15 using standard peptide coupling
conditions (TBTU, DIEA), producing the fully protected syn and anti-phosphonates.
Removal of the silyl ethers via treatment with HF and pyridine in acetonitrile[17]
produced the final substrates 1 and 2.
1 and 2 both inhibited TE activity. Pretreatment of TE15-283 with 1 and 2 for 18 hours
abolished enzymatic activity for hydrolysis of the N-acetyl cysteamine thioester of 3-
hydroxyheptanoate 4 (Figure 2). This data suggests that the substrates can access the
active site and modify the active site Ser as expected.
16

ACCEPTED MANUSCRIPT
Soaking, co-crystallization, and mass spectrometry experiments with 1 and 2
Based on these encouraging results, the structural characterization of the DEBS TE15-283
in complex with 1 and 2 was initiated. Crystals were soaked in crystallization solutions
supplemented with 5 mM of 1 and 2 individually. Soaking times between 1 and 14 days
were investigated, and X-ray diffraction datasets were collected and processed. Analysis
of the electron density maps did not shown any additional densities in the active site.
We next attempted to form the acyl-enzyme intermediate in solution prior to
crystallization. TE15-283 was incubated with 5 mM 1 and 2 individually, which completely
inhibited TE15-283 activity. The ligand-enzyme complex was purified by gel filtration,
crystallized using the same conditions as for the unmodified protein, and the structure
was determined. Once again, no additional electron density was observed in the active
site, despite the fact that the complex between the protein and the phosphonate was
initially formed in solution.
To evaluate the stability of the complex, TE15-283 was incubated with 1 to achieve full
inhibition. The complex was then exchanged into a buffer containing all the components
present during crystallization, with a reduced concentration of PEG to allow TE15-283 to
remain soluble, and the complex was kept at room temperature for 14 days, which is the
approximate length of time required for crystal growth. Activity assays were then
performed with TE15-283 that had and had not been modified with 1, before and after the
14 day incubation in crystallization-like conditions. Surprisingly, TE15-283 regained
17

ACCEPTED MANUSCRIPT
complete activity after the 14 day incubation in crystallization-like conditions, indicating
that the alkylphosphonate had been lost from the active site Ser (Figure 7). Replacing
individual components of the crystallization-like matrix (buffer: MES pH=6.5, HCl/Tris
pH=8 – Precipitants: CaCl2, Na-acetate, K-acetate, PEG-400, PEG-500 MME) did not
suppress phosphonate hydrolysis.
18

ACCEPTED MANUSCRIPT
Figure 7. DEBS TE loses the phenyl alkylphosphono diester modification over time.
Panel A shows how diphenyl phosphonate inhibitors can proceed through immature
phenyl phosphono diester attached to Ser142, to give either to mature, stable phosphono
monoester attached to Ser142 (though loss of a phenyl) or to an unmodified protein
(through hydrolysis of the phenyl phosphono diester). Panel B shows the hydrolysis
activity of TE15-283 without inhibitor (circles; active for hydrolysis), of TE15-283 treated
with inhibitor, washed, and incubated in acylation buffer for 14 days (squares, largely
inhibited), and of TE15-283 treated with inhibitor in crystallization-like buffer for 14 days
(diamonds, inhibition relieved). A control of TE15-283 that was not modified and was
incubated crystallization-like buffer for 14 days (triangle, active in hydrolysis) is also
shown. Panel C shows mass spectra of intact TE15-283 that was not treated with 1 (red),
TE15-283 that was treated with 1 (green), and TE15-283 that was treated with 1 and incubated
for 7 days in crystallization-like solution (blue). The MS data indicate progression from
unmodified protein (apo-TE15-283) to phenyl phosphono diester-TE TE15-283, back to
unmodified portein. Traces have been scaled to equal the same area under the curve as for
the untreated (red) protein.
The putative mechanism for phosphonate formation on active site Ser occurs via two
steps (Figure 7). The first step involves reversible phosphorylation of the Ser by the
diphenyl phosphonate ester, such as 1, 2, and 3, releasing a single equivalent of
phenol.[31] In the second step, a second equivalent of phenol is released, generating the
mature Ser phosphonomonoester, which is thought to be non-labile.[32,33] Based on this
mechanism paradigm, and the observation of the monophenyl phosphonate ester, we
propose that following has occurred in our experiments: The immature Ser phenyl
phosphonodiester is formed when TE15-283 is treated with 1 or 2. This immature
intermediate is then very slow to mature to the non-hydrolyzable Ser
phosphonomonoester and under crystallization conditions, the faster, competing
hydrolysis reaction leads to the removal of the immature diester to generate free TE15-283
and the monophenyl phosphonate esters of 1 and 2.
19

ACCEPTED MANUSCRIPT
To test whether this competition is a likely reason for the lack of observed electron
density in crystallography experiments with 1 and 2, we performed whole protein mass
spectrometric analyses of unmodified and modified TE15-283 (Figure 7C, S4). Upon
incubation with 1, there is an increase in observed m/z for the protein of ~446 Da, the
mass of the monophenyl phosphonate ester of 1 (C24H33NO5P). After 7 days in
crystallization-like buffer, most of TE15-283 losses the monophenyl phosphonate ester and
returns to the unmodified TE15-283 m/z. No peak corresponding to the m/z of the mature
non-hydrolysable phosphonomonoester is observed at the expected m/z of 29440
(increase of 370 Da to the unmodified TE15-283 mass). This indicates that hydrolysis of the
immature diester is favoured over maturation to the phosphonomonoester of 1 and 2. In
contrast, the diphenyl allylphosphonate derived immature TE15-283 complex appears to
undergo maturation to the non-hydrolyzable phosphonomonoester more rapidly and can
be clearly seen electron density maps (Figure 5). Use of the more reactive
phosphofluoridates may circumvent this limitation of diphenyl phosphonates by more
rapidly accessing the non-hydrolyzable seryl-phosphonate monoester in both simple and
complex inhibitors.[34–36]
Conclusion:
Obtaining high-resolution structural information on the enzyme-substrate interactions that
govern the TE-mediated release of products in polyketide biosynthesis is essential to
understand how these powerful catalysts function. We used phosphonate-based non-
hydrolyzable acyl-enzyme intermediates in an attempt to gain insight into this product
release. Our strategy led to a new construct of the DEBS TE with uncomplexed structure
20

ACCEPTED MANUSCRIPT
determined at much higher resolution than those previously published and a high
resolution structure of a simple allyl-enzyme intermediate. Using multistep stereo-
selective organic synthesis, we were able to obtain the complex phosphonate esters 1 and
2 that were close structural homologs of substrates known to undergo macrocyclization
and hydrolysis. These phosphonate esters effectively inhibited the enzyme, but did not
lead to co-complex structures. We suggest that the maturation of the labile phenyl seryl-
phosphonate diester intermediate into the non-hydrolyzable seryl-phosphonate monoester
product was slower than hydrolysis of the phenyl seryl-phosphonate diester, thus
liberating free enzyme during crystallization. This result indicates that for successful
application of phosphonate esters in characterizing acyl-enzyme intermediates, there is a
delicate balance between maturation of the initial intermediate and its hydrolysis.
Methods
Synthesis of novel inhibitors the TE domain of DEBS
Reactions were carried out under argon or nitrogen atmosphere with dry solvents
and dried glassware under anhydrous conditions unless specified otherwise. Reaction
solvents THF, Et2O and DCM were dried through activated alumina columns. Reagents
were purchased at the highest commercial quality and used without further purification.
Reactions were monitored by thin-layer chromatography (TLC) carried out on 0.25 mm
E. Merck silica gel plates (60F-254) using UV light for visualization and/or ceric
ammonium molybdate (CAM), p-anisaldehyde (PA) and potassium permanganate
(KMnO4) as stating solutions. Flash chromatography was performed with E. Merck silica
21

ACCEPTED MANUSCRIPT
gel (60, particle size 0.040-0.063 mm) and preparative thin-layer chromatography
(PTLC) separations were carried out on 0.25 mm E. Merck silica gel plates (60F-254).
1H NMR, 13C NMR, 31P NMR spectra were recorded on the Bruker DPX-300,
AMX-300, AMX-300II, and AMX-400 spectrometers and calibrated using residual
undeuterated solvents as internal references. Data is reported as follows: chemical shifts
in ppm (δ ); multiplicities are indicated by the following abbreviations: s = singlet, d =
doublet, t = triplet, q = quartet, p = pentet, dd = doublet of doublets, dt = doublet of
triplets, m = multiplet, br = broad; coupling constants (J) in Hz. Infrared spectra (IR)
were collected on a Shimadzu FTIR-8400S in solution and reported in wavenumbers (cm-
1). HRMS were obtained on a Kratos Analytical Concept instrument (University of
Ottawa Mass Spectrum Center).
HPLC data for enantiopurity were collected with an Agilent 1100 serices HPLC
with a DAD detector and Chiralpak IB column. The conditions for the separation were
98:2 Hexane:isopropyl alcohol at a rate of 0.8 mL/min and detection at 220 nm. MALDI-
TOF traces were collected with a Bruker microflex LT with sinnapic acid sandwich.
Spectrophotometric measurements were collected on a Thermo Scientific Evolution 300
UV-Vis spectrophotometer.
General procedure for formation of alkylphosphonates (3, 5-7)
To diphenyl phosphoryl chloride (1 equiv.) dissolved in ether (0.2 M) was added
an alkyl Grignard (2.5 equiv.) at -78 ºC. The reaction was stirred for 3 hours and allowed
to warm to room temperature. The reaction mixture was quenched with saturated
22

ACCEPTED MANUSCRIPT
ammonium chloride (15 mL) and diluted with water. The aqueous layer was then
extracted with EtOAc (3 × 10 mL) and the resulting organic layer washed with brine (2 ×
10 mL), dried with MgSO4 and concentrated in vacuo. The product was purified by flash
column chromatography.
Diphenyl allylphosphonate (3):
The procedure outlined above was employed for the reaction between diphenyl
phosphoryl chloride (0.5 mL, 2.41 mmol, 1 equiv.) and allylmagnesium bromide (6.0
mL, 6.03 mmol, 2.5 equiv) yielding 0.106 g (0.38 mmol, 16%) of diphenyl
allylphosphonate 3. Rf = 0.24 (silica gel, 20% EtOAc/Hex); 1H NMR (400 MHz, CDCl3):
δ 7.32-7.28 (m, 4H), 7.16-7.12 (m, 6H), 6.02-5.94 (m, 1H), 5.34-5.42-5.33 (m, 2H), 3.01
(dt, J = 7.4, 1.4 Hz) ppm; 13C NMR (100 MHz, CDCl3): δ 129.8, 126.3, 125.2, 125.2,
121.4, 120.6, 32.4, 31.0 ppm. NMR spectra are consistent with those previously
reported.[37]
Diphenyl hexylphosphonate (5)
The procedure outlined above was employed for the reaction between diphenyl
phosphoryl chloride (0.5 mL, 2.41 mmol, 1 equiv.) with hexylmagnesium bromide (3.0
mL, 6.03 mmol, 2.5 equiv) yielding 0.143 g (0.46 mmol, 19%) of diphenyl
hexylphosphonate 5. Rf: 0.35 (silica gel, 15% (EtOAc/Hex); 1H NMR (400 MHz,
CDCl3): δ 7.33-7.27 (m, 4H), 7.16-7.12 (m, 6H), 2.11-1.98 (m, 2H), 1.83-1.72 (m, 2H),
1.50-1.18 (m, 9H) ppm; 13C NMR (100 MHz, CDCl3): δ 150.2, 130.0, 124.8, 121.1, 36.0,
23

ACCEPTED MANUSCRIPT
29.2, 26.8, 26.0, 25.8, 25.7, 25.6 ppm. NMR spectra are consistent with those previously
reported.[38]
Diphenyl cyclohexylphosphonate (6)
The procedure outlined above was employed for the reaction between diphenyl
phosphoryl chloride (0.5 mL, 2.41 mmol, 1 equiv.) with cyclohexylmagnesium bromide
(3.0 mL, 6.03 mmol, 2.5 equiv) yielding 0.136 g (0.42 mmol, 15%) of diphenyl
cyclohexylphosphonate 6. Rf: 0.36, (silica gel, 15% (EtOAc/Hex); 1H NMR (400 MHz,
CDCl3): δ 7.31-7.26 (m, 4H), 7.16-7.12 (m, 6H), 2.21-2.09 (m, 1H), 1.92-1.84 (m, 4H),
1.50-1.32 (m, 6H) ppm. 13C NMR (100 MHz, CDCl3): δ 150.4, 129.6, 124.7, 120.6, 36.0,
29.2, 25.8, 25.7, 25.6 ppm. NMR spectra are consistent with those previously
reported.[39]
Diphenyl phenylphosphonate (7)
The procedure outlined above δ (0.5 mL, 2.41 mmol, 1 equiv.) with phenyl
lithium (2.4 mL, 6.03 mmol, 2.5 equiv) yielding 0.105 g (0.34 mmol, 14%) of diphenyl
phenylphosphonate 7. Rf: 0.25, (silica gel, 20% (EtOAc/Hex); 1H NMR (400 MHz,
CDCl3): δ 7.32-7.27 (m, 4H), 7.16-7.12 (m, 6H) ppm; 13C NMR (100 MHz, CDCl3): δ
150.4, 150.4, 133.2, 132.3, 132.2, 129.7, 128.7, 128.6, 125.1, 120.7, 120.6 ppm. NMR
spectra are consistent with those previously reported.[40]
24

ACCEPTED MANUSCRIPT
1-bromo-6-(tert-butoxycarbonylamino)-hexane (9)
To 6-aminohexan-1-ol (0.500 g, 2.82 mmol, 1 equiv.) dissolved in a 1:1
dioxane/water mixture (12 mL) was added K2CO3 (1.870 g, 13.56 mmol, 4 equiv.) and
mixed at room temperature for 10 minutes. Di-tert-butyl dicarbonate (0.7 mL, 3.39
mmol, 1.2 equiv.) was added to the mixture and allowed to mix overnight. The reaction
was diluted with EtOAc (12 mL) and the organic layer separated. The layer was washed
with 10 % HCl (2 × 10 mL), dried over MgSO4 and concentrated in vacuo. The reaction
yield was 0.570 g (2.62 mmol, 93%) and the product was taken onto the next step without
further purification. Rf: 0.50, (silica gel, 60% (EtOAc/Hex); 1H NMR (400 MHz, CDCl3):
δ 4.51 (br s, 1H), 3.61 (t, J = 6.5 Hz, 2H), 3.09 (t, J = 6.7 Hz, 2H), 1.42 (s, 9H), 1.55-1.30
(m, 8H) ppm; 13C NMR (100 MHz, CDCl3): δ 156.1, 79.0, 62.7, 40.4, 32.6, 30.1, 28.4,
26.4, 25.3 ppm. NMR spectra are consistent with those previously reported.[41]
To 6-(tert-butoxycarbonylamino)hexan-1-ol (0.600 g, 2.76 mmol, 1 equiv.) in dry
dichloromethane (14 mL) was added CBr4 (0.920 g, 2.76 mmol, 1 equiv.) and PPh3
(0.720 g, 2.76 mmol, 1 equiv.) at 0 ºC. The resulting mixture was allowed to warm to
room temperature over 2 hours. The reaction was concentrated in vacuo and purified by
column chromatography to yield 0.696 g of 9 (2.48 mmol, 90%). Rf: 0.57, (silica gel,
20% (EtOAc/Hex); 1H NMR (400 MHz, CDCl3): δ 4.48 (br s), 3.66 (t, J = 6.8 Hz, 2H),
3.09 (t, J = 6.5 Hz, 2H), 1.88-1.82 (m, 2H), 1.47 (s, 9H), 1.54-1.18 (m, 6H) ppm; 13C
NMR (100 MHz, CDCl3): δ 156.1, 62.9, 51.1, 32.8, 28.5, 26.4, 25.4 ppm. NMR spectra
are consistent with those previously reported.[42]
25

ACCEPTED MANUSCRIPT
6-(tert-butoxycarbonylamino)-1-diphenoxyphoryl hexane (10)
To diphenylphosphite (0.2 mL, 1.04 mmol, 1 equiv.) in acetonitrile (4 mL) was
added DBU (0.18 mL, 1.15 mmol, 1.1 equiv.) at 0 ºC. This solution was mixed for 10
minutes and 1-bromo-6-(di-tert-butoxycarbonylamino)hexane (0.18 mL, 1.15 mmol, 1.1
equiv.) was added slowly over 10 minutes. The resulting mixture was allowed to warm to
room temperature and stirred overnight. The reaction was diluted with H2O (3 mL),
extracted with EtOAc (3 × 5 mL) and washed with 10% HCl (2 × 5 mL) and brine (2 × 5
mL). The resulting organic layer was dried over MgSO4, concentrated in vacuo and
purified by column chromatography to yield 0.126 g of 10 (0.29 mmol, 28%). Rf: 0.50,
(silica gel, 60% (EtOAc/Hex); 1H NMR (400 MHz, CDCl3): δ 7.32-7.28 (m, 4H), 7.16-
7.13 (m, 6H), 4.48 (s, 1H, -NH), 3.09 (t, J = 6.0 Hz, 2H), 2.11-2.00 (m, 2H), 1.83-1.73
(m, 2H), 1.57-1.42 (m, 3H), 1.36 (s, 9H), 1.35-1.33 (m, 3H) ppm; 13C NMR (100 MHz,
CDCl3): δ 156.0, 150.4, 129.8, 125.1, 120.6, 120.5, 77.23, 30.2, 30.0, 26.5, 26.2, 25.2,
22.3, 22.3 ppm; 31P NMR (121 MHz, CDCl3): δ 25.80 (tt, J = 17.4, 12.5 Hz) ppm. HRMS
(EI) m/z calcd for C23H32NO5P [M+H]+ 434.2096, found 434.1998.
(R)-2-(tert-Butyldimethylsilyloxy)-1-phenylpent-4-ene (12):
To (-)-B-methoxydisopinocamphenylborane (5.0 g, 15.81 mmol, 1.25 equiv.)
dissolved in ether (15.2 mL) was added allylmagnesium bromide (15.2 mL, 15.2 mmol,
1.20 equiv.) and stirred vigorously at 0 ºC for 1 hour. A large amount of white solid
(presumably MgBrOMe) precipitates upon addition of the allylmagnesium bromide. The
26

ACCEPTED MANUSCRIPT
reaction was allowed to warm to room temperature over 2 hours. A solution of
phenylacetylaldehyde (1.5 mL, 12.65 mmol, 1 equiv.) in ether (9.7 mL) and the reaction
are cooled to -100ºC. The phenylacetylaldehyde solution was added dropwise to the
borane solution ensuring that the temperature remains at -100 ºC. The ether/liquid N2
bath was maintained for 1 hour and then allowed to warm to room temperature. A
solution of 15 % NaOH (6 mL) and 30 % H2O2 (9 mL) was premixed and added to the
reaction dropwise at 0 ºC to ensure the vigorous reaction does not overflow. This mixture
was equipped with a water condenser and refluxed for 45 minutes. The reaction was
diluted with EtOAc (12 mL) and the aqueous layer extracted with EtOAc (2 × 20 mL)
The organic layer was washed with NaCl (2 × 15 mL), dried over MgSO4 and
concentrated in vacuo. The product was purified by flash column chromatography (2%
EtOAc/Hexanes) to yield 1.26 g of (R)-1-phenylpent-4-en-2-ol (7.5 mmol, 60%) Chiral
HPLC analysis showed the compound to have 85% ee (See SI for data). Rf: 0.38, (silica
gel, 10% (EtOAc/Hex); 1H NMR (400 MHz, CDCl3): δ 7.33-7.19 (m, 5H), 5.92-5.79 (m,
1H), 5.18-5.14 (m, 1H), 5.12 (t, J = 1.2 Hz, 1H), 3.92-3.83 (m, 1H), 2.84-2.68 (m, 2H),
2.37-2.29 (m, 1H), 2.25-2.16 (m, 1H), 1.62 (br s,1H) ppm; 13C NMR (100 MHz, CDCl3):
δ 138.4, 134.7, 129.4, 128.6, 126.5, 118.2, 71.7, 43.3, 41.2 ppm. NMR spectra are
consistent with those previously reported.[17]
To (R)-1-phenylpent-4-en-2-ol (2.13g, 13.13 mmol, 1 equiv.) dissolved in
anhydrous DMF (13.1 mL) was added imidiazole (1.97 g, 28.89 mmol, 2.2 equiv.) and
TBSCl (2.97 g, 19.70 mmol, 1.5 equiv.) at room temperature and mixed overnight. After
checking by TLC for completion, the reaction mixture was quenched with saturated
27

ACCEPTED MANUSCRIPT
ammonium chloride (15 mL) and diluted with EtOAc. The aqueous layer was then
extracted with EtOAc and the resulting organic layer washed with water whilst increasing
the concentration of brine to 100%. The organic layer was dried with MgSO4 and
concentrated in vacuo. The product was purified by flash column chromatography (2%
EtOAc/Hexanes) to yield 3.34 g of 12 (12.08 mmol, 92%). Rf: 0.22, (silica gel, 10%
(EtOAc/Hex); 1H NMR (400 MHz, CDCl3): δ 7.27-7.14 (m, 5H), 5.88-5.80 (m, 1H),
5.07-5.01 (m, 2H), 3.88-3.84 (m, 1H), 2.78-2.63 (m, 2H), 2.34-2.18 (m, 2H), 0.82 (s, 9H)
-0.08 (s, 3H) -0.25 (s,3H) ppm; 13C NMR (100 MHz, CDCl3): δ 139.3, 135.1, 129.9,
128.1, 126.1, 117.1, 73.5, 43.6, 41.7, 41.7, 26.0, 18.1, -4.8, -5.1 ppm. HRMS (EI) m/z
calcd for C16H26O2Si [M+H]+ 219.1200, found 219.1192.
(S)-3-tert-Butyldimethylsilyl-5-phenylpentan-1-al (13)
To 12 (1.26 g, 4.56 mmol, 1 equiv.) dissolved in anhydrous 1:1 tBuOH:H2O (46
mL) was added K2CO3 (1.89 g, 13.68 mmol, 3 equiv.) and K3Fe(CN)6 (4.50 g, 13.68
mmol, 3 equiv.) at room temperature. The reaction was stirred at room temperature until
most of the solid was dissolved followed by the addition of K2OsO7•(H2O)2 (0.020 g,
0.05 mmol, 0.01 equiv.) in 15% NaOH (1.5 mL), which was mixed overnight. After
checking by TLC for completion, the reaction mixture was diluted with H2O and
extracted with EtOAc (3 × 30 mL). The pooled organic layer was washed with brine,
dried with MgSO4 and concentrated in vacuo. The product was then dissolved in DCM
(18 mL) and Pd(OAc)4 (2.43 g, 5.47 mmol, 1.2 equiv.) and stirred for 1 hour. The
reaction was then concentrated in vacuo, redissolved in EtOAc, silica gel was added, and
the solvent evaporated in vacuo. The dry silica was loaded onto a flash column and the
28

ACCEPTED MANUSCRIPT
product purified by flash column chromatography (5% EtOAc/Hexanes) to yield 0.686 g
of 13 (2.46 mmol, 54%). Rf: 0.45, (silica gel, 10% (EtOAc/Hex); 1H NMR (400 MHz,
CDCl3): δ 9.76 (t, J = 2.6 Hz, 1H), 7.29-7.13 (m, 5H), 4.36 (p, J = 6.2 Hz, 1H), 2.88-2.74
(m, 2H), 2.49 (dd, J = 5.7, 2.5 Hz, 2H), 0.84 (s,9H) -0.01 (s,3H) -0.12 (s, 3H) ppm. 13C
NMR (100 MHz, CDCl3): δ 202.0, 137.8, 129.7, 128.4, 69.6, 50.6, 44.6, 25.8, 18.0, -4.8,
-4.9 ppm; HRMS (EI) m/z calcd for C16H26O2Si [M+H]+ 277.1701, found 277.1610.
(4S,6S)-4,6-Bis(tert-butyldimethylsilyloxy)-7-phenylhept-1-ene (14)
To (-)-B-methoxydisopinocamphenylborane (2.33 g, 7.36 mmol, 1.25 equiv.) dissolved
in ether (7.4 mL) was added allylmagnesium bromide (7.0 mL , 7.0 mmol, 1.20 equiv.)
and the reaction was stirred vigorously at 0 ºC for 1 hour. A large amount of white solid
(presumably MgBrOMe) precipitated upon addition of the allylmagnesium bromide. The
reaction was allowed to warm to room temperature over 2 hours. A solution of 13 (1.64
g, 5.89 mmol, 1 equiv.) in ether (4.5 mL) and the borane reaction were both cooled to -
100 ºC. The aldehyde solution was added dropwise to the borane reaction ensuring that
the temperature remained at -100 ºC. An ether/liquid N2 bath was maintained for 1 hour
and then the reaction was allowed to warm to room temperature. A solution of 15%
NaOH (6 mL) and 30% H2O2 (9 mL) was premixed and added to the reaction dropwise at
0 ºC to ensure the vigorous reaction does not overflow. This mixture was equipped with a
water condenser and refluxed for 45 minutes. The reaction was diluted with EtOAc (12
mL) and the aqueous layer extracted with EtOAc (2 × 20 mL) The organic layer was
washed with NaCl (2 × 15 mL), dried over MgSO4 and concentrated in vacuo. The
product was purified by flash column chromatography (2% EtOAc/Hexanes) to yield
29