Small-molecules that covalently react with a human prolyl hydroxylase-towards activity modulation and substrate capture

Article abstract of DOI:10.1039/c8cc07706a

We describe covalently binding modulators of the activity of human prolyl hydroxylase domain 2 (PHD2) and studies towards a strategy for photocapture of PHD2 substrates. Reversible active site binding of electrophile bearing compounds enables susbsequent covalent reaction with a lysine residue (K408) in the flexible C-terminal region of PHD2 to give a modified protein that retains catalytic activity.

Full text of DOI:10.1039/c8cc07706a

ChemComm
View Article Online
View Journal
COMMUNICATION
Small-molecules that covalently react with a
human prolyl hydroxylase – towards activity
Cite this:DOI: 10.1039/c8cc07706a
modulation and substrate capture†
Received 26th September 2018,
Accepted 1st November 2018
a
Jacob T. Bush,
Robert K. Lesniak,a Tzu-Lan Yeh,a Roman Belle,b
´
Holger Kramer,a Anthony Tumber,a Rasheduzzaman Chowdhury,
a
Emily Flashman,a Jasmin Mecinovic
*
and Christopher J. Schofield
*
abc
a
DOI: 10.1039/c8cc07706a
´
rsc.li/chemcomm
We describe covalently binding modulators of the activity of human HIF-1a target, whereas vascular endothelial growth factor
prolyl hydroxylase domain 2 (PHD2) and studies towards a strategy (VEGF) is a HIF-2a target. We envisaged covalent modification
for photocapture of PHD2 substrates. Reversible active site binding of a non-catalytically essential PHD component may alter the
of electrophile bearing compounds enables susbsequent covalent oxygen dependence of the PHDs, ultimately enabling safer
reaction with a lysine residue (K408) in the flexible C-terminal region medicines than those working by complete inhibition. Covalent
of PHD2 to give a modified protein that retains catalytic activity.
inhibition may provide a strategy for selective inhibition by
reacting with non-conserved residues. The ability to install a
The hypoxia inducible factor (HIF) prolyl hydroxylases (PHD1-3) reactive functional group near the active site, without inhibition,
play a central role in cellular responses to hypoxia in humans may also enable substrate capture. Here we communicate initial
and other animals.1,2 PHD-catalysed prolyl hydroxylation findings demonstrating the feasibility of selective covalent
of HIF-a isoforms (Pro402, Pro564 in human HIF-1a) signals modification of human PHD2.
for proteasomal degradation, hence hindering formation of
Covalent conjugation of small molecules to proteins is often
transcriptionally active a,b-HIF. When oxygen levels are limiting, achieved by reaction of an electrophile with a nucleophilic
PHD catalysis slows and HIF-a levels rise, so enabling a,b-HIF residue.10–13 PHD2 structures suggest that its active site does
mediated upregulation of a gene array working to alleviate not contain Lys- or Cys-residues; however, Lys-residues are
hypoxic effects.3 The PHDs are Fe(II) and 2-oxoglutarate (2OG) present in a loop (K246, K251) and the C-terminal section
dependent oxygenases. PHD inhibitors are in clinical trials for (K400, K402, K408, K416, K423), both of which are flexible
anaemia treatment because erythropoietin (EPO) is a HIF target and associated with the active site entrance (Fig. 1A).14,15 We
gene. Clinical candidate PHD inhibitors, like most reported 2OG envisioned appropriate functionalisation of a PHD2 inhibitor
oxygenase inhibitors, are 2OG competitors coordinating to the with an electrophile may enable covalent reaction with one of
active site Fe(II).4–7 Since the identification of HIF-a as a PHD these Lys-residues. We proposed that the flexibility of these
substrate, there have been reports of non-HIF PHD substrates, regions might enable the covalently bound inhibitor to associate
though the physiological roles of these are uncertain.8
and dissociate from the active site, causing activity modulation
Inspired by the mode of action of aspirin in modulating, but rather than complete inhibition.
not ablating, cyclooxygenase activity by covalent modification,9
Salicylic and picolinic acid derivatives coupled to glycine are
we are interested in developing molecules that change PHD reversible PHD2 inhibitors.5 Modelling based on structures,
activity rather than blocking catalysis. Such compounds have including of the catalytic domain of PHD2 with N-oxalylglycine
potential to alter PHD substrate selectivity and oxygen depen- (NOG) (Fig. 1A), indicated that an electrophile at the 3, 4 or 5
dence. The former is medicinally relevant because HIF-1a and positions of the salicylic acid aryl ring may be positioned to react
HIF-2a have different gene targets; e.g. anaerobic glycolysis is a with the targeted flexible regions. Initially, two aldehyde deriva-
tives of glycine-coupled salicylic acid, 1 and 2, along with several
a Chemistry Research Laboratory, University of Oxford, 12 Mansfield Road, Oxford,
other analogues, were synthesised (Fig. 1B and Fig. S1, ESI†).
The interaction of the ligands with PHD2181–426 (hereafter
PHD2) was examined by non-denaturing ESI-MS, using cone
voltage variation to distinguish covalent and non-covalent
complexes (non-covalent interactions are typically preserved
at cone voltages of 50–80 V, and dissociated at 200 V).16–18
Aldehydes 1 and 2 were incubated with equimolar PHD2 and
OX1 3TA, UK. E-mail: christopher.schofield@chem.ox.ac.uk
b Institute for Molecules and Materials, Radboud University, Heyendaalseweg 135,
6525 AJ Nijmegen, The Netherlands
c Department of Physics, Chemistry and Pharmacy, University of Southern Denmark,
Campusvej 55, 5230 Odense, Denmark. E-mail: mecinovic@sdu.dk
† Electronic supplementary information (ESI) available: Synthesis, mass spectro-
metry, inhibition assays, mutagenesis. See DOI: 10.1039/c8cc07706a
This journal is ©The Royal Society of Chemistry 2018
Chem. Commun.

View Article Online
Communication
ChemComm
Fig. 1 (A) View from a crystal structure of PHD2 in complex with Mn
(purple, substituting for Fe) docked with a bidentate inhibitor, highlighting
the flexible C-terminal tail (blue) and loop (red). Inset: Close-up active site
view. (B) Salicylic acid derivatives as potential covalent inhibitors of PHD2,
functionalised either with an aldehyde (1 and 2) or a Michael acceptor (3),
and a control, non-covalently binding inhibitor (4).
Fe(II) (37 1C; 5 or 20 min). After 20 min, analysis at 200 V cone
voltage indicated 1/2 react with PHD2 to give a peak with mass
of PHD2 + 1/2 À H2O, suggesting condensation (Fig. 2). Analysis
of 1 after 5 min indicated complete condensation; with 2, reaction
was slower, with an apparent non-covalent adduct predominating
after 5 min, with B30% condensation (Fig. S2, ESI†). No peaks
were observed for addition of more than one molecule of 1/2 to
PHD2, implying site selective crosslinking (Fig. 2B). Without Fe(II),
neither covalent nor non-covalent adducts were observed, implying
covalent conjugation involves active site Fe(II) binding, with sub-
sequent condensation. Consistent with this, preincubation with the
reversible iron-binding inhibitor FG2216 (IC50 o 1 mM),15 followed
by addition of 1/2 did not manifest covalent adducts. No covalent
modification was observed with the 5-formyl derivative (S1) or an
analogue lacking the glycine sidechain (S2) (Fig. S3 and S4, ESI†).
Fig. 2 (A) Scheme showing species formed on PHD2 incubation with 1/2,
and FG2216. (B) Non-denaturing MS spectra indicating 1/2 covalently
crosslink with PHD2 (20 mM, 37 1C, 5 or 20 min, cone voltage 200 V) with
equimolar Fe(II) and 1/2. (C) FG2216 binds to PHD2 covalently modified by
1/2. PHD2 was incubated (37 1C, 20 min) with equimolar Fe(II) and 1/2 to
enable covalent reaction. Subsequently, FG2216 (20 mM) was added,
followed by a further incubation (37 1C, 5 min, cone voltage 50 V).
(D) Evidence of CODD hydroxylation by covalently modified PHD2.
Some apparent covalent modification was observed with a ketone and so limiting hydrolysis). Controls were carried out without
analogue of 1 (S3) and with a pyridinyl analogue (S4), but to a lesser 1/2 and with FG2216 (30 mM) (Fig. 2D). CODD hydroxylation
extent than for 1/2 (Fig. S4, ESI†).
was fastest without any inhibitors, and the activity was com-
When FG221615 (20 mM) was added to the pre-formed pletely inhibited by FG2216. Interestingly, both PHD2ÁFeÁ1 and
covalent adducts PHD2ÁFeÁ1 and PHD2ÁFeÁ2, MS analysis PHD2ÁFeÁ2 were apparently active, producing hydroxylated
indicated formation of a tertiary complex PHD2Á1/2ÁFG2216 CODD slightly more slowly than unmodified PHD2. PHD2ÁFeÁ1
(with Fe(II)), suggesting the covalently-linked inhibitors can was more active than PHD2ÁFeÁ2, consistent with the observation
dissociate from the active site so enabling FG2216 binding, that 1 is more easily displaced by FG2216 (Fig. S5, ESI†).
while maintaining the covalent crosslink (Fig. 2C and Fig. S5,
We next sought to create more stable covalent adducts
ESI†). The observation of a tertiary complex suggested that by incorporation of an alternative electrophile. An analogous
covalently crosslinked PHD2 may bind its substrate(s) and inhibitor with an a,b-unsaturated amide at C-3 of the salicylic
retain catalytic activity. The activity of covalently modified acid ring was synthesised (3); we envisaged this may irreversibly
PHD2 was examined using MALDI-MS to monitor hydroxylation react with a lysine residue by Michael addition (and so rule out
of the HIF-1a C-terminal oxygen dependent degradation domain the possibility of active PHD2 being produced by reversible
(CODD, residues 556–574). Covalently modified PHD2 was covalent reaction). Compound 4, without the amide, was synthe-
prepared with complete (495%) crosslinking as confirmed by sised as a negative control (Fig. S6, ESI†). 3/4 were incubated
MS and added to a solution of 2OG, ascorbate, CODD peptide with equimolar PHD2/Fe(II) and analysed by non-denaturing MS.
and 1/2 (to maintain concentration of the inhibitor on dilution After 5 min, spectra obtained at 50 V indicated the presence of
Chem. Commun.
This journal is ©The Royal Society of Chemistry 2018

View Article Online
ChemComm
Communication
PHD2ÁFeÁ3 and PHD2ÁFeÁ4, which were assigned as non-covalent
adducts by analysis at 200 V (Fig. S6, ESI†). Incubation for 1 and
2 h manifested B1/3 and 2/3 conversion, respectively, to the
PHD2ÁFeÁ3 covalent adduct. No covalent adduct was observed
with 4 after up to 2 h, supporting the proposed crosslinking to 3
by addition to the a,b-unsaturated amide (Fig. S6, ESI†). The
slower rate of formation of the covalent adduct with 3 compared
with aldehydes 1/2 may be due to the normally lower reactivity of
a,b-unsaturated amides compared to aldehydes with amines.
In contrast to the likely imine adducts formed with 1/2, the
PHD2ÁFeÁ3 covalent adduct was stable to the acidic conditions
of denaturing MS. This method was used for a timecourse
study (Fig. S7, ESI†); near complete conversion to crosslinked
PHD2ÁFeÁ3 occurred after 3 h (37 1C). The crosslinking could
be fitted by a one-site decay function (R2 = 0.99, GraphPad
Prism), consistent with a quasi intramolecular reaction of
the PHD2ÁFeÁ3 non-covalent species to give a covalent adduct
(Fig. S5, ESI†). The reaction rate was slower at lower tempera-
tures, with little conversion at 4 1C (Fig. S8, ESI†).
To investigate whether the covalent PHD2ÁFeÁ3 adduct is
Fig. 3 (A) Non-denaturing ESI-MS data on the flexible loop truncated
(top) and C-terminal tail tuncated (bottom) versions of PHD2 in the
active, PHD2 and 3 were pre-incubated (0, 1, 2 and 4 h) prior to
assays. The extent of crosslinking was determined to be 0%, 50%, presence of equimolar Fe(II) and 3 (3 hour incubation, 37 1C. Cone voltage:
200 V). A = [PHD2D + Fe], B = [PHD2D + Fe + 3], C = [PHD2181–402 + Fe],
D = [PHD2181–402 + Fe + 3]. (B) MS/MS spectrum of the trypsin derived
peptide 403YLTGEKGVR411 crosslinked to 3 identified K408 as the site
of crosslinking.
66% and 90%, respectively (Fig. S9, ESI†). In controls lacking 3
during pre-incubation, resultant PHD2 activity decreased to the
extent that fully inactive protein was manifested after 4 h pre-
incubation. By contrast, the activity of PHD2 that was pre-
incubated with 3, remained as active as the non pre-incubated,
uninhibited PHD2 for all pre-incubation time points, despite the R411M implied that R411 is not directly involved in cross-
fact that after 4 h pre-incubation with 3 the PHD2 was B90% linking (Fig. S11, ESI†).
covalently modified. This observation suggests that the covalently-
linked ligand in PHD2ÁFeÁ3 reversibly (or does not) associates the and LC-MS/MS. The imine adducts formed by 1/2 were fixed by
active site, and can leave to allow the CODD substrate to bind. reduction using sodium cyanoborohydride (Fig. S12, ESI†).
The site of crosslinking was investigated by trypsin digest
PHD2 inhibition by 1, 2 and 3 was measured using an Modified protein samples were digested (trypsin) and analysed
antibody based assay to obtain IC50 values (Table S1, ESI†); by ‘nano LC-MS/MS’.19 In all 3 samples, modification by
all were weak inhibitors (IC50 4 100 mM), with 3 not reaching the respective inhibitor was identified on the trypsin-derived
50% inhibition at up to 2 mM. Interestingly, the IC50 values are peptide 403YLTGEKGVR411, and localised to K408 (Fig. 3B and
410 fold higher than the 30 mM used for covalent conjugation. Fig. S13, ESI†). This observation is supported by experiments
IC50 values for 1 and 2 were found to drop by B2 fold on with PHD2 K408I that did not undergo apparent covalent
preincubation, possibly reflecting inhibition of non-covalently modification (o10%) with 3 (Fig. S11, ESI†). Consistent with
versus covalently modified PHD2.
these results, the human PHD isoforms PHD1 and PHD3, which
The site of crosslinking of 3 was investigated using protein possess Gly and Arg residues at the analogous position to K408
variants. First, 3 was incubated with C-terminal truncated in PHD2, did not covalently react with 3 (Fig. S14, ESI†).
PHD2 (PHD2181–402) and a loop deletion variant (deletion of
Cellular PHD inhibition in cells increases HIF-a levels, so
residues 238–250, PHD2D). Crosslinking was maintained with mimicking the hypoxic response. The effects of 1, 2 and 3 on
the loop truncate, but near completely absent in C-terminally HIF-1a regulation in HeLa cells were studied, but no upregula-
truncated PHD2, suggesting a Lys-residue (K408, K416 or K423) tion of HIF-1a was observed (Fig. S15, ESI†), possibly reflecting
in the C-terminal region is involved in crosslinking (Fig. 3A). poor permeability of these carboxylic acids. Ester derivatives 5
Thus, 3 was incubated with PHD2 constructs truncated at and 6 were synthesised and caused upregulation of HIF-1a
residues 402, 410, 414 and 418. Both the 418 (PHD2181–418
)
(Fig. S16, ESI†). All tested doses caused a small upregulation of
and 414 (PHD2181–414) truncated proteins were covalently mod- HIF-1a, which was apparently independent of tested concentra-
ified by 3, suggesting K423 and K416 are not involved in tions, and much smaller than observed for FG2216. This HIF-1a
crosslinking (Fig. S10, ESI†). Crosslinking did not occur with induction may be due formation of covalently modified PHD2,
the 410 truncate; however, there are no Lys-residues between which as described above is less active than WT PHD2 (covalent
411 and 414, so it is possible K408 is involved and that the PHD2 modification of PHD2 may also promote its degradation).
truncation at 410 alters the conformation such that crosslinking High concentrations of 5 (4250 mM) induced more pro-
does not occur. Crosslinking experiments with PHD2 nounced HIF-1a upregulation, which may involve inhibition
This journal is ©The Royal Society of Chemistry 2018
Chem. Commun.

View Article Online
Communication
ChemComm
2OG oxygenases, by covalent reaction of an enzyme region
involved in substrate binding. Azide functionalisation of
one of the covalently reacting molecules enabled subsequent
conjugation with biotin, offering application in activity-based
protein profiling. This azide was also found to be photoreactive,
suggesting a strategy for covalent capture of PHD2 substrates.
Our work was supported by the Newton-Abraham Fund
(J. M.), the British Heart Foundation (R. K. L.), the Wellcome Trust
(WT 106244/Z/14/Z) and Cancer Research UK (C8717/A18245).
Conflicts of interest
Fig. 4 Schematic representation of proposed applications of azide 7 for
activity-based protein profiling and substrate capture. Left: Modification of
PHD2 by 7 after 2.5 h incubation, middle: subsequent strain promoted
cycloaddition reaction enables conjugation of the protein with a biotinyl
cyclooctyne 8, right: UV irradiation of PHD2 + 7 leads to loss of nitrogen
(N2), indicating modified protein is photoreactive.
There are no conflicts to declare.
Notes and references
1 C. J. Schofield and P. J. Ratcliffe, Nat. Rev. Mol. Cell Biol., 2004, 5, 343–354.
2 C. J. Schofield and P. J. Ratcliffe, Biochem. Biophys. Res. Commun.,
2005, 338, 617–626.
3 W. G. Kaelin Jr., Annu. Rev. Biochem., 2005, 74, 115–128.
4 M. C. Chan, J. P. Holt-Martyn, C. J. Schofield and P. J. Ratcliffe, Mol.
Aspects Med., 2016, 47–48, 54–75.
5 N. R. Rose, M. A. McDonough, O. N. F. King, A. Kawamura and
C. J. Schofield, Chem. Soc. Rev., 2011, 40, 4364–4397.
6 S. E. Wilkins, M. I. Abboud, R. L. Hancock and C. J. Schofield,
ChemMedChem, 2016, 11, 773–786.
7 T.-L. Yeh, Thomas M. Leissing, M. I. Abboud, C. C. Thinnes,
O. Atasoylu, J. P. Holt-Martyn, D. Zhang, A. Tumber, K. Lippl, C. T.
Lohans, I. K. H. Leung, H. Morcrette, I. J. Clifton, T. D. W. Claridge,
A. Kawamura, E. Flashman, X. Lu, P. J. Ratcliffe, R. Chowdhury,
C. W. Pugh and C. J. Schofield, Chem. Sci., 2017, 8, 7651–7668.
8 B. W. Wong, A. Kuchnio, U. Bruning and P. Carmeliet, Trends
Biochem. Sci., 2013, 38, 3–11.
of (covalently modified) PHD2 by the free 1, consistent with IC50
values for inhibition of isolated PHD2ÁFeÁ1 by 1 (Table S1, ESI†).
Ligands that covalently modify proteins are used for activity-
based profiling by functionalisation with an appropriate handle.20
The observation that covalently modified PHD2 is active suggests
an additional application, i.e. incorporation of a photoreactive
group in the ligand to photocrosslink with substrates, which
could then be isolated and identified. This method for incorpora-
tion of a photoreactive group near the active site of a protein to
enable substrate capture is complementary to ‘photoreactive site
directed mutagenesis’.21,22 Azide 7 was designed for use as a tag to
9 J. R. Vane and R. M. Botting, Thromb. Res., 2003, 110, 255–258.
allow ‘click’ conjugation with a handle (e.g. biotin) for activity- 10 T. A. Baillie, Angew. Chem., Int. Ed., 2016, 55, 13408–13421.
11 R. A. Bauer, Drug Discovery Today, 2015, 20, 1061–1073.
12 R. Lonsdale and R. A. Ward, Chem. Soc. Rev., 2018, 47, 3816–3830.
13 S. E. Dalton, L. Dittus, D. A. Thomas, M. A. Convery, J. Nunes,
based profiling, or as a photoreactive aryl azide for substrate
capture. 7 covalently modified PHD2 at a similar rate to 3 (Fig. 4).
To explore the potential of 7 in protein profiling, biotinyl cyclo-
octyne (8) was added to the modified protein, causing near
complete conversion of covalent PHD2ÁFeÁ7 to biotinylated
PHD2ÁFeÁ7.8 (Fig. 4). The photoreactivity of the covalently modified
protein was investigated by irradiation of ‘covalent’ PHD2ÁFeÁ7
with UV light. MS analysis indicated near complete conversion
of the adduct to a species with a mass of 28 Da less, implying
nitrogen loss and insertion of the resulting nitrene into a
protein bond (Fig. 4). There is scope for optimisaiton and
broader application of the approach described here. We found
7 inhibits a subclass of Fe/2OG histone demethylases, albeit
to a lesser extent than the established inhibitors IOX1 and
J. T. Bush, J. P. Evans, T. Werner, M. Bantscheff, J. A. Murphy and
S. Campos, J. Am. Chem. Soc., 2018, 140, 932–939.
14 R. Chowdhury, M. A. McDonough, J. Mecinovic, C. Loenarz,
´
E. Flashman, K. S. Hewitson, C. Domene and C. J. Schofield,
Structure, 2009, 17, 981–989.
15 M. A. McDonough, V. Li, E. Flashman, R. Chowdhury, C. Mohr, B. M. R.
´
Lienard, J. Zondlo, N. J. Oldham, I. J. Clifton, J. Lewis, L. A. McNeill,
R. J. M. Kurzeja, K. S. Hewitson, E. Yang, S. Jordan, R. S. Syed and
C. J. Schofield, Proc. Natl. Acad. Sci. U. S. A., 2006, 103, 9814–9819.
´
16 J. Mecinovic, R. Chowdhury, E. Flashman and C. J. Schofield, Anal.
Biochem., 2009, 393, 215–221.
´
´
17 J. Mecinovic, R. Chowdhury, B. M. R. Lienard, E. Flashman,
M. R. G. Buck, N. J. Oldham and C. J. Schofield, ChemMedChem,
2008, 3, 569–572.
´
18 J. Mecinovic, C. Loenarz, R. Chowdhury and C. J. Schofield, Bioorg.
Med. Chem. Lett., 2009, 19, 6192–6195.
2,4-pyridinedicarboxylic acid (Table S2, ESI†). Notably, we did 19 S. Tyanova, T. Temu and J. Cox, Nat. Protoc., 2016, 11, 2301–2319.
20 B. F. Cravatt, A. T. Wright and J. W. Kozarich, Annu. Rev. Biochem.,
2008, 77, 383–414.
21 D. Rotili, M. Altun, R. B. Hamed, C. Loenarz, A. Thalhammer,
not accrue evidence that the human histone demethylase
KDM4A undergoes covalent reaction with 3 (Fig. S14, ESI†).
Work toward the application of 7 and derivatives in protein
profiling and substrate capture in cells is ongoing.
Overall the results reveal the potential for the modulation
(either inhibition or activation) of PHD2, and more generally
R. J. Hopkinson, Y.-M. Tian, P. J. Ratcliffe, A. Mai, B. M. Kessler
and C. J. Schofield, Chem. Commun., 2011, 47, 1488–1490.
22 D. Rotili, M. Altun, A. Kawamura, A. Wolf, R. Fischer, I. K. H. Leung,
M. M. Mackeen, Y.-M. Tian, P. J. Ratcliffe, A. Mai, B. M. Kessler and
C. J. Schofield, Chem. Biol., 2011, 18, 642–654.
Chem. Commun.
This journal is ©The Royal Society of Chemistry 2018