Human riboflavin kinase: Species-specific traits in the biosynthesis of the FMN cofactor

Article abstract of DOI:10.1096/fj.202000566R

Human riboflavin kinase (HsRFK) catalyzes vitamin B₂ (riboflavin) phosphorylation to flavin mononucleotide (FMN), obligatory step in flavin cofactor synthesis. HsRFK expression is related to protection from oxidative stress, amyloid-β toxicity, and some malignant cancers progression. Its downregulation alters expression profiles of clock-controlled metabolic-genes and destroys flavins protection on stroke treatments, while its activity reduction links to protein-energy malnutrition and thyroid hormones decrease. We explored specific features of the mechanisms underlying the regulation of HsRFK activity, showing that both reaction products regulate it through competitive inhibition. Fast-kinetic studies show that despite HsRFK binds faster and preferably the reaction substrates, the complex holding both products is kinetically most stable. An intricate ligand binding landscape with all combinations of substrates/products competing with the catalytic complex and exhibiting moderate cooperativity is also presented. These data might contribute to better understanding the molecular bases of pathologies coursing with aberrant HsRFK availability, and envisage that interaction with its client-apoproteins might favor FMN release. Finally, HsRFK parameters differ from those of the so far evaluated bacterial counterparts, reinforcing the idea of species-specific mechanisms in RFK catalysis. These observations support HsRFK as potential therapeutic target because of its key functions, while also envisage bacterial RFK modules as potential antimicrobial targets.

Full text of DOI:10.1096/fj.202000566R

Received: 9 March 2020
DOI: 10.1096/fj.202000566R
Revised: 26 May 2020
Accepted: 5 June 2020
|
|
R E S E A R C H A R T I C L E
Human riboflavin kinase: Species-specific traits in the
biosynthesis of the FMN cofactor
Ernesto Anoz-Carbonell1,2
Maribel Rivero1
Victor Polo2,3
|
|
|
Adrián Velázquez-Campoy1,2,4,5,6
Milagros Medina1,2
|
1Departamento de Bioquímica y Biología
Molecular y Celular, Facultad de Ciencias,
Universidad de Zaragoza, Zaragoza, Spain
Abstract
Human riboflavin kinase (HsRFK) catalyzes vitamin B2 (riboflavin) phosphoryla-
tion to flavin mononucleotide (FMN), obligatory step in flavin cofactor synthesis.
HsRFK expression is related to protection from oxidative stress, amyloid-β toxicity,
and some malignant cancers progression. Its downregulation alters expression pro-
files of clock-controlled metabolic-genes and destroys flavins protection on stroke
treatments, while its activity reduction links to protein-energy malnutrition and thy-
roid hormones decrease. We explored specific features of the mechanisms underly-
ing the regulation of HsRFK activity, showing that both reaction products regulate it
through competitive inhibition. Fast-kinetic studies show that despite HsRFK binds
faster and preferably the reaction substrates, the complex holding both products is
kinetically most stable. An intricate ligand binding landscape with all combinations
of substrates/products competing with the catalytic complex and exhibiting moderate
cooperativity is also presented. These data might contribute to better understand-
ing the molecular bases of pathologies coursing with aberrant HsRFK availability,
and envisage that interaction with its client-apoproteins might favor FMN release.
Finally, HsRFK parameters differ from those of the so far evaluated bacterial coun-
terparts, reinforcing the idea of species-specific mechanisms in RFK catalysis. These
observations support HsRFK as potential therapeutic target because of its key func-
tions, while also envisage bacterial RFK modules as potential antimicrobial targets.
2Instituto de Biocomputación y Física de
Sistemas Complejos (GBsC-CSIC and
BIFI-IQFR Joint Units), Universidad de
Zaragoza, Zaragoza, Spain
3Departamento de Química Física,
Universidad de Zaragoza, Zaragoza, Spain
4Fundación ARAID, Diputación General de
Aragón, Zaragoza, Spain
5Aragon Institute for Health Research (IIS
Aragon), Zaragoza, Spain
6Biomedical Research Networking
Centre for Liver and Digestive Diseases
(CIBERehd), Madrid, Spain
Correspondence
Milagros Medina, Departamento de
Bioquímica y Biología Molecular y Celular,
Facultad de Ciencias, Pedro Cerbuna 12,
Universidad de Zaragoza, 50009-Zaragoza,
Spain.
Email: mmedina@unizar.es
Funding information
Spanish Ministry of Economy, Industry and
Competitiveness, Grant/Award Number:
BIO2016-75183-P AEI/FEDER; Spanish
Ministry of Science and Innovation, Grant/
Award Number: PID2019-103901GB-I00
AEI/FEDER; Government of Aragon-
FEDER, Grant/Award Number: E35_20R
K E Y W O R D S
calorimetry, kinetics limiting step, ligand binding and cooperativity, pre-steady-state kinetics,
product inhibition, riboflavin kinase, therapeutic target
Abbreviations: ANP, adenine nucleotide (ATP or ADP); Ca, Corynebacterium ammoniagenes; CD, circular dichroism; FADS, FAD synthase; FLV,
flavinic nucleotide (FMN or FAD); FMNAT, ATP:FMN adenylyltransferase; GST, glutathione S-transferase; HPLC, high-performance liquid
chromatography; Hs, Homo sapiens; ITC, isothermal titration calorimetry; Ka, association constant; kcat, catalytic constant; Kd, dissociation constant; Ki,
Inhibition constant; KM, Michaelis constant; kobs, observed rate constant; koff, dissociation rate constant; kon, association rate constant; PIPES, 1,4-piperazine
diethane sulfonic acid; rpm, revolutions per minute; RF, riboflavin, Vitamin B2; RFK, ATP:riboflavin kinase; SDS-PAGE, sodium dodecyl sulfate-
polyacrylamide gel electrophoresis; Sp, Schizosaccharomyces pombe; Spn, Streptococcus pneumonia; α, heterotropic interaction constant; ε, extinction
coefficient.
© 2020 Federation of American Societies for Experimental Biology
The FASEB Journal. 2020;34:10871–10886.
wileyonlinelibrary.com/journal/fsb2
10871
|

10872
ANOZ-CARBONELL Et AL.
|
1
INTRODUCTION
demonstrated for the RFK cycles of the prokaryotic FADSs
of Corynebacterium ammoniagenes and Streptococcus
pneumoniae (CaFADS and SpnFADS, respectively).19,20
Furthermore, the substrate binding and catalytic motifs—
PTAN and GxY—despite conserved, exhibit species-specific
arrangements and traits (Figure1 and SP2).2,18
|
Human riboflavin kinase (HsRFK; ECnonbreakingsp
ace2.7.1.26) is an essential enzyme that catalyzes the biosyn-
thesis of the flavin mononucleotide (FMN) cofactor using as
substrates riboflavin (RF, vitamin B2) and ATP, therefore, ex-
hibiting an ATP:riboflavin kinase activity (ECnonbreakingsp
ace2.7.1.26).1,2 Besides its key role in the synthesis of flavin
cofactors, HsRFK is predicted as involved in a protein-pro-
tein association network that at the system level affects to
different cellular processes (Figure SP1). In this context, the
orthologous protein in mice has been shown essential for
embryonic development, while, in both human and mouse
cells, its binding to the Tumor Necrosis Factor Receptor 1
enhances assembly of the flavin adenine dinucleotide (FAD)
cofactor to NADPH oxidases activating the production of re-
active oxygen species as defence and signaling molecules.3
HsRFK expression is also related to cellular protection from
progression of some malignant prostate cancers.4 Recently,
RFK, as well as its FMN metabolic product, have also been
reported as critical modulators of amyloid-β toxicity.5 On the
contrary, its downregulation alters the expression profiles of
cryptochromes and other clock-controlled metabolic genes,
being also a potential risk factor for stroke by destroying the
protective effect of flavin treatments.6,7 Finally, insufficient
conversion of RF into flavin cofactors by reduction of the
RFK activity is reported in protein-energy malnutrition sit-
uations coursing with reduced thyroid hormone concentra-
tions.8-10 Altogether such observations point to HsRFK as a
potential therapeutic target.
Part of the FMN produced by HsRFK is directly used as
cofactor of FMN-dependent flavoproteins, but most of it is
subsequently adenylylated by a FAD synthase (HsFADS),
that exhibits ATP:FMN adenylyltransferase activity
(FMNAT, ECnonbreakingspace2.7.7.2) and provides the
flavin adenine dinucleotide (FAD) cofactor.11,12 RF to FMN
conversion seems to be the major rate limiting step in FAD
synthesis, being therefore HsRFK a key element for the pro-
duction of the FAD cofactor required by more than 60 human
flavoenzymes.13,14 RFK and FMNAT activities located in
two independent enzymes is a trait of non-photosynthetic
eukaryotes, but in most prokaryotes a single bifunctional
enzyme known as type I FAD synthase (FADS) carries out
both activities.15 Noticeably, prokaryotic FADSs fold in two
modules, being the N-terminal module associated to the
FMNAT activity and the C-terminal to the RFK one.15,16 The
eukaryotic monofunctional RFKs and the C-terminal mod-
ules of prokaryotic FADSs are structurally homologs (both
folding into a six-stranded β barrel core with Greek key to-
pology), but species-specific differences in secondary struc-
ture elements are observed.1,2,16-18 These variations, although
apparently minor, utterly modulate conformational reorgani-
zations occurring during substrate binding and catalysis, as
As FMN and FAD are essential cofactors, not only in cell
energetics metabolism, but also for the action of a plethora
of flavoenzymes with very different cellular functions, their
biosynthesis must be thoroughly regulated to maintain cellu-
lar and flavoproteome homeostasis.13,21 In prokaryotes, we
recently learned that FMN biosynthesis is regulated through
mechanisms that, depending on the species, can include the
inhibition of the RFK activity by the RF substrate and/or
products, the redox state of the RF substrate, the formation
of transient assemblies coupled to catalysis or the subsequent
FMNAT activity occurring at the protein N-terminal mod-
ule.19,20,22-27 Eukaryotic cells provide additional levels of reg-
ulation by making use of different isoforms or intracellular
compartmentalization. The mechanisms underlying FMN ho-
meostasis in eukaryotes remain still poorly understood,28,29
but those found in prokaryotic FADSs might also apply. In
this context, the available crystallographic structures of
HsRFK and of RFK from the yeast Schizosaccharomyces
pombe (SpRFK) also envisage large conformational changes
upon ligand binding and catalysis, particularly in the surface
loops termed FlapI and FlapII (Figure 1 and SP2).1,2,17
In the present study, we focus on the intrinsic mechanisms
underlying the regulation of HsRFK enzymatic activity. We
have used the apo-HsRFK form, to avoid substrates and prod-
ucts of the reaction interfering our experimental approach.
Our steady-state data indicate that HsRFK is inhibited by the
reaction products, but not by substrates. Stopped-flow spec-
trophotometry allows ascertaining sequential conformational
changes upon ligand binding and catalysis. Finally, ligand
binding studies using isothermal titration calorimetry (ITC)
display an intricate thermodynamic landscape. Our results
are discussed in the context of the here modeled structure
for apo-HsRFK, the available crystallographic structures of
monofunctional RFKs and previous structure-function stud-
ies concerning RFK modules of prokaryotic FADSs.
2
MATERIALS AND METHODS
Cloning, expression, and purification
|
2.1
|
of recombinant HsRFK
The coding sequence of HsRFK (accession number
Q969G6) was codon optimized for its heterologous expres-
sion in Escherichia coli and synthetized by GenScript. This
DNA sequence was cloned between the NcoI and BamHI re-
striction sites of a pET28a(+)-derived vector, in which the

ANOZ-CARBONELL Et AL.
10873
|
(C)
(A)
(D)
(E)
(B)
FIGURE 1 Structural properties of RFKs. Cartoon representation of the crystallographic structure of HsRFK in complex with FMN and ADP
in either the (A) open (PDB ID 1P4M) or the (B) closed conformation (PDB ID 1Q9S) of the flavin binding site. (C), Cartoon representation of
SpRFK as crystallized in the apo form (PDB ID 1N05). FlapII is disordered and depicted as a dashed loop. Comparison of the organization of the
active site in the closed HsRFK:FMN:ADP complex relative to the structures of the (D) open HsRFK:FMN:ADP complex and of the (E) apo-
SpRFK (shown in grey). In all panels FMN and ADP ligands are shown as CPK colored sticks with carbons in green and magenta, respectively,
and Mg2+ is shown as a green sphere. Unless otherwise stated, FlapI, FlapII, and L6c are highlighted in green, violet and pink, respectively, the
consensus PTAN motif at the active site is in blue for HsRFK and grey for SpRFK. Relevant side-chains are in sticks
thrombin cleavage site was substituted by the PreScission
Protease cleavage signal. The sequence-verified plasmid
was transformed into E coli BL21(DE3) strain, and trans-
formed cells were grown at 37°C in LB medium (1% (w/v)
of tryptone, 0.5% (w/v) of yeast extract, and 1% (w/v) of
NaCl) supplemented with 50 mg/mL of kanamycin. At an
OD600 nm of 0.7, cultures were induced with 0.2 mM of iso-
propyl β-d-1-thiogalactopyranoside, incubated under the
same conditions and, after 24 hours, harvested. Cells were
resuspended in 20 mM sodium phosphate, pH 7.4, 500 mM
NaCl, and 10 mM imidazole, containing lysozyme (1 mg/
mL), DNase (0.1 mg/mL), and protease inhibitors (0.2 mM
of Phenylmethanesulfonyl fluoride and 10 mM of benzami-
dine), and subsequently sonicated at 4°C. After centrifuga-
tion to remove cell debris, the supernatant containing the
soluble protein was loaded into a HisTrap HP affinity col-
umn (GE Healthcare) and the protein was eluted applying
a gradient from 10 to 500 mM of imidazole in 20 mM of
sodium phosphate, pH 7.4, 500 mM NaCl. Buffer was ex-
changed to 50 mM Tris/HCl, pH 7.4, 150 mM NaCl, using a
HiPrep Desalting Column (GE Healthcare). The His6-Tag
was removed by 24 hours incubation with the PreScission
protease (GE Healthcare) at 4°C, in ratio 1:10 (w/w), and
the protein was then loaded into the HisTrap HP column in
tandem coupled to a GSTrap 4B column (GE Healthcare)
to eliminate both the remained His6-Tagged HsRFK and
the PreScission protease (a GST fused product). The re-
sulting yellowish unbound fraction was fractioned with
20% of ammonium sulfate and centrifuged for 30 minutes
at 4000 g. The supernatant was loaded onto a Phenyl-
Sepharose High performance (GE Healthcare) column
equilibrated with 50 mM of Tris/HCl, pH 8.0, 20% of am-
monium sulfate. The column was washed with the same
buffer until most of the yellow color washed out. Finally,
the protein was eluted using a 20→0% reversed-gradient of
ammonium sulfate in the same buffer. Protein purity was

10874
ANOZ-CARBONELL Et AL.
|
assessed by 17% of SDS-PAGE and buffer was finally ex-
changed by dialysis in 20 mM PIPES, pH 7.0. Pure protein
solution was stored at −80°C until used.
the Michaelis-Menten kinetic model, obtaining kcat and Km
values. The inhibition mechanism of the products of the RFK
reaction -FMN and ADP- was determined by evaluating their
effect on Km and kcat by individual fitting of data sets to the
Michaelis-Menten model. Additionally, data sets acquired
at variable concentrations of the product acting as inhibitor
were globally fitted to the Lineweaver-Burk equation for
competitive inhibition
2.2
Spectroscopic analysis
|
UV-visible spectra were recorded in a Cary 100 spectro-
photometer (Agilent Technologies) and ligand-free HsRFK
was quantified using the theoretical extinction coeffi-
cient ε279 nm = 21 430 M−1cm−1 and a molecular weight
of 17.76 kDa (ProtParam). Fluorescence emission spectra
were recorded in a Cary 100 spectrophotometer (Agilent
Technologies) in 20 mM of PIPES, pH 7.0, 0.3 mM of
MgCl2 at 25°C, exciting the protein aromatic residues at
280 nm. Fluorescence excitation scans were recorded at the
maximum emission wavelength (323 nm) and in the same
experimental conditions. Circular dichroism (CD) spectra
were recorded at 25°C using a Chirascan spectropolarime-
ter (Applied Photosystem Ltd.). Samples containing ~5 and
~20 µM of HsRFK in 20 mM of PIPES, pH 7.0, 0.3 mM
of MgCl2 were used in the far-UV (cuvette path length,
0.1 cm) and near-UV CD (cuvette path length, 0.4 cm),
respectively.
(
)
Km
[e]
[I]
1
1
(1)
=
1+
+
Ki kcat [S] kcat
v0
where [S] is the concentration of the varying substrate (RF or
ATP) and [I] that of the inhibitor (FMN or ADP), and Ki is
the corresponding inhibition constant. All experiments were
performed in triplicate. Estimated errors in kcat, KM and Ki
values were, in general, within 10% of their values, assumed
to be larger than the standard deviation between replicates
and the numerical error after fitting analysis.
2.4
Pre-steady-state kinetics
|
Kinetic experiments in the pre-steady state were registered
using stopped-flow spectroscopy on an Applied Photophysics
SX17.MV spectrophotometer, using the ProData SX soft-
ware (Applied Photophysics Ltd.) for fluorescence data ac-
quisition and kinetic data analysis. Fast kinetic measurements
were carried out at 25°C in PIPES 20 mM, pH 7.0, 0.3 mM
of MgCl2 as previously described.19,20 In short, 0.2 µM of
HsRFK was mixed with reaction samples containing increas-
ing concentrations of the FLV ligands in the absence or pres-
ence of saturating concentrations of ANP ligands. Evolution
of flavin fluorescence after mixing was measured with an
excitation wavelength of 445 nm, while fluorescence emis-
sion was recovered using a >530 nm cut-off filter. Control
experiments, recorded in the same buffer but in absence of
MgCl2, produced similar profiles with significantly smaller
amplitudes in the fluorescence change. All the concentra-
tions indicated for these experiments are the final ones in
the stopped-flow observation cell. Three to five reproducible
traces were acquired at each time and concentration condi-
tion assayed. These kinetic traces were then fitted to expo-
nential functions,
2.3
Steady-state RFK activity
|
The RFK activity of HsRFK was measured at 25°C in 500 µL
of 20 mM of PIPES, pH 7.0 and 0.3 mM of MgCl2, contain-
ing variable concentrations of RF (0.2-15 µM) and ATP (2-
600 µM), as previously described.30,31 The inhibitory effect
of the substrates and products of the reaction was also ana-
lyzed by measuring the RFK activity in reaction mixtures at
increasing concentrations of FMN (0-5 µM), varying the ATP
concentration and keeping the RF constant for the determina-
tion of the FMN inhibitory effect; and at increasing concen-
trations of ADP (0-120 µM), varying the RF concentration
and keeping ATP fixed. In all cases, reactions were initiated
by addition of 40 nM HsRFK to reaction mixtures preincu-
bated at 25°C. After 1 minute of incubation at 25°C reactions
were stopped by boiling the samples at 100°C for 5 minutes.
The flavin composition of the supernatant was analyzed
using an Alliance HPLC system (Waters) equipped with a
2707 autosampler and an HSST3 column (4.6 × 50 mm,
3.5 mm; Waters) preceded by a pre-column (4.6 × 20 mm,
3.5 mm; Waters) as previously described.31 Flavin concen-
trations were quantified using RF or FMN standard curves
acquired under the same conditions, and the observed steady-
state rates for FMN production (v0) were determined in units
of nmoles of flavin transformed per min per nmol of enzyme
(v0/[e]). The kinetic data obtained for one substrate at satu-
rating concentrations of the second substrate were fitted to
ꢀ
(
)
(2)
y=
Ai ⋅exp −kobsi ⋅t
where Ai and kobsi are the amplitude of the fluorescence
change and the observed kinetic constant for a particular
spectroscopic process i that contributes to the overall time-
dependent fluorescence change. The previous equation was
corrected by the addition of a linear term when a process was
not finished in the timeframe of the measurement.

ANOZ-CARBONELL Et AL.
10875
|
ꢀ
(
)
Cooperative ligand binding was assessed by ITC employ-
ing an exact analysis of the heterotropic interactions con-
sidering a ternary equilibrium between the protein and two
ligands approximate analysis as previously described.34-36
A quasi-binary equilibrium between the protein and one li-
gand, with the influence of the other ligand being implicit
within the apparent ligand binding parameters, was also used
to confirm parameters.20 Briefly, ANP ligands were used to
titrate mixtures of HsRFK (10 µM) with different FLV con-
centrations (0-100 µM) in the calorimetric cell, and the coop-
erativity constant for the heterotropic interaction (α) between
ANP and FLV ligands was determined from any single ti-
tration employing the exact methodology.34-36 Additionally,
titrations of HsRFK:FLV mixtures (at different FLV concen-
trations) with ANP ligands were analyzed as binary titrations,
from which apparent association constants for ANP ligands
at certain FLV concentrations were determined. These data
were fitted to a cooperativity model that considers explicitly
the influence of the FLV ligand in the protein binding affinity
for the ANP ligand:19,20,34-36
y=
Ai ⋅exp −kobsi ⋅t +b⋅t
(3)
It should be noticed that when more than one process is
detected, accuracy in determination Ai and kobsi for processes
after the initial one decreases with the FLV concentration
assayed.
kobs values showing a linear dependence on the FLV
concentration were fitted to a one-step model associated
to the binding equilibrium of the flavin ligand to HsRFK
(HsRFK+FLV⇆HsRFK:FLV), whose kinetics can be repre-
sented by the following equation
k
obs =kon ⋅[FLV]+koff
(4)
where kon and koff are the kinetic rate constants for complex
formation and dissociation, respectively.
When kobs values show hyperbolic dependence on the
FLV concentration, data can be fitted to an induced fit
model (HsRFK+FLV⇆HsRFK:FLV*→HsRFK:FLV)32 that
parametrizes both the ligand binding (Kd, koff/kon) and the
subsequent conformational change (kr, as the kinetic rate con-
stant for the structural rearrangement).
1+ꢀKFaLV [FLV]
Kaapp,ANP =KAa NP
(6)
1+KFaLV [FLV]
[FLV]
k
obs =kr ⋅
(5)
where Kaapp,ANP is the apparent association constant for the
ANP ligand as a function of the FLV concentration, KANP
is the intrinsic association constant for ANP (that is, in athe
absence of FLV ligand), KaFLV is the association constant for
FLV, and [FLV] is the concentration of flavin in the calori-
metric cell, from which α between ANP and FLV ligands can
also be determined.
ITC experiments were performed in duplicate or tripli-
cate. The errors considered in the measured parameters (in
general 15% in Kd and Ka values, 0.3 kcal mol−1 in ΔH,
ΔG and −TΔS, and 20% in α) were assumed to be larger
than the standard deviation between replicates and the nu-
merical error after fitting analysis.
Kd +[FLV]
When detected, flavin photobleaching within the stop-
flow observation chamber was evaluated as previously
described.20
2.5
Isothermal titration calorimetry
|
ITC measurements were performed using a high precision
Auto-iTC200 MicroCal calorimeter (Malvern Panalytical)
thermostated at 25°C. Up to 19 injections of 2 µL of the titrat-
ing solution were added to the calorimetric sample cell and
mixed using a stirring speed of 750 rpm. Ligands and protein
were both dissolved in 20mM of PIPES, pH 7.0, in absence
and presence of 0.3mM of MgCl2 (cation concentration for
optimal enzymatic activity). Typically, 50µM of RF, 70µM
of FMN, and 80µM of ANP (ATP or ADP) in PIPES 20mM,
pH 7.0 were used to titrate HsRFK (10 µM) in a 200µL sam-
ple cell. Additionally, preformed complexes of the enzyme
with the FLV or ANP ligands at saturating concentrations
were titrated with ANP or FLV ligands, respectively. The as-
sociation constant (Ka), the enthalpy change (ΔH), and the
binding stoichiometry (N) were obtained through nonlinear
least squares regression of the data using a homemade model
for one binding site, which was implemented in Origin 7.0
(OriginLab) as previously described.33 The Kd, the free en-
ergy change (ΔG), and the entropy change (ΔS) were ob-
tained from essential thermodynamic relationships.
2.6
Molecular dynamics simulations
|
The initial model of apo-HsRFK, excluding any ligand, was
built by removing ADP and Mg2+ from the crystallographic
structure with PDB ID 1NB0. PROPKA software was used
to assign pKa values to structures, which were protonated
to pH 7.0.37 PyMOL was used for structural manipulations
and figures production.38 All-atom molecular dynamics
(MD) simulations were performed using GROMACS 5.1.539
and AMBER ff03 parameters.40 The system was placed in
the center of a rhombic dodecahedron box, solvated with a
TIP3P water model and neutralized by adding sodium ions.
Final system consisted in 22345 total atoms. A steepest de-
scent minimization was performed to avoid close contacts

10876
ANOZ-CARBONELL Et AL.
|
or clashes. Desired conditions were achieved after a 100 ps
simulation with NVT ensemble and the generation of ran-
dom initial velocities, and a 100 ps simulation with NPT en-
semble, restraining the movement of atoms of protein with a
1000 kJ·mol−1·nm−1 harmonic potential. Longer NPT (300 K)
simulations with positions unrestrained were then performed,
collecting the data every 10 ps. Time step of 2 fs and leap-
frog integrator, periodic boundary conditions, Particle Mesh
Ewald method for long range electrostatic interaction,
Parrinello-Rahman method for pressure control, modified
Berendsen method for temperature equilibration and LINCS
to restrain bonds including hydrogen atoms were used. Five
replicates were made for each trajectory, which were ana-
lyzed using VMD41 and GROMACS package tools.39
data: around 27% of α-helix, 30% of β-sheet and a large con-
tent of loops. The near-UV CD spectrum (Figure SP3D) of
HsRFK showed a broad 265-295nm negative band with min-
ima at 285 and 292nm related to the protein tertiary structure
organization. Collectively, these data indicate that the purified
HsRFK is correctly folded and in the apo-form.
3.2
The HsRFK activity is modulated
|
by the products of the reaction
In advance to determining the HsRFK kinetic parameters,
we quantitatively evaluated the influence of the reducing
environment (by using sodium dithionite) as well as of the
presence of Mg2+. Under saturating concentrations of both
substrates (RF and ATP), the RFK activity resulted tightly
regulated by the cation concentration, with maximal rates at
a concentration of 0.3mM (Figure SP4A). Herein, a MgCl2
concentration of 0.3mM was used. On the contrary, the rate
for RF transformation was independent on the concentration
of the reducing agent (Figure SP4B). This later behavior re-
sulted similar to that described for CaFADS and SpnFADS,
but differed for other species, as Listeria monocytogenes,
whose turnover and efficiency highly depended on the redox
status of the media.19,22,23,26,27
3
RESULTS
|
3.1
HsRFK is purified free of flavin and
|
adenine ligands
E coli BL21 (DE3) cells transformed with the recombinant
pET28a-HsRFK vector produced a high level of expression of
active HsRFK enzyme, with a typical yield after purification of
3.2 mg of protein per liter of culture. HsRFK fractions were yel-
low colored, suggesting the protein was expressed in complex
with a tightly bound flavinic ligand that kept bound throughout
the purification process. Hence, initial purified fractions of the
recombinant protein showed the typical flavoprotein bands-I
and -II in the Vis absorbance spectrum (Figure SP3A). HPLC
analysis of the supernatant obtained upon thermal denaturation
of the enzyme demonstrated that the bound cofactor was FMN.
Additionally, the other product of the RFK activity, ADP, was
also detected as trapped into the purified enzyme. This was evi-
denced by the displacement of the aromatic absorption peak of
protein toward 260 nm (characteristic of nucleotides) in some
flavin-free enzyme samples and by the absorption properties of
its supernatant after thermal denaturation. High affinity of the
adenine nucleotide was previously reported, probably being the
reason why all HsRFK structures so far reported in the PDB
contain it.1,2 Therefore, despite the protein was pure, a phenyl-
sepharose hydrophobic chromatographic step was required to
remove the FMN and ADP products from HsRFK. To the best
of our knowledge, this will be the first purification of HsRFK in
its apo-form to homogeneity (spectrum shown in Figure SP3A).
Excitation of HsRFK at 280nm produced a fluorescence
emission band centered at 332nm (Figure SP3B), in agreement
with Trp residues embedded within the protein environment.
The negative couplet (195-210 nm) in the far-UV CD spectrum
of HsRFK (Figure SP3C) indicated secondary α-helix struc-
ture, and the shoulder at 218nm also pointed to an important
content of β-sheet. Such spectral properties are consistent with
the secondary structure content derived from crystallographic
Steady-state rates for the HsRFK activity showed satura-
tion profiles for both substrates (Figure 2A). These profiles
fitted to the Michaelis-Menten model allowing determination
of kcat, KMRF, and KMATP, in 102 min−1, 2.5 µM and 30 µM
(Table 1), respectively. Furthermore, no inhibition by excess
of RF substrate was detected, contrary to that reported for the
RFK activity of CaFADS31 but in line with previous data for
HsRFK,42 SpnFADS,19 and LmFADS-1.27 Nonetheless, our
HsRFK data showed high efficiency for the synthesis of FMN
when compared not only to bifunctional FADSs,19,20,27 but
also to data previously reported for HsRFK43 (Table 1). In the
case of HsRFK, such differences can be associated to different
protein isolation procedures, in agreement with early reports
showing different RFK pools with distinct enzymatic activity
co-existing in rat liver cells. Such differences are probably due
to dissimilarities in the content of ligands bound to the enzyme.
Since inhibition of the RFK activity by the reaction
products contributes to maintain the flavin and flavopro-
teome homeostasis in prokaryotes, it is worthy to evaluate
the occurrence of such effects in the HsRFK activity. The
Michaelis-Menten and Lineweaver-Burk plots at increasing
concentrations of either ADP or FMN products showed that
while kcat remained constant, both KAMTP and KRF increased
(Figure 2). This pointed out to a competitiveMmechanism
of inhibition for both reaction products, allowing determi-
nation of the apparent inhibition constants, KAI DP and KIFMN
(Table 1). The similar values of KAI DP and KMATP as well as of
KFI MN and KRMF indicated that ADP and FMN bind to the free

ANOZ-CARBONELL Et AL.
10877
|
(B)
(A)
(C)
(D)
FIGURE 2 Inhibition of HsRFK activity by FMN and ADP products. (A) and (C) Michaelis-Menten plots and, (B and D) their corresponding
Lineweaver-Burk representations, as a function of different concentrations of the (A and B) ATP and (C and D) RF substrates in the presence
of different concentrations of the products (A and B) ADP and (C and D) FMN. Lines correspond to global fits to the equation for competitive
inhibition (Equation 1)
TABLE 1 Steady-state kinetic parameters for the HsRFK activity
kcat/KATP
kcat/KRF
KFMN
M
(min−1 µM−1
M
(min−1 µM−1
I
RFK enzyme
kcat (min−1
)
KATP (µM)
)
KADP (µM)
KRF (µM)
)
(µM)
M
I
M
HsRFKa
102
18
7
30
8
3.4 1.1
33
6
2.5 0.4
117
36
6.9 0.4
1.2 0.3
0.5 0.1
41
9
2.5 0.8
HsRFK(−)b
0.16
0.08
19
HsRFK(+)b
30
130 30
CaFADSc RFK module
SpnFADSd
40 12
3.2 1.7
0.7 0.1
0.8 0.1
7.9 0.9
17
3
5
1.4 0.2
1.3 0.4
7.1 1.3
55
33
95
2
2
7
75
41
12
7
2
1
130 16
844 95
46 13
LmFADS-1(−)e
LmFADS-1(+)e
66
95
4
1
10
1
aExperiments performed at 25°C in 20mM of PIPES, pH 7.0, 0.3 mM of MgCl2 (n = 3, mean SD). Presented data calculated by global fitting.
bData from (43). Experiments performed at 37°C in 50 mM of potassium phosphate buffer, pH 7.5, 12mM of MgCl2, both in absence (−) and presence (+) of 24 mM
of sodium dithionite.
cData from (20). Experiments performed at 25°C in 20mM of PIPES, pH 7.0, 0.8 mM of MgCl2.
dData from (19). Since both ADP and FMN act as mixed inhibitors two different KI values are shown for each. Experiments performed at 25°C in 20 mM of PIPES, pH
7.0, 0.8 mM of MgCl2.
eData from (27), experiments performed in 20mM of PIPES and 0.8 mM of MgCl2, pH 7.0, at 25°C, both in absence (−) and presence (+) of 24 mM of sodium
dithionite.
HsRFK enzyme with alike affinity to the substrates of the
reaction. Also, the significantly smaller inhibition constant
for FMN in comparison with ADP (KADP/KIFMN = 13.2) en-
visages FMN as a more potent inhibitoIr.
CaFADS and SpnFADS also showed product inhibition
of the RFK activity, but their inhibition mechanisms dif-
fered from that of HsRFK as well as among them: mixed for
both products in SpnFADS versus competitive for FMN and

10878
ANOZ-CARBONELL Et AL.
|
uncompetitive for ADP in CaFADS. FMN is a more potent
inhibitor in CaFADS (KRF/KFI MN = 7.1) than in SpnFADS or
HsRFK (0.8 and 1, respeMctively); and also, RF is a very strong
inhibitor of the RFK activity in CaFADS but not in the other
enzymes.19,20,27 Therefore, FMN biosynthesis in Homo sapi-
ens will be regulated by the products of the RFK reaction.
In addition, and considering that transport across membranes
appears favored for the RF substrate over FMN and FAD,
intracellular compartmentalization with different substrate/
product concentrations and/or the presence of different iso-
forms might also apply in the in vivo HsRFK regulation.11,28,44
its binding and/or dissociation to the protein, and/or as con-
sequence of conformational changes in its environment,19,20
while RF transformation into FMN is not observed due to
the same fluorescence spectra and quantum yields of both
flavins. When we mixed HsRFK with either RF or FMN, we
only detected very slow and linear fluorescence decays con-
sistent with previously reported flavin photobleaching.19,20
These observations suggested that the apo-form of HsRFK
is not able to bind RF or FMN (herein both referred as FLV)
ligand, or at least to internalize its isoalloxazine ring in the
expected catalytically competent enclosed conformation.
A fast and intense exponential decay in fluorescence in
the 2 seconds time frame was, however, detected when mix-
ing HsRFK with all combinations of ATP or ADP (denoted
herein as ANP) and FLV ligands (Figure 3A). We related the
fluorescence decay to FLV binding and/or internalization in
the protein matrix by FlapII displacement (Figure 3A), con-
cluding that the ANP presence/binding prones HsRFK to
bind and internalize the FLV ligand. Noticeably, no subse-
quent recover of fluorescence was observed for the assayed
3.3
Binding of HsRFK substrates is the
|
kinetically preferred process
We used stopped-flow spectrophotometry to kinetically
differentiate individual processes during the HsRFK reac-
tion. This technique allowed us to detect small changes in
the dielectric environment of the flavin isoalloxazine upon
FIGURE 3 Pre-steady-state stopped-flow kinetics of the binding of RF and FMN to HsRFK in the presence of adenine nucleotides. (A)
Normalized evolution of kinetic changes in fluorescence upon mixing HsRFK (0.2 µM) with all possible FLV-ANP ligand combinations (0.125
and 250 µM, respectively). (B) Example of the fittings of kinetic traces (in this case, corresponding to mixtures of HsRFK with RF-ATP), and
residuals of the fitting of the 1 µM RF-250 µM ATP data to a biexponential function. Evolution of (C) kobs1 and (D) kobs2 as a function of the FLV
concentrations. Insets show schemes representing the corresponding processes

ANOZ-CARBONELL Et AL.
10879
|
combinations of ligands, differing this behavior from the
one reported when similarly evaluating the RFK modules of
CaFADS and SpnFADS. For these two enzymes, the initial
flavin fluorescence decay was followed by fluorescence re-
cover related to an ATP-induced conformational change that
re-opens the flavin binding site making the isoalloxazine ac-
cessible to the solvent after the reaction has taken place.19,20
Kinetic traces corresponding to mixtures of HsRFK with
the RF-ADP ligands fitted to a single exponential decay,
whereas two and up to three independent processes were
identified when, respectively, evaluating the binding kinetics
in the FLV-ATP and FMN-ADP combinations (Figure 3B).
Noticeably, while the amplitude of the first process (A1)
dominates the fluorescence decay for RF binding, A1 and
A2 became similar when assaying the binding kinetics of the
FMN product (no shown). Therefore, we identified the ini-
tial process, generally accounting for most of the amplitude
decay, as FLV binding/internalization in HsRFK by FlapII
displacement (up to ~14Å reorganization of amino acids com-
prising the loop), similarly to that reported for CaFADS and
SpnFADS.19,20 The succeeding fluorescence decays have to
relate to subsequent conformational changes in HsRFK loops
further contributing to additional changes in the isoalloxazine
environment after the initial binding. Considering these extra
processes are not observed when mixing the protein with RF-
ADP, they appear related to the extra phosphates of ATP and
FMN, over ADP and RF, respectively, influencing some con-
formational flexibility.
ATP) were the least favored processes from the kinetic point
of view (smaller kon), though the combination of FMN-ADP
products showed an amplitude comparable to that of sub-
strates and a considerably lower koff. As consequence, the
complex of HsRFK with the products of its activity exhibits
the smaller Kd and appears, therefore, as the most stable one.
On their side, when detected, kobs2 showed a saturation
profile on the FLV concentration (Figure 3D). When fitting
these data to an induced fit model representing changes in
the protein conformation induced by binding of the ligand,32
we were able to determine an equilibrium reorganization
constant as well as the kinetic constant to achieve the final
state (Table 2B). Noticeably, this process presented very par-
ticular features when evaluating the FMN-ADP products; it
was considerably slower (kr), A2 was comparable to A1, and
the product of its reorganization was the most stable (Kreorg).
Therefore, binding of the products of the RFK activity to
the enzyme occurs through stabilization of a transient inter-
mediate in the isoalloxazine internationalization by FlapII
displacement, while these intermediate is hardly populated,
or not at all, in the binding of other ligand combinations.
Noticeably, this FMN-ADP combination is the only one for
which a third considerably slower process (80-100 min−1 at
the FMN concentrations assayed), likely independent of the
flavin concentration, is observed when binding to HsRFK.
Interestingly, kon and kr values for the binding of RF-ATP
substrates are substantially faster than the kcat, while koff for
the FMN-ADP products is situated in its range (Tables 1 and
2). Collectively, these observations indicate that kinetics of
products release limits the HsRFK catalytic activity.
kobs1 values showed a linear dependence on the FLV con-
centration (Figure 3C) that permitted to determine kon and
koff for flavin binding and, as a consequence the process dis-
sociation constant (Kd) (Table 2). These data indicated that
binding processes containing the RF substrate are the fastest
(Table 2A). In addition, the largest amplitude in fluorescence
decay is observed for RF and ATP (Figure 3A). Therefore,
HsRFK binds preferably the substrates of the RFK reaction,
RF and ATP, over other combinations of substrates/products,
similarly to SpnFADS but contrary to CaFADS. Noticeably,
binding of the FMN product in presence of ANP (particularly
These data point to differences in the regulation of the
catalytic activity of HsRFK when compared to CaFADS
and SpnFADS. In HsRFK, as in SpnFADS, the binding of
substrates of the RFK reaction -RF and ATP- is the kineti-
cally favored process,19 whereas in CaFADS binding of any
other combination of ligands is faster.20 Moreover, while in
CaFADS and SpnFADS the ATP substrate activates FLV
ligand internalization as well as the subsequent cavity re-
opening to make this ligand again solvent accessible, such
TABLE 2 Pre-steady-state kinetic parameters for the binding and dissociation of flavins to HsRFK in the presence of adenine nucleotides.
Experiments were performed at 25°C in 20 mM of PIPES, pH 7.0, 0.3 mM of MgCl2. (n = 5, mean SEM) in and stopped-flow equipment
kobs1 (flavin binding)
kobs2 (conformational rearrangement)
Ligands
ΔGreorg
combination
kon (min−1 µM−1
)
koff (min−1
)
Kd (µM)
ΔG (kcal mol−1
)
kr (min−1
)
Kreorg (µM)
(kcal mol−1
)
RF-ATP
RF-ADP
FMN-ATP
FMN-ADP
1760 50
1670 90
420 40
700 40
610 10
110 10
0.40 0.02
0.36 0.03
0.26 0.04
0.05 0.01
−8.6 0.1
−8.7 0.1
−8.9 0.1
−9.9 0.1
530 80
–
127 14
0.38 0.08
–
0.83 0.41
0.16 0.1
−8.8 0.1
–
−8.3 0.3
−9.2 0.4
a
a
a
1140 30
59
8
12
2
aProcess not observed for this combination of ligands.

10880
ANOZ-CARBONELL Et AL.
|
FIGURE 4 Scheme of the conformational and ligand binding spaces along the HsRFK cycle. The diagram summarizes the different HsRFK
species envisaged considering kinetic, inhibition, binding, and structural data available. All presented structures correspond to different crystal
structures for HsRFK (PDB IDs are indicated), with the only exception of the apo-form that has been produced by MD simulations of the 1NB0
pdb after removing the ADP:Mg2+ ligand. Crystal structures are represented by B-factor of backbone atoms, with higher radius of ribbons and
warmer colors indicating higher fluctuations, being all snapshots normalized according the color code. Those states lacking structural data are
represented by circles (nonproductive states are shown in red, alternative paths in violet and competent state in blue start). Processes leading to the
formation of sub-stoichiometric complexes (HsRFK:RF) are highlighted with blue arrows
final exposure is not detected with HsRFK. In addition, this
enzyme is the only from the three for which koff for the FMN-
ADP ligands is considerably slower than kon (nearly 20-fold),
making the binding of the products stronger and envisaging
different mechanisms for the FMN product release from the
RFK site among these three proteins.

ANOZ-CARBONELL Et AL.
10881
|
3.4
Thermodynamics modulates the ligand
of a KRF value in the micromolar range, but the low interac-
tion stdoichiometry observed envisages very low occupancy
(N around 0.16). Mg2+ further hindered RF binding due to
a higher increase in the entropic contribution to the binding
than in the enthalpic one (2.32- and 1.37-fold, respectively)
(Table SP2, Figure SP6).
|
binding landscape of HsRFK
Subsequently, we performed ITC experiments to assess if the
kinetically detected processes were significant in reaching the
thermodynamic equilibrium. Binary and ternary interactions
of HsRFK with ANP and/or FLV ligands were analyzed at
pH 7.0 and 25°C both in absence and in presence of 0.3 mM
of Mg2+. The corresponding determined thermodynamic pa-
rameters are summarized in Table SP1, while some examples
of the experimental thermodynamic dissections are displayed
in Figure SP5.
Direct titrations allowed the determination of the intrinsic
binding parameters of the interaction of HsRFK with sub-
strates and products of the RFK reaction. For ANP ligands,
KdANP values were in the low micromolar range and the stoi-
chiometry of the interaction, around 0.6, was consistent with a
unique ANP-binding site, with occupancy below unity being
probably associated to protein conformational heterogeneity.
This agrees with low B-factors of the bound ADP-Mg2+ in
the available crystallographic structures (Figure 4). The pres-
ence of Mg2+ resulted in the reduction of the favorable en-
thalpic contribution to binding, as well as of the unfavorable
entropic one (Figure SP6). Nonetheless, ΔG remained mostly
insensitive to the cation through entropy/enthalpy compensa-
tion. A similar situation was previously reported for the RFK
module of CaFADS.33 On the contrary, FLV (RF and FMN)
ligands were hardly able to directly bind HsRFK. No inter-
action heat was detected for the protein titration with FMN,
suggesting either lack of interaction, very slow binding or in-
teraction occurring without appreciable exchange of heat. In
titrations with the RF substrate, data allowed for estimation
Titrations of ANP:HsRFK or FLV:HsRFK binary mix-
tures with, respectively, FLV or ANP permitted to further
unravel the complete thermodynamic landscape of ligand
binding. Figure 5 summarizes all possible binary and tertiary
interactions of the enzyme with substrates and products in
absence (Figure 5A) and presence of Mg2+ (Figure 5B), in-
cluding the fraction of binding-competent protein (N) in each
case as the thickness of the arrows. Titrations involving both
RF and ATP in the presence of the divalent cation (catalytic
conditions) were not measured, since the heat of the catalytic
reaction masked the interaction heat. As shown in the fig-
ure, pathways leading to non-competent tertiary complexes
compete with formation of the HsRFK:ATP:RF catalytically
complex (orange pathways in Figure 5). Nevertheless, there
was no thermodynamically preferred binding pathway, both
in terms of final complex stability and production probability
(N, interaction stoichiometry). The presence of Mg2+ slightly
increased the fraction of protein prone to interact and, con-
sequently, the probability of a particular path to occur, but,
contrary to CaFADS and SpnFADS,19,20 hardly modulated
the binding landscape (Figure 5A,B, SP5, Table SP1). The
only exception to this behavior was observed for the FMN-
ADP products combination, where the cation presence made
enthalpic as well as entropic contributions to the HsRFK
binding favorable. Therefore, differences in conformation
of ternary HsRFK complexes as a consequence of the cation
FIGURE 5 Gibbs free energy flow for the interaction of HsRFK with substrates and products. Diagrams summarize the thermodynamics
of the interaction of HsRFK with different combination of its ligands as obtained by ITC (Table SP1) at 25°C (A) in 20 mM of PIPES, pH 7.0,
0.3 mM of MgCl2, and (B) in 20 mM of PIPES, pH 7.0. HsRFK is represented as blue spheres, RF and FMN as orange and yellow hexagons, and
ATP and ADP as green and blue triangles, respectively. The length of the arrows is proportional to the ΔG for the interaction (value in kcal mol−1
are indicated in numbers), and its thickness is representative to the fraction of HsRFK binding the titrating ligand. Processes not directly observed
by ITC (interaction of HsRFK with FMN) are shown as dotted arrows. Paths leading to the formation of the tertiary catalytic complex (HsRFK with
ATP and RF substrates) are highlighted with orange arrows. NM indicates processes that could not be measured in the presence of Mg2+ since the
reaction heat would mask the interaction heat

10882
ANOZ-CARBONELL Et AL.
|
TABLE 3 Cooperativity coefficients for the binding of the
different combinations of FLV and ANP ligands to HsRFK in presence
and absence of Mg2+. Experiments were performed at 25°C in 20mM
of PIPES, pH 7.0, both in absence and presence of 0.3mM of MgCl2.
(n = 5, mean SEM)
slightly negative for the RF-ATP substrates combination
(α < 1). In general, the magnitudes of the cooperativity con-
stants for ligand binding to HsRFK are moderated and in the
range of those for SpnFADS, while those for CaFADS are
considerably larger, particularly in the presence of MgCl2.
Cooperativity in RF and ATP substrates binding to HsRFK
and SpnFADS is slightly negative and positive, respectively.
Noticeably, in the case of CaFADS its sign and magnitude are
highly influenced by the RF substrate concentration that in
this case also acts as inhibitor.19,20 Thus, differences are also
found in the cooperation of substrates and products binding
to RFK enzymes from different organisms.
[MgCl2] Ligands
α
N
Δh kcal/mol
0.3 mM
RF-ATP
RF-ADP
FMN-ATP
FMN-ADP
RF-ATP
RF-ADP
FMN-ATP
FMN-ADP
N.M.a
N.M.a
N.M.a
2.2 0.2
1.4 0.1
2.0 0.2
0.62 0.01 −1.3 0.2
0.49 0.01 6.2 1.0
0.50 0.01 −10 0.3
0 mM
0.87 0.20 0.68 0.01 5.3 0.4
8.3 1.2 0.46 0.01 1.8 0.6
0.96 0.10 0.61 0.01 2.7 0.2
6.9 1.2 0.56 0.01 −5.9 0.6
4
DISCUSSION
|
aN.M., not measured. When mixing RF and ATP in presence of Mg2+, the
catalytic reaction heat conceals the interaction heat.
4.1
Conformational landscape in the
|
HsRFK catalytic cycle
presence are only predicted for the formation of the ternary
complex containing the FMN and ADP products, contrary
to that reported for CaFADS and SpnFADS.19,20 The energy
diagram in the absence of MgCl2 (Figure 5B) showed two
alternative pathways leading to HsRFK:ATP:RF “pseudo-re-
active” complexes, which in addition are among the most
probable. Therefore, the HsRFK behavior is more similar to
SpnFADS than to CaFADS, enzyme that favors all the other
nonproductive CaFADS:ANP:FLV complexes against the
CaFADS:ATP:RF one.19,20
To date, there is no an available 3D crystal structure of
HsRFK in the absence of any ligand, either ANP or FLV,
or both. To gain insight into such conformation, we gener-
ated a model of apo-HsRFK by removing Mg2+:ADP from
the HsRFK-Mg2+:ADP crystal structure (PDB ID 1NB0).
This model was minimized and relaxed by MD simulations
(5 replicas) (Figure6 and SP7). Trajectories for Cα root mean
square deviation (RMSD), energy, solvent accessible surface
(SAS) and radius of gyration (GyR) indicate that apo-HsRFK
keeps overall folding along simulations (Figure SP7). The
most remarkable fact was the transient breaking of the
Lys20-Asp88 salt bridge, with the consequent displacement
of FlapI and loop 5 (L5c), and the observation of a dynamic
opening/closing of the ADP/ATP-binding cavity (Figure 6).
On the contrary, the conformation of the active site, formed
by the consensus PTAN motif and Glu78, retained confor-
mations similar to those observed in the crystal binary- and
ternary- HsRFK complexes (Figure 6C and SP8). Such con-
formations resemble those in the apo-forms of SpRFK and
RFK module of SpnFADS (Figure 1 and SP7). This leaves
apo-CaFADS as the only RFK showing a different PTAN
conformation due to its Thr exhibiting considerably differ-
ent Φ and ψ values (Figure SP8A). In conclusion, our MD
data show that FlapI and L5c adopt different conformations
in apo-HsRFK with respect to binary HsRFK:ADP com-
plexes.2 Additionally, our simulations predict an open Flavin
binding site for apo-HsRFK, while crystal structures indicate
that closed conformations must be populated in ternary com-
plexes due FlapII displacement toward this cavity (Figure 1
and SP2).17 Therefore, it is accepted that HsRFK must un-
dergo a series of sequential conformational changes during
the catalytic cycle (Figure 4).
3.5
ANP and FLV ligands cooperate
|
in their binding to HsRFK
Although direct FLV binding to the free protein was hardly
observed by ITC or stopped-flow spectrophotometry, the
presence of FLV increased HsRFK affinity for ANP ligands,
particularly when the cation is present (compare ΔG values
for the titrations of free HsRFK and binary mixtures, Table
SP1, Figure 5A and SP6). These observations suggest that, as
in the bacterial RFK modules, (a) FLV ligands have a slow-
binding mode to HsRFK that permits to indirectly estimate
their binding parameters 34,45,46 and (b) ANP and FLV show
cooperativity in their binding.
To evaluate cooperativity, we titrated HsRFK:FLV binary
mixtures with ANP and fitted the resulting thermograms
to a model for heterotrophic interactions applying, as ex-
plained in Materials and Methods section, two complemen-
tary methodologies.34,36,46 Our data, summarized in Table 3,
show that Mg2+ modulates ligand binding cooperativity to
HsRFK. In its presence, FLV and ANP ligands show positive
cooperativity (α > 1). When Mg2+ is absent, RF-ADP and
FMN-ADP binding cooperativity increases (up to 4-fold and
3-fold, respectively), while binding cooperativity becomes
An ordered bi-bi mechanism for mammalian RFKs was
previously proposed, in which RF binding was followed by

ANOZ-CARBONELL Et AL.
10883
|
(B)
(A)
(C)
(D)
FIGURE 6 The apo-HsRFK structural model. (A) Cartoon overlapping and (B) surfaces around the ADP/ATP binding site of apo-HsRFK
structural models for the starting structure (violet) and the snapshot after 5 ns of MD (brown). L5c and FlapI are, respectively, colored in salmon
and olive. Side chains for Lys20 and Asp88 are shown in sticks. (C) Stick representation of the conformation of the PTAN motif and the catalytic
residue Glu78 in the starting structure (violet) and the snapshot after 5 ns of MD (brown). Panels A-C show snapshots of replica 1, and initial
distances among selected atoms are shown as black dashed black lines, while corresponding distances at the end of the simulation are shown as
grey dashed lines. (D) Trajectories for the evolution of the relative distances between residues Lys20 and Asp88, as well as among Thr26 at the
PTAN motif and Ser19 and Glu78 along 5ns MD simulations of apo-HsRFK. Data are shown for the 5 replicates. Simulations carried out at 300 K

10884
ANOZ-CARBONELL Et AL.
|
ATP binding, being ADP and FMN subsequently released
after catalysis.1,2,47 However, our ITC and stopped-flow
experiments demonstrate that FLV (RF and FMN) ligands
hardly interact with HsRFK in the absence of ANP ligands.
The interaction of RF detected by ITC—a minority pathway
in the whole interaction landscape—might be associated to
the slow binding of a few molecules, probably associated to
RF molecules recognizing some motives of the large flavin
binding site in the open conformation expected in apo-Hs-
RFK (Figure 4) (as observed in SpRFK17). Nonetheless, RF
recognition in this state appears scarce, slow and unable
to trigger by itself the structural reorganization of FlapII.
This is confirmed by the absence of changes indicative of
internalization of the isoalloxazine ring in our stopped-
flow experiments as well as in the crystallographic struc-
tures (differences in FlapI and FlapII disposition relative
to HsRFK:ADP:FMN complex, Figures 1, 4 and SP2).
Nonetheless, the larger magnitudes for the entropic and
enthalpic contributions for ATP binding to apo-HsRFK
when compared to the HsRFK-RF mixture (Table SP1 and
Figure SP6) point to structural rearrangements associated
to the RF presence in its binding cavity that favors ANP
binding. Nonetheless, even if RF favors the initial binding
of ANP, the accommodation of the ANP in the cavity and
the establishment of new FlapI-ANP interactions probably
elapses the initial nonproductive RF interaction mode to a
new one. Thus, conformational changes in FlapI have also
an effect in FlapII conformation (see FlapI and FlapII in
HsRFK:ADP:FMN crystal structures, Figure 1 and SP2),
further contributing to place RF in an arrangement com-
patible with catalysis.1,2 Our transient kinetic experiments
support such mechanism, since FlapII displacements can be
indirectly perceived by the changes in flavin fluorescence.
Thus, for mixtures of HsRFK with FMN-ANP, we observed
a slower reorganization process consistent with additional
isoalloxazine burial into the binding pocket by FlapII re-
organization. However, in the case of RF-ATP mixes, k2 is
relatively faster and shows lower amplitude, probably only
reflecting the protein dynamics during the catalytic turnover.
Noticeably, our stopped-flow data also envisage that after
catalysis the flavin binding site of HsRFK remains blocked
by FlapII (Figure 4), making FMN release the rate-limiting
step in FMN production and envisaging that this reaction
process might be controlled by factors different from the pro-
tein itself. In this context, it is worth to note that HsFADS,
besides its FMNAT activity, also operates as a FAD chaper-
one for flavin delivery to its client apoproteins.29,48 Our data
envisage that a similar mechanism might apply in HsRFK
for FMN transfer to HsFADS as well as to FMN dependent
client apoproteins, with direct protein-protein interaction
favoring FMN release from HsRFK. Such tight regulation
agrees with FMN, as well as FAD, being crucial cofactors in
a pletora of enzymes devoted to manage cell bioenergetics.
4.2
Different organisms, different
|
regulatory strategies
Our previous hypothesis of a species-specific inhibition and
activity modulation of the RFK activity in bifunctional FADS
is here reinforced, as well as extended to monofunctional
proteins, by the HsRFK data. Despite the overall structural
similarity among eukaryotic RFK enzymes and prokaryotic
RFK-modules (RMSD of core Cα positions are only 1.2, 1.6,
and 1 Å when comparing HsRFK with SpRFK, CaFADS,
and SpnFADS, respectively), differences among species
occur in the conformation of several structural elements, in-
cluding the FlapI and FlapII loops, and, particularly, the cata-
lytic PTAN motif (Figure SP8A).18 In this structural context,
activity and binding studies reflect some common regulatory
mechanisms, as well as highly relevant differences among
HsRFK and the RFK-modules of CaFADS and SpnFADS.
These variations modulate ligand binding and, consequently,
catalytic cycles, resulting in a variety of species-specific
mechanisms regulating the biosynthesis of flavin cofactors.
Thus, the inhibition of the RFK activity by the products of
the reaction seems to be common for all of them, although in-
hibition potency and mechanism varies with the enzyme. For
example, CaFADS is more strongly inhibited by both reaction
products as deduced from the KIFMN/KMRF and KIADP/KMATP ra-
tios (0.2 vs 1 and 1, and 0.42 vs 1.16 and 11.2, for HsRFK and
SpnFADS, respectively).19,20 Kinetic data for binding are con-
sistent with these differences, since binding of the RFK reac-
tion products is kinetically preferred in CaFADS, while HsRFK
and SpnFADS bind the substrates faster. Thermodynamics also
shows that pathways leading to the catalytic RFK complex are
un-favored respect to those leading to “pseudo-reactive” com-
plexes in CaFADS,20 whereas these differences either do not
exist or favor formation of the catalytic complex in HsRFK
and SpnFADS, respectively (Table SP1).19 Differences are
also observed in heterotropic ligand binding cooperativity.
Cooperativity seems to be determinant in regulating RFK
activity in CaFADS, while in general the HsRFK behavior
is more modest and resembles that of SpnFADS. Thus, our
data indicate that HsRFK, CaFADS, and SpnFADS achieve
the catalytic RFK:RF:ATP complex through mechanisms ex-
hibiting relevant differences. HsRFK and SpnFADS follow a
random sequential binding of the RFK substrates, while sub-
strates binding to CaFADS is concerted. Differences in the
substrates cooperative behavior and magnitude might relate
also to the different conformations of the PTAN motif among
species (Figure SP8A), which points to specific conformational
changes during the RFK activity. In CaFADS, the occupation
of the ANP binding site by RF—when it is in excess—might
prevent the ligand-induced conformational change of this
motif, which is necessary for ATP binding.18 This structural re-
arrangement is not expected to be necessary neither for HsRFK
(Figure 1D,E) nor for SpnFADS.19 Therefore, the absence of

ANOZ-CARBONELL Et AL.
10885
|
inhibition by RF in HsRFK and SpnFADS might be associated
to the minimal rearrangement of PTAN motif during the cata-
lytic cycle of these enzymes.
Medina; formal analysis, E. Anoz-Carbonell, M. Rivero, V.
Polo and M. Medina; investigation, E. Anoz-Carbonell and
M. Rivero; data curation, E. Anoz-Carbonell and M. Medina;
writing—original draft preparation, E. Anoz-Carbonell and
M. Medina; writing—review and editing, E. Anoz-Carbonell
and M. Medina; project administration, M. Medina; funding
acquisition, M. Medina.
For all three proteins release of FMN and ADP products
appears to be the reaction limiting step, but clear differ-
ences are also envisaged in the conformation of such ter-
nary complex in solution. Thus, our stopped-flow analyses
suggest that FMN is not accessible to the solvent when the
HsRFK:FMN:ADP complex is in solution, while it is accessi-
ble when similarly evaluating the RFK modules of CaFADS
and SpnFADS. This might implicate different modes for the
transfer of the newly synthesized FMN to the client proteins.
In this context, we must also consider that while eukaryotic
RFKs are relatively small, monofunctional and monomeric
proteins, their bacterial counterparts have an additional
FMNAT module that duplicates its size.16 Moreover, these
bifunctional enzymes can stabilize quaternary assemblies
with direct interaction of RFK and FMNAT ligand binding
cavities, which potentially will contribute to FMN release
from the RFK module by direct transfer to the FMNAT mod-
ule as well as to regulate FlapII conformation and flavin ac-
cessibility to the solvent.16,25
REFERENCES
1. Karthikeyan S, Zhou Q, Mseeh F, Grishin NV, Osterman AL, Zhang
H. Crystal structure of human riboflavin kinase reveals a beta bar-
rel fold and a novel active site arch. Structure. 2003;11:265-273.
2. Karthikeyan S, Zhou Q, Osterman AL, Zhang H. Ligand binding-in-
duced conformational changes in riboflavin kinase: structural basis
for the ordered mechanism. Biochemistry. 2003;42:12532-12538.
3. Yazdanpanah B, Wiegmann K, Tchikov V, et al. Riboflavin kinase cou-
ples TNF receptor 1 to NADPH oxidase. Nature. 2009;460:1159-1163.
4. Hirano G, Izumi H, Yasuniwa Y, et al. Involvement of riboflavin
kinase expression in cellular sensitivity against cisplatin. Int J
Oncol. 2011;38:893-902.
5. Chen X, Ji B, Hao X, et al. FMN reduces Amyloid-β toxicity in
yeast by regulating redox status and cellular metabolism. Nat
Commun. 2020;11:867.
6. Zou YX, Zhang XH, Su FY, Liu X. Importance of riboflavin kinase
in the pathogenesis of stroke. CNS Neurosci Ther. 2012;18:834-840.
7. Hirano A, Braas D, Fu YH, Ptáček LJ. FAD regulates
CRYPTOCHROME protein stability and circadian clock in mice.
Cell Rep. 2017;19:255-266.
8. Capo-chichi CD, Guéant JL, Lefebvre E, et al. Riboflavin and ribo-
flavin-derived cofactors in adolescent girls with anorexia nervosa.
Am J Clin Nutr. 1999;69:672-678.
9. Capo-Chichi CD, Feillet F, Guéant JL, et al. Concentrations of ri-
boflavin and related organic acids in children with protein-energy
malnutrition. Am J Clin Nutr. 2000;71:978-986.
10. Lee SS, McCormick DB. Thyroid hormone regulation of flavoco-
enzyme biosynthesis. Arch Biochem Biophys. 1985;237:197-201.
11. Barile M, Giancaspero TA, Brizio C, et al. Biosynthesis of flavin
cofactors in man: implications in health and disease. Curr Pharm
Des. 2013;19:2649-2675.
12. Barile M, Brizio C, Valenti D, De Virgilio C, Passarella S. The
riboflavin/FAD cycle in rat liver mitochondria. Eur J Biochem.
2000;267:4888-4900.
13. Lienhart WD, Gudipati V, Macheroux P. The human flavopro-
teome. Arch Biochem Biophys. 2013;535:150-162.
14. Patel MV, Chandra TS. Metabolic engineering of Ashbya gossypii
for enhanced FAD production through promoter replacement of
FMN1 gene. Enzyme Microb Technol. 2020;133:109455.
15. Yruela I, Arilla-Luna S, Medina M, Contreras-Moreira B.
Evolutionary divergence of chloroplasts FAD synthetase proteins.
BMC Evol Biol. 2010;10:311.
16. Herguedas B, Martinez-Julvez M, Frago S, Medina M, Hermoso
JA. Oligomeric state in the crystal structure of modular FAD syn-
thetase provides insights into its sequential catalysis in prokary-
otes. J Mol Biol. 2010;400:218-230.
In conclusion, we present here an integrated thermody-
namic and kinetic description of the catalytic mechanism
of HsRFK that might contribute to the better understanding
of the molecular bases of certain pathologies coursing with
changes in the expression or catalytic efficiency of this pro-
tein. We also report key thermodynamic, kinetic and struc-
tural differences of the regulation of the HsRFK catalytic
cycles relative to bacterial modules performing the same
activity. To date, antimicrobials only targeting the FMNAT
activity of prokaryotic FADSs have been investigated,49
probably because of the overall sequence and structural sim-
ilarity of their RFK module to their eukaryotic counterparts
envisaged specificity compromise and as a consequence dele-
terious effects to the host.50 However, the here presented data
also foresee that the bacterial RFK activity might be consider
a potential antimicrobial target for some bacterial pathogens.
ACKNOWLEDGMENTS
Authors would like to acknowledge the use of Servicios
Generales de Apoyo a la Investigación-SAI, Universidad de
Zaragoza.
CONFLICT OF INTEREST
The authors declare no conflict of interest. The funders had
no role in the design of the study; in the collection, analyses,
or interpretation of data; in the writing of the manuscript, or
in the decision to publish the results.
17. Bauer S, Kemter K, Bacher A, Huber R, Fischer M, Steinbacher
S. Crystal structure of Schizosaccharomyces pombe riboflavin ki-
nase reveals a novel ATP and riboflavin-binding fold. J Mol Biol.
2003;326:1463-1473.
AUTHOR CONTRIBUTIONS
Conceptualization, M. Medina; methodology, E. Anoz-
Carbonell, M. Rivero, V. Polo, A. Velazquez-Camoy and M.

10886
ANOZ-CARBONELL Et AL.
|
18. Herguedas B, Lans I, Sebastián M, Hermoso JA, Martínez-Júlvez
M, Medina M. Structural insights into the synthesis of FMN in
prokaryotic organisms. Acta Crystallogr D Biol Crystallogr.
2015;71:2526-2542.
19. Sebastián M, Velázquez-Campoy A, Medina M. The RFK cata-
lytic cycle of the pathogen Streptococcus pneumoniae shows spe-
cies-specific features in prokaryotic FMN synthesis. J Enzyme
Inhib Med Chem. 2018;33:842-849.
20. Sebastián M, Serrano A, Velázquez-Campoy A, Medina M.
Kinetics and thermodynamics of the protein-ligand interactions
in the riboflavin kinase activity of the FAD synthetase from
Corynebacterium ammoniagenes. Sci Rep. 2017;7:7281.
21. Walsh CT, Wencewicz TA. Flavoenzymes: versatile catalysts in
biosynthetic pathways. Nat Prod Rep. 2013;30:175-200.
22. Sebastián M, Lira-Navarrete E, Serrano A, et al. The FAD syn-
thetase from the human pathogen Streptococcus pneumoniae: a
bifunctional enzyme exhibiting activity-dependent redox require-
ments. Sci Rep. 2017;7:7609.
23. Matern A, Pedrolli D, Großhennig S, Johansson J, Mack M. Uptake
and metabolism of antibiotics roseoflavin and 8-demethyl-8-ami-
noriboflavin in riboflavin-auxotrophic Listeria monocytogenes.
J Bacteriol. 2016;198:3233-3243.
24. Marcuello C, Arilla-Luna S, Medina M, Lostao A. Detection of a
quaternary organization into dimer of trimers of Corynebacterium
ammoniagenes FAD synthetase at the single-molecule level and at
the in cell level. Biochim Biophys Acta. 2013;1834:665-676.
25. Lans I, Seco J, Serrano A, et al. The dimer-of-trimers assembly pre-
vents catalysis at the transferase site of prokaryotic FAD synthase.
Biophys J. 2018;115:988-995.
26. Solovieva IM, Tarasov KV, Perumov DA. Main physicochemical
features of monofunctional flavokinase from Bacillus subtilis.
Biochemistry (Mosc). 2003;68:177-181.
27. Sebastián M, Arilla-Luna S, Bellalou J, Yruela I, Medina M. The
biosynthesis of flavin cofactors in Listeria monocytogenes. J Mol
Biol. 2019;431:2762-2776.
28. Barile M, Giancaspero TA, Leone P, Galluccio M, Indiveri C.
Riboflavin transport and metabolism in humans. J Inherit Metab
Dis. 2016;39:545-557.
of cooperative ligand binding by isothermal titration calorimetry.
Biophys J. 2006;91:1887-1904.
36. Vega S, Abian O, Velazquez-Campoy A. A unified framework
based on the binding polynomial for characterizing biological sys-
tems by isothermal titration calorimetry. Methods. 2015;76:99-115.
37. Olsson MHM, Søndergaard CR, Rostkowski M, Jensen JH.
PROPKA3: consistent treatment of internal and surface residues in
empirical pKa predictions. J Chem Theory Comput. 2011;7:525-537.
38. Delano WL. PyMOL: an open-source molecular graphics tool.
CCP4 Newsletter Pro Crys. 2002;40:82-92.
39. Abraham MJ, Murtola T, Schulz R, et al. GROMACS: High perfor-
mance molecular simulations through multi-level parallelism from
laptops to supercomputers. SoftwareX. 2015;1-2:19-25.
40. Duan Y, Wu C, Chowdhury S, et al. A point-charge force field
for molecular mechanics simulations of proteins based on con-
densed-phase quantum mechanical calculations. J Comput Chem.
2003;24:1999-2012.
41. Humphrey W, Dalke A, Schulten K. VMD: visual molecular
dynamics. J Mol Graph. 1996;14:33-38, 27-38.
42. Lee SS, McCormick DB. Effect of riboflavin status on hepatic activities
of flavin-metabolizing enzymes in rats. J Nutr. 1983;113:2274-2279.
43. Pedrolli DB, Nakanishi S, Barile M, et al. The antibiotics roseo-
flavin and 8-demethyl-8-amino-riboflavin from Streptomyces
davawensis are metabolized by human flavokinase and human
FAD synthetase. Biochem Pharmacol. 2011;82:1853-1859.
44. Jin C, Yao Y, Yonezawa A, et al. Riboflavin transporters RFVT/
SLC52A mediate translocation of riboflavin, rather than FMN or
FAD, across plasma membrane. Biol Pharm Bull. 2017;40:1990-1995.
45. Bollen YJ, Westphal AH, Lindhoud S, van Berkel WJ, van Mierlo
CP. Distant residues mediate picomolar binding affinity of a pro-
tein cofactor. Nat Commun. 2012;3:1010.
46. Martínez-JúlvezM,MedinaM,Velázquez-CampoyA.Bindingther-
modynamics of ferredoxin:NADP+ reductase: two different pro-
tein substrates and one energetics. Biophys J. 2009;96:4966-4975.
47. Yamada Y, Merrill AH Jr, McCormick DB. Probable reaction
mechanisms of flavokinase and FAD synthetase from rat liver.
Arch Biochem Biophys. 1990;278:125-130.
48. Torchetti EM, Bonomi F, Galluccio M, et al. Human FAD synthase
(isoform 2): a component of the machinery that delivers FAD to
apo-flavoproteins. FEBS J. 2011;278:4434-4449.
49. Sebastián M, Anoz-Carbonell E, Gracia B, et al. Discovery of an-
timicrobial compounds targeting bacterial type FAD synthetases.
J Enzyme Inhib Med Chem. 2018;33:241-254.
50. Serrano A, Ferreira P, Martínez-Júlvez M, Medina M. The pro-
karyotic FAD synthetase family: a potential drug target. Curr
Pharm Des. 2013;19:2637-2648.
29. Giancaspero TA, Colella M, Brizio C, et al. Remaining challenges
in cellular flavin cofactor homeostasis and flavoprotein biogenesis.
Front Chem. 2015;3:30.
30. Serrano A, Sebastián M, Arilla-Luna S, et al. The trimer inter-
face in the quaternary structure of the bifunctional prokaryotic
FAD synthetase from Corynebacterium ammoniagenes. Sci Rep.
2017;7:404.
31. Serrano A, Frago S, Herguedas B, Martinez-Julvez M, Velazquez-
Campoy A, Medina M. Key residues at the riboflavin kinase
catalytic site of the bifunctional riboflavin kinase/FMN adenylyl-
transferase from Corynebacterium ammoniagenes. Cell Biochem
Biophys. 2013;65:57-68.
32. Vogt AD, Di Cera E. Conformational selection or induced fit?
A critical appraisal of the kinetic mechanism. Biochemistry.
2012;51:5894-5902.
SUPPORTING INFORMATION
Additional Supporting Information may be found online in
the Supporting Information section.
33. Frago S, Velázquez-Campoy A, Medina M. The puzzle of ligand
binding to Corynebacterium ammoniagenes FAD synthetase. J
Biol Chem. 2009;284:6610-6619.
34. Martinez-Julvez M, Abian O, Vega S, Medina M, Velazquez-
Campoy A. Studying the allosteric energy cycle by isothermal ti-
tration calorimetry. Methods Mol Biol. 2012;796:53-70.
35. Velázquez-Campoy A, Goñi G, Peregrina JR, Medina M. Exact
analysis of heterotropic interactions in proteins: Characterization
How to cite this article: Anoz-Carbonell E, Rivero
M, Polo V, Velázquez-Campoy A, Medina M. Human
riboflavin kinase: Species-specific traits in the
biosynthesis of the FMN cofactor. The FASEB
Journal. 2020;34:10871–10886. https://doi.
org/10.1096/fj.202000566R