Histone demethylase LSD1 is a folate-binding protein

Article abstract of DOI:10.1021/bi200247b

Methylation of lysine residues in histones has been known to serve a regulatory role in gene expression. Although enzymatic removal of the methyl groups was discovered as early as 1973, the enzymes responsible for their removal were isolated and their mechanism of action was described only recently. The first enzyme to show such activity was LSD1, a flavin-containing enzyme that removes the methyl groups from lysines 4 and 9 of histone 3 with the generation of formaldehyde from the methyl group. This reaction is similar to the previously described demethylation reactions conducted by the enzymes dimethylglycine dehydrogenase and sarcosine dehydrogenase, in which protein-bound tetrahydrofolate serves as an accepter of the formaldehyde that is generated. We now show that nuclear extracts of HeLa cells contain LSD1 that is associated with folate. Using the method of back-scattering interferometry, we have measured the binding of various forms of folate to both full-length LSD1 and a truncated form of LSD1 in free solution. The 6R,S form of the natural pentaglutamate form of tetrahydrofolate bound with the highest affinity (K _d = 2.8 μM) to full-length LSD1. The fact that folate participates in the enzymatic demethylation of histones provides an opportunity for this micronutrient to play a role in the epigenetic control of gene expression.

Full text of DOI:10.1021/bi200247b

ARTICLE
pubs.acs.org/biochemistry
Histone Demethylase LSD1 Is a Folate-Binding Protein
Zigmund Luka,† Frank Moss,‡ Lioudmila V. Loukachevitch,† Darryl J. Bornhop,*,‡,§ and Conrad Wagner*,†
†Department of Biochemistry, Vanderbilt University Medical Center, ‡Department of Chemistry, and
§The Vanderbilt Institute for Chemical Biology, Vanderbilt University, Nashville, Tennessee 37232, United States
ABSTRACT: Methylation of lysine residues in histones has been known to
serve a regulatory role in gene expression. Although enzymatic removal of
the methyl groups was discovered as early as 1973, the enzymes responsible
for their removal were isolated and their mechanism of action was
described only recently. The first enzyme to show such activity was
LSD1, a flavin-containing enzyme that removes the methyl groups from
lysines 4 and 9 of histone 3 with the generation of formaldehyde from the
methyl group. This reaction is similar to the previously described
demethylation reactions conducted by the enzymes dimethylglycine
dehydrogenase and sarcosine dehydrogenase, in which protein-bound
tetrahydrofolate serves as an accepter of the formaldehyde that is gener-
ated. We now show that nuclear extracts of HeLa cells contain LSD1 that is
associated with folate. Using the method of back-scattering interferometry,
we have measured the binding of various forms of folate to both full-length LSD1 and a truncated form of LSD1 in free solution. The
6R,S form of the natural pentaglutamate form of tetrahydrofolate bound with the highest affinity (Kd = 2.8 μM) to full-length LSD1.
The fact that folate participates in the enzymatic demethylation of histones provides an opportunity for this micronutrient to play a
role in the epigenetic control of gene expression.
olate cofactors in eukaryotic cells have been considered to be
distributed between both the cytosol and mitochondria
folate as substrates. They suggested that the replitase complex
was involved in the metabolic control of DNA replication. In a
recent series of publications, the Stover laboratory reported
strong evidence of the nuclear localization of the most important
folate-dependent enzymes in the thymidylate biosynthesis path-
way: serine hydroxymethyltransferase (SHMT), thymidylate
synthase (TS), and dihydrofolate dehydrogenase (DHFR).9,10
SHMT uses tetrahydrofolate (THF) for synthesis of 5,10-
methylene tetrahydrofolate, which is then used by TS for
synthesis of thymidylate with dihydrofolate as a product.
The mechanism of histone H3 demethylation by demethylase
LSD1 attracted our attention, because of its similarity to the
mechanism of demethylation of dimethylglycine by DMGDH
and sarcosine by SDH.2À4,11,12 In all three enzymes, the first step
of demethylation is the oxidation of N-methyl groups to an imine
intermediate. The latter is nonenzymatically hydrolyzed to the
demethylated amine and formaldehyde. Most importantly, both
DMGDH and SDH contain tightly bound THF, which was
shown to form 5,10-methylene tetrahydrofolate, thereby serving
as a trap for formaldehyde to prevent cross-linking of proteins.4
This, and the preszence of folate in eukaryotic nuclei, sug-
gested that LSD1 might also use tetrahydrofolate to form 5,
10-methylene tetrahydrofolate as a result of histone demethyla-
tion. Methylation on the N-terminal tails of histone lysines serves
as an epigenetic control mechanism.13 It should be noted that
F
where they are used for transfer of one-carbon units between
numerous metabolic pathways.1 All of the natural folate cofactors
are polyglutamylated intracellularly and bind to the respective
enzymes more tightly than the corresponding forms that contain
only a single glutamate residue. In most cases, however, the
monoglutamate forms perform the same reaction as the natural,
polyglutamated forms.1 Early studies from our laboratory
showed that the polyglutamated form of tetrahydrofolate was
tightly bound to the mitochondrial enzymes, dimethylglycine
dehydrogenase (DMGDH) and sarcosine dehydrogenase (SDH).
Both of these enzymes conduct demethylation reactions in which
the N-methyl group is first oxidized to the imine followed by
hydrolysis to liberate formaldehyde. We showed that bound
tetrahydrofolate reacts with the formaldehyde that is generated
to produce N-5,10-methylenetetrahydrofolate, presumably to
protect the enzymes from cross-linking by the formaldehyde
formed at the active site.2À4
Several early studies showed that, in addition to cytosol and
mitochondria, a small amount of folate was also found in the
nuclei.5,6 Earlier studies from our laboratory showed that 24 h
after injection of radioactive folate into rats and subsequent
analysis of the liver cytosol, mitochondrial, and nuclear fractions,
∼2.5% of the radioactivity was located in the nuclei.7 The role of
folate in the nucleus was unknown until the work of Prem veer
Reddy and Pardee.8 They identified a multienzyme complex that
they named “replitase” that contained the enzymes thymidylate
synthase and dihydrofolate reductase that used different forms of
Received: February 17, 2011
Revised:
April 19, 2011
Published: April 21, 2011
r
2011 American Chemical Society
4750
dx.doi.org/10.1021/bi200247b Biochemistry 2011, 50, 4750–4756
|

Biochemistry
ARTICLE
enzymatic demethylation of histones was first reported in 1973,14
but the enzymes responsible were not isolated. Recently, a group
of histone demethylases were discovered in the nucleus.15À20
There are two classes of these enzymes that perform the removal
of a methyl group from different methylated lysine residues in the
histones. One class of these lysine-specific demethylases, LSD1
and LSD2, consists of amine oxidases containing FAD as the
electron acceptor to first oxidize the N-methyl amine to an imine.
They catalyze the demethylation of mono- and dimethylated
lysine residues, 4 and 9, on histone H3 (H3 K4 and H3 K9,
respectively). The second class of histone demethylases is a JmjC
family of iron (II)-R-ketoglutarate-dependent histone demethy-
lases for conducting the oxidation of the N-methylamine to the
imine. The properties of these enzymes are discussed in recent
reviews.19,20
We hypothesized that LSD1 contains bound polyglutamylated
THF that could play a similar role in trapping formaldehyde as in
the case of DMGDH and SDH. To test this hypothesis, we
studied the localization of folate in nuclei and its association with
LSD1 as well as the binding of various forms of tetrahydrofolate
with LSD1. We now show that LSD1 is associated with folate in
the nuclei of HeLa cells and that various forms of folate bind
tightly in vitro to purified LSD1. The natural form of THF
polyglutamate bound with the highest affinity.
Figure 1. Chemical structures of folates used in the binding study. The
monoglutamate and pentaglutamate forms of tetrahydrofolate are
shown. In the monoglutamate form, R indicates the position of an H
atom at the N-5 position in THF, a CH3 group in 5-methyl-THF, and a
CHO group in 5-formyl-THF.
for thrombin (LVPRGS). The predicted molecular mass of that
protein with an N-terminal methionine is 95190 Da. By analysis
of a tryptic digest of LSD1 using MALDI and LCÀMS/MS
methods at the Proteomics Facility of the Vanderbilt University
Mass Spectrometry Center, it was determined that N-terminal
methionine in the expressed protein had been removed; there-
fore, the molecular mass of the expressed LSD1 was 95059 Da.
The plasmid for the N-terminally truncated LSD1 was a
generous gift from P. A. Cole (Johns Hopkins University,
Baltimore, MD). In this plasmid, part of LSD1 cDNA (amino
acids 171À852) was cloned into a pGEX-6P1 expression vector
from which it was expressed as a fusion protein with glutathione
S-transferase (GST) and a site for PreScission protease.
Methods. Size-Exclusion Chromatography of Nuclear Protein
Extracts. A nuclear extract (75 μL) from HeLa cells (BIOMOL)
containing 695 μg of protein was mixed with column buffer
[50 mM CHES, 50 mM HEPES, 100 mM NaCl, and 20 mM
β-mercaptoethanol (pH 7.85)] and the mixture applied to a
Superose-12 column (Pharmacia) equilibrated with column buffer.
Protein was eluted using an AKTA-Purifier system (Amersham)
at a flow rate of 0.5 mL/min, and fractions of 0.5 mL were
collected. Protein elution was monitored by absorbance at
280 nm. Each fraction was also analyzed for the presence of
folate by using a microbiological method (Lactobacillus
casei24) and for the presence of LSD1 by dot blotting with
antibodies against LSD1. The Superose column was cali-
brated with cytochrome c, lysozyme, chymotrypsin, carbonic
anhydrase, ovalbumin, BSA, alcohol dehydrogenase, potato
β-amylase, and aldolase without added urea and also with
column buffer containing 8 M urea. The void volume, Vo, was
determined with blue dextran, and the total solvent-accessible
volume, Vt, was determined with acetone as reported previo-
usly.25
’ EXPERIMENTAL PROCEDURES
Materials. Formaldehyde dehydrogenase, peroxidase, NAD,
4-aminoantipurine, 3,5-dichloro-2-hydroxybenzenesulfonic acid,
and all general chemicals for buffers and microbiological media
were from Sigma. Dimethylhistone H3(Lys4) peptide and anti-
bodies against LSD1 were from Millipore. PreScission protease
was from GE Healthcare.
The following were gifts from EPROVA: the stereoisomers of
tetrahydrofolate monoglutamate [(6R,S)-THF-Glu1] and the
natural 6S stereoisomers of tetrahydrofolate monoglutamate
[(6S)-THF-Glu1, (6S)-5-methyl-THF-Glu1, and (6S)-5-formyl-
THF-Glu1]. Tetrahydrofolate pentaglutamate [(6R,S)-THF-
Glu5] was synthesized from pteroylpenta-γ-L-glutamic acid
(Schircks Laboratory) by reduction with NaBH4 in the presence
of Pb(NO3)2 according to a published procedure.21,22 The product
was spectrophotometrically pure THF (λmax = 298 nm at
pH 7.0). It was desalted on a Bio-Gel P-2 column equilibrated
with 40 mM ammonium acetate (pH 7.0) and 100 mM
β-mercaptoethanol. After the sample had been desalted,
β-mercaptoethanol was added to a final concentration of 0.4 M
and THF preparations were kept in small aliquots at À20 °C
under argon. Solutions of other folates were prepared in 0.4 M
β-mercaptoethanol and were kept at À20 °C. Folate concentra-
tions were determined spectrophotometrically in 20 mM potas-
sium phosphate buffer (pH 7.0) and 14 mM β-mercaptoethanol
by using the following extinction coefficients: 29.1 mMÀ1 cmÀ1
at 298 nm for all forms of THF, 31.7 mMÀ1 cmÀ1 at 290 nm for
(6S)-5-methyl-THF-Glu1, and 37.2 mMÀ1 cmÀ1 at 285 nm for
(6S)-5-formyl-THF-Glu1.23 The chemical structures of folates
used in this work are shown in Figure 1.
The plasmid for full-length LSD1 expression was a generous
gift from Y. Shi (Harvard University, Cambridge, MA). In this
plasmid, a full-length human cDNA for LSD1 was cloned into the
pET-15b expression vector. The cloned cDNA contained an
additional sequence at the N-terminus, MGSSHHHHHHSS-
GLVPRGSNF, which included six histidines and a cleavage site
Full-Length Protein. For expression of full-length LSD1,
Escherichia coli BL21(DE3) competent cells were transformed
with a plasmid containing full-length human LSD1 with an
attached N-terminal His tag in the pET-15b expression vector
4751
dx.doi.org/10.1021/bi200247b |Biochemistry 2011, 50, 4750–4756

Biochemistry
ARTICLE
using a standard heat-shock protocol. It was cultured overnight in
30 mL of LB medium with 100 mg/L ampicillin. The over-
night culture was inoculated into 1 L of Terrific Broth containing
100 mg/L ampicillin. It was cultured at 37 °C until the absor-
bance at 600 nm reached 1.4À1.6. At this time, IPTG was added
to a final concentration of 0.5 mM and incubation continued for
7 h at 25 °C. The cells were harvested by centrifugation, washed
with cold 0.1 M Tris-HCl (pH 7.5), and kept at À20 °C until the
protein was purified.
was eluted by 30 mM reduced GSH in 50 mM HEPES (Na salt)
(pH 8.0). Fractions containing GST-LSD protein were pooled
and concentrated to 0.5À1 mL.
GST from the fusion protein was removed by using PreScis-
sion protease. The buffer in the sample was first changed to
PreScission buffer [50 mM HEPES (Na salt) (pH 7.8), 150 mM
NaCl, and 1 mM DTT]. The fusion protein was treated in that
buffer with PreScission protease for 16À20 h at 3 °C. After the
protease treatment, the N-terminally truncated LSD1 was easily
separated from GST and the PreScission protease by a small DE-
52 column equilibrated with 25 mM HEPES (Na salt) (pH 7.8).
Under these conditions, LSD1 did not bind to the column but
other proteins did. The final purity of LSD1 was at least 97% as
determined by sodium dodecyl sulfate (SDS) electrophoresis
with Coomassie staining.
Purification of Full-Length LSD1. The initial procedure used
for purification of full-length LSD1 based on the use of a
Ni-agarose affinity chromatography resulted in a relatively low
yield.15 A much better purification of LSD1 could be achieved by
combination of ammonium sulfate precipitation and ion-exchange
chromatography on DE-52 cellulose. In this protocol, the collec-
ted E. coli cells were sonicated in homogenization buffer in which
the components were 20 mM Tris-HCl buffer (pH 7.8), 14 mM
β-mercaptoethanol, and protease inhibitors (protease inhibitor
cocktail from Sigma or combination of leupeptin, pepstatin, and
PMSF). Cell debris was removed by centrifugation in a Sorvall RC-5
centrifuge in an SS-34 rotor at 18000 rpm for 30 min at 4 °C.
Proteins in the supernatant were fractionated by ammonium
sulfate precipitation. The protein fraction that precipitated
between 26 and 36% ammonium sulfate saturation contained
most of the LSD1 protein. This protein fraction was dissolved in
DE-52 buffer [20 mM Tris-HCl (pH 7.8) containing 5 mM
β-mercaptoethanol] and desalted on a 20 mL column of Bio-Gel
P-30 equilibrated with DE-52 buffer. The desalted protein
fraction was loaded onto a 20 mL Whatman DE-52 column
equilibrated with DE-52 buffer. The column was washed with
DE-52 buffer containing 35 mM NaCl. After being washed, the
LSD1 protein was eluted with DE-52 buffer, containing 65 mM
NaCl. Fractions containing purified LSD1 were pooled and
concentrated using Millipore concentrators (YM-50). This method
routinely provided protein with a purity of >96% with a yield of
2.5 mg of LSD1/L of culture. For storage, glycerol was added to
LSD1 samples to a final concentration of 40%. This solution was
kept at À20 °C in small aliquots and was stable for months
without loss of enzyme activity.
LSD1 Activity Assay. LSD1 activity was assayed by using
coupled assays for hydrogen peroxide or for formaldehyde as
described elsewhere.17 When activity was determined by mea-
suring the level of hydrogen peroxide, the concentrations of the
components in the 100 μL reaction mixture were as follows:
25 mM potassium phosphate buffer (pH 7.2), 0.1 mM 4-ami-
noantipurine, 1.0 mM 3,5-dichloro-2-hydroxybenzenesulfonic
acid, 1 unit of horseradish peroxidase, 17 μM dimethyl-histone
H3(Lys4) peptide as a substrate, and 1À3 μM LSD1. The
course of reaction was monitored by absorbance at 505 nm in
Ultra Micro Cells on a Shimadzu 2401-PC spectrophotometer.
When LSD1 activity was monitored by formaldehyde produc-
tion, the concentrations of the components in the 100 μL
reaction mixture were as follows: 20 mM Tris-HCl (pH 7.5),
17 μM dimethyl-histone H3(Lys4) peptide as the LSD1 sub-
strate, 2 mM NAD, 1À3 μM LSD1, and 0.1 unit of formalde-
hyde dehydrogenase. The course of reaction was monitored by
the increase of absorbance at 340 nm in 70 μL Ultra Micro Cells
on a Shimadzu 2401-PC spectrophotometer.
LSD1ÀFolate Binding. LSD1Àfolate binding was studied by
two methods. Preliminary data were obtained by using separa-
tion of unbound folate from a solution of LSD1 and folate with
centrifugal devices (concentrators). In initial binding experi-
ments, aliquots of folate solutions were mixed with LSD1 in a
total volume of 300 μL and incubated at room temperature in the
dark for 1 h. Controls were prepared with the same volume of
folate solutions in the protein buffer but without LSD1. After
incubation, the LSD1Àfolate reaction mixtures were loaded on
centrifugal filters (Centricon YM-50, Millipore) and centrifuged
in an Eppendorf microcentrifuge at 5000 rpm for 10À13 min,
yielding 100À120 mL of filtrate. Control samples were treated
the same way. Concentrations of various folates in the filtrates
were determined by absorption spectra in the 240À400 nm
range on a Shimadzu 2401-PC spectrophotometer using Ultra
Micro Cells (Shimadzu) with a 70 μL working volume.
N-Terminally Truncated LSD1. Expression of N-terminally
truncated LSD1 was similar to that used for the full-length
protein with some minor differences in IPTG concentration,
temperature, and time of incubation. Collected cells were stored
at À20 °C until the protein was purified.
The protocol for protein purification was based on the
published method18 with some substantial differences. Use of
ammonium sulfate precipitation greatly reduces the procedure
time with no effect on the final enzyme preparation activity. In
our protocol, cells were sonicated in homogenization buffer
[50 mM HEPES (K salt) (pH 7.8), 150 mM LiCl, 1 μg/mL
leupeptin and pepstatin, 1 mM PMSF, 5 mM EDTA, 1 mg/mL
lysozyme, 1 μg/mL DNase, and 10 mM DTT]. The homogenate
was centrifuged at 18000 rpm in an SS-34 rotor on a Sorvall
centrifuge, and the supernatant was fractionated by ammonium
sulfate. Proteins that precipitated between 24 and 36% ammo-
nium sulfate saturation were dissolved in the glutathione (GSH)
buffer [50 mM HEPES (K salt) (pH 7.8), 150 mM LiCl, 1 μg/mL
leupeptin and pepstatin, 0.5 mM PMSF, 5 mM EDTA, and 5 mM
β-mercaptoethanol], and the solution was clarified by centrifugation.
The crude extract was loaded on the GSH-agarose (Sigma)
column equilibrated with GSH buffer, and unbound proteins
were washed out with GSH buffer. The GST-LSD fusion protein
Measurement of LSD1ÀFolate Binding by Back-Scattering
Interferometry (BSI). The majority of the binding experiments
were performed using BSI, described previously.26 Briefly, the
instrument is composed of a simple optical train that consists of a
HeNe laser, a microfluidic chip with a channel etched in boro-
silicate glass, and a CCD camera. The laser impinges on the
samples in the microfluidic channel, producing a high-contrast
interference pattern, which is reflected onto the CCD array.
Changes in the refractive index (RI) of the solution contained
within the channel cause spatial shifts in the interference pattern,
which are monitored and recorded using software designed
in-house.
4752
dx.doi.org/10.1021/bi200247b |Biochemistry 2011, 50, 4750–4756

Biochemistry
ARTICLE
Solutions of LSD1 and folates used in BSI experiments were
prepared in 50 mM Tris-HCl (pH 7.5), 100 mM NaCl, and 35 mM
β-mercaptoethanol, concentrations required to prevent oxidation of
oxygen-labile forms of folate. All solutions were filtered through a
0.2 μm filter and degassed prior to binding experiments.
Binding experiments were performed in an end point format.27,28
For the binding samples, a constant amount of LSD1 was mixed
off-line with increasing concentrations of folates and incubated at
room temperature in the dark for at least 2 h. This assures that
equilibrium has been reached prior to the samples being injected
into the BSI instrument.
Specific binding is quantified in BSI experiments by measuring
the difference in the signal between the control or blank and the
binding pair. Here each form of folate served as its own blank;
therefore, a calibration curve of concentration versus BSI signal
for the ligand alone was constructed, by preparing samples over
the same concentration range that was used in the binding experi-
ments. Then these values were used to correct the binding curves for
bulk RI changes that occur due to the increasing concentration of
the folates, particularly at higher ligand concentrations.
Figure 2. Folate and LSD1 in HeLa nuclear extract. The solid line
represents the absorbance at 280 nm. Fractions of 0.5 mL were collected.
The dashed line represents the concentration of folate in each fraction.
The presence of LSD1 in the fractions was determined immunologically
by dot blot. This is positioned to coincide with the fractions that were
collected and are an indication of the presence of LSD1 in the eluted
fractions.
The difference between the calibration and binding signals was
plotted versus ligand concentration, yielding a saturation binding
isotherm. The corrected binding data were analyzed with GraphPad
Prism (GraphPad Software Inc., San Diego, CA). A one-site
binding algorithm was chosen because LSD1 is a monomer that
contains a single bound FAD. Therefore, it seems logical that
there is one THF binding site per LSD1 molecule. To test this,
we performed a statistical comparison of the one-site binding and
two-site binding models for the natural ligand, THF-Glu5, using
both an F-test and an Akaike’s information criterion (AIC) test.
Both tests supported the one-site binding model (94.0% con-
fidence using the F-test and 99.3% confidence using the AIC
test). This was further supported by the approximately 10-fold
increase in the standard error of the calculated Kd values when the
two-site binding model was applied. Using the one-site binding
model, the Kd was found to be 2.77 ( 0.46 μM, while the two-site
model calculates Kd1 to be 0.00 ( 4.39 μM and Kd2 to be 3.77 (
4.96 μM. The same trend was observed for the other forms of folate.
Folate Assay. Folate was assayed by using the L. casei method
as described by Horne et al.24 Briefly, the protein samples were
mixed with 10 volumes of extraction buffer [50 mM HEPES,
50 mM CHES (pH 7.85), 28 mM β-mercaptoethanol, and 2%
ascorbate] and heated for 10 min in boiling water. Precipitated
proteins were separated by centrifugation, and the supernatant
was treated with conjugase for 3 h at 37 °C. Reaction mixtures
were boiled again and centrifuged, and the supernatant was used
for analysis according to the original protocol.
purification of histone demethylases.29 Therefore, we used nuclei
from HeLa cells to look for the possible presence of complexes of
LSD1 and folate coenzymes. A commercial preparation of
dialyzed nuclear extract (BIOMOL) was fractionated by size-
exclusion chromatography. Measurement of total folate in the
extract indicated that it contained 2.2 nmol of total folate/mg
of nuclear protein. The extract was applied to a Superose-12
column as described in Methods. Elution was monitored by
absorbance at 280 nm; fractions were assayed for the presence of
folate by a microbiological assay and for the presence of LSD1 by
an immunoassay. The results are shown in Figure 2. Eluted
proteins were detected in almost all fractions, from the void
volume (Vo, 8.0 mL) to the total liquid volume (Vt, 20.1 mL).
Most of the folate eluted in a broad band from 16 to 22 mL,
indicating it was unbound folates. Because the nuclear extract had
been dialyzed prior to chromatography, this indicated that most
of the folates that eluted between 16 and 22 mL either had been
loosely protein-bound and then dissociated during passage
through the column or had been unbound in the nuclear extract
and not removed by dialysis. Control experiments showed that
THF pentaglutamate (961 Da) eluted from this column at
17.5 mL. This corresponds to a globular protein of ∼10 kDa
[cytochrome c (12400 Da) eluted at 16 mL]. Because of the
asymmetry of the folates, they behave in solution as larger
molecules and require dialysis membranes with a larger pore size
through which to pass. It is therefore highly likely that all folate
polyglutamates were not removed by dialysis prior to passing
through the Superose column because the preparation from
BIOMOL was dialyzed using a membrane with a 10 kDa cutoff.
In addition to the major folate peak, there was a smaller folate
peak at 12À15 mL. This corresponds to the elution volume of
globular proteins of 30À100 kDa as determined by calibration of
the column. The folate eluted in this region was probably tightly
protein bound.
Other Methods. The concentration of protein samples was
determined by the BCA method (BCA Protein Assay kit, Pierce)
with bovine serum albumin as a standard. Protein spectra were
recorded on a Shimadzu 2401-PC spectrophotometer. Protein
purity was determined by SDS electrophoresis with Coomassie
staining. Immunoblotting was conducted according to a standard
immunoblotting protocol using a nitrocellulose membrane,
primary and secondary antibody binding, and visualization using
a SuperSignal West Femto Maximum Sensitivity Substrate Kit
(Pierce).
The presence of LSD1 in these fractions was determined
immunologically and appears most abundantly in the void
volume where multimeric protein complexes of LSD1 would
be found.16 A smaller but clearly distinct amount of LSD1 was
eluted at 12À13 mL that contained the shoulder of the first peak
of the early eluting folate. Globular proteins of 100À120 kDa
would elute in this region, and LSD1 appears to be ∼100 kDa as
’ RESULTS
Presence of Folate and LSD1 in Nuclear Protein Extracts
from HeLa Cells. HeLa cells have been used as a source for
4753
dx.doi.org/10.1021/bi200247b |Biochemistry 2011, 50, 4750–4756

Biochemistry
ARTICLE
Figure 3. Representative binding curves obtained by BSI. The BSI signal is plotted vs ligand concentration for (A) (6R,S)-THF-Glu1 and (B) (6R,S)-
THF-Glu5.
determined by SDS electrophoresis. These data are interpreted
to indicate that LSD1 in the nuclear protein extract exists mainly
as a part of a multimeric protein complex but also as a free
protein. An important conclusion is that fractions in which free
LSD1 is eluted also contain folate (Figure 2). It is reasonable to
speculate that an LSD1Àfolate complex eluted at this volume. To
verify this conclusion, we conducted a series of experiments to
measure the in vitro binding of selected forms of folate to LSD1.
LSD1ÀTHF Interaction under Steady-State Conditions.
Interaction of folates with LSD was studied by two methods
using, first, membrane separation of free folate and, second, BSI
as described in Methods. Preliminary data, using membrane
separation, indicated an LSD1Àfolate interaction with very high
binding constants (data not shown). After LSD1Àfolate binding
had been established using this method, we used the more
sophisticated BSI method for a detailed characterization of
LSD1Àfolate interaction in solution without any physical se-
paration methods, eliminating the potential of a biased result due
to nonequilibrium conditions.
In both methods, binding was analyzed in 50 mM Tris-HCl
(pH 7.5), 100 mM NaCl, and 35 mM β-mercaptoethanol. That
concentration of reducing agent was established in separate
experiments to be sufficient to protect THF from oxidation for
at least for 6 h as determined spectrophotometrically.
We used four forms of folate [THF-Glu1, THF-Glu5, 5-for-
myl-THF-Glu, and 5-methyl-THF-Glu1 (Figure 1)] and deter-
mined the relative binding affinities for these species. Typical
binding isotherms are shown in Figure 3. The first form used was
a mixture of stereoisomers, (6R,S)-THF-Glu5. It was found that
the Kd for that form of folate was 2.8 μM (Table 1). To determine
whether polyglutamation of THF affects affinity, we studied
binding of (6R,S)-THF-Glu1 to LSD1. Indeed, the binding
affinity of the monoglutamate form was significantly lower than
that of the polyglutamate species with a Kd value of 30 μM, as
compared to a Kd of 2.8 μM.
Table 1. Binding of Different Forms of Folate to LSD1a
folate
Kd (μM)
(6R,S)-THF-Glu5
2.77 ( 0.46
30.3 ( 3.8
19.5 ( 5.0
46.3 ( 12.3
72.9 ( 16.2
4.02 ( 0.92
65.1 ( 14.0
(6R,S)-THF-Glu1
(6S)-THF-Glu1
(6S)-5-CH3-THF-Glu1
(6S)-5-CHO-THF-Glu1
(6R,S)-THF-Glu5 with truncated LSD1
(6S)-5-CHO-THF-Glu1 without BME
a Values are the means ( the standard error of the mean of dissociation
constants.
There are a variety of folate coenzymes that participate in the
transfer of one-carbon units, but only THF has the ability to
combine nonenzymatically with formaldehyde. A second indica-
tion of biological significance would be that these other forms of
folate would bind to LSD1 with a lower affinity. To confirm this,
we performed folateÀLSD1 binding experiments with (6S)-
5-formyl-THF-Glu1. As expected, the affinity of LSD1 for
(6S)-5-formyl THF-Glu1 was much lower than that for (6S)-
THF-Glu1 with a Kd value of 70 μM (Table 1). The order of
affinity for the folate species with full-length LSD is as follows:
(6R,S)-THF-Glu5 > (6S)-THF-Glu1 > (6R,S)-THF-Glu1 > (6S)-
5-CH3-THF-Glu1 > (6S)-5-formyl-THF-Glu1.
An important methodological question in the study of LSD1À
THF interaction is whether the use of a high concentration of
reducing reagent to prevent the extremely labile THF from being
oxidized could influence the interaction. The high concentration
(35 mM) of β-mercaptoethanol used in our study also might
affect the conformation of LSD1 that could ultimately affect the
interaction with folate. We addressed this question by comparing
binding of (6S)-5-formyl THF-Glu1, which is completely stable
in air, to LSD1 in the presence and absence of β-mercaptoetha-
nol. It was found that use of 35 mM β-mercaptoethanol does not
change the Kd value for (6S)-5-formyl-THF-Glu1 binding
(73 μM in the presence and 65 μM in the absence of
β-mercaptoethanol).
If binding of THF to LSD1 is specific (indicating biological
function), then two things should be observed. First, the natural
isomer, (6S)-THF, should bind to LSD1 with a higher affinity.
We verified that by analysis of the interaction of LSD1 with (6S)-
THF-Glu1. It was found (Table1) that the affinity of the natural
6S isomer is greater, with a Kd almost 2-fold lower (19 μM) than
that of the 6R,S mixture (30 μM).
Having established that (6R,S)-THF-Glu5 binds with high
specificity, we found it was important to identify the binding
site in the protein. A definitive answer to this question can be
4754
dx.doi.org/10.1021/bi200247b |Biochemistry 2011, 50, 4750–4756

Biochemistry
ARTICLE
obtained with a high-resolution crystal structure, work that is in
progress in our laboratory. While it is less definitive, some insight
into the binding site of folate can be gained by analysis of the
interaction of folate with selectively mutated LSD1 species that
have been used to obtain a crystal structure of the enzyme. Here,
we compared (6R,S)-THF-Glu5 binding to full-length LSD1
with binding to its N-terminally truncated variant lacking the 170
N-terminal amino acid residues. Results from this binding study
showed that the N-terminally truncated protein binds (6R,S)-
THF-Glu5 with the same affinity as the full-length protein
(Table 1), indicating that the first 170 amino acid residues do
not participate in folate binding.
Author Contributions
Z.L. and F.M. contributed equally to this work.
Funding Sources
Support for this study was, in part, provided by National
Institutes of Health Grants DK15289 and DK080010 to C.W.
and National Science Foundation Grant CHE-0848788 to
D.J.B. and F.M.
’ ACKNOWLEDGMENT
We thank Dr. Yang Shi and Dr. Philip A. Cole for the generous
gift of plasmids for LSD1. We thank EPROVA for the generous
gift of folates.
’ DISCUSSION
The data presented here indicate that LSD1 is present in
nuclei obtained from HeLa cells. It appears to be present pri-
marily as a complex.16 There is also a small amount of LSD1 that
is not present in a complex but appears to be associated with a
folate coenzyme. Binding experiments using recombinant full-
length LSD1 showed that THF pentaglutamate binds with high
affinity, suggesting that the form of folate associated with LSD1
in HeLa cell nuclei was THF pentaglutamate. Furthermore, it is
shown that BSI can be used to rank the folate species according to
their binding affinity. This is the first time that direct binding,
label-free and in solution, has been quantified for this class of
molecules. We believe that the biological function of the bound
folate is to serve as an acceptor of the formaldehyde that is
generated during the oxidative demethylation of histones. For-
maldehyde reacts with THF nonenzymatically producing 5,
10-CH2-THF;30,31 therefore, there is a little doubt that the
formaldehyde released in the course of the histone demethyla-
tion reaction by LSD1 will react with THF. This would be similar
to the role played by THF in similar reactions conducted by
dimethylglycine dehydrogenase and sarcosine dehydrogenase.2À4
The function of THF is to serve as a carrier of one-carbon units
that are transferred between enzymes for use as building blocks in
a number of metabolic pathways.1,32 In the nuclei of mammalian
cells, at least one metabolic pathway that uses THF has recently
been established. The Stover lab has provided strong evidence
that in the nuclei a pathway for thymidylate synthesis utilizes
THF. In the first step of this synthetic pathway, it is thought that
serine hydroxymethyltransferase transfers a methylene group
from serine to tetrahydrofolate producing methylenetetrahydrofo-
late (5,10-CH2-THF). Once formed, 5,10-CH2-THF is used by
thymidylate synthase for synthesis of dTMP. It would be reasonable
to suggest that 5,10-CH2-THF synthesized by oxidative demethyla-
tion might be used as a substrate for thymidylate synthase as well.
The fact that folate participates in the enzymatic demethyla-
tion of histones provides an opportunity for this micronutrient to
play a role in the epigenetic control of gene expression.
’ ABBREVIATIONS
LSD1, lysine-specific histone demethylase 1; BSI, back-scattering
interferometry; SHMT, serine hydroxymethyltransferase;TS, thy-
midylate synthase; DHFR, dihydrofolate reductase (folate refers
to the general class of folate cofactors); THF, tetrahydrofolate;
THF-Glu1, tetrahydrofolate monoglutamate; 5-methyl-THF-
Glu1, 5-methyltetrahydrofolate monoglutamate; 5-formyl-THF-
Glu1, 5-formyltetrahydrofolate monoglutamate; THF-Glu5, tet-
rahydrofolate pentaglutamate; JmjC, Jumonji-containing; DMGDH,
dimethylglycine dehydrogenase; SDH, sarcosine dehydrogenase.
’ REFERENCES
(1) Cook, R. J. (2001) Folate Metabolism. In Homocysteine in Health
and Disease (Carmel, R., and Jacobsen, D. W., Eds.) pp 113À134,
Cambridge University Press, Cambridge, U.K.
(2) Wittwer, A. J., and Wagner, C. (1980) Identification of folate
binding protein of mitochondria as dimethylglycine dehydrogenase.
Proc. Natl. Acad. Sci. U.S.A. 77, 4484–4488.
(3) Wittwer, A. J., and Wagner, C. (1981) Identification of the folate-
binding proteins of rat liver mitochondria as dimethylglycine dehydro-
genase and sarcosine dehydrogenase. Purification and folate-binding
characteristics. J. Biol. Chem. 256, 4102–4108.
(4) Wittwer, A. J., and Wagner, C. (1981) Identification of the folate-
binding proteins of rat liver mitochondria as dimethylglycine dehydro-
genase and sarcosine dehydrogenase. Flavoprotein nature and enzy-
matic properties of the purified proteins. J. Biol. Chem. 256, 4109–4115.
(5) Siva Sankar, D. V., Geisler, A., and Rozsa, P. W. (1969) Intra-
cellular Distribution of Folic Acid in Mouse Liver. Experientia 25,
691–692.
(6) Shin, Y. S., Chan, C., Vidal, A. J., Brody, T., and Stokstad, E. L.
(1976) Subcellular localization of γ-glutamyl carboxypeptidase and of
folates. Biochim. Biophys. Acta 444, 795–801.
(7) Zamierowski, M. M., and Wagner, C. (1977) Identification of
folate binding proteins of rat liver. J. Biol. Chem. 252, 933–938.
(8) Prem veer Reddy, G., and Pardee, A. B. (1980) Multienzyme
complex for metabolic channeling in mammalian DNA replication. Proc.
Natl. Acad. Sci. U.S.A. 77, 3312–3316.
(9) Woeller, C. F., Anderson, D. D., Szebenyi, D. M., and Stover, P. J.
(2007) Evidence for small ubiquitin-like modifier-dependent nuclear
import of the thymidylate biosynthesis pathway. J. Biol. Chem. 282,
17623–17631.
(10) Anderson, D. D., Woeller, C. F., and Stover, P. J. (2007) Small
ubiquitin-like modifier-1 (SUMO-1) modification of thymidylate synthase
and dihydrofolate reductase. Clin. Chem. Lab. Med. 45, 1760–1763.
(11) Cook, R. J., Misono, K. S., and Wagner, C. (1984) Identification
of the covalently bound flavin of dimethylglycine dehydrogenase and
sarcosine dehydrogenase from rat liver mitochondria. J. Biol. Chem.
259, 12475–12480.
’ AUTHOR INFORMATION
Corresponding Author
*C.W.: 620 Light Hall, Vanderbilt University Medical Center,
Nashville, TN 37232; telephone, (615) 343-9866; fax, (615) 343-
0704; e-mail, conrad.wagner@vanderbilt.edu. D.J.B.: 5419
Stevenson, Vanderbilt University Department of Chemistry
and Vanderbilt Institute for Chemical Biology, Nashville, TN
37232; telephone, (615) 322-4226; e-mail, darryl.bornhop@
vanderbilt.edu.
4755
dx.doi.org/10.1021/bi200247b |Biochemistry 2011, 50, 4750–4756

Biochemistry
ARTICLE
(12) Porter, D. H., Cook, R. J., and Wagner, C. (1985) Enzymatic
properties of dimethylglycine dehydrogenase and sarcosine dehydro-
genase from rat liver. Arch. Biochem. Biophys. 243, 396–407.
(13) Anand, R., and Marmorstein, R. (2007) Structure and mechanism
of lysine-specific demethylase enzymes. J. Biol. Chem. 282, 35425–35429.
(14) Paik, W. K., and Kim, S. (1973) Enzymatic Demethylation of
Calf Thymus Histones. Biochem. Biophys. Res. Commun. 51, 781–788.
(15) Shi, Y., Lan, F., Matson, C., Mulligan, P., Whetstine, J. R., Cole,
P. A., Casero, R. A., and Shi, Y. (2004) Histone demethylation mediated
by the nuclear amine oxidase homolog LSD1. Cell 119, 941–953.
(16) Shi, Y. J., Matson, C., Lan, F., Iwase, S., Baba, T., and Shi, Y.
(2005) Regulation of LSD1 histone demethylase activity by its asso-
ciated factors. Mol. Cell 19, 857–864.
(17) Forneris, F., Binda, C., Vanoni, M. A., Mattevi, A., and Battaglioli,
E. (2005) Histone demethylation catalysed by LSD1 is a flavin-dependent
oxidative process. FEBS Lett. 579, 2203–2207.
(18) Yang, M., Culhane, J. C., Szewczuk, L. M., Gocke, C. B.,
Brautigam, C. A., Tomchick, D. R., Machius, M., Cole, P. A., and Yu,
H. (2007) Structural basis of histone demethylation by LSD1 revealed
by suicide inactivation. Nat. Struct. Mol. Biol. 14, 535–539.
(19) Lan, F., Nottke, A. C., and Shi, Y. (2008) Mechanisms involved
in the regulation of histone lysine demethylases. Curr. Opin. Cell Biol.
20, 316–325.
(20) Culhane, J. C., and Cole, P. A. (2007) LSD1 and the chemistry
of histone demethylation. Curr. Opin. Chem. Biol. 11, 561–568.
(21) Matthews, R. G. (1986) Preparation and analysis of pteroylpo-
lyglutamate substrates and inhibitors. Methods Enzymol. 122, 333–339.
(22) Walkup, A. S., and Appling, D. R. (2005) Enzymatic character-
ization of human mitochondrial C1-tetrahydrofolate synthase. Arch.
Biochem. Biophys. 442, 196–205.
(23) Temple, C., and Montgomery, J. A. (1984) Chemical and
Physical Properties of Folic Acid and Reduced Derivatives. In Folates
and Pterins (Blakley, R. L., and Benkovic, S. J., Eds.) pp 61À120, John
Wiley & Sons, Inc., New York.
(24) Horne, D. W., and Patterson, D. (1988) Lactobacillus casei
microbiological assay of folic acid derivatives in 96-well microtiter plates.
Clin. Chem. 34, 2357–2359.
(25) Luka, Z., and Wagner, C. (2003) Human glycine N-methyl-
transferase is unfolded by urea through a compact monomer state. Arch.
Biochem. Biophys. 420, 153–160.
(26) Bornhop, D. J., Latham, J. C., Kussrow, A., Markov, D. A., Jones,
R. D., and Sorensen, H. S. (2007) Free-solution, label-free molecular
interactions studied by back-scattering interferometry. Science 317,
1732–1736.
(27) Morcos, E. F., Kussrow, A., Enders, C., and Bornhop, D. (2010)
Free-solution interaction assay of carbonic anhydrase to its inhibitors
using back-scattering interferometry. Electrophoresis 31, 3691–3695.
(28) Kussrow, A., Baksh, M. M., Bornhop, D. J., and Finn, M. G.
(2011) Universal Sensing by Transduction of Antibody Binding with
Backscattering Interferometry. ChemBioChem 12 (3), 367–370.
(29) Tsukada, Y., and Zhang, Y. (2006) Purification of histone
demethylases from HeLa cells. Methods 40, 318–326.
(30) Blakley, R. L. (1958) Interaction of formaldehyde and tetra-
hydrofolic acid and its relation to the enzymic synthesis of serine. Nature
182, 1719–1722.
(31) Kallen, R. G., and Jencks, W. P. (1966) The mechanism of the
condensation of formaldehyde with tetrahydrofolic acid. J. Biol. Chem.
241, 5851–5863.
(32) Wagner, C. (1995) Biochemical Role of Folate in Cellular
Metabolism. In Folate in Health and Disease (Bailey, L. B., Ed.) pp
23À42, Marcel Dekker, New York.
4756
dx.doi.org/10.1021/bi200247b |Biochemistry 2011, 50, 4750–4756