L-2-Hydroxyglutarate production arises from noncanonical enzyme function at acidic pH

Article abstract of DOI:10.1038/nchembio.2307

The metabolite 2-hydroxyglutarate (2HG) can be produced as either a D-R- or L-S- enantiomer, each of which inhibits α-ketoglutarate (αKG)-dependent enzymes involved in diverse biologic processes. Oncogenic mutations in isocitrate dehydrogenase (IDH) produce D-2HG, which causes a pathologic blockade in cell differentiation. On the other hand, oxygen limitation leads to accumulation of L-2HG, which can facilitate physiologic adaptation to hypoxic stress in both normal and malignant cells. Here we demonstrate that purified lactate dehydrogenase (LDH) and malate dehydrogenase (MDH) catalyze stereospecific production of L-2HG via 'promiscuous' reduction of the alternative substrate αKG. Acidic pH enhances production of L-2HG by promoting a protonated form of αKG that binds to a key residue in the substrate-binding pocket of LDHA. Acid-enhanced production of L-2HG leads to stabilization of hypoxia-inducible factor 1 alpha (HIF-1α) in normoxia. These findings offer insights into mechanisms whereby microenvironmental factors influence production of metabolites that alter cell fate and function.

Full text of DOI:10.1038/nchembio.2307

article
puBLIsHed OnLIne: 6 MarCH 2017 | dOI: 10.1038/nCHeMBIO.2307
L-2-Hydroxyglutarate production arises from
noncanonical enzyme function at acidic pH
andrew M Intlekofer1,2, Bo Wang1, Hui Liu3, Hardik shah3, Carlos Carmona-Fontaine4, ariën s rustenburg4,
salah salah5, M r Gunner5, John d Chodera4, Justin r Cross3 & Craig B thompson1*
The metabolite 2-hydroxyglutarate (2HG) can be produced as either a D-R- or L-S- enantiomer, each of which inhibits
a-ketoglutarate (aKG)-dependent enzymes involved in diverse biologic processes. Oncogenic mutations in isocitrate dehy-
drogenase (IDH) produce D-2HG, which causes a pathologic blockade in cell differentiation. On the other hand, oxygen
limitation leads to accumulation of L-2HG, which can facilitate physiologic adaptation to hypoxic stress in both normal and
malignant cells. Here we demonstrate that purified lactate dehydrogenase (LDH) and malate dehydrogenase (MDH) catalyze
stereospecific production of L-2HG via ‘promiscuous’ reduction of the alternative substrate aKG. Acidic pH enhances produc-
tion of L-2HG by promoting a protonated form of aKG that binds to a key residue in the substrate-binding pocket of LDHA.
Acid-enhanced production of L-2HG leads to stabilization of hypoxia-inducible factor 1 alpha (HIF-1a) in normoxia. These
findings offer insights into mechanisms whereby microenvironmental factors influence production of metabolites that alter
cell fate and function.
he discovery of oncogenic mutations in IDH enzymes demon-
Genetic evidence suggests that LDH and MDH enzymes are
strated the profound impact that altered metabolism can have major sources of hypoxia-induced L-2HG16–18,21, but there has been
on cell identity and function1,2. Mutant IDH enzymes effi- no confirmation that these enzymes directly catalyze L-2HG produc-
T
ciently catalyze reduction of αKG to the ‘oncometabolite’ D-2HG3,4. tion. Here we demonstrate that purified LDH and MDH enzymes
D-2HG inhibits a large family of >70 different αKG-dependent can catalyze stereospecific reduction of αKG to L-2HG and that
enzymes that regulate chromatin modifications, stability of HIFs, acidic reaction conditions dramatically enhance LDH- and MDH-
extracellular-matrix maturation, and DNA repair1. In particular, mediated production of L-2HG in vitro and in cells. Mechanistically,
D-2HG-mediated inhibition of chromatin-modifying enzymes acidic pH enhances lactate dehydrogenase A (LDHA)-mediated
impairs induction of gene expression programs that are required reduction of αKG by driving equilibrium toward a protonated form
for normal cell differentiation and ‘locks’ malignant cells in an of αKG that binds more stably to the LDHA enzyme. In living cells,
undifferentiated stem-cell-like state5–7. The relative contributions of the pH-dependent induction of L-2HG acts as a potent stabilizer of
different D-2HG targets to oncogenesis likely vary depending on HIF-1α in normoxia, representing a previously unknown pathway of
the cellular context8,9.
HIF-1α stabilization that has potential relevance to human disease
While intensive efforts have been directed at investigating the states. These findings offer new insights into the mechanisms that
molecular and cellular effects of IDH-mutant-derived D-2HG, regulate noncanonical substrate use by metabolic enzymes and the
potential sources and functions of the mirror-image enantiomer consequences of promiscuous L-2HG production in cell physiology.
L-2HG are less well understood10. Importantly, biochemical assays
demonstrated that L-2HG functions as a more potent inhibitor RESULTS
of αKG-dependent enzymes than D-2HG; thus, cells may be quan- LDH and MDH enzymes use aKG as an alternative substrate
titatively more sensitive to changes in L-2HG than to changes Hypoxic cells undergo a number of metabolic changes (Fig. 1a)
in D-2HG11–13. Indeed, deregulated L-2HG disposal causes sub- including an increased NADH to NAD+ ratio18,22, an increased intra-
stantial developmental pathology and is associated with brain and cellular concentration of αKG18,23, and acidification of the extracel-
kidney cancers14–16
.
lular and intracellular environments24,25. Prior reports implicated
It was recently reported that hypoxia induces production of LDHA and malate dehydrogenase 1 and 2 (MDH1 and MDH2) as
L-2HG in both normal and malignant cells as a physiologic response being contributors to hypoxia-induced L-2HG in cells (Fig. 1a)17,18
.
to oxygen limitation17,18. Hypoxia-induced production of L-2HG To determine whether these enzymes could directly produce L-2HG,
occurs independently of HIF but acts to reinforce the hypoxic we tested the ability of purified LDH and MDH to catalyze NADH-
response, at least in part, through stabilization of HIF protein13,17,18
.
dependent reduction of their canonical substrates as well as αKG.
Accumulation of L-2HG slows glycolysis and mitochondrial As expected, LDH rapidly consumed NADH in the presence of its
respiration by reducing the rate of regeneration of NAD+ canonical substrate pyruvate (Fig. 1b). LDH also consumed NADH
(ref. 18). Moreover, hypoxia-induced L-2HG promotes the same in the presence of αKG, confirming its capacity to utilize alterna-
repressive chromatin marks that characterize the differentia- tive, larger α-ketoacid substrates (Fig. 1b)26,27. Likewise, MDH effi-
tion blockade of IDH-mutant malignancies8,17,19, consistent with ciently catalyzed reduction of its canonical substrate oxaloacetate
the well-established association between hypoxic niches and (OAA; Fig. 1c), while also demonstrating an ability to use αKG as a
stem-cell populations20.
noncanonical substrate, albeit less efficiently (Fig. 1c).
1cancer biology and Genetics program, Memorial Sloan Kettering cancer center, new York, new York, uSA. 2department of Medicine, Memorial Sloan
Kettering cancer center, new York, new York, uSA. 3the donald b. and catherine c. Marron cancer Metabolism center, Memorial Sloan Kettering
cancer center, new York, new York, uSA. 4computational biology program, Memorial Sloan Kettering cancer center, new York, new York, uSA. 5physics
department, city college of new York, new York, new York, uSA. *e-mail: thompsonc@mskcc.org
nature CHeMICaL BIOLOGY | AdvAnce online publicAtion | www.nature.com/naturechemicalbiology
1

NATURE cHEMIcAL bIOLOGy dOI: 10.1038/nCHeMBIO.2307
article
a
LDH and MDH catalyze reduction of aKG to L-2HG
Glucose
LDHA
Standardmetabolitederivatizationandchromatographymethodsdo
not distinguish between enantiomeric species. In order to separate
α-hydroxyacid enantiomers, we adapted a liquid chromatography–
mass spectrometry (LC–MS) protocol involving derivatization of
metabolites with the chiral compound diacetyl-L-tartaric anhydride
(Supplementary Results and Supplementary Fig. 1)28. The iden-
tity of α-hydroxyacid enantiomers was confirmed by comparison
to similarly derivatized standards (Fig. 2a–c). As expected, chiral
derivatization showed that LDH catalyzed reduction of pyruvate
to L-lactate, whereas MDH catalyzed this reaction poorly (Fig. 2a).
MDH catalyzed its canonical reaction involving reduction of OAA to
L-malate (Fig. 2b), while LDH catalyzed this reaction less effectively
(Fig. 2b). Both LDH and MDH catalyzed stereospecific reduction
+
MDH1
OAA
Lactate
–
Pyr
Malate
HIF-1α
Citrate
Citrate
Hypoxia:
NADH
OAA
MDH2
+
NAD
αKG
αKG
Malate
αKG
pH
Glutamine
LDH
MDH
b
c
1.0
0.8
0.6
0.4
0.2
0.0
1.0
0.8
0.6
0.4
0.2
0.0
α-Ketoacid substrate:
None
Pyruvate (Pyr)
Oxaloacetate (OAA)
α-Ketoglutarate (αKG)
of αKG to L-2HG (Fig. 2c and Supplementary Fig. 2). To ensure
that L-2HG production was not facilitated by an unknown contami-
nant of the enzyme preparations (Supplementary Fig. 3a), we also
studied recombinant human LDHA, MDH1, and MDH2 enzymes.
The recombinant enzymes also catalyzed stereospecific reduction of
αKG to L-2HG (Supplementary Fig. 3b–f). Collectively, these find-
ings substantiate genetic evidence suggesting that LDH and MDH
0
100 200 300 400 500
Time (min)
0
100 200 300 400 500
Time (min)
d
e
α-Hydroxyacid
No enzyme
LDH
MDH
product:
2.0
1.6
1.2
0.8
0.4
0
NAD+
NADH
Lactate
Pyr
O
(m/z 190)
OH
O–
enzymes contribute to cellular production of L-2HG16–18,21
.
O–
O
O
6.6 6.7 6.8 6.9
6.6 6.7 6.8 6.9
6.6 6.7 6.8 6.9
Wild-type IDH and PHGDH enzymes produce D-2HG
3
2
1
NAD+
Priorreportsdemonstratedthatwild-typeIDHenzymescancatalyze
reduction of αKG to 2HG, but the stereochemistry of the 2HG prod-
uct was not determined29,30. Therefore, we sought to assess whether
wild-type IDH enzymes might represent another cellular source of
D- or L-2HG. Using a cell-based assay, we observed that transfec-
tion with empty vector had no effect on D- or L-2HG levels, whereas
transfection with wild-type IDH1 or IDH2 selectively increased
D-2HG levels (Supplementary Figs. 4a,b and 5). Supplementation
with cell-permeable αKG substrate resulted in a further increase in
the amount of D-2HG in cells transfected with wild-type IDH1 or
IDH2 (Supplementary Figs. 4a,b and 5), albeit the increase was
~50- to 100-fold less than that observed in cells transfected with an
oncogenic IDH1 R132H mutant (Supplementary Fig. 4c).
NADH
Malate
OAA
O
(m/z 335)
O
O
OH
O–
O–
–O
–O
O
O
0
12.5
12.6
12.5
12.6
12.5
12.6
f
1.2
0.8
0.4
0
NAD+
NADH
2HG
αKG
O
(m/z 247)
OH
–O
O–
–O
O–
O
O
O
O
13.5
13.6
13.5
13.6
13.5
13.6
Retention time (min)
Figure 1 | LDH and MDH enzymes catalyze reduction of the alternative
substrate aKG. (a) Schematic of metabolic changes that occur in hypoxic
cells. lactate dehydrogenase A (ldHA) and malate dehydrogenase
1 and 2 (MdH1 and MdH2) represent potential enzymatic sources of
l-2-hydroxyglutarate (l-2HG) production in cells. αKG, α-ketoglutarate;
HiF-1α, hypoxia-inducible factor 1 alpha; oAA, oxaloacetate. (b) enzyme
assay measuring nAdH consumption by ldH purified from bovine heart
(5 units/ml) in reactions with no substrate (gray), 0.3 mM pyruvate
(green), or 1 mM αKG (blue). (c) nAdH consumption by MdH purified
from porcine heart (5 units/ml) in reactions with no substrate (gray),
0.3 mM oAA (orange), or 3 mM αKG (blue). data in b and c are shown as
the mean ± 95% confidence interval (c.i.) for triplicate reactions.
(d–f) α-Hydroxyacid product formation catalyzed by dehydrogenases.
Gc–MS analysis of lactate production from pyruvate (d), malate production
from oAA (e), and 2HG production from αKG (f) in 6-h reactions with no
enzyme, ldH purified from bovine heart (10 μg/ml), or MdH purified from
porcine heart (10 μg/ml). Reactions were conducted in 33 mM potassium
phosphate buffer, pH 7.0, with 0.22 mM nAdH. Results throughout the
figure are representative of ≥3 independent experiments.
Commercially available preparations of IDH from porcine
heart were impure and thus unsuitable for enzymatic assays,
prompting us to use recombinant human IDH1 as an alternative
(Supplementary Fig. 3a). Employing previously reported reaction
conditions with Tris-based buffers31, we detected substantial nonen-
zymatic production of 2HG that was equally distributed as D- and
L- enantiomers (Supplementary Fig. 6). By contrast, nonenzymatic
2HG production was negligible in phosphate-based reaction buffers
(Supplementary Fig. 6). Using phosphate-based buffers, we found
that recombinant human IDH1 selectively catalyzed production of
D-2HG (Supplementary Fig. 4d). Taken together, the cell-based
and recombinant enzyme assays demonstrate that wild-type IDH
enzymes can catalyze production of D-2HG, an intrinsic enzymatic
property that is substantially enhanced by the active site arginine
residue mutations in IDH that are observed in cancer29,30
The enzyme 3-phosphoglycerate dehydrogenase (PHGDH) was
recently identified as another potential cellular source of D-2HG32,33
Indeed, we found that recombinant human PHGDH catalyzed ste-
.
.
We performed gas chromatography–mass spectrometry (GC– reospecific reduction of αKG to D-2HG in vitro (Supplementary
MS) to determine the identities of the α-hydroxyacid reaction prod- Fig. 4e). The locus encoding PHGDH is frequently amplified in
ucts generated by LDH and MDH enzymes. Analysis of reaction breast cancer and melanoma34,35, and elevated 2HG levels have been
mixtures without enzyme demonstrated that there was no significant detected in primary breast tumors with aggressive features but no
nonenzymatic production of lactate, malate, or 2HG (Fig. 1d–f). IDH mutation36. Thus, determining the chirality of 2HG should
LDH, but not MDH, catalyzed reduction of pyruvate to lactate help elucidate which metabolic pathway contributes to 2HG pro-
(Fig. 1d). Both MDH and LDH catalyzed reduction of OAA to duction in various cancer settings.
malate, with MDH catalyzing this reaction more efficiently than
LDH, as expected (Fig. 1e). Notably, both LDH and MDH enzymes Acidic pH enhances the alternative enzymatic activity of LDH
were capable of promiscuously catalyzing NADH-dependent reduc- While investigating the enzymatic properties of LDH, we noted a
tion of the alternative substrate αKG to 2HG (Fig. 1f).
striking enhancement in NADH-dependent reduction of αKG
2
nature CHeMICaL BIOLOGY | AdvAnce online publicAtion | www.nature.com/naturechemicalbiology

NATURE cHEMIcAL bIOLOGy dOI: 10.1038/nCHeMBIO.2307
article
a
b
c
Lactate (m/z 89)
L-Lactate D-Lactate
Malate (m/z 133)
L-Malate D-Malate
2HG (m/z 147)
L-2HG D-2HG
Acidity enhances reduction of a subset of -ketoacids
4
3
2
1
6
4
2
10
8
6
4
2
0
To determine whether acidic pH might increase accessibility to the
substrate-binding pocket of LDH, we tested the effect of pH on LDH-
mediated reduction of α-ketoacids with bulky tail groups, including
phenylpyruvate (PP) and hydroxyphenylpyruvate (HPP) (Fig. 4a).
We observed that LDH reduced both PP and HPP at a rate at least
as great as that of αKG, and there were no substantial pH-depen-
dent changes in the rate of reduction for either substrate (Fig. 4a).
These findings suggest that acidic pH does not function merely to
enhance access for substrates larger than pyruvate, such as αKG, to
the substrate-binding pocket of LDH.
0
0
4.5
5.0
5.5
2.0
2.2
2.4
3.0 3.5 4.0 4.5
12
8
8
6
4
2
0
10
8
6
4
2
0
No enzyme
Pyr
No enzyme
OAA
No enzyme
αKG
4
0
We investigated whether the nature of the chemical moiety pres-
ent in the tail end of the α-ketoacid substrate might dictate the pH-
dependent properties of the LDH enzymatic reaction. We identified
pairs of α-ketoacid substrates of identical carbon chain length that
differed only in whether there was a carboxylate or methyl group
present in the final position (Fig. 4b): OAA (4-carbon), αKG
(5-carbon), and oxoadipic acid (OAdA; 6-carbon) along with their
respective methylated counterparts α-ketobutyrate (αKB; 4-car-
bon), oxopentanoic acid (OPA; 5-carbon), and ketohexanoic acid
(KHA; 6-carbon). For each substrate pair, only the α-ketoacid with
a carboxylate group in the final position exhibited pH-dependent
enhancement of enzymatic reduction (Fig. 4b).
4.5
5.0
5.5
2.0
2.2
2.4
3.0 3.5 4.0 4.5
8
6
4
2
0
10
8
6
4
2
0
12
8
LDH
Pyr
LDH
OAA
LDH
αKG
4
0
4.5
4.5
5.0
5.0
5.5
2.0
2.0
2.2
2.2
2.4
3.0 3.5 4.0 4.5
12
8
8
6
4
2
0
10
8
6
4
2
0
MDH
Pyr
MDH
OAA
MDH
αKG
4
0
5.5
2.4
3.0 3.5 4.0 4.5
Retention time (min)
Retention time (min)
Retention time (min)
Acidification enhances interaction of aKG with LDHA Q100
We performed Monte Carlo protonation state modeling simulations
with virtually docked substrates to determine whether pH-depen-
dent changes in the protonation states of αKG, NADH, or LDHA
might be responsible for the acid-enhanced enzymatic reduction
of αKG. The docked structure of αKG demonstrated that the car-
boxylate group in the tail of the substrate is positioned immediately
adjacent to glutamine 100 (Q100) in the substrate-binding pocket of
LDHA (Fig. 5a). Prior reports demonstrated that the corresponding
residue in bacterial LDH plays an important role in determining
Figure 2 | LDH and MDH selectively produce L--hydroxyacids, including
L-2HG. (a–c) chiral lc–MS analysis resolving enantiomers of lactate
produced from pyruvate (a), malate produced from oAA (b), and 2HG
produced from αKG (c) in 6-h reactions with no enzyme, ldH purified
from bovine heart (5 units/ml), or MdH purified from porcine heart (5
units/ml) in 33 mM potassium phosphate buffer, pH 7.0, with 0.22 mM
nAdH and 1 mM substrate. the top panels show mixtures of standards for
l- and d- enantiomers of lactate, malate, and 2HG that were derivatized
and analyzed by the same method in parallel. vertical blue and red bars
in the background serve as visual references to facilitate the identification
of l- and d- enantiomers, respectively. Results throughout the figure are
representative of ≥3 independent experiments.
substrate specificity39,40
.
Analysis of the Monte Carlo simulations demonstrated a change
of the predominant protonation state of αKG upon binding LDHA
with NADH. The carboxylate tail of αKG has a predicted pKa of 4.28
± 1.27 (closely matching the experimental pKa of 4.4)41, indicating
that it would be mostly deprotonated in solution at pH values rang-
at more acidic pH (Fig. 3a). Specifically, the Km of LDH for αKG ing from 6.0 to 7.4. However, upon binding to LDHA, the carboxy-
decreased from >15 mM at pH 7.4 to 3.86 ± 0.97 mM at pH 6.0 late tail of αKG becomes protonated, whereupon it is predicted to
(Fig. 3b), a value that approximates the concentration of αKG in form a hydrogen bond with Q100 (Fig. 5b). This is associated with a
hypoxic cells18. We found that the enhanced rate of αKG-dependent free-energy penalty for protonating the carboxylate tail that opposes
NADH consumption at more acidic pH resulted in increased LDH- binding, which for a given pH can be expressed as
dependentproductionofL-2HG(Fig.3c).Acidificationalsoenhanced
ΔG =k T(pH − pK )(ln10)
the ability of MDH to catalyze reduction of αKG to L-2HG, albeit to a
lesser degree (Fig. 3d–f). Although the rate of nonenzymatic NADH
pH
B
a
degradation increased at acidic pH, there was no detectable nonenzy- where kB is the Boltzmann constant and T equals the absolute tem-
matic production of L-2HG (Supplementary Fig. 7).
perature. Protonating the carboxylate tail at pH 7.4 would cost 3.2
In contrast to the effects on noncanonical substrate usage, kBT more than protonating it at pH 6.0 (Fig. 5c). The expected fold
acidic pH had no substantial effect on the ability of LDH to change in the aqueous concentration of the deprotonated form is
reduce its canonical substrate pyruvate at a near physiologic calculated as follows.
pyruvate concentration of 0.3 mM (Fig. 3g)37. At concentrations
−ΔG /k T
6.0
B
e
e
of pyruvate greater than 0.3 mM, acidic pH impaired the ability
of LDH to reduce pyruvate because of enhanced substrate inhibi-
tion, as previously described (Fig. 3g,h)38. Together, these results
demonstrate that acidification enhances the alternative substrate
−ΔG /k T
7.4
B
If the enzyme greatly prefers to bind the protonated form, then
utilization of LDH such that the noncanonical substrate αKG is the predicted Km of αKG would be approximately 25 times higher
relatively favored at lower pH, resulting in increased production (corresponding to lower affinity) at pH 7.4 than at 6.0, lacking addi-
of L-2HG.
tional pH-dependent effects. These predictions are consistent with
We also examined the effects of acidic pH on the ability of wild- the increased Km measured experimentally for αKG at pH 7.4 as
type IDH1 and PHGDH to reduce αKG to D-2HG. Wild-type IDH1 compared to pH 6.0 (Fig. 3c).
exhibited minimal pH-dependent change in D-2HG production
Mutation of Q100 to a positively charged amino acid might alter
(Supplementary Fig. 8a). In contrast, PHGDH exhibited enhanced the substrate preference for LDHA such that it would favor αKG
D-2HG production in acidic pH (Supplementary Fig. 8b). with the carboxylate tail in the deprotonated (negatively charged)
nature CHeMICaL BIOLOGY | AdvAnce online publicAtion | www.nature.com/naturechemicalbiology
3

NATURE cHEMIcAL bIOLOGy dOI: 10.1038/nCHeMBIO.2307
article
Enzyme: LDH
Substrate: αKG
Enzyme: LDH
Substrate: αKG
Enzyme: LDH
Substrate: αKG
Enzyme: MDH
Substrate: αKG
a
b
c
d
4
3
2
1
1.0
0.8
0.6
0.4
0.2
0.0
20
15
10
5
1.0
0.8
0.6
0.4
0.2
0.0
L
-2HG
D-2HG
pH 7.4
pH 7.0
pH 6.6
pH 6.0
pH 7.4
pH 7.0
pH 6.6
pH 6.0
pH 7.4
pH 7.0
pH 6.6
pH 6.0
pH 7.4
pH 6.0
0
0
0
100 200 300 400 500
Time (min)
0
2
4
6
8
3.0 3.5 4.0 4.5
Retention time (min)
0
100 200 300 400 500
Time (min)
αKG (mM)
e
Enzyme: MDH
Substrate: αKG
f
Enzyme: MDH
Substrate: αKG
g
Enzyme: LDH
Substrate: Pyr
h
Enzyme: LDH
Substrate: Pyr
25
20
15
10
5
0.4
0.3
0.2
0.1
0.5
0.4
0.3
0.2
0.1
12
8
L
-2HG
D-2HG
pH 7.4
pH 7.0
pH 6.6
pH 6.0
L
-Lactate D-Lactate
pH 7.4
pH 6.0
pH 7.4
pH 6.0
pH 7.4
pH 7.0
pH 6.6
pH 6.0
4
0.0
0
0.0
0
4.5
0
2
4
6
8
3.0 3.5 4.0 4.5
Retention time (min)
0.1
1
10
5.0
5.5
αKG (mM)
Pyruvate (mM)
Retention time (min)
Figure 3 | Acidic pH enhances L-2HG production by LDH and MDH. (a–c) ldH- and (d–f) MdH-catalyzed conversion of αKG to l-2HG as a function of pH.
(a,d) increased rates of nAdH consumption in acidic reaction conditions with purified ldH (a) or MdH (d) at 5 units/ml and 3 mM αKG. data in a and d
are shown as the mean ± 95% c.i. for triplicate reactions. (b,e) Reaction velocities as a function of αKG concentration for purified ldH (b) or MdH (e) at 5
units/ml. (c,f) chiral lc–MS analysis of 2HG enantiomers from 6-h reactions with purified ldH (c) or MdH (f) at 5 units/ml and 1 mM αKG. (g) Reaction
velocities as a function of pyruvate concentration for purified ldH at 1 μg/ml. (h) chiral lc–MS analysis of lactate enantiomers from 6-h reactions with
purified ldH at 5 units/ml and 1 mM pyruvate. Reactions were done in 33 mM potassium phosphate buffer at the indicated pH with 0.22 mM nAdH. data
for b, e, and g are shown as the mean ± s.d. for triplicate reactions. Results throughout the figure are representative of ≥3 independent experiments.
state, thus abolishing the pH trend observed with the wild-type of enzyme for in vitro reactions with addition of αKG and NADH,
LDHA enzyme. Indeed, mutation of the corresponding residue in and 2HG production was assessed by GC–MS. As the reaction pH
bacterial LDH from glutamine to arginine (Q102R) enhanced the was lowered, there was a progressive increase in the amount of 2HG
ability of the enzyme to reduce OAA39,40. Therefore, we generated produced such that there was >3-fold more 2HG at pH 6.0 than there
vectors expressing either wild-type LDHA (LDHA WT) or LDHA was at pH 7.5 (Fig. 6a). Assessment of chirality by LC–MS confirmed
with a point mutation resulting in replacement of glutamine 100 that acidification of cell-lysate reactions potently induced production
with arginine (LDHA Q100R). The constructs were transfected of L-2HG (Fig. 6b). Acid-enhanced L-2HG production was blocked
into 293T cells, and the FLAG-tagged enzymes were purified by by addition of oxamate, an LDH inhibitor that is a structural analog
immunoprecipitation (Supplementary Fig. 9). Equal quantities of of pyruvate (Fig. 6b)42. Thus, acid-enhanced production of L-2HG
enzyme were used for in vitro reactions with NADH cofactor and depends, at least in part, on the enzymatic activity of LDH.
the substrates αKG, OPA, pyruvate (Pyr), and OAA (Fig. 5d–g).
Hypoxia results in acidification of the cellular environment and
In contrast to the acid-enhanced reduction of αKG mediated by an increase in the intracellular concentration of αKG (Fig. 1a)18,23–25
,
wild-type LDHA, the LDHA Q100R mutant exhibited a relatively but it is unclear to what extent these conditions might be sufficient
faster rate of αKG reduction at higher pH values, with no enhance- to induce production of L-2HG. Therefore, we tested the effects of
ment of activity as the pH decreased (Fig. 5d). Thus, the positive acidified medium and/or exogenous αKG on cells cultured in nor-
charge of the Q100R mutant appears to eliminate the energetic pen- moxia (Fig. 6c). As previously reported, addition of cell-permeable
alty normally required to protonate the carboxylate tail of αKG. By αKG resulted in a modest increase in 2HG production in cells cul-
contrast, the LDHA Q100R mutant exhibited relatively impaired tured in standard medium in normoxia (Fig. 6c)17. However, we
enzymatic reduction of OPA, which has a methyl group in the tail observed a synergistic enhancement of 2HG production in cells
position (Fig. 5f). Likewise, the LDHA Q100R mutant exhibited a cultured in acidified medium supplemented with cell-permeable
relative preference for OAA (Fig. 5e) and a relatively impaired ability αKG (Fig. 6c). Quantification of the intracellular concentrations of
to reduce pyruvate (Fig. 5g). Taken together, these findings suggest L- and D-2HG demonstrated that L-2HG levels exceeded 0.3 mM in
that acidic pH enhances the ability of wild-type LDHA to reduce acidified media supplemented with cell-permeable αKG (Fig. 6d),
substrates with a carboxylate group in the final position by diminish- approximating previous measurements in hypoxic cells17,18. In
ing the energetic penalty of protonating the substrate tail to accom- contrast, D-2HG concentrations were approximately tenfold lower
modate binding with Q100 in the LDHA substrate-binding pocket. (Fig. 6d). Measurement of intracellular pH with fluorescent probes
The LDHA Q100R mutant introduces a positively charged residue at demonstrated a value of pH 6.53 ± 0.29 for cells cultured in acidic
the Q100 site, which alters substrate preference such that α-ketoacids medium plus αKG (Supplementary Fig. 10).
with carboxylate tails need not pay the energetic cost of being proto-
nated to bind and are relatively favored at higher pH values.
Cells cultured in acidic medium plus αKG demonstrated a strik-
ing stabilization of HIF-1α despite the presence of atmospheric
oxygen levels (Fig. 6c and Supplementary Fig. 11). The observed
stabilization of HIF-1α by αKG is paradoxical given the fact that
Acid-enhanced production of L-2HG stabilizes HIF-1a
Given the striking enhancement of LDH-catalyzed L-2HG produc- αKG functions as a substrate for the egg-laying deficiency protein
tion observed in vitro, we sought to determine the effects of acidi- nine-like (EGLN) enzymes that hydroxylate proline residues on HIF
fication on 2HG production in cells. Glioblastoma cells grown in and target it for proteasomal degradation43. Therefore, we examined
normoxic conditions were lysed in nondenaturing buffer with pH the effects of acidification plus αKG on the ratio of hydroxyl-HIF-1α
ranging from 7.5 to 6.0 (Fig. 6a). Cell lysates were used as a source to total HIF-1α in 293T cells cultured in the presence of the
4
nature CHeMICaL BIOLOGY | AdvAnce online publicAtion | www.nature.com/naturechemicalbiology

NATURE cHEMIcAL bIOLOGy dOI: 10.1038/nCHeMBIO.2307
article
a
HO
a
b
O
O
O
–O
O–
PP
HPP
O–
O–
αKG
NADH
NADH
R169
H193
R169
O
O
O
O
4
3
2
1
pH 7.4
pH 7.0
pH 6.6
pH 6.0
H193
αKG
(protonated)
2.7 Å
Q100
1.9 Å
Q100
αKG
(deprotonated)
0
αKG
O
PP
HPP
O
c
–O
O–
HO
O–
∆GpH 7.4 = 7.2 KBT
∆GpH 6.0 = 4.0 KBT
b
Ketoacid
pairs: 4-carbon
OAA
O
O
O
O
5-carbon
6-carbon
αKG deprotonated tail
αKG protonated tail
OAdA
αKG
O
O
O
O
O
Tail:
–O
O–
O–
O–
Carboxylate
–O
–O
d
e
O
O
αKG
O
OAA
–O
O–
O
O
–O
O–
O
O
O
O
O
O
O
αKB
OPA
O
KHA
O
0.5
0.4
0.3
0.2
0.1
25
O–
O–
O– Methyl
20
15
10
5
pH 7.4
pH 7.0
pH 6.6
pH 6.0
O
O
5
4
3
2
1
pH 7.4
pH 7.0
pH 6.6
pH 6.0
0.0
0
LDHA: WT
Q100R
LDHA: WT
Q100R
f
g
O
O
OPA
Pyr
0
O–
O–
OAA
αKB
αKG
OPA
OAdA
KHA
12
4
3
2
1
O
O
pH 7.4
pH 7.0
pH 6.6
pH 6.0
Figure 4 | LDH exhibits pH-sensitive reduction of a-ketoacids with
carboxylate tails. (a) Relative rates of nAdH consumption at the indicated
pH with purified ldH at 5 units/ml, 0.22 mM nAdH, and 1 mM substrate.
For each substrate, the rate at pH 7.4 was normalized to 1 to illustrate the
degree of rate enhancement with acidification. Structures for substrates
αKG, phenylpyruvate (pp), and hydroxyphenylpyruvate (Hpp) are shown.
(b) Relative rates of nAdH consumption at the indicated pH with purified
ldH at 5 units/ml, 0.22 mM nAdH, and 1 mM substrate. For each
substrate, the rate at pH 7.4 was normalized to 1 to illustrate the degree of
rate enhancement with acidification. Structures are shown for α-ketoacid
substrates with carboxylate tails, oAA, αKG, and oxoadipic acid (oAdA),
along with their respective methylated counterparts, α-ketobutyrate (αKb),
oxopentanoic acid (opA), and ketohexanoic acid (KHA). data are shown
as the mean ± s.d. for triplicate reactions. Results throughout the figure are
representative of ≥3 independent experiments.
8
4
0
0
LDHA: WT
Q100R
LDHA: WT
Q100R
Figure 5 | Acidity-induced rate enhancement arises from preference of
LDHA for the protonated form of aKG. (a) virtually docked structure
of deprotonated αKG (yellow) bound to ldHA, showing a potentially
unfavorable interaction between the carbonyl of the Q100 residue and
the tail of αKG. (b) virtually docked structure of protonated αKG (yellow)
bound to ldHA, showing a potential hydrogen bond interaction between
the carbonyl of the Q100 residue and the protonated carboxylate tail
of αKG, predicted to be the dominant bound species by Monte carlo
simulations. (c) the equilibrium of protonation states of the tail of αKG
in solution. Shown on the right is the theoretical free-energy penalty of
protonating the tail at pH 7.4 (ΔGpH 7.4) and pH 6.0 (ΔGpH 6.0) in units of kbT,
in which kb is the boltzmann constant and T is the absolute temperature.
(d–g) Wild-type (Wt) or mutant (Q100R) ldHA enzymes were
purified and used in enzymatic reactions with α-ketoglutarate (αKG) (d),
oxaloacetate (oAA) (e), oxopentanoic acid (opA) (f), or pyruvate (pyr) (g).
Reactions were performed in 33 mM potassium phosphate buffer at the
indicated pH with 0.22 mM nAdH and 1 mM substrate and with enzyme
at 7 μg/ml (d,f), 0.7 μg/ml (e), or 0.07 μg/ml (g). data for d–g are shown
as the mean ± s.d. for triplicate reactions. Results throughout the figure are
representative of ≥3 independent experiments.
proteasome inhibitor MG132 (Fig. 6e and Supplementary Fig. 11)44.
Cells cultured in standard medium plus MG132 accumulated
hydroxyl-HIF-1α relative to total HIF-1α in response to cell-
permeable αKG (Fig. 6e and Supplementary Fig. 11). In contrast,
cells cultured in acidified medium plus MG132 did not accumulate
hydroxyl-HIF-1α despite a substantial increase in total HIF-1α pro-
tein (Fig. 6e and Supplementary Fig. 11). Thus, we conclude that
acid-enhanced L-2HG production stabilizes HIF-1α by inhibiting
the αKG-dependent prolyl hydroxylases that target HIF-1α for deg-
radation, in agreement with previously observed effects of L-2HG
on prolyl hydroxylases in vitro11,13
.
To further explore the possibility that acid-enhanced L-2HG lies DIScUSSION
upstream of HIF-1α, we tested the effects of ablating LDHA and Herein we demonstrate that LDH and MDH enzymes catalyze pro-
MDH2,thetwoenzymesprimarilyresponsibleforL-2HGproductionin miscuous enzymatic reduction of the alternative substrate αKG
cells17,18. Cells treated with nontargeting control small interfering RNA to L-2HG. Decades-old studies demonstrated alternative substrate
(siRNA) exhibited normoxic stabilization of HIF-1α when cultured in use by dehydrogenase enzymes but lacked techniques to accurately
acidic medium supplemented with αKG (Fig. 6f and Supplementary assess the identity and chirality of metabolite products26,27. Our find-
Fig. 11). Ablation of either LDHA or MDH2 alone partially impaired ings indicate that there are multiple enzymes that can reduce αKG to
stabilization of HIF-1α in response to acidification (Supplementary 2HG and that the chirality of 2HG is dictated by the specific enzyme
Figs. 12a,b and 13). However, combined targeting of both LDHA and involved. For example, LDH and MDH enzymes catalyze stereospe-
MDH2 abrogated L-2HG production (Supplementary Fig. 12c) and cific production of L-2HG, whereas wild-type (and mutant) IDH
abolished the ability of cells to stabilize HIF-1α in response to acidi- and PHGDH catalyze stereospecific production of D-2HG.
fication (Fig. 6f and Supplementary Fig. 11). These findings demon-
The findings presented here substantiate prior genetic evidence
strate that acid-enhanced conversion of αKG to L-2HG by LDHA and implicating LDH and MDH as cellular sources of L-2HG16–18,21
.
MDH2 is sufficient to stabilize HIF-1α in the absence of hypoxia.
In hypoxia, cells selectively produce L-2HG in a manner primarily
nature CHeMICaL BIOLOGY | AdvAnce online publicAtion | www.nature.com/naturechemicalbiology
5

NATURE cHEMIcAL bIOLOGy dOI: 10.1038/nCHeMBIO.2307
article
+
a
NADH
αKG
pH 7.0
NAD
acidification might also explain, at least in part, the increased L-2HG
observed in cells with electron-transport-chain dysfunction46.
In hypoxia, L-2HG can function as a metabolic signaling mol-
ecule that mediates physiologic responses to help cells adapt to
2HG (m/z 247)
pH 7.5
pH 6.5
pH 6.0
2
1
oxygen limitation17,18
. L-2HG acts to stabilize HIF by inhibit-
ing the αKG-dependent prolyl hydroxylases that target HIF for
degradation11,13. Whether HIF-1α stabilization is a qualitatively or
quantitatively unique property of L-2HG as compared to D-2HG
remains controversial13,47,48. We have identified acidosis as a previ-
ously unappreciated physiologic stimulus that enhances conver-
sion of αKG to L-2HG, resulting in robust stabilization of HIF-1α
in normoxia. HIF-1α facilitates cellular adaptation to an acid load
by inducing expression of carbonic anhydrases49, and our findings
suggest that acid-enhanced production of L-2HG may be part of this
signaling axis. Indeed, we find that ablation of LDHA and MDH2
abrogates the ability of acidotic cells to stabilize HIF-1α. It is also
possible that acidification and L-2HG cooperate in the stabilization
of HIF-1α (ref. 50). In comparison to L-2HG, we observed approxi-
mately tenfold lower intracellular concentrations of D-2HG. Given
that L-2HG functions as a more potent inhibitor of αKG-dependent
enzymes than D-2HG11–13, we conclude that D-2HG derived from
wild-type IDH and/or PHGDH is negligible and likely does not play
a physiologic role in this setting.
0
14.4
14.5
14.6 14.4
+
14.5
14.6 14.4
14.5
14.6 14.4
14.5
14.6
Retention time (min)
b
c
d
Normal pH Veh
Normal pH αKG
Acid pH Veh
NADH NAD
αKG
20
15
10
5
8
6
4
2
0
L
-2HG
Acid pH αKG
(m/z 147)
400
300
200
100
0
0
Oxamate:
+
+
+
+
HIF-1α
pH: 7.5
7.0
6.5
6.0
α-tubulin
Acid:
αKG:
L-2HG
D-2HG
+
+
+
+
MG132
e
f
Acid:
αKG:
+
+
+
+
+
+
+
+
LDHA/
MDH2
siRNA: Control
Acid:
αKG:
HIF-1α
+
+
+
+
+
+
+
+
HIF-1α
Ezh2
Promiscuous enzymatic activity results in the production of
numerous metabolites that have been described as ‘metabolite dam-
age’ or mistakes10. However, emerging evidence suggests that L-2HG,
a metabolite produced by promiscuous enzyme activity, might rep-
resent a conserved metabolic response to multiple environmental
stimuli including hypoxia and acidosis. L-2HG appears to medi-
ate its effects through both HIF-dependent and HIF-independent
mechanisms17,18. Future investigations will be directed at elucidating
the mechanisms whereby L-2HG regulates physiologic responses to
hypoxia and acidosis, as well as how deregulation of L-2HG metabo-
Hydroxyl-
HIF-1α
Ezh2
Figure 6 | Acid-enhanced production of L-2HG stabilizes HIF-1a.
(a) SFXl cells were lysed in nondenaturing conditions at the indicated pH.
cell lysates were incubated for 16 h with 1 mM αKG and 0.22 mM nAdH.
total 2HG was analyzed by Gc–MS. (b) lysate reactions were prepared as
in a with addition of either vehicle (water) or 1 mM oxamate. l-2HG was
measured by chiral lc–MS. data are shown as the mean ± s.d. for triplicate
reactions. (c) SFXl cells were cultured for 24 h in standard dMeM or
dMeM titrated to pH 6.0 (‘acid’) with vehicle or 5 mM cell-permeable
dimethyl-αKG. total 2HG was measured by Gc–MS. data are shown
as the mean ± s.d. for triplicate wells. (d) Quantification of intracellular
l- and d-2HG concentrations by lc–MS (online Methods). data are
shown as the mean ± s.d. of a total of six replicates per condition from
two independent experiments. veh, vehicle. (e) 293t cells were cultured
for 24 h as in c with addition of vehicle or 10 μM MG132 for the final 1 h.
Western blotting of nuclear extracts shows HiF-1α, hydroxyl-HiF-1α, and
ezh2 (loading control). (f) ldHA and MdH2 were targeted by siRnAs in
SFXl cells followed by culture for 24 h as in c. the Western blots for c used
whole-cell lysates, while those in e and f used nuclear extracts. Full blot
images for c, e, and f are shown in Supplementary Figure 11. Results in a–c
and e are representative of ≥3 independent experiments. Results in f are
representative of two independent experiments.
lism might contribute to oncogenesis14,15
.
received 16 June 2016; accepted 15 December 2016;
published online 6 March 2017
METHODS
Methods, including statements of data availability and any associated
accession codes and references, are available in the online version
of the paper.
references
1. Losman, J.A. & Kaelin, W.G. Jr. What a difference a hydroxyl makes:
mutant IDH, (R)-2-hydroxyglutarate, and cancer. Genes Dev. 27,
836–852 (2013).
2. Pavlova, N.N. & ompson, C.B. e emerging hallmarks of cancer
metabolism. Cell Metab. 23, 27–47 (2016).
3. Dang, L. et al. Cancer-associated IDH1 mutations produce
2-hydroxyglutarate. Nature 462, 739–744 (2009).
4. Ward, P.S. et al. e common feature of leukemia-associated IDH1 and IDH2
mutations is a neomorphic enzyme activity converting α-ketoglutarate to
2-hydroxyglutarate. Cancer Cell 17, 225–234 (2010).
dependent on LDHA and MDH2 (refs. 17,18). The relative contribu-
tions of LDHA and MDH2 to hypoxia-induced L-2HG appear to vary
depending on the type of cell, which may be related to enzyme levels
and subcellular pools of substrate. Hypoxic cells undergo a variety
5. Figueroa, M.E. et al. Leukemic IDH1 and IDH2 mutations result in a
hypermethylation phenotype, disrupt TET2 function, and impair
hematopoietic differentiation. Cancer Cell 18, 553–567 (2010).
6. Lu, C. et al. IDH mutation impairs histone demethylation and results in a
block to cell differentiation. Nature 483, 474–478 (2012).
7. Sasaki, M. et al. IDH1(R132H) mutation increases murine haematopoietic
progenitors and alters epigenetics. Nature 488, 656–659 (2012).
8. Kats, L.M. et al. Proto-oncogenic role of mutant IDH2 in leukemia initiation
and maintenance. Cell Stem Cell 14, 329–341 (2014).
of metabolic changes that might favor L-2HG production18,22–25,45
.
Here we identify acidification and increased αKG concentration as
factors that are sufficient to induce cellular production of L-2HG
in normoxia. Indeed, in vitro enzymatic assays demonstrate that
the promiscuous enzymatic activity of LDH, and to a lesser extent
MDH, is enhanced by acidic pH, resulting in more efficient reduc-
tion of αKG to L-2HG. Acidic pH appears to enhance LDHA-
mediated reduction of αKG by driving equilibrium toward the
protonated state of the carboxylate tail, which permits docking of
αKG via interaction with the Q100 residue in LDHA. Intracellular
9. Saha, S.K. et al. Mutant IDH inhibits HNF-4α to block hepatocyte
differentiation and promote biliary cancer. Nature 513, 110–114 (2014).
10. Linster, C.L., Van Schafingen, E. & Hanson, A.D. Metabolite damage and its
repair or pre-emption. Nat. Chem. Biol. 9, 72–80 (2013).
11. Chowdhury, R. et al. e oncometabolite 2-hydroxyglutarate inhibits histone
lysine demethylases. EMBO Rep. 12, 463–469 (2011).
6
nature CHeMICaL BIOLOGY | AdvAnce online publicAtion | www.nature.com/naturechemicalbiology

NATURE cHEMIcAL bIOLOGy dOI: 10.1038/nCHeMBIO.2307
article
12. Xu, W. et al. Oncometabolite 2-hydroxyglutarate is a competitive inhibitor of
α-ketoglutarate-dependent dioxygenases. Cancer Cell 19, 17–30 (2011).
13. Koivunen, P. et al. Transformation by the (R)-enantiomer of
2-hydroxyglutarate linked to EGLN activation. Nature 483, 484–488 (2012).
14. Haliloglu, G. et al. L-2-hydroxyglutaric aciduria and brain tumors in children
with mutations in the L2HGDH gene: neuroimaging findings. Neuropediatrics
39, 119–122 (2008).
38. Vesell, E.S. pH dependence of lactate dehydrogenase isozyme inhibition by
substrate. Nature 210, 421–422 (1966).
39. Wilks, H.M. et al. A specific, highly active malate dehydrogenase by redesign
of a lactate dehydrogenase framework. Science 242, 1541–1544 (1988).
40. Dunn, C.R. et al. Design and synthesis of new enzymes based on the lactate
dehydrogenase framework. Phil. Trans. R. Soc. Lond. B 332, 177–184 (1991).
41. Tokonami, N. et al. α-Ketoglutarate regulates acid–base balance through an
intrarenal paracrine mechanism. J. Clin. Invest. 123, 3166–3171 (2013).
42. Novoa, W.B., Winer, A.D., Glaid, A.J. & Schwert, G.W. Lactic dehydrogenase.
V. Inhibition by oxamate and by oxalate. J. Biol. Chem. 234, 1143–1148
(1959).
43. Schofield, C.J. & Ratcliffe, P.J. Oxygen sensing by HIF hydroxylases. Nat. Rev.
Mol. Cell Biol. 5, 343–354 (2004).
44. Finley, L.W. et al. SIRT3 opposes reprogramming of cancer cell metabolism
through HIF1α destabilization. Cancer Cell 19, 416–428 (2011).
45. Metallo, C.M. et al. Reductive glutamine metabolism by IDH1 mediates
lipogenesis under hypoxia. Nature 481, 380–384 (2011).
46. Mullen, A.R. et al. Oxidation of α-ketoglutarate is required for reductive
carboxylation in cancer cells with mitochondrial defects. Cell Rep. 7,
1679–1690 (2014).
47. Zhao, S. et al. Glioma-derived mutations in IDH1 dominantly inhibit IDH1
catalytic activity and induce HIF-1α. Science 324, 261–265 (2009).
48. Tarhonskaya, H. et al. Non-enzymatic chemistry enables 2-hydroxyglutarate-
mediated activation of 2-oxoglutarate oxygenases. Nat. Commun. 5, 3423
(2014).
49. Chiche, J. et al. Hypoxia-inducible carbonic anhydrase IX and XII promote
tumor cell growth by counteracting acidosis through the regulation of the
intracellular pH. Cancer Res. 69, 358–368 (2009).
50. Mekhail, K., Gunaratnam, L., Bonicalzi, M.E. & Lee, S. HIF activation by
pH-dependent nucleolar sequestration of VHL. Nat. Cell Biol. 6, 642–647
(2004).
15. Shim, E.H. et al. L-2-Hydroxyglutarate: an epigenetic modifier and putative
oncometabolite in renal cancer. Cancer Discov. 4, 1290–1298 (2014).
16. Rzem, R. et al. A mouse model of L-2-hydroxyglutaric aciduria, a disorder of
metabolite repair. PLoS One 10, e0119540 (2015).
17. Intlekofer, A.M. et al. Hypoxia induces production of L-2-hydroxyglutarate.
Cell Metab. 22, 304–311 (2015).
18. Oldham, W.M., Clish, C.B., Yang, Y. & Loscalzo, J. Hypoxia-mediated
increases in L-2-hydroxyglutarate coordinate the metabolic response to
reductive stress. Cell Metab. 22, 291–303 (2015).
19. Chen, C. et al. Cancer-associated IDH2 mutants drive an acute myeloid
leukemia that is susceptible to Brd4 inhibition. Genes Dev. 27, 1974–1985
(2013).
20. Simon, M.C. & Keith, B. e role of oxygen availability in embryonic
development and stem cell function. Nat. Rev. Mol. Cell Biol. 9, 285–296
(2008).
21. Rzem, R., Vincent, M.F., Van Schafingen, E. & Veiga-da-Cunha, M.
L-2-Hydroxyglutaric aciduria, a defect of metabolite repair. J. Inherit. Metab.
Dis. 30, 681–689 (2007).
22. Garofalo, O., Cox, D.W. & Bachelard, H.S. Brain levels of NADH and NAD+
under hypoxic and hypoglycaemic conditions in vitro. J. Neurochem. 51,
172–176 (1988).
23. Wise, D.R. et al. Hypoxia promotes isocitrate dehydrogenase–dependent
carboxylation of α-ketoglutarate to citrate to support cell growth and viability.
Proc. Natl. Acad. Sci. USA 108, 19611–19616 (2011).
24. Bright, C.M. & Ellis, D. Intracellular pH changes induced by hypoxia and
anoxia in isolated sheep heart Purkinje fibres. Exp. Physiol. 77, 165–175
(1992).
25. Yan, G.X. & Kléber, A.G. Changes in extracellular and intracellular pH in
ischemic rabbit papillary muscle. Circ. Res. 71, 460–470 (1992).
26. Meister, A. Reduction of αγ-diketo and α-keto acids catalyzed by muscle
preparations and by crystalline lactic dehydrogenase. J. Biol. Chem. 184,
117–129 (1950).
27. Schatz, L. & Segal, H.L. Reduction of α-ketoglutarate by homogeneous lactic
dehydrogenase X of testicular tissue. J. Biol. Chem. 244, 4393–4397 (1969).
28. Struys, E.A., Jansen, E.E., Verhoeven, N.M. & Jakobs, C. Measurement of
urinary D- and L-2-hydroxyglutarate enantiomers by stable-isotope-dilution
liquid chromatography–tandem mass spectrometry afer derivatization with
diacetyl-L-tartaric anhydride. Clin. Chem. 50, 1391–1395 (2004).
29. Pietrak, B. et al. A tale of two subunits: how the neomorphic R132H IDH1
mutation enhances production of αHG. Biochemistry 50, 4804–4812 (2011).
30. Rendina, A.R. et al. Mutant IDH1 enhances the production of
2-hydroxyglutarate due to its kinetic mechanism. Biochemistry 52, 4563–4577
(2013).
31. Leonardi, R., Subramanian, C., Jackowski, S. & Rock, C.O. Cancer-associated
isocitrate dehydrogenase mutations inactivate NADPH-dependent reductive
carboxylation. J. Biol. Chem. 287, 14615–14620 (2012).
32. Fan, J. et al. Human phosphoglycerate dehydrogenase produces the
oncometabolite D-2-hydroxyglutarate. ACS Chem. Biol. 10, 510–516 (2015).
33. Mattaini, K.R. et al. An epitope tag alters phosphoglycerate dehydrogenase
structure and impairs ability to support cell proliferation. Cancer Metab. 3, 5
(2015).
34. Locasale, J.W. et al. Phosphoglycerate dehydrogenase diverts glycolytic flux
and contributes to oncogenesis. Nat. Genet. 43, 869–874 (2011).
35. Possemato, R. et al. Functional genomics reveal that the serine synthesis
pathway is essential in breast cancer. Nature 476, 346–350 (2011).
36. Terunuma, A. et al. MYC-driven accumulation of 2-hydroxyglutarate is
associated with breast cancer prognosis. J. Clin. Invest. 124, 398–412 (2014).
37. Zhu, A., Romero, R. & Petty, H.R. A sensitive fluorimetric assay for pyruvate.
Anal. Biochem. 396, 146–151 (2010).
acknowledgments
We thank members of the Thompson laboratory for helpful discussions. We thank
M. Isik, S. Hanson, and A. Rizzi from the Chodera laboratory for assistance. A.M.I.
was supported by the NIH/NCI (K08 CA201483-01A1), the Leukemia & Lymphoma
Society (Special Fellow Award 3356-16), the Burroughs Wellcome Fund (Career Award
for Medical Scientists 1015584), the Conquer Cancer Foundation of ASCO, the Susan
and Peter Solomon Divisional Genomics Program, and the Steven A. Greenberg Fund.
The work was also supported, in part, by the Leukemia & Lymphoma Society Specialized
Center of Research Program (7011-16), the Starr Cancer Consortium (I6-A616), and
grants from the NIH, including R01 CA168802-02 and R01 CA177828-02 (C.B.T.),
K99 CA191021-01A1 (C.C.-F.), and the Memorial Sloan Kettering Cancer Center
Support Grant (NIH P30 CA008748). M.R.G. and S.S. received financial support from
the National Science Foundation (MCB 1022208) and infrastructure support from the
National Institute on Minority Health and Health Disparities (8G12MD007603-29).
author contributions
A.M.I. and C.B.T. conceived the project, designed the experiments, analyzed the data,
and wrote the manuscript. A.M.I. and B.W. performed all of the experiments. H.L., H.S.,
and J.R.C. assisted with liquid chromatography–mass spectrometry. C.C.-F. assisted with
intracellular pH measurement. A.S.R., S.S., M.R.G., and J.D.C. performed the computa-
tional modeling. All authors read and approved the manuscript.
Competing financial interests
The authors declare competing financial interests: details accompany the online version
of the paper.
additional information
Any supplementary information, chemical compound information and source data are
available in the online version of the paper. Reprints and permissions information is
available online at http://www.nature.com/reprints/index.html. Correspondence and
requests for materials should be addressed to C.B.T.
nature CHeMICaL BIOLOGY | AdvAnce online publicAtion | www.nature.com/naturechemicalbiology
7

ONLINE METHODS
wavelength of 535 nm. Background fluorescence (cells without probes) was
Reagents. Purified enzymes were purchased from Sigma, including LDH subtracted before all calculations. Intracellular pH was quantified by a standard
from bovine heart (L2625), LDH from rabbit muscle (L2500), and MDH curve ranging from pH 5.5 to pH 7.5 using DMEM with 10 μM nigericin and
from porcine heart (M1567). Recombinant human enzymes were purchased
from Abcam, including LDHA (ab93699), MDH1 (ab99244), MDH2
10 μM valinomycin to equilibrate intracellular pH with extracellular pH.
(ab99238), IDH1 (ab113858), and PHGDH (ab198455). Substrates, cofactors, Enzyme assays. Enzyme reactions were conducted in 33 mM potassium phos-
and inhibitors were purchased from Sigma, including pyruvate, oxaloacetate, phate buffer titrated to pH 7.4, 7.0, 6.6, or 6.0. Purified enzymes were used at
α-ketoglutarate, dimethyl-α-ketoglutarate, L-lactate, D-lactate, L-malate,
D-malate, L-2-hydroxyglutarate, D-2-hydroxyglutarate, α-ketobutyrate, oxo-
5 units/ml, 1 μg/ml, or 10 μg/ml as indicated, and recombinant enzymes were
usedat15μg/mlunlessotherwiseindicated.FLAG-taggedwild-typeandQ100R
pentanoic acid, oxoadipic acid, ketohexanoic acid, phenylpyruvate, hydroxy- mutant LDHA enzymes were purified from transfected 293T cells using Anti-
phenylpyruvate, NADH, NADPH, and oxamate. FLAG-tagged wild-type FLAG M2 Affinity Gel (Sigma) according to the manufacturer’s instructions.
and Q100R mutant LDHA constructs were cloned into the PCDNA3.1(+) FLAG-tagged enzymes were quantified by BCA assay and gel electrophoresis
vector (Addgene) by standard site-directed mutagenesis and verified by
Sanger sequencing.
withCoomassiestaining.NADHandNAPDHwereusedat0.22mMor0.30mM,
respectively, throughout unless otherwise indicated. Substrates were used at
1 mM unless otherwise indicated. For NADH consumption assays, reactions
Cell culture. Adherent cell lines 293T (purchased from ATCC) and SFXL were conducted in UV-transparent 96-well plates (Corning) with reaction vol-
(SF188 cells with stable expression of Bcl-XL; generated as previously
described51) were maintained at low passage number in high-glucose DMEM
with 10% FBS, 25 mM glucose, 4 mM glutamine, 100 units/ml penicillin, and
umes of 200 μl. A SpectraMax Plus 384 Microplate Reader (Molecular Devices)
was used to monitor the absorbance at 340 nm every 30 s or 12 s through-
out the course of the reaction. For each condition, the mean rate of NADH
100 μg/ml streptomycin and split every 2–3 d before reaching confluence. consumption for triplicate control reactions without enzyme was subtracted
Cell lines were authenticated by short tandem repeat (STR) profiling. Cell lines
repeatedly tested negative for mycoplasma throughout the experimental period.
from the rate of NADH consumption for triplicate experimental reactions with
enzyme. Reaction velocities were calculated using an extinction coefficient for
For siRNA experiments, cells were reverse-transfected with siRNA mixed NADH at ε340 of 6,220 M−1cm−1 and a path length of 0.56 cm for a 200-μl reac-
with Lipofectamine RNAiMAX (Life Technologies) in Opti-MEM Reduced tion volume in a standard 96-well plate. For NADH consumption assays with
Serum Medium (Life Technologies) as described by the manufacturer. The purified LDH and pyruvate, reactions were conducted in UV-transparent 3-ml
following siRNAs (Thermo Fisher) were used: siLDHA = J-008201-06, target cuvettes (BrandTech; path length 1 cm) with measurement of the absorbance
sequence: GGCAAAGACUAUAAUGUAA; siMDH2 = J-008439-12, target at 340 nm every 2 s. For cell-lysate enzyme assays, SFXL cells were cultured in
sequence: CGCCUGACCCUCUAUGAUA. For experiments with acidified normoxic conditions in regular DMEM and harvested at subconfluence. The
medium, standard DMEM was titrated to pH 6.0 at room temperature and harvested cells were divided into four equal fractions, and nondenaturing cell
atmospheric oxygen and then filtered through a sterile 0.2-μm filter before use lysates were prepared with PBS plus 0.1% Triton X titrated to pH 7.5, 7.0, 6.5,
in tissue culture.
or 6.0 and used at either 50 or 100 μl per 200-μl reaction with αKG and NADH.
The reaction mixtures were stopped after 16 h, at which time metabolites were
extracted, derivatized, and analyzed by GC–MS or LC–MS.
Western blotting. Cell lysates were extracted in 1× RIPA buffer (Cell
Signaling), sonicated, and centrifuged at 21,000g at 4 °C, and supernatants were
then collected. Nuclear extracts were prepared by harvesting cells in 50 mM Metabolite extraction and analysis. Metabolites were extracted with ice-cold
NaCl, 0.5 M sucrose, 0.5% Triton X plus protease–phosphatase inhibitor
cocktail (Thermo Fisher). Isolated nuclei were washed with 10 mM KCl and
then lysed in 500 mM NaCl and 0.1% NP-40. Cleared cell lysates or nuclear
extracts were quantified by BCA assay (Thermo Fisher) and normalized for
total protein concentration. Samples were separated by SDS–PAGE, transferred
to nitrocellulose membranes (Life Technologies), blocked in 5% milk prepared
80:20 methanol:water containing 2 μM deuterated 2-hydroxyglutarate (D-2-
hydroxyglutaric-2,3,3,4,4-d5 acid; deuterated-2HG) as an internal standard.
After overnight incubation at −80 °C, cell extract was harvested, sonicated,
and centrifuged at 21,000g for 20 min at 4 °C to precipitate protein. Extracts
were then dried in an evaporator (Genevac EZ-2 Elite). For GC–MS, metabo-
lites were resuspended by addition of 50 μl of methoxyamine hydrochloride
in Tris-buffered saline with 0.1% Tween 20 (TBST), incubated with primary (40 mg/ml in pyridine) and incubated at 30 °C for 90 min with agitation.
antibodies overnight at 4 °C and then incubated with horseradish peroxidase Metabolites were further derivatized by addition of 80 μl of MSTFA + 1%
(HRP)-conjugated secondary antibodies (GE Healthcare; anti-mouse, NA931V, TCMS (Thermo Scientific) and 70 μl of ethyl acetate (Sigma) and incubated
sheep, 1:5,000; anti-rabbit, NA934V, donkey, 1:5,000) for 1 h the following
day. After incubation with ECL (Thermo Fisher or GE Healthcare), imaging
was performed using the Amersham Imager 600 (GE Healthcare). Primary
at 37 °C for 30 min. Samples were diluted 1:2 with 200 μl of ethyl acetate, then
analyzed using an Agilent 7890A GC coupled to an Agilent 5975C mass selec-
tive detector. The GC was operated in splitless mode with constant helium car-
antibodies used included anti-Ezh2 (Cell Signaling, 5246P; rabbit; 1:1,000), rier gas flow of 1 ml/min and with an HP-5MS column (Agilent Technologies).
anti-HIF-1α (BD Biosciences, 610959; mouse; 1:200), anti-hydroxyl-HIF-1α The injection volume was 1 μl, and the GC oven temperature was ramped from
(Cell Signaling, 3434P; rabbit; 1:1,000), anti-IDH1 (Proteintech, 12332-1-AP; 60 °C to 290 °C over 25 min. Peaks representing compounds of interest were
rabbit; 1:1,000), anti-IDH2 (Abcam, ab55271; mouse; 1:1,000), anti-LDHA extracted and integrated using MassHunter software (Agilent Technologies)
(Cell Signaling, 2012S; rabbit; 1:1,000), anti-MDH2 (Abcam, ab96193; rabbit; and then normalized to both the internal standard (deuterated-2HG) peak area
1:1,000), anti-alpha-tubulin (Sigma, T9026: mouse; 1:5,000), and anti-vinculin and protein content as applicable. Ions used for quantification of metabolite
(Abcam, ab18058; mouse; 1:1,000).
levels were 2HG m/z 247 (confirmatory ion m/z 349), deuterated-2HG m/z
252 (confirmatory ion m/z 354), malate m/z 335 (confirmatory ions m/z 233,
245), and lactate m/z 190 (confirmatory ion m/z 219). Peaks were manually
Intracellular pH measurement. Cells were loaded with either 5-(and-6)-
carboxy, acetoxymethyl ester, acetate (SNARF-1; Thermo) or 2′,7′-bis-(2- inspected and verified relative to known spectra for each metabolite.
carboxyethyl)-5-(and-6)-carboxyfluorescein, acetoxymethyl ester (BCECF;
Thermo) ratiometric pH probes according to the manufacturer’s instructions.
For LC–MS, dried samples were derivatized with 100 μl of freshly prepared
50mg/ml(+)-diacetyl-L-tartaricanhydride(DATAN, Sigma)indichloromethane-
After loading with pH probe, cells were incubated in DMEM without phenol acetic acid (v/v = 4:1) at 75 °C for 30 min. After cooling to room temperature,
red that was either left unmanipulated or titrated to pH 6.0 with addition of
either DMSO (vehicle) or 5 mM dimethyl-αKG. After 4 h of incubation at
37 °C with 5% CO2, fluorescence was measured using a plate reader (Tecan,
derivatized samples were dried under nitrogen at room temperature and resus-
pended in 200 μl of UltraPure water (18.2 MΩ, PureLab) before LC–MS/MS
analysis. Analysis was performed on a Thermo Vantage triple-quadrupole mass
Infinite M1000). The SNARF-1 probe was detected with a single excitation at spectrometer operating in SRM and negative ionization modes. LC separa-
514 nm and a dual-emission ratio (580 nm and 640 nm). The BCECF probe tion was using an Acquity UPLC HSS T3 analytical column (2.1 × 100 mm,
was detected with a dual excitation (490 nm and 440 nm) and a fixed emission 1.8 μm, Waters) with an Agilent 1260 infinity binary pump. Mobile phase A
nature CHeMICaL BIOLOGY
doi:10.1038/nchembio.2307

was 125 mg/l ammonium formate in water adjusted to pH 3.5 with formic acid,
mobile phase B was methanol, and flow rate was 0.3 ml/min. Initial conditions the Poisson–Boltzmann equation to obtain the pairwise interactions between
were 3% B for 5 min, then increased to 80% B at 5.5 min and held for a further charged and/or polar groups and the solvation energy with an internal dielectric
480,000. MCCE2 explores all side chain rotamers. DelPhi58 was used to solve
2.5 min. 10 min of re-equilibration time was used to ensure retention time constant of 4.0 and 80 for the solvent, with 0.15 salt concentration. The atomic
stability. The column temperature was held at 40 °C. Samples were kept at 4 °C partial charges and radii for the amino acids were obtained from PARSE59, while
and the injection volume was 5 μl. MS source parameters were as follows: spray
voltage: 2,500 V; capillary temperature: 300 °C; vaporizer temperature: 250 °C;
AM1-BCC60,61 charges for the ligands were generated by QUACPAC (OpenEye
Toolkit 2016-Jun.1). MCCE2 calculates how the protein modifies the solution
sheath gas pressure: 50 psi; aux gas pressure: 40 psi. Compound-specific pKa (pKa,sol) of amino acids and substrates upon transfer into its position in the
S-lens values were as follows: 37 V (2HG), 36 V (malate), 34 V (lactate), 40 V protein. NADH and αKG protonation states and tautomers, including their
(α-ketoglutarate), and 41 V (deuterated-2HG). Individual reactions were
relative populations, were generated and performed using Epik (Schrödinger
monitored and collision energies (CE) were as follows: 2HG m/z 363.0 → release 2015-3). Relative Epik energies were calculated at pH 4.0–10.0 at inter-
147.1 (CE: 12 V)*, 129.1 (CE: 27 V); malate m/z 349.0 → 133.0 (CE: 14 V)*, vals of 1.0 pH unit, as well as pH 6.6 and 7.4. The energetic description for
115.0 (CE: 28 V); lactate m/z 305.0 → 89.1 (CE: 14 V)*, α-ketoglutarate m/z amino acids or standard ligands includes the pKa,sol as well as AMBER62 torsion
145.1 → 57.0 (CE: 12 V); 113.0 (CE: 16V); deuterated-2HG m/z 368.0 → 152.1 energies and van der Waals self-interactions. For the substrate, values were set
(CE: 13 V)*, 132.9 (CE: 22 V), with * indicating the primary transition used to to zero and the relative Epik energy of each tautomer and protonation state pro-
quantify each metabolite. Pure standards of L-2HG, D-2HG, L-malate, D-malate,
L-lactate, and D-lactate were derivatized and analyzed in parallel for each chi-
vided the starting reference state in solution. MCCE2 then calculated the shift
in relative energy within the protein and thus showed the change in their dis-
ral determination experiment. The identities of metabolite enantiomers were tribution. The protein side chain rotamers and protonation states were sampled
determined by comparing to the retention times of the derivatized pure stand- along with the ligand tautomers and protonation states at the given pH.
ards and additionally confirmed by spike-ins of derivatized pure standards into
the experimental sample. Absolute metabolite quantification was performed
using an external calibration curve with deuterated-2HG internal standard
and the resulting concentrations corrected for the total cell volume extracted.
Data availability. The docking and MCCE2 calculation data, includ-
ing scripts, have been deposited in “figshare” (https://dx.doi.org/10.6084/
m9.figshare.4289894.v3). Inside the zip file, see the ‘Docking’ folder for files
Chromatograms were acquired and processed with XCalibur and TraceFinder used in the docking procedures. Docking was performed using Maestro as
software (Thermo Fisher).
part of Schrödinger release 2015-3. See the ‘MCCE’ folder for files used for
the MCCE2 calculations for lactate dehydrogenase A. MCCE calculations
were performed using MCCE2 (https://sites.google.com/site/mccewiki/install-
Aqueous solution pKa prediction. The aqueous solution pKa values for the tail
carboxylate moiety of αKG were predicted using Epik (Schrödinger release mcce). Epik calculations were performed using Schrödinger release 2015-3.
2015-3)52 using default settings.
Referenced accession codes can be found in the Protein Data Bank under
accession numbers 4OKN and 1I10.
Virtual docking of αKG to LDHA. Virtual docking of αKG was performed
using Glide XP (Schrödinger release 2015-3)53,54 to chain A of the 2.1-Å struc-
ture of lactate dehydrogenase A (LDHA; PDB identifier: 4OKN; ref. 55). Missing
side chains were automatically constructed, amino acids were assigned proto-
nation states compatible with a pH of 7.0 using the protein preparation wizard
(Schrödinger release 2015-3), and crystallographic waters were removed. Bond
orders for NADH were manually verified to avoid misassignment. Oxalate
was used as a reference ligand to define the binding site, after adding ficti-
tious atoms to ensure the minimum atom requirement for binding-site defini-
tion was met. Hydroxyl and thiol groups were allowed to rotate during grid
generation. The tail-protonated form of αKG was docked in the presence of
crystallographic NADH. Docking poses were minimized post-docking. The
resulting conformations were clustered within a 0.5-Å r.m.s. deviation cutoff.
This resulted in a single representative pose. All steps used default parameters
unless otherwise noted. Poses and predicted hydrogen bonds were visualized
using PyMOL.
51. Wise, D.R. et al. Myc regulates a transcriptional program that stimulates
mitochondrial glutaminolysis and leads to glutamine addiction. Proc. Natl.
Acad. Sci. USA 105, 18782–18787 (2008).
52. Shelley, J.C. et al. Epik: a sofware program for pKa prediction and
protonation state generation for drug-like molecules. J. Comput. Aided Mol.
Des. 21, 681–691 (2007).
53. Friesner, R.A. et al. Glide: a new approach for rapid, accurate docking and
scoring. 1. Method and assessment of docking accuracy. J. Med. Chem. 47,
1739–1749 (2004).
54. Halgren, T.A. et al. Glide: a new approach for rapid, accurate docking and
scoring. 2. Enrichment factors in database screening. J. Med. Chem. 47,
1750–1759 (2004).
55. Kolappan, S. et al. Structures of lactate dehydrogenase A (LDHA) in apo,
ternary and inhibitor-bound forms. Acta Crystallogr. D Biol. Crystallogr. 71,
185–195 (2015).
56. Song, Y., Mao, J. & Gunner, M.R. MCCE2: improving protein pKa calculations
with extensive side chain rotamer sampling. J. Comput. Chem. 30, 2231–2247
(2009).
57. Read, J.A., Winter, V.J., Eszes, C.M., Sessions, R.B. & Brady, R.L. Structural
basis for altered activity of M- and H-isozyme forms of human lactate
dehydrogenase. Proteins 43, 175–185 (2001).
58. Yang, A.S., Gunner, M.R., Sampogna, R., Sharp, K. & Honig, B. On the
calculation of pKas in proteins. Proteins 15, 252–265 (1993).
59. Cornell, W.D. et al. A second generation force field for the simulation of
proteins, nucleic acids, and organic molecules. J. Am. Chem. Soc. 117,
5179–5197 (1995).
Monte Carlo protonation state modeling using MCCE2 simulations.
MCCE2 (ref. 56) was used to perform Monte Carlo simulations of amino acid
side chain position and side chain and substrate protonation at equilibrium.
The 2.3-Å structure of lactate dehydrogenase A (LDHA; PDB identifier 1I10;
ref. 57) was used as input for all calculations (~60 computational hours on
two Intel Xeon 2.40 GHz Six Core CPUs). αKG was modeled using a virtu-
ally docked pose from previous work17. This structure was prepared using the
same docking procedure as detailed in this paper. The protonation states were
compared between structures with the coenzyme and substrate (NADH and
αKG) bound to LDHA or removed. The same starting protein structure was
used for both bound and unbound types of simulation. The simulations were
performed between pH 4.0–10.0 at intervals of 1.0 pH unit, and at pH 6.6 and
7.4. The total number of Monte Carlo steps for equilibration at each pH was
60. Jakalian, A., Bush, B.L., Jack, D.B. & Bayly, C.I. Fast, efficient generation of
high-quality atomic charges. AM1-BCC model: I. Method. J. Comput. Chem.
21, 132–146 (2000).
61. Jakalian, A., Jack, D.B. & Bayly, C.I. Fast, efficient generation of high-quality
atomic charges. AM1-BCC model: II. Parameterization and validation. J.
Comput. Chem. 23, 1623–1641 (2002).
62. Levitt, M. & Lifson, S. Refinement of protein conformations using a
macromolecular energy minimization procedure. J. Mol. Biol. 46, 269–279
(1969).
nature CHeMICaL BIOLOGY
doi:10.1038/nchembio.2307