Derivatization of the tricarboxylic acid intermediates with O-benzylhydroxylamine for liquid chromatography-tandem mass spectrometry detection

Analytical Biochemistry 465 (2014) 134–147

Contents lists available at ScienceDirect

Analytical Biochemistry

journal homepage: w ww.elsevier.com/locate/yabio

Derivatization of the tricarboxylic acid intermediates

with O-benzylhydroxylamine for liquid chromatography–tandem

mass spectrometry detection

b

a

Bo Tan ^a,, Zhaohai Lu , Sucai Dong , Genshi Zhao , Ming-Shang Kuo

⇑

^aTailored Therapeutics, Lilly Research Laboratories, Lilly Corporate Center, Indianapolis, IN 46285, USA

^bCancer Research, Lilly Research Laboratories, Lilly Corporate Center, Indianapolis, IN 46285, USA

a r t i c l e i n f o

a b s t r a c t

Article history:

Received 29 April 2014

Received in revised form 24 July 2014

Accepted 25 July 2014

Available online 4 August 2014

The tricarboxylic acid (TCA) cycle is an interface among glycolysis, lipid metabolism, and amino acid

metabolism. Increasing interest in cancer metabolism has created a demand for rapid and sensitive meth-

ods for quantifying the TCA cycle intermediates and related organic acids. We have developed a liquid

chromatography–tandem mass spectrometry (LC–MS/MS) method to quantify the TCA cycle intermedi-

ates in a 96-well format after O-benzylhydroxylamine (O-BHA) derivatization under aqueous conditions.

This method was validated for quantitation of all common TCA cycle intermediates with good sensitivity,

Keywords:

Tricarboxylic acid cycle intermediates

Derivatization

O-benzylhydroxylamine

LC–MS/MS

including

a-ketoglutarate, malate, fumarate, succinate, 2-hydroxyglutarate, citrate, oxaloacetate,

pyruvate, isocitrate, and lactate using a 8-min run time in cancer cells and tissues. The method was used

to detect and quantify changes in metabolite levels in cancer cells and tumor tissues treated with a phar-

macological inhibitor of nicotinamide phosphoribosyl transferase (NAMPT). This method is rapid, sensi-

tive, and reproducible, and it can be used to assess metabolic changes in cancer cells and tumor samples.

The tricarboxylic acid (TCA)¹cycle plays an important role in

cancer cell metabolism and has been the subject of intensive

research. The ﬁnding that IDH1 mutant enzymes can convert a-keto-

glutarate to 2-hydroxyglutarate in human malignant gliomas [1] has

led to a renewed effort in this area. Therefore, development of ana-

lytical tools for assessment of the TCA cycle intermediates in cancer

cells and tumor tissues is critical to further our understanding of

cancer metabolism and its role in cancer development. A rapid and

sensitive method able to separate and detect all TCA cycle interme-

diates from both cells and tissues would be ideal. Mass spectrometry

(MS) represents one of the most sensitive and versatile analytical

technologies [2–5] for the detection of TCA cycle intermediates.

Gas chromatography is the earliest chromatographic technique

applied to the analysis of selected TCA cycle intermediates. The

method used acylation [6,7] or silylation [4,8–13] with low pg/ml

limits of detection (LODs) and cycle times of 13 min [11] to 60 min

[8]. Capillary electrophoresis (CE)/MS analysis of all TCA cycle inter-

mediates after derivatization [14] demonstrated good sensitivity in

the low micromolar range. However, the difﬁculty of maintaining

CE and limits on injection volumes are limitations of this approach.

Liquid chromatographic approaches, including ion exchange/exclu-

sion chromatography [2,13,15–17], ion pairing [3,18–20], hydro-

philic interaction chromatography (HILIC) [5,21–23], and reverse

phase [24], were also applied broadly in the TCA cycle intermediate

analysis. Ion exchange chromatography coupled with a suppressor

or ion exclusion chromatography can separate many anions, includ-

ing TCA cycle intermediates, with good sensitivity (low ng/ml) and

15- to 75-min run times coupled with MS [2,13,15–17]. Ion pairing

reagents [3] coupled with nano high-performance liquid chromatog-

raphy (HPLC) can be used to separate low molecular weight polar

analytes with LODs around 1 ng/ml. Without nano HPLC, the detec-

tion limits are at the low ng/ml range with a 23- to 36-min cycle

⇑

Corresponding author. Fax: +317 433 6400.

E-mail address: tanbo76@hotmail.com (B. Tan).

Abbreviations used: TCA, tricarboxylic acid; MS, mass spectrometry; LOD, limit of

1

detection; CE, capillary electrophoresis; HILIC, hydrophilic interaction chromatogra-

phy; HPLC, high-performance liquid chromatography; MS/MS, tandem mass

spectrometry; MPEA, N-methyl-2-phenylethanamine; 4-APEBA, 4-(2-((4-bromophen-

ethyl)dimethylammonio)ethoxy)benzenaminium dibromide; EDC, 1-ethyl-3-(3-

dimethylaminopropyl) carbodiimide; O-BHA, O-benzylhydroxylamine; RPMI-1640,

Roswell Park Memorial Institute medium; DMEM, Dulbecco’s modiﬁed Eagle’s

medium; FBS, fetal bovine serum; 2-hydroxyglutarate, ^D-a-hydroxyglutarate

disodium salt; NAMPT, nicotinamide phosphoribosyl transferase; LC, liquid chroma-

tography; NMR, nuclear magnetic resonance; IS, internal standard; QC, quality

control; PCR, polymerase chain reaction; ESI, electrospray ionization; MRM, multiple

reaction monitoring; CE, collision energy; GS1, nebulizer gas; GS2, auxiliary gas; CUR,

curtain gas; CAD, collision gas; DP, declustering potential; EP, entrance potential; CXP,

cell exit potential; S/N, signal/noise; LOQ, limit of quantiﬁcation; RSD, relative

standard deviation.

http://dx.doi.org/10.1016/j.ab.2014.07.027

Derivatization of TCA intermediates for LC–MS/MS / B. Tan et al. / Anal. Biochem. 465 (2014) 134–147

135

time [18–20,25]. This strategy provided good sensitivity for anions

Preparation of standard solutions

but limited the use of the instruments for positive mode detection

due to the ion suppression of residual ion pairing reagents [25].

HILIC and reverse phase methods have been applied to metabolo-

mics [22–24] and show good sensitivity for some TCA cycle interme-

diates [5,21,23]. Although all of those methods have their own

merits and ﬁt their original goals, they were not ideal for our

purpose.

Stock solutions of each TCA cycle intermediate (10 mg/ml) were

prepared by dissolving powdered standards in water, including

lactate, pyruvate, succinate, fumarate, malate,

a-ketoglutarate,

2-hydroxyglutarate, oxaloacetate, citrate, and isocitrate. The stan-

dard solutions were mixed and then diluted in water to provide

a set of standard solutions ranging from 1 ng/ml to 500 lg/ml.

Theoretically our task can be achieved by using reverse phase

HPLC–tandem mass spectrometry (MS/MS) with a derivatization

method to improve chromatographic behavior, ionization efﬁ-

ciency, and sensitivity. Some of these beneﬁts have been shown

by many derivatization reagents [26] such as 4-dimethylamino-

benzylamine [27], 2-picolylamine, 2-hydrazinopyridine, hydrazi-

noquinoline [28], N-methyl-2-phenylethanamine (MPEA) [25],

4-(2-((4-bromophenethyl)dimethylammonio)ethoxy)benzenaminium

dibromide (4-APEBA) [29], 3-hydroxy-1-methylpiperidine [30],

p-dimethylaminophenacyl bromide [31,32], aniline [33], 2-nitro-

phenylhydrazine [34], and 2,2,2-triﬂuoroethylamine hydrochloride

(TFEA) [35]. Most of the reagents reported were not designed for

high throughput, good sensitivity, broad coverage, and ease of

use [25,27–30,32–35] for measurement of the TCA cycle interme-

diates in complex matrices. It is also desirable to perform analysis

of complex samples in a 96-well format in aqueous solution in

order to avoid phase transfer that can cause analyte loss.

In this work, we report for the ﬁrst time that a reagent can be

used to derivatize all TCA cycle intermediates, including lactate

and pyruvate, in aqueous solution. When analyzed with reverse

phase HPLC and triple quadruple mass spectrometer, this method

showed good sensitivity and good separation for all TCA cycle

intermediates. This method was also successfully applied to detect

and quantify metabolites in cancer cell extracts and tumor extracts

in a 96-well format by means of stable isotope-labeled internal

standard addition experiments [36]. This method is able to fulﬁll

drug discovery research needs and has the potential to be used

for global proﬁling of many other known endogenous metabolites

that contain carboxylic acid groups.

Internal standard (IS) solutions were made in 80% methanol/water

containing ﬁxed concentrations of IS (5000 ng/ml citrate-d₄,

1000 ng/ml fumarate-¹³C₄, 1000 ng/ml 2-hydroxyglutarate-d₄,

500 ng/ml

a-ketoglutarate-d₆, 300 ng/ml succinate-d₄, 5000 ng/ml

pyruvate-¹³C₃, and 5000 ng/ml lactate-¹³C₃). Methanol/water

(80%) was used to make IS solutions in order to keep the ﬁnal

extraction solution close to 80%. During the validation period,

standard solutions and IS solutions were freshly made. Standard

solutions (10

ll) and IS solutions (10

ll) were mixed, dried,

reconstituted in 100

ll of water, and used as standard curves in

simple aqueous solution.

Preparation of QC samples for validation and TCA metabolite

measurement in cells

Cancer cells (A2780) were grown in 96-well plates as described

in the literature [37]. Brieﬂy, cells were cultured in RPMI-1640 in

the presence of 10% FBS, seeded into a 96-well culture plate at a

density of 5 Â 10⁴cells per well, and incubated in 100

ll of DMEM

supplemented with 10% FBS and 25 mM glucose at 37 °C in 5% CO₂

for 4 h. The medium was removed, and some of the samples were

spiked with standard solutions for use as quality control (QC) sam-

ples. For each concentration of QC samples, 10

ll of standard solu-

tions and 10 l of IS solutions were spiked into cell samples (n = 3).

l

Unspiked cell samples were saved as matrices for the standard

curves and other experiments. The process was repeated for a total

of three 96-well plates, and the plates were frozen at À80 °C. Each

day, one set of QCs was thawed. A freshly made standard curve

solution (10

into the previously unspiked wells. Then all wells were extracted

by the addition of 100 l ice-cold 80% methanol. After incubation

ll) and a freshly made IS solution (10 ll) were spiked

Materials and methods

l

on wet ice for 15 min, the resulting extracts were transferred into

a 96-well cluster tube plate (Analytical Sales and Services, Pomp-

ton Plains, NJ, USA), dried under a stream of nitrogen at 40 °C using

Materials

MS/HPLC-grade formic acid, 1-ethyl-3-(3-dimethylaminopropyl)

carbodiimide (EDC), glucose, O-benzylhydroxylamine (O-BHA),

Roswell Park Memorial Institute medium (RPMI-1640), Dulbecco’s

modiﬁed Eagle’s medium (DMEM), fetal bovine serum (FBS),

ammonium bicarbonate, ammonium hydroxide (28–30%), hydro-

chloric acid (12.1 M HCl), succinate, fumarate, malate, sodium

a homemade 96-well plate dryer, and reconstituted in 100

water. The samples were derivatized and analyzed.

ll of

For the method sensitivity test, ISs with various concentrations

(10 l) were added into the previously unspiked cell samples

(n = 3). These samples were combined with 100 l of ice-cold

l

80% methanol/water and incubation on wet ice for 15 min. The

extracts were transferred, dried, and reconstituted in the same

way as QC samples.

pyruvate, sodium lactate, oxaloacetate, maleate,

a-ketoglutarate-d₆,

sodium pyruvate-¹³C₃,

D

-a

-hydroxyglutarate disodium salt

(2-hydroxyglutarate), citrate-d₄, and succinate-d₄were purchased

from Sigma–Aldrich (St. Louis, MO, USA). Fumarate-¹³C₄and

lactate-¹³C₃were purchased from Cambridge Isotope Laboratories

(Tewksbury, MA, USA). Citrate and isocitrate trisodium salts and

Optima liquid chromatography (LC)/MS-grade water were

purchased from Thermo Fisher Scientiﬁc (Pittsburgh, PA, USA).

2-Hydroxyglutarate-d₄and FK866 (nicotinamide phosphoribosyl

transferase [NAMPT] inhibitor) were custom-synthesized and

conﬁrmed by nuclear magnetic resonance (NMR) and LC/MS.

For the TCA intermediate measurement after FK866 treatment,

cells were treated for 24 h [37] and the cell medium was removed.

IS solution (10

ll) was added, and the samples were combined

with 100 l of ice-cold 80% methanol/water and incubated on

l

wet ice for 15 min. The extracts were transferred, dried, and recon-

stituted in the same way as QC samples.

Preparation of QC samples for validation and TCA intermediate

measurement in tumors

Methylmalonate sodium salt and the lysing matrix

A were

purchased from MP Biomedicals (Santa Ana, CA, USA). Ethyl

acetate and HPLC-grade methanol were purchased from VWR

International (Plainview, NY, USA). Cancer cell lines (A2780 and

HCT116) were purchased from American Type Culture Collection

(ATCC, Manassas, VA, USA). Fox chase SCID mice were purchased

from Charles River Laboratories (Chicago, IL, USA).

A2780 tumors were grown as described in the literature [37].

The tumor samples (ꢀ50 mg) were combined with ice-cold 80%

methanol/water (1 ml/50 mg tissue) using lysing matrix A. Tumors

were homogenized for 30 s with a FastPrep FP120 homogenizer

(Santa Ana, CA, USA). The homogenates were pooled and vortexed

for 10 min, and 10 ll of homogenate was added into all wells of

136

Derivatization of TCA intermediates for LC–MS/MS / B. Tan et al. / Anal. Biochem. 465 (2014) 134–147

three 96-well polymerase chain reaction (PCR) plates. For each

concentration of QC samples, 10 l of standard solution and 10 ll

of IS solution were spiked into tumor homogenate samples

(n = 3). Unspiked samples were saved as matrices for the standard

curves and other experiments. The process was repeated for a total

of three 96-well plates, and the plates were frozen at À80 °C. Each

day, one set of QCs was thawed. A freshly made standard curve

were kept at À20 °C for 2 weeks or at room temperature for 2 days.

The stability was determined by the peak area ratios between the

stored samples and the freshly made ones after LC–MS/MS analysis

as described in the next section.

l

LC–MS/MS analysis of derivatized samples

solution (10

into the unspiked wells and used as standard curves. All samples

were combined with 100 l of ice-cold 80% methanol. The plate

was shaken for 10 min using a VWR Multi-Tube Vortexer (VWR

International, Radnor, PA, USA). The 96-well plate was centrifuged

at 4000g for 10 min at 5 °C on a 5810R Benchtop Centrifuge

(Eppendorf, Westbury, NY, USA), and the supernatants were col-

lected into a 96-well cluster tube plate. The supernatants were

dried under a stream of nitrogen at 40 °C and reconstituted in

ll) and a freshly made IS solution (10 ll) were spiked

The derivatized samples (10 ll) were injected onto an LC–triple

quadrupole mass spectrometer (QTRAP 5500, AB Sciex, Framing-

ham, MA, USA), coupled with a pair of Shimadzu 10ADVP pumps,

and a Shimadzu SIL-20A autosampler (Shimadzu, Columbia, MD,

USA). Samples were separated by HPLC using a Waters XBridge

C18 column (2.1 Â 50 mm, Waters, Milford, MA, USA) with a

two-solvent system (A: water with 0.1% formic acid; B: methanol)

and a ﬂow rate of 0.6 ml min^À1. The gradient program was as fol-

lows: 0 to 1.0 min, 0 to 30% B; 1.0 to 5.5 min, 30 to 85% B; 5.5 to

5.6 min, 85 to 98% B; 5.6 to 6.5 min, 98% B; 6.5 to 8.0 min, 0% B.

l

100

For the method sensitivity test, ISs with various concentrations

(10 l) were added to 10 l of tumor homogenates (n = 3). These

samples were combined with 100 l of ice-cold 80% methanol/

ll of water. The samples were derivatized and analyzed.

The injection volume was 10 ll. The MS acquisition methods used

l

positive electrospray ionization (ESI) and analysis in multiple reac-

tion monitoring (MRM) mode. MRM transitions and collision

energy (CE) parameters for all analytes are listed in Table 1. Ion-

spray voltage was 4500 V. Nebulizer gas (GS1), auxiliary gas

(GS2), curtain gas (CUR), and collision gas (CAD) were 60, 60, 30,

and 5 (arbitrary units), respectively. The ion source temperature

was maintained at 600 °C. The curtain gas and collision gas were

nitrogen. Declustering potential (DP), entrance potential (EP), and

cell exit potential (CXP) were 80, 10, and 10, respectively. For the

separation of methylmalonate/succinate and maleate/fumarate, a

longer HPLC method was employed using a Phenomenex Synergi

l

water, shaken, and centrifuged in the same way as QC samples.

The extracts were transferred, dried, and reconstituted in the same

way as QC samples.

For the TCA intermediate measurement after treatment with

FK866 (10 mg/kg) for 10 days as described in the literature [37],

the tumor samples (ꢀ50 mg) were homogenized after the addition

of ice-cold 80% methanol/water (1 ml/50 mg tissue) using lysing

matrix A and a FastPrep FP120 homogenizer in the same way as

QC samples. Then 10

were added into a 96-well PCR plate. These samples were com-

bined with 100 l of ice-cold 80% methanol/water, shaken, and

ll of homogenate and 10 ll of IS solution

C18 Fusion-RP column (250 Â 2.0 mm, 4

lM, Torrance, CA, USA).

l

This method used a 0.3-ml min^À1ﬂow rate and the following gra-

dient program: 0 to 0.5 min, 10 to 50% B; 0.5 to 3.0 min, 50% B; 3.0

to 10.0 min, 50 to 80% B; 10.0 to 10.5 min, 80 to 98% B; 10.5 to

13.5 min, 98% B; 13.5 to 14.0 min, 98 to 10% B; 14.0 to 17.0 min,

10% B.

centrifuged in the same way as QC samples. The extracts were

transferred, dried, and reconstituted in the same way as QC

samples.

Sample derivatization

Pyridine buffer was prepared by combining 5.4 ml of HCl

(12.1 M), 8.6 ml of pyridine, and 86 ml of water. The pH was mea-

sured with pH paper to be around 5.0. Using a Biomek FX worksta-

tion (Beckman Coulter, Brea, CA, USA), reconstituted standard

curves (in water, in cell matrix, and in tumor matrix) and samples

Measurement of cell and tumor lysates using HILIC method

A HILIC LC–MS/MS method was used as a comparison with the

derivatization method. Cell samples were cultured as described in

the validation section above (‘‘Preparation of QC samples for vali-

dation and TCA metabolite measurement in cells’’). The medium

were mixed with 50

pyridine buffer. After 1 h at room temperature, ethyl acetate

(300 l) was added and the plates were shaken for 10 min using

ll of 1 M O-BHA and 50 ll of 1 M EDC in the

was discarded, and the cells were extracted with 100

ll of 80%

l

methanol after IS addition (10 l). The extract was centrifuged at

l

a VWR Multi-Tube Vortexer. The organic layer was taken into a

96-well 2-ml plate using a Biomek FX workstation. Then the aque-

ous layer was extracted again with 300 ll of ethyl acetate using the

Vortexer and Biomek FX workstation. The organic layers were

combined into a 96-well 2-ml plate. The 2-ml plate was dried using

a stream of nitrogen at 40 °C and reconstituted in 1 ml of 50%

methanol/water. The samples were analyzed by LC–MS/MS.

For optimization of the derivatization procedure, 5 ml of 1 M

EDC and 5 ml of 1 M O-BHA were added into 10 ml of a 1-mg/ml

standard solution for each analyte. At different time points (0.25,

4000g for 10 min at 5 °C on a 5810R Benchtop Centrifuge and

injected onto the column. Tumor samples were collected and

homogenized as in the validation section above (‘‘Preparation of

QC samples for validation and TCA intermediate measurement in

tumors’’). The homogenates (10

ll) were combined with 10-ll IS

solutions and 100 l of 80% methanol/water. The mixtures were

l

centrifuged at 4000g for 10 min at 5 °C on a 5810R Benchtop Cen-

trifuge. The top layer was dried under a stream of nitrogen at 40 °C

and reconstituted in 100 ll of 90% acetonitrile/water solutions.

Then samples were directly injected onto the column. The HILIC

method used a Phenomenex Luna Amino column (2.1 Â 30 mm)

with a two-solvent system (A: 20 mM ammonium bicarbonate in

water with 0.2% ammonium hydroxide; B: acetonitrile) and a ﬂow

rate of 1 ml min^À1. The gradient program was as follows: 0 to

0.5 min, 90% B; 0.5 to 2.0 min, 90 to 40% B; 2.0 to 3.0 min, 40% B;

0.5, 0.75, 1, 2, 4, and 24 h), 50-ll aliquots were taken out, diluted

3-fold with mobile phase A of the ion pairing method, and ana-

lyzed by the ion pairing method described below (see ‘‘Measure-

ment of cell and tumor lysates using HILIC method’’ section) to

monitor the reaction completion. The derivatization patterns of

TCA cycle intermediates with multiple functional groups were

obtained from examination of MS spectra of the infused reaction

mixtures. The MS analysis was also optimized by infusion of the

reaction mixture into the MS instrument.

3.0 to 4.5 min, 90% B. The injection volume was 10 ll. The MS

acquisition methods used the MRM method in negative ESI mode.

The MRM transitions, CE, and DP parameters are listed in Supple-

mental Table 1 of the online Supplementary material. Ionspray

voltage was À4500 V. GS1, GS2, CUR, and CAD were 70, 70, 30,

and 5 (arbitrary units), respectively. The ion source temperature

For the stability test of the derivatives, 10,000 ng/ml O-BHA

derivatives of the TCA cycle intermediates in 50% methanol/water

Derivatization of TCA intermediates for LC–MS/MS / B. Tan et al. / Anal. Biochem. 465 (2014) 134–147

137

Table 1

MS parameters, retention times, ISs, and detection limits for derivatized TCA cycle intermediates.

Analyte name

MRM ion

pair Q1/Q3

CE Retention

time (min)

IS

LOQ

Linear range

LOQ in cells

(ng/50 K cells) matrix

(ng/mg tumor)

LOQ in tumor

(pg on-column) in water

solution

(pg on-column)

Citrate

508/385

327/204

345/222

448/325

462/339

329/206

299/181

196/124

359/236

327/204

329/206

333/210

512/389

331/208

466/343

363/240

302/181

199/124

13 4.44

13 4.30

15 3.90

15 3.18

15 4.90

17 5.00

15 3.50

10 4.80

25 1.22

13 3.40

15 3.90

15 3.47

15 3.44

13 4.44

15 3.90

17 5.00

13 4.20

15 4.80

10 1.22

Citrate-d₄

5

10

5

5–1000

NA

1

NA

5

NA

2

Isocitrate

Fumarate

Malate

Citrate-d₄

10–3000

5–3000

10–1000

0.3–300

3–100

10–3000

300–50,000

3–1000

N/A

Fumarate-¹³C₄

NA

6

NA

1

N/A

0.6

1

2-Hydroxyglutarate-d₄10

Oxaloacetate

a

-Ketoglutarate

a-Ketoglutarate-d₆

0.3

3

10

300

3

30

NA

0.5

N/A

0.3

0.5

0.03

N/A

0.3

3

Succinate

Succinate-d₄

Pyruvate

Pyruvate-¹³C₃

Lactate^a

Lactate-¹³C₃

2-Hydroxyglutarate

Maleate

2-Hydroxyglutarate-d₄

Fumarate-¹³C₄

Methylmalonate

Succinate-d₄

Citrate-d₄

Succinate-d₄

N/A

30

N/A

Fumarate-¹³C₄

1

a-Ketoglutarate-d₆

0.06

N/A

0.6

6

2-Hydroxyglutarate-d₄

Pyruvate-¹³C₃

Lactate-¹³C₃

a

The CE parameter for lactate was adjusted higher to generate lower peaks in order to avoid peak saturation.

was maintained at 700 °C. The CUR and CAD were nitrogen. EP and

CXP were À10 and À10, respectively.

Results and discussion

Derivatization reaction

Measurement of cell and tumor lysates using ion pairing method

In searching for an ideal derivatization agent, we aimed for

three critical success factors: (i) high throughput, (ii) high sensitiv-

ity of TCA metabolites for cell and tumor extracts, and (iii) general

applicability to organic acids with and without other functional

groups. Ideally, the derivatization should be carried out in an aque-

ous environment, have a nonpolar group to facilitate chromatogra-

phy, and have another functional group to enhance ionization in

the ESI mode. Several carboxylic acid derivatization approaches

used EDC to form nonpolar amide derivative in order to facilitate

chromatography [25,29,33–35]. EDC is widely used as a coupling

reagent for the carboxylic acid group and can be used under aque-

ous conditions and at room temperature. In the ﬁrst step of the

EDC coupling reaction, the carboxylic acid group is activated and

forms an electrophilic O-acylisourea. Then the intermediate is

attacked by nucleophilic amines or water. Under aqueous condi-

tions, water is present in a large excess. Reaction with water leads

to deactivated carboxylic acids and failed derivatization. To mini-

mize hydrolysis, the reaction rate with amines needs to be opti-

mized, which is affected greatly by the pH of reaction [25,38]. At

pH 3.5 to 5.0, more than 80% of carboxylic groups are activated

by EDC in seconds [38] with a 1:1 ratio of carboxylic acid and

amine. Protonation of the amine inhibits amide formation, so that

pH needs to be at or above the pK_aof the amine. Therefore, an

amine with a pK_aaround 5.0 is ideal for EDC coupling. This rules

out the amines APEBA, 4-dimethylamino-benzylamine, 2-picolyl-

amine, 2-hydrazinopyridine, hydrazinoquinoline, and MPEA

because they have pK_avalues higher than 7.0. Aniline has a pK_a

around 5.0, but it was not able to react with pyruvate [25,33]

and the negative mode ion pairing method was used after deriva-

An ion pairing method was used as a comparison with the

derivatization method and also to monitor unreacted TCA cycle

intermediates for optimization of the reaction time. Cell and tumor

samples were prepared and extracted as in the previous section,

dried, and reconstituted in 100 ll of water. The ion pairing method

was modiﬁed based on the literature [20] and used a Waters T3

column (2.1 Â 50 mm) with a two-solvent system (A: 10 mM trib-

utyl amine and 15 mM acetic acid in 3% methanol/water; B: 20 mM

tributyl amine and 30 mM acetic acid in 90% acetonitrile/water)

and a ﬂow rate of 0.4 ml min^À1. The gradient program was as fol-

lows: 0 to 0.5 min, 0% B; 0.5 to 1.8 min, 0 to 70% B; 1.8 to

3.0 min, 90% B; 3.0 to 6.0 min, 0% B. The injection volume was

10 ll, and the method used negative ESI mode. The MRM transi-

tions, DP, and CE parameters are listed in Supplemental Table 1.

Ionspray voltage was À4500 V. GS1, GS2, CUR, and CAD were 50,

50, 30, and 5 (arbitrary units), respectively. The ion source temper-

ature was maintained at 600 °C. The CUR and CAD were nitrogen.

EP and CXP were À10 and À10, respectively.

Statistical analysis

Data analysis was performed using MultiQuant 2.1 (AB Sciex).

Calibration curves were calculated by least-squares linear regres-

sion with 1/x weighting. The TCA cycle intermediates were quanti-

ﬁed using the analyte/IS area ratios from the standard curves.

Comparisons between groups were made with the one-way analy-

sis of variance (ANOVA) followed by Dunnett’s t test using JMP 9.0

software. P < 0.05 was considered as the signiﬁcant level of

difference.

tization. We suspect that the close proximity of the

a-carbonyl

group in pyruvate reduces the reactivity of the carboxylic acid

group. To overcome that, the reaction needs to use a more nucleo-

philic amine. O-BHA (pK_aof 4.8) has been reported to stoichiomet-

rically react with the carboxylic acid groups of N-protected b-

Institutional compliance statement

Mice were housed in an Association for Assessment and Accred-

itation of Laboratory Animal Care (AAALAC)-accredited facility at

Lilly Research Laboratories (Indianapolis, IN, USA). Study protocols

were approved by the Eli Lilly institutional animal care and use

committee.

chloro-L-alanine [39] using carbodiimide as coupling reagent under

aqueous conditions. O-BHA is more nucleophilic than aniline and

has better solubility at pH 5.0 compared with MPEA [25]. A higher

ratio of amine versus carboxylic acid groups can be maintained to

138

Derivatization of TCA intermediates for LC–MS/MS / B. Tan et al. / Anal. Biochem. 465 (2014) 134–147

Fig.1. Mechanism of O-BHA derivatization and MS/MS spectra of O-BHA derivatives. (A) Mechanism of the derivatization reaction of

mechanism of derivatized -ketoglutarate. (C–N) Positive ESI MS/MS spectra of O-BHA derivatives of TCA cycle intermediates (C: malate; D: succinate; E: citrate; F: 2-

hydroxyglutarate; G: -ketoglutarate; H: pyruvate; I: isocitrate; J: oxaloacetate; K: lactate; L: fumarate; M: maleate; N: methylmalonate).

a-ketoglutarate. (B) Fragmentation

a

fully label all carboxylic acid groups. Furthermore, the carbonyl

groups can form oxime ethers with O-BHA to increase the mass

even more. The derivatized acids would possess benzyl groups to

facilitate reverse phase chromatography and amide functional

groups to enhance ESI detection. Therefore, O-BHA with EDC

coupling was chosen to derivatize TCA cycle intermediates with

or without carbonyl under aqueous conditions in our method.

We conﬁrmed that O-BHA derivatizes carboxylic acid and carbonyl

groups on important TCA cycle intermediates, including pyruvate,

lactate, succinate, citrate, isocitrate,

a-ketoglutarate, 2-hydroxy-

Derivatization of TCA intermediates for LC–MS/MS / B. Tan et al. / Anal. Biochem. 465 (2014) 134–147

139

Fig.2. (A) HPLC–MS/MS spectra of O-BHA derivatized TCA cycle metabolite standards using XBridge C18 column (2.1 Â 50 mm). (B) Time course of derivatization reaction.

The percentage of the TCA metabolites reacted is 100% minus the unreacted portions. The unreacted portions were measured by the ion pairing method.

glutarate, oxaloacetate, fumarate, malate, methylmalonate, and

maleate. After the reaction, the solution was extracted with ethyl

acetate. The excess reagents and salts were left in the aqueous

phase. Therefore, the organic phase produced a clean matrix for

LC–MS/MS analysis after it was dried down and reconstituted in

50% methanol/water.

and 2-hydroxyglutarate gain a mass of 210 Da (two units of

105 Da) because they have either two carboxylic groups or one car-

boxylic group and one carbonyl. Citrate, oxaloacetate, and a-keto-

glutarate gain a mass of 315 Da (three units of 105 Da). An

examination of the MS/MS spectra (Fig. 1) of all derivatized prod-

ucts under positive ESI conditions showed that the fragment ion

with m/z 91 (benzyl from O-BHA) had the highest intensity for

all analytes. However, in derivatized samples, there were many

interference peaks in MRM channels using this product ion. All

derivatized organic acids, except lactate and pyruvate, lose

123 Da in the collision cell, corresponding to the loss of the

O-BHA moiety (NH2OCH2C6H5) (Fig. 1). This neutral loss provides

more speciﬁcity for derivatized TCA cycle intermediates. Lactate

gives an O-BHA positive fragment ion of m/z 124. Derivatives of

pyruvate and ¹³C-labeled pyruvate both produce a fragment ion

of m/z 181, which is generated through gas phase rearrangement

and does not contain the pyruvate backbone. These alternate frag-

ment ions were selected (Table 1) because they had improved

selectivity compared with the m/z 91 fragment.

Optimization of the derivatization reaction was conducted at

room temperature and pH 5.0 because the pK_aof O-BHA was 4.8

and the EDC-activated carboxylic acids hydrolyzed more rapidly

at higher temperatures [29]. Pyridine/HCl was used as a buffer

because the tertiary amine of pyridine does not participate in

amide formation. The reaction time was optimized by monitoring

the disappearance of the unreacted organic acids at various time

points using the ion pairing method. The MS spectra of infused

reaction mixtures at various time points were examined for the

derivatization species with multiple carboxylic acid and carbonyl

groups. We found that the reaction was complete in less than

60 min with a yield of more than 90% (see Fig. 2 below) and all car-

boxylic acid and carbonyl groups were labeled. Therefore, 1 h was

chosen as the reaction time. A comparison of the intensities of

freshly prepared and stored derivatives showed them to be stable

at À20 °C for 2 weeks and at room temperature for 2 days, similar

to other amide derivatives of carboxylic acids [25].

The reaction scheme (Fig. 1) showed that both the carboxylic

acid and carbonyl groups are derivatized with O-BHA, but with dif-

ferent mechanisms. The former groups generate amides through

EDC activation, and the latter groups form oxime ethers through

direct reaction with O-BHA. The mass gain is the same for both

mechanisms. Each analyte can gain one to three groups with a

mass of 105 Da, depending on the number of acids and carbonyls.

For example, lactate has one carboxylic group, so it gains a mass of

105 Da after derivatization. Fumarate, malate, succinate, pyruvate,

The chromatographic elution order was determined by the

polarity and molecular weight (Fig. 2). Analytes with hydroxyl

groups or lower molecular weights elute ﬁrst. Without the hydro-

xyl group, succinate and fumarate elute later than malate. Among

the derivatives with a molecular weight greater than 400 Da, cit-

rate and 2-hydroxylglurate with hydroxyl groups are more polar

than oxaloacetate and

following: lactate, malate, 2-hydroxylglutarate, succinate, malate,

fumarate, isocitrate, citrate, oxaloacetate, pyruvate, and -ketoglu-

a-ketoglutarate. So the elution order is the

a

tarate. Citric acid and isocitric acid were separated using an

XBridge C18 column. Derivatization of pyruvate generated two

peaks presumed to be the E and Z isomers of oxime [40,41]. The

latter peak was used for quantitation due to a better signal/noise

140

Derivatization of TCA intermediates for LC–MS/MS / B. Tan et al. / Anal. Biochem. 465 (2014) 134–147

α

-Ketoglutarate-d₆

A

D

B

C

2-Hydroxyglutarate

1.0E+5

5.0E+4

1.0E+4

5.0E+3

1.0E+4

5.0E+3

Citrate-d₄

0.0E+0

1.0E+4

0.0E+0

2.0E+4

0.0E+0

2.0E+4

2

3

4

5

6

4

5

6

Time,min

E

F

Fumarate ¹³C₄

Isocitrate

Pyruvate ¹³C₃

1.0E+4

0.0E+0

1.0E+4

0.0E+0

1.0E+4

5.0E+3

0.0E+0

5.0E+3

0.0E+0

2.0E+4

4

5

6

3

4

5

4

5

6

Time,min

Time, min

G

H

I

4.0E+4

3.0E+4

2.0E+4

1.0E+4

0.0E+0

Oxaloacetate

Succinate-d₄

Lactate d₃

1.0E+4

0.0E+0

0.5

1.5

2.5

4

5

6

3

4

5

Time, min

Time,min

Fig.3. LC–MS/MS chromatograms of O-BHA derivatized nonspiked cell extract (lower) and isotope-labeled or nonlabeled TCA cycle intermediates (upper) at the lowest

quantiﬁcation level spiked into the cells: (A) 2-hydroxyglutarate (5 pg); (B) a

-ketoglutarate-d₆(0.3 pg); (C) citrate-d₄(5 pg); (D) isocitrate (10 pg); (E) fumarate-¹³C₄(5 pg);

(F) pyruvate-¹³C₃(3 pg); (G) lactate-¹³C₃(30 pg); (H) succinate-d₄(3 pg); (I) oxaloacetate (50 pg). Numbers in parentheses are the amounts injected into the HPLC. Labeled

malate was not able to be purchased.

(S/N) ratio (see Figs. 3 and 4 below). This method does not distin-

guish between the -hydroxyglutarate and -hydroxyglutarate

mined for all analytes and are listed in Table 1. The LOQ of lactate

was limited by the background signals present in HPLC-grade water.

All other analytes had LOQs of 0.3 to 30 pg on-column. Detection

limits of TCA cycle metabolites in biological matrices such as cell

and tumor extracts were difﬁcult to determine due to the endoge-

nous presence of these metabolites. The use of nonlabeled standards

in surrogate matrix and the use of surrogate analytes in biological

matrix are the two common approaches [37]. A suitable surrogate

matrix could not be found for TCA cycle intermediate measurement

in cancer cells and tumors. It is possible to use isotope-labeled stan-

dards as surrogate analytes. Without the second set of isotope-

labeled standards as ISs, the assay may show high variation. The

use of ¹³C-labeled cell culture extracts [42,43] can avoid using the

second set of ISs. However, they were not used because validation

of those approaches can be difﬁcult. Instead, a stable labeled stan-

dard addition experiment without IS was conducted to estimate

the LOQs. All stable labeled standards and nonlabeled standards

with low endogenous levels, including 2-hydroxyglutarate, isoci-

D

-a

L-a

forms of 2-hydroxyglutarate. Other isobaric pairs, methylmalo-

nate/succinate and maleate/fumarate, were not separated on the

XBridge column and were separated by a Phenomenex Synergi

C18 Hydro RP column using the same HPLC mobile phase with a

slower gradient (see Supplemental Fig. 1 in Supplementary mate-

rial). Methylmalonate and maleate were not observed in cancer

cell and tumor extracts. Because both methylmalonate and male-

ate are not produced at detectable levels in cancer cells, they are

not discussed further. Under these circumstances, the XBridge

C18 column was sufﬁcient for our purposes.

Detection limits

The limit of quantiﬁcation (LOQ) is deﬁned as the lowest amount

of analyte that can be quantitatively determined with less than 20%

variation for precision and accuracy. The LOQs in water were deter-

Derivatization of TCA intermediates for LC–MS/MS / B. Tan et al. / Anal. Biochem. 465 (2014) 134–147

141

α-Ketoglutarate-d₆

A

B

C

2-Hydroxyglutarate

1.0E+5

5.0E+4

1.0E+4

5.0E+3

0.0E+0

2.0E+4

Citrate-d₄

5.0E+3

0.0E+0

1.0E+4

2

3

4

5

6

4

5

6

Time,min

Time, min

D

E

F

Fumarate ¹³C₄

Pyruvate ¹³C₃

2.0E+4

1.0E+4

0.0E+0

Isocitrate

1.0E+4

5.0E+3

0.0E+0

1.0E+4

0.0E+0

2.0E+4

3

4

5

4

5

6

5

6

Time, min

G

H

I

2.0E+4

Oxaloacetate

Succinate-d₄

Lactate d₃

1.0E+4

0.0E+0

5.0E+3

0.0E+0

1.0E+4

0.0E+0

0.5

1.5

2.5

3

4

5

6

Time,min

Time, min

Fig.4. LC–MS/MS chromatograms of O-BHA derivatized nonspiked tumor extract (lower) and isotope-labeled or nonlabeled TCA cycle intermediates (upper) at the lowest

quantiﬁcation level spiked into the tumor extracts: (A) 2-hydroxyglutarate (5 pg); (B)

a-ketoglutarate-d₆(0.3 pg); (C) citrate-d₄(5 pg); (D) isocitrate (10 pg); (E)

fumarate-¹³C₄(5 pg); (F) pyruvate-¹³C₃(3 pg); (G) lactate-¹³C₃(30 pg); (H) succinate-d₄(3 pg); (I) oxaloacetate (30 pg). Numbers in parentheses are the amounts injected into

the HPLC. Labeled malate was not able to be purchased.

trate, and oxaloacetate in various concentrations, were spiked into

cells and tumor extracts. LC–MS/MS spectra of cells and tumor

extracts with or without spiked standards at the lowest quantitation

levels are shown in Figs. 3 and 4. Those LOQs are 0.5- to 1-fold of the

LOQs in the water except that they are within a factor of 3 for 2-

hydroxyglutarate and oxaloacetate. This indicated that the LOQs

determined in the water can be used to estimate the sensitivity of

the derivatization/reverse phase method. Without another set of

stable labeled standards to use as ISs to compensate for adsorption,

degradation, and extraction efﬁciency, the accuracy and precision

for the interday experiment were more than 20%. This reinforced

the importance of ISs for optimal assay performance.

oxaloacetate. With the help of those standards, it is possible to val-

idate the use of authentic standards in water as calibration curves.

Parallelism needs to be demonstrated between calibration curves

in water, cancer cells, and tumors [37]. The method also needs to

be validated for the interday and intraday studies. All of these

experiments were conducted by means of the standard addition

in cancer cell and tumor matrices. Two different matrices (A2780

cancer cell and A2780 tumor extracts), shown in Tables 2 and 3,

were selected for the studies. Standard concentrations added ran-

ged from endogenous levels to 30 times endogenous levels. Colli-

sion energy for some analytes such as lactate was adjusted away

from the optimized value to avoid saturation (Table 1).

The calibration curves were plotted for three different matrices

(water, cancer cell extract, and tumor extract) shown in Fig. 5 (for

equations, see also Supplemental Table 2 in Supplementary mate-

rial). 2-Hydroglutarate-d₄was chosen as the IS for malate and oxa-

loacetate because it can compensate the adsorption and extraction

Assay validation

For the validation of the assays, the isotope-labeled standards

were purchased or synthesized for all analytes except malate and

142

Derivatization of TCA intermediates for LC–MS/MS / B. Tan et al. / Anal. Biochem. 465 (2014) 134–147

Table 2

Validation results for the quantiﬁcation of TCA cycle intermediates in A2780 cell.

Analyte name

Amount of standards

spiked (ng/50 K cells)

Intrabatch

accuracy (n = 3)

(%)

Intrabatch

precision (n = 3)

(%RSD)

Interbatch

accuracy (n = 3)

(%)

Interbatch

precision (n = 3)

(%RSD)

Endogenous levels in

cells (ng/50 K cells)

Citrate

10.0

30.0

50.0

100.0

85.1

95.2

101.0

87.5

12.6

5.4

3.9

86.7

93.9

92.4

87.6

5.4

16.3

12.0

11.6

7.0

3.4

Isocitrate

Fumarate

3.0

5.0

97.5

103.7

97.6

86.3

95.3

94.2

86.2

10.5

11.2

2.3

14.7

4.5

97.8

88.1

87.0

89.2

89.7

87.3

4.9

10.2

9.7

3.5

9.6

BLD^*

10.0

30.0

50.0

100.0

300.0

7.7

13.8

9.4

14.9

5.0

100.1

101.4

107.5

93.6

103.1

6.5

5.6

4.5

7.1

3.9

89.0

90.5

87.9

91.4

85.2

85.7

7.1

4.2

11.1

7.8

11.3

5.0

8.2

10.0

30.0

50.0

100.0

300.0

Malate

30.0

50.0

100.0

300.0

105.2

93.1

95.5

86.5

7.1

4.2

6.1

3.7

110.2

97.4

89.2

85.8

13.5

9.4

8.6

39.1

2.2

8.5

Oxaloacetate

1.0

3.0

5.0

10.0

30.0

50.0

100.0

110.4

97.6

93.3

87.1

89.1

9.4

8.4

5.4

6.8

9.5

5.8

4.9

112.2

109.6

102.8

110.5

97.5

5.6

10.4

5.8

9.7

6.8

4.5

7.3

100.1

101.3

95.8

96.4

a-Ketoglutarate

3.0

5.0

10.0

30.0

91.2

97.0

91.5

85.6

3.6

2.9

3.8

1.2

86.0

87.2

87.7

85.3

7.2

9.7

10.1

5.2

3.2

Succinate

Pyruvate

3.0

5.0

10.0

91.1

91.5

88.9

0.9

8.3

3.2

88.5

90.1

87.8

4.3

3.4

5.2

3.3

2.4

1.0

3.0

5.0

10.0

30.0

50.0

100.0

300.0

104.2

91.1

91.3

93.6

97.8

102.8

103.9

89.7

8.7

3.6

0.8

7.5

11.4

13.7

10.0

11.4

109.8

94.7

89.5

91.9

92.5

99.2

92.7

86.9

3.6

14.9

5.9

5.1

8.9

11.3

7.3

5.4

Lactate

100.0

300.0

500.0

1000.0

3000.0

5000.0

99.4

93.1

100.4

102.5

91.4

93.6

11.3

4.8

4.2

8.9

7.9

5.5

100.9

106.4

108.1

96.9

99.9

95.1

9.3

8.4

13.1

7.6

5.7

3.0

131.3

2-Hydroxyglutarate

1.0

3.0

5.0

10.0

30.0

100.0

109.9

104.1

93.5

96.8

92.2

3.6

3.2

5.8

7.2

4.4

2.6

97.2

95.3

97.5

92.1

94.1

90.9

8.1

6.5

8.4

6.4

8.3

11.4

1.1

100.9

*

BLD is deﬁned as below detection limits.

factors for malate and oxaloacetate. Any matrix effect is compen-

sated for by the IS as demonstrated by the overlapping slopes for

the calibration curves in these matrices. The relative IS responses

(Supplemental Table 3), deﬁned as the intensity ratios of IS

between standards in matrices and the ones in water, were within

15% variation from 100%, indicating that the matrix effect is mini-

mal. Some QCs showed recoveries greater than 100%, but this was

most likely due to variation in the method as demonstrated by the

fact that the difference from 100% was smaller than the coefﬁcient

of variation (%CV). With the help of IS, the curves in all matrices

showed good linearity (r²> 0.98). The x-intercepts of those equa-

tions were less than 0, showing the existence of endogenous ana-

lytes in those matrices, and are listed in Tables

2 and 3.

Therefore, calibration curves in water can be used to measure

TCA cycle intermediates in cancer cells and tumors. For the varia-

tion estimation, the accuracy and precision were determined by

analysis of QCs in triplicate in 1 day for intraday assessment and

in 3 days for interday assessment. The mean accuracies for

Derivatization of TCA intermediates for LC–MS/MS / B. Tan et al. / Anal. Biochem. 465 (2014) 134–147

143

Table 3

Validation results for the quantiﬁcation of TCA cycle intermediates in A2780 tumors.

Analyte name

Amount of standards

spiked in tumor (ng/mg)

Intrabatch

accuracy (n = 3)

(%)

Intrabatch

precision (n = 3)

(%RSD)

Interbatch

accuracy (n = 3)

(%)

Interbatch

precision (n = 3)

(%RSD)

Endogenous levels in

tumor (ng/mg)

Citrate

20

60

100

200

96.5

110.1

99.7

1.9

6.6

4.0

7.1

95.0

7.4

6.3

2.5

4.3

57.2

101.5

105.6

101.4

107.8

Isocitrate

6

10

20

60

100

200

600

90.2

97.3

97.1

103.5

90.0

95.6

90.7

13.4

9.3

13.2

2.8

9.4

7.0

108.6

91.4

110.5

101.7

103.0

99.2

13.2

8.1

6.1

8.8

7.7

BLD^*

4.3

13.4

2.6

108.0

Fumarate

Malate

60

95.5

97.5

102.9

99.1

14.1

2.8

8.1

106.2

104.2

104.3

92.1

8.4

7.7

5.7

3.4

44.8

100

200

600

8.2

60

87.3

89.7

88.4

86.7

8.4

6.2

2.5

6.3

98.5

108.2

90.3

87.1

6.8

7.3

7.5

8.1

125.6

BLD^*

100

200

600

Oxaloacetate

2

6

114.3

100.6

110.1

85.2

100.3

93.3

5.4

5.1

5.3

4.2

6.2

6.9

5.9

113.7

99.8

100.1

98.8

99.8

112.6

95.4

6.8

13.5

7.5

10

20

60

100

200

6.8

11.4

10.4

8.5

94.2

a-Ketoglutarate

2

6

10

20

60

92.9

3.8

6.1

3.2

9.2

0.9

92.2

5.5

7.0

2.5

2.1

1.1

4.6

104.0

106.3

104.2

87.0

106.2

107.0

106.1

88.1

Succinate

Pyruvate

2

6

10

20

108.4

107.1

98.4

6.5

0.9

6.0

3.5

105.6

101.4

99.2

7.5

9.2

5.5

2.8

5.9

97.8

100.4

20

60

100

200

600

108.6

111.0

112.2

99.4

8.6

2.3

6.8

3.5

14.2

106.4

111.9

108.6

100.0

104.3

4.1

3.8

4.5

8.9

10.5

23.8

100.3

Lactate

2000

6000

10000

102.5

91.4

93.6

10.4

5.4

7.9

96.9

99.9

95.1

4.0

3.9

2.5

3798.0

BLD^*

2-Hydroxyglutarate

0.6

1

2

104.3

112.1

110.4

105.7

102.9

96.5

11.8

7.8

4.6

1.6

3.6

2.2

2.6

10.7

108.7

110.5

114.2

104.6

103.1

94.3

10.5

6.4

8.7

6.3

7.4

6.8

3.6

3.2

6

10

20

60

200

100.9

102.9

109.9

104.1

*

BLD is deﬁned as below detection limits.

interday and intraday analysis were determined to be 85 to 112%

(Tables 2 and 3). The relative standard deviations (RSDs) deﬁning

precision were shown to be less than 15% (Tables 2 and 3). In the

end, this method is suitable for both cell and tumor extracts.

High levels of lactate in both cancer cells and tumors were

observed, consistent with the Warburg effect. Pyruvate, fumarate,

malate, and citrate were present at moderate amounts in both

(Fig. 3). Oxaloacetate was also detectable in cells because it was

stabilized by derivatization before signiﬁcant degradation had

occurred [44]. The inability to detect oxaloacetate in tumors could

be due to its intrinsic low levels or due to the degradation [44] by

the heat generated during homogenization of tumor samples. Iso-

citrate was undetectable in both cells and tumors.

extracts. Although succinate and

a-ketoglutarate were at low

Comparison of sensitivity of HILIC, ion pairing, and derivatization

methods

amounts, the LC–MS/MS peaks for them are high due to higher

response factors from the derivatives. Isocitrate, 2-hydroxygluta-

rate, and oxaloacetate were either below detection limits or around

detection limits. The sensitivity of the new method allows us to

detect 2-hydroxyglutarate even in cells without the IDH1 mutation

The derivatization/reverse phase method has an 8-min cycle

time, which is faster than most of the gas chromatography and

ion exchange/exclusion chromatography methods. It is also faster

144

Derivatization of TCA intermediates for LC–MS/MS / B. Tan et al. / Anal. Biochem. 465 (2014) 134–147

2-Hydroxyglutarate

a-Ketoglutarate

Citrate

6

5

4

3

2

1

0

6

5

4

3

2

1

0

2.5

2.0

1.5

1.0

0.5

0.0

A

B

C

0.0

2.5

5.0

7.5

10.0

12.5

0

1

2

3

4

0.0

2.5

5.0

7.5

10.0

12.5

Concentration, μg/ml

Fumarate

Pyruvate

Lactate

5

4

3

2

1

0

3.5

3.0

2.5

2.0

1.5

1.0

0.5

0.0

3

2

1

0

D

E

F

0

10

20

30

40

0

10

20

30

40

0

100

200

300

400

500

600

Concentration, μg/ml

Oxaloacetate

Succinate

10.0

7.5

5.0

2.5

0.0

12

G

H

Water

Cell matrix

Tumor matrix

10

8

6

4

2

0

0.0

2.5

5.0

7.5

10.0

12.5

0

1

2

3

4

Concentration, μg/ml

Malate

Isocitrate

4

2.0

1.5

1.0

0.5

0.0

I

J

3

2

1

0

0.0

2.5

5.0

7.5

10.0

12.5

0

10

20

30

40

Concentration, μg/ml

Fig.5. Standard addition plots of different TCA cycle metabolites in simple aqueous solution, in cell matrix, and in tumor matrix using O-BHA derivatization method/reverse

phase method: (A) 2-hydroxyglutarate; (B) -ketoglutarate; (C) citrate; (D) fumarate; (E) pyruvate; (F) lactate; (G) succinate; (H) oxaloacetate; (I) malate; (J) isocitrate.

Squares indicate aqueous solutions. Upward triangles indicate cell matrix. Downward triangles indicate tumor matrix.

a

than most of the HILIC and ion pairing methods in the literature.

However, those methods were designed for proﬁling. For a fair

comparison among HILIC, ion pairing, and our method, the ﬁrst

two methods were optimized to have shorter run times and better

sensitivities. The run times among HILIC, ion pairing, and the deriv-

atization method are comparable. Then the sensitivities among

those methods in water were compared. The on-column LOQ of

the derivatization method was 10- to 1000-fold lower than those

of the other two methods (Table 1 and Supplemental Table 1).

For the comparison of the performance in complex matrices

(Fig. 6), the same cell number and amount of tumor extract were

used for all three methods. The cell number is the maximum to

ﬁt in a single well of a 96-well plate. The ﬁnal sample volumes

before LC–MS/MS analysis were 100 ll for the HILIC and ion pair-

ing methods and 1 ml for the derivatization method to avoid MS

signal saturation. So samples from the derivatization method are

10-fold diluted compared with the ones from the other two meth-

ods. Most analytes had peak intensities increased more than 10-

fold, pyruvate increased 100-fold, and

a-ketoglutarate increased

10,000-fold. The ion paring and HILIC methods were not able to

quantify six or seven of the TCA cycle intermediates (Fig. 3) when

such a low amount of cells and tumor samples were used. On the

contrary, the derivatization method was able to detect and quan-

tify all analytes except isocitrate in cells and 2-hydroxyglutarate,

oxaloacetate, and isocitrate in tumor extracts. These results

indicated that even when samples were diluted 10-fold after

Derivatization of TCA intermediates for LC–MS/MS / B. Tan et al. / Anal. Biochem. 465 (2014) 134–147

145

2.88

3.1

5.0E+6

A

B

1.2E+6

1.0E+6

8.0E+5

6.0E+5

4.0E+5

2.0E+5

0.0E+0

2.53

2.88

X10 Succinate

3.0

2.90

4.0E+6

3.0E+6

2.0E+6

1.0E+6

0.0E+0

X10 Succinate

X10 Lactate

2.95

3.1

2.92

2.95

Lactate

2.92

X1000 Pyruvate

Fumarate

X500 Pyruvate

Fumarate

2.95

X1000 α-Ketoglutarate

Citrate

Malate

X10 Citrate

Malate

5

2

3

4

5

2

3

4

Time,min

Time, min

2.95

2.53

2.95

2.53

1.2E+6

C

D

6.0E+6

4.0E+6

2.0E+6

0.0E+0

X10 Succinate

X10 Lactate

X100 Succinate

X10 Lactate

X1000 Pyruvate

3.1

2.71

2.88

2.71

8.0E+5

4.0E+5

0.0E+0

2.88

X100 Pyruvate

2.92

X1000 α-Ketoglutarate

Fumarate

3.0

Fumarate

Malate

X10 Citrate

Malate

Citrate

1

2

3

4

5

1

3

Time, min

5

Time, min

1.0E+7

8.0E+6

6.0E+6

4.0E+6

2.0E+6

0.0E+0

E

F

8.0E+6

6.0E+6

4.0E+6

2.0E+6

0.0E+0

3.3

X50 2-Hydroxyglutarate

3.9 4.8

1.2

3.18

1.2

X10 Pyruvate

X10 Lactate

3.2

Lactate

4.9

3.4

5.0

X100 Oxaloacetate

Succinate

Pyruvate

Fumarate

4.9

X10 Fumarate

3.9 4.8

X100 Oxaloacetate

5.0

Succinate

4.3

4.4

α-Ketoglutarate

4.44

4

α-Ketoglutarate

Citrate

X1000 Isocitrate

X10 Citrate

Malate

1

2

3

4

5

6

1

2

3

5

6

Time,min

Fig.6. LC–MS/MS chromatograms of A2780 cell extract (A) and tumor extract (B) using the ion pairing method, of A2780 cell extract (C) and tumor extract (D) using the HILIC

method, and of A2780 cell extract (E) and tumor extract (F) using the O-BHA derivatization method/reverse phase method.

derivatization, this method is signiﬁcantly more sensitive than the

other two methods.

S/N ratio is good enough to provide a reliable estimate of the

metabolite levels when they are lowered by FK866, Fig. 7 shows

the dose–response curves and chromatograms of various analytes

in cells treated with FK866. We conﬁrmed that the 24-h treatment

Quantiﬁcation of organic acids in biological matrices

with FK866 depleted a-ketoglutarate levels and increased oxaloac-

etate levels (Fig. 7) in HCT116 cells. Derivatization enabled us to

quantify not only stable organic acids but also unstable analytes

(oxaloacetate) and low-level analytes (a-ketoglutarate). To further

conﬁrm this, we also tested FK866-treated A2780 tumors from the

mouse xenograft models in which the metabolites were extracted

and derivatized. We showed that lactate and a-ketoglutarate were

As discussed in the previous section, we expect that this

method can determine the concentration of the TCA cycle interme-

diates at concentrations well below endogenous levels. We tested

this using a NAMPT inhibitor (FK866) that reduces the glucolytic

ﬂux and causes a signiﬁcant reduction in the concentrations of

TCA cycle intermediates in cells [37]. To demonstrate that the

146

Derivatization of TCA intermediates for LC–MS/MS / B. Tan et al. / Anal. Biochem. 465 (2014) 134–147

A

B

30

15

10

5

20

10

0

0.01

0.1

1

10

0.01

0.1

1

10

FK866, nM

C

D

E

6.0×10⁶

1.0E+6

2000

4

2

0

Control

Treated

3.0×10⁶

5.0E+5

1000

0

0.0

10.0

0.0

10.0

4.0

4.5

5.0

Time, min

5.5

FK866, mg/kg

Fig.7. Levels of TCA cycle intermediates in cells and tumors detected using O-BHA derivatization method/reverse phase method. (A, B) Levels of

a-ketoglutarate (A) and

oxaloacetate (B) in HCT116 cells treated with FK866. (C) HPLC–MS/MS chromatogram of a-ketoglutarate in HCT116 cells. (D, E) Levels of a-ketoglutarate (D) and lactate (E) in

A2780 tumors with or without FK866 (10 mg/kg). ^⁄P < 0.05.

Heiden, S.M. Su, Cancer-associated IDH1 mutations produce 2-

hydroxyglutarate, Nature 462 (2009) 739–744.

[2] K. Burgess, D. Creek, P. Dewsbury, K. Cook, M.P. Barrett, Semi-targeted analysis

of metabolites using capillary-ﬂow ion chromatography coupled to high-

resolution mass spectrometry, Rapid Commun. Mass Spectrom. 25 (2011)

3447–3452.

signiﬁcantly decreased (Fig. 7), consistent with the ﬁnding [37]

that inhibition of NAMPT signiﬁcantly disrupted the TCA cycle.

Conclusion

[3] P. Kiefer, N. Delmotte, J.A. Vorholt, Nanoscale ion-pair reversed-phase HPLC–

MS for sensitive metabolome analysis, Anal. Chem. 83 (2011) 850–855.