1. Introduction
Cancer is the top two cause of death among people under the age of 70 in
most countries, which has long been a serious threat to human life and
health [1]. Cancer is caused by spontaneous or environmental factors
induced genetic mutations which involves abnormal cell growth with the
potential to invade or spread to other parts of the body [2]. As the
direct and indirect consequences of carcinogenic mutations,
tumorigenesis largely depends on the reprogramming of cell metabolism
with metabolic changes into six hallmarks such as dysregulated nutrient
uptake kinetics, enhanced opportunistic modes of nutrient acquisition,
increased used of central intermediates for biosynthesis and
nicotinamide adenine dinucleotide phosphate (NADPH) production,
increased demand for nitrogen, altered metabolite-driven gene
regulation, and metabolic interactions with the microenvironment
[3]. For example, Cairns et al found that the glucose uptake rate of
tumor cells could reach about 10 times that of normal cells [4], and
Patra et al. found that the expression of hexokinase 2 (HK2) was
significantly up-regulated and HK2 ablation inhibited the neoplastic
phenotype of human lung and breast cancer cells in vitro and in vivo
[5]. Spinelli et al. found that breast cancer cells can recycle
waste by-product ”ammonia” as nitrogen source to promote tumor growth
[6]. As a result, these specific hallmarks exhibited by an
individual tumor can confer a druggable metabolic vulnerability
sufficient to open a therapeutic window. For example, the expression of
PUMILIO 1/2 (PUM1/2) was increased in colorectal cancer (CRC). Gong et
al. found that knockdown of Pum1 and/or Pum2 in human CRC cells resulted
in a significant reduction in tumorigenicity and intravenous injection
of nanoparticle-encapsulated anti-Pum1 and Pum2 siRNA reduced the growth
of CRC cells. These findings revealed the potential of PUM proteins as
therapeutic targets for CRC [7]. Non-small-cell lung cancer (NSCLC)
patients with activating epidermal growth factor receptor (EGFR)
mutations developed resistance to tyrosine kinase inhibitor (TKI). Wang
et al. found that EGFR C797S mutation was closely related to AXL through
transcriptome and proteome analysis, and inhibition of AXL could
effectively slow down the growth of NSCLC cells harboring EGFR C797S.
This finding suggested that AXL inhibition might be a potential adjuvant
therapy for NSCLC harboring the EGFR C797S mutation [8].
Metabolome defines a set of metabolites in a biological sample, which
provides snapshots of tumor cells in response to genetic and/or
therapeutic perturbations. Metabolic footprinting has been used to
identify tumor biomarkers for cancer diagnosis and prognosis, and such
biomarkers can be assayed in non-invasively collected biofluids like
blood or serum [9-11]. Recently, the combination of cancer models
with metabolomic analysis has been shown to be a powerful approach to
generate new hypotheses about predicting drug susceptibility, resistance
and mode of action [12]. For example, JAIN et al. found that
interfering with glycine metabolism slowed the growth of rapidly
proliferating cancer cells and suggested that the increased dependence
of rapidly growing cancer cells on glycine might provide a potential
target for therapeutic intervention [13]. Newman et al. found that
cancer cells could utilize serine as a major source of carbon units and
it allowed cancer cells to maintain high proliferation rates [14].
Tao et al. observed 270 dysregulated lipids in serum exosomes between
pancreatic cancer (PC) patients and healthy controls through
non-targeted lipidomic analysis, which might become diagnostic
biomarkers [15]. Through metabolomic analysis, Bermudez et al.
assessed proliferation of a collection of cancer cells following
inhibition of the mitochondrial electron transport chain (ETC) and found
that cell lines least sensitive to ETC inhibition maintained
intracellular aspartate levels by importing it. This suggested that
aspartate was the limiting metabolite for tumor growth and may be a
target for cancer therapy [16].
In order to exploit pivotal intracellular mechanisms associated with
tumor growth and drug resistance requires an insightful investigation of
intracellular metabolites. To achieve this, sample preparation and
analytical protocols should be developed and evaluated in terms of
reproducibility and number of metabolite features obtained [17].
Nevertheless, acquiring metabolome changes is experimentally demanding,
such that dynamic data in the development, progression and therapeutic
treatment of cancer are extremely scarce. Meanwhile, achieving the
absolute quantification of global metabolite is further hampered by
sample pretreatment with loss and biased mass spectrometry-based
analyses. To obtain true snapshots of intracellular metabolites,
leakage-free sample pretreatment and unbiased metabolite determination
are required. For cell acquisition, suspension cells could be directly
centrifuged to obtain cell pellets while the adherent cells must be
detached before metabolite extraction. The adherent growth cells can be
harvested by scraping or using trypsin. Traditionally, trypsin was often
used to harvest cells. However, previous studies have concluded that
harvesting cells by trypsinization lead to the loss of a large amount of
metabolites [18, 19]. Therefore, direct quenching of cells with
physical scraping was used as an alternative. However, this was not
conducive to cell counting. To address this, measuring the protein
content of cell pellets to correct the number of cells has been reported
as a promising method for standardizing cell culture systems [20].
Metabolite quenching is the first critical step in the sample
preparation process. The quenching process should meet several
requirements: the enzyme activity must be quenched as quickly as
possible to avoid metabolite degradation and changes in sample
composition. Some metabolites are extremely unstable with a short
turnaround time, such as adenosine triphosphate (ATP) and glucose
6-phosphate (G6P) [21], so rapid quenching procedure is essential to
ensure the true state of intracellular metabolites [22]. Quenching
can usually be achieved by freezing samples in liquid nitrogen [23],
or in cold methanol solution [24]. It has been reported that cold
methanol quenching caused serious metabolite leakage in cell, resulting
in an overall loss of metabolite recovery [25]. However, in cancer
cells, a systematic evaluation of the extent of metabolite leakage among
different quenchers and the influential factors giving rise to it has
rarely been reported.
Besides leakage-free quenching method, a complete extraction of
intracellular metabolite is required. In previous studies, a lot of
efforts have been devoted to the development of extraction methods for
intracellular metabolites in different mammalian cells. For instance,
Sellick et al. used 100% methanol extraction twice and then water once
to extract the metabolites, which could recover the largest range of
intracellular metabolites from Chinese hamster ovary (CHO) cells
[26]. However, at the same time, other extraction method should be
required to maximize the non-polar metabolites recovered. Dettmer et al.
used methanol/water to scrap cells harvest and extract metabolites from
adherently growing SW480 cells [18]. Wamelink et al. used 50%
acetonitrile to extract intermediates in pentose phosphate pathway in
fibroblasts and lymphoblasts [27]. However, these extraction methods
have not yet been evaluated together with quenching procedures to
develop a generalized protocol for acquiring intracellular metabolome in
mammalian cells. Furthermore, obtaining accurate concentrations of
intracellular metabolites were of great significance for the validation
of metabolic fluxes and the establishment of metabolic models [22, 28,
29]. Mass spectrometry such as Gas Chromatography-Mass Spectrometry
(GC-MS) and Liquid Chromatography tandem Mass Spectrometry (LC-MS/MS)
has been widely used in the determination of metabolites due to its high
accuracy and high resolution [30, 31]. However, sample recovery
needs to be determined experimentally for each analyte. To circumvent
this, the stable isotope internal standard of the analyte was the best
choice, because the substance could be distinguished by mass
spectrometry, and IDMS could be used to track and correct the loss of
the sample during pretreatment and analysis [32]. The stable isotope
internal standards could be cultivated on a medium with uniformly
labeled 13C-glucose as the sole carbon source, and the13C-labeled internal standard of intermediate
metabolites and target metabolites could be obtained through the
self-metabolism of microbial cells [33]. By combining GC-MS/LC-MS/MS
with the IDMS method, it is possible to track losses during sample
pretreatment and analytical process [32, 33] and greatly increase
the accuracy of metabolite quantitation. Furthermore, previously
developed sample preparation methods were merely developed for 2D
adherent cells, which cannot be directly applied to 3D tumor model
without further evaluation. As is known, the 2D cells cannot simulate
the complex structure and heterogeneity of tumor tissue in vivo [34,
35]. 3D multicellular tumor spheroids (3D MTSs) are a tumor model
between 2D monolayer cell culture and animal models [36]. Compared
with traditional 2D model, the internal concentration gradient of
nutrients and oxygen led to outer proliferation of living cells, inner
accumulation of necrotic cells and living quiescent cells in the
transition zone [37]. This inhomogeneity is an important feature of
solid tumors in vivo. Many anticancer drugs exert cytotoxic effects on
proliferating cells, while quiescent cells escape treatment [38,
39]. Therefore, model systems such as 3D MTS involving quiescent cells
are critical for preclinical drug screening and drug resistance
investigations.
In this study, we systematically evaluated 12 sample preparation
protocols for accurate determination of intracellular metabolites in
both 2D and 3D tumor cell models. Further, based on the optimal sample
preparation protocol, quantitative metabolomics analysis was used to
discover the differential metabolome features of both 2D and 3D tumor
models in response to DOX treatment.