Proliferative Glioblastoma Cancer Cells Exhibit Persisting Temporal
Control of Metabolism and Display Differential Temporal Drug
Susceptibility in Chemotherapy
Paula M. Wagner
1,2
& Lucas G. Sosa Alderete
1,2,3
& Lucas D. Gorné
4,5
& Virginia Gaveglio
6
& Gabriela Salvador
6
&
Susana Pasquaré
6
& Mario E. Guido
1,2
Received: 1 December 2017 /Accepted: 24 May 2018
#
Springer Science+Business Media, LLC, part of Springer Nature 2018
Abstract
Even in immortalized cell lines, circadian clocks regulate physiological processes in a time-dependent manner, driving transcrip-
tional and metabolic rhythms, the latter being able to persist without transcription. Circadian rhythm disruptions in modern life
(shiftwork, jetlag, etc.) may lead to higher cancer risk. Here, we investigated whether the human glioblastoma T98G cells
maintained quiescent or under proliferation keep a functional clock and whether cells display differential time responses to
bortezomib chemotherapy. In arrested cultures, mRNAs for clock (Per1, Rev-erbα) and glycerophospholipid (GPL)-synthesizing
enzyme genes,
32
P-GPL labeling, and enzyme activities exhibited circadian rhythmicity; oscillations were also found in the redox
state/peroxiredoxin oxidation. In proliferating cells, rhythms of gene expression were lost or their periodicity shortened whereas
the redox and GPL metabolisms continued to fluctuate with a similar periodicity as under arrest. Cell viability significantly
changed over time after bortezomib treatment; however, this rhythmicity and the redox cycles were altered after Bmal1 knock-
down, indicating cross-talk between the transcriptional and the metabolic oscillators. An intrinsic metabolic clock continues to
function in proliferating cells, controlling diverse metabolisms and highlighting differential states of tumor suitability for more
efficient, time-dependent chemotherapy when the redox state is high and GPL metabolism low.
Keywords Circadian rhythm
.
Tumor cell
.
Glioblastoma
.
Clock gene
.
Glycerophospholipid metabolism
.
Redox state
Introduction
In mammals, the circadian timing system generates periodic
oscillations in many physiological processes and behaviors,
allowing the organism to anticipate and adapt to daily envi-
ronmental changes and thus be in synchrony with the solar
cycle. Circadian clocks are present in most tissues examined
and even in immortalized cell lines and primary cell cultures
Paula M. Wagner and Lucas G. Sosa Alderete contributed equally to this
work.
Electronic supplementary material The online version of this article
(https://doi.org/10.1007/s12035-018-1152-3) contains supplementary
material, which is available to authorized users.
* Mario E. Guido
mguido@fcq.unc.edu.ar
1
CIQUIBIC-CONICET, Departamento de Química Biológica,
Facultad de Ciencias Químicas, Universidad Nacional de Córdoba,
Haya de la Torre s/n, Ciudad Universitaria, 5000 Córdoba, Argentina
2
Departamento de Química Biológica “Ranwel Caputto”, Facultad de
Ciencias Químicas, Universidad Nacional de Córdoba,
5000 Córdoba, Argentina
3
Present address: Department of Molecular Biology, UNRC,
CONICET, Río Cuarto, Argentina
4
Facultad de Ciencias Exactas Físicas y Naturales, Universidad
Nacional de Córdoba, Córdoba, Argentina
5
Consejo Nacional de Investigaciones Científicas y Técnicas,
CONICET, Instituto Multidisciplinario de Biología Vegetal (IMBiV),
Córdoba, Argentina
6
Departamento de Biología, Bioquímica y Farmacia, Instituto de
Investigaciones Bioquímicas de Bahía Blanca (INIBIBB,
UNS-CONICET), Universidad Nacional del Sur, 8000 Bahía
Blanca, Argentina
Molecular Neurobiology
https://doi.org/10.1007/s12035-018-1152-3
displaying self-sustained rhythms in gene expression and met-
abolic activities [1–4]. At the molecular level, the clock is
controlled by a transcriptional/translational feedback circuitry
involving a set of so-called clock genes such as Clock, Bmal1,
Periods (Per),andCryptochromes (Cry), among others,
encoding for the corresponding clock proteins acting as regu-
latory transcription factors (activators or repressors) able to
generate rhythms of clock gene and clock-controll ed gene
expression under a circadian base [5]. Through their rhythmic
transcription, circadian clock genes regulate the metabolism
and respiration of cells, therefore becoming very important
regulators of the cell division cycle [6]. Loss of clock gene
function or misalignment of circadian rhythms may therefore
lead to diverse metabolic disorders (obesity, diabetes, hyper-
lipidemia, etc.) [7] and higher risk of cancer by inducing ma-
lignant cell growth and tumor development [6].
Carcinogenesis is a process resulting in the accumulation
of genetic alterations primarily in genes involved in the regu-
lation of signaling pathways relevant to the regulation of cell
growth and division (reviewed in [8]). Characteristics typical
of neoplastic processes include sustained proliferative activa-
tion, growth suppressor evasion, cell death resistance, replica-
tive immortality, induction of angiogenesis, and invasiveness
and metastasis, all of which are based on genome instability
and inflammation. In this context, knowledge at the molecular
level of circadian biology can contribute significantly to the
understanding and potential treatment of human pathologies,
in particular cancer and metabolic and behavioral disorders [6,
9]. Moreov er, recent transcriptomic, proteomic, and
metabolomic studies in mammalian tissues have shown a ro-
bust crosslink between the circadian clock and cellular metab-
olism in general and glycerophospholipid (GPL) homeostasis
and metabolism in particular [10–13].
In addition, a redox/metabolic clock that drives
peroxiredoxin (PRX) oxidation cycles has recently been re-
ported to operate even in the absence of transcription and
shown to be highly conserved through evolution and present
in all different kingdoms of life [14, 15]. On the basis of this
new evidence, it can be inferred that there is an intrinsic clock
at the cellular level, comprising the transcriptional clock and
metabolic oscillators that temporally control a plethora of cel-
lular processes. GPLs constitute a fundamental group of lipids
with important roles as structural components of all biological
membranes and key cellular components involved in cell sig-
naling, energy balance, vesicular transport, cell division, apo-
ptosis, and cell-to-cell communication [16, 17].
Phosphatidylcholine (PC) and phosphatidylethanolamine
(PE) are the most abundant GPLs present in all eukaryotic
cells; they are first synthesized from glycerol-3-phosphate
via a de novo pathway described by Kennedy and Weiss
[17, 18] and subsequently remodeled by the Lands Cycle,
involving the sequential activity of phospholipases A (PLA)
and lysophospholipid acyltransferases (LPLAT) [19, 20]. For
PC biosynthesis, the Kennedy pathway involves three enzy-
matic steps catalyzed by choline kinase (ChoK), CTP/
phosphocholine cytidylyltranferase (CCT), and CDP-cho-
line/1,2-diacylglycerol choline phosphotransferase (CPT) in
which CCT activity is considered the rate-limiting and regu-
latory step under most metabolic conditions [21].
Nevertheless, it has been demonstrated that the availability
of diacylglycerol (DAG) and regulation of ChoK also influ-
ence PC biosynthesis [22–24]. In most mammals, there are
two genes encoding for ChoK: Chka codes for ChoKα
1/2
an
d Chkb codes for ChoKβ [25, 26]. Mice lacking ChoKα
die early in embryogenesis [25, 26], whereas ChoK overex-
pression has been implicated in human carcinogenic processes
[25, 26]. In addition, ChoKα expression and/or activity are
subject to circadian control in both cultured fibroblasts and
mice liver after synchronization [3, 4, 27]. I n mammalian
cells, PE is mainly de novo synthesized from the CDP-
ethanolamine Kennedy pathway in which ethanolamine
(Etn) is phosphorylated to Etn-P by the Etn kinase (EK) and
then converted to CDP-Etn by CTP/phosphoethanolamine
cytidylyltranferase (Pcyt2). In the last step, CDP-Etn is trans-
ferred to DAG to produce PE by the CDP-Etn/1,2-
diacylglycerol ethanolamine phosphotransferase (EPT).
Pcyt2 is the main regulatory enzyme in the de novo biosyn-
thesis of PE (for a review, see [28]). In mammals, the disrup-
tion of genes encoding GPL biosynthetic enzymes has severe
physiological consequences during development and can be
lethal [25, 26].We have previously reported that de novo syn-
thesis of whole GPLs in different cell types from mammalian
and non-mammalian vertebrates is controlled by a circadian
clock, as observed in chicken retinal neurons in vivo or in
vitro [29–31], in the liver of mice synchronized to light-dark
cycles [4], as well as in quiescent murine fibroblasts after
synchronization by a serum shock [3, 32]. Furthermore, the
biosynthesis of membrane GPLs including PC and PE is of
crucial significance for cell growth and progression through
the cell cycle, whereas deregulated proliferation is a hallmark
of cancer cells [8]. Although circadian clocks influence the
cell division cycle through complex regulatory circuits, dereg-
ulation of core clock gene expression and loss of circadian
homeostasis may promote cancer development [6, 33]. At
present, little is known about the temporal regulation of GPL
biosynthesis in immortalized tumor cells. We therefore inves-
tigated whether human glioblastoma T98G cells subject to
proliferation in the presence of serum or maintained quiescent
(arrested) in a serum-free medium retain a functional clock
capable of regulating gene expression at the molecular level
and redox and GPL metabolism under a circadian base. To this
end, we first examined the expression of clock core genes
(Per1 and Rev-erbα) and GPL-synthesizing enzyme genes
(ChoKα and Pcyt-2) for PC and PE biosynthetic pathways,
respectively, in proliferating and quiescent cells at different
times after dexamethasone (DEX, 100 nM) synchronization.
Mol Neurobiol
We also assessed whether redox metabolism (redox state and
peroxiredoxin oxidatio n cycles), the metabolic labeling of
32
P-GPLs, and the activities of GPL-synthesizing enzyme
phosphatidate phosphohydrolase ( PAP) and LPLATs were
temporally regulated in synchronized T98G cells after DEX
synchronization under both proliferative conditions. We fur-
ther assessed the susceptibility of proliferating cells over time
to treatment with the proteasome inhibitor bortezomib, a
known chemotherapeutic agent. Lastly, we investigated the
role of the transcriptional clock in redox cycle rhythms and
chemotherapeutic susceptibility in proliferating cells in which
Bmal1 expression was knocked-down by CRISPR/cas9
technology.
Results
Characterization of T98G Cell Culture Conditions
T98G glioblastoma cells display typical cancer cell character-
istics and are thus subject to continuous proliferation, as a
result of which no inhibition by contact is observed when cells
are maintained in culture. One round of division along the cell
cycle takes close to 24–28 h.
To investigate the temporal regulation of GPL synthesis,
redox metabolism and expression of clock and clock-
controlled genes in T98G cells, we first characterized the ex-
perimental culture conditions in relation to proliferation
levels. To this end, cel ls grown to 50% of confl ue nce in
10% fetal bovine serum (FBS)-Dulbecco’s modified Eagle’s
medium (DMEM) were synchronized with a 20-min 100 nM
DEX shock and then maintained in the presence of a 5%
serum medium (proliferative) or in an FBS-free medium
(arrested) to achieve quiescence for 24 h. Irrespective of the
cell cycle itself, a synchronization protocol is essential to ad-
just individual cells within the whole culture population to the
same phase. We examined whether cells undergo a significant
rate of cell division after DEX synchronization under both
FBS conditions. We found by flow cytometry that within the
first 24 h, most cells in this serum-free state were arrested at
the G
0
/G
1
phases (~ 71–80%) while < 12–17% of cells
reached the G
2
/M and S phases (Supplementary Fig. 1a).
The distribution of cell populations throughout the cell cycle
remained constant at all times examined. These two major
features—the basal proliferative condition and synchroniza-
tion by an external factor—make T98G cell cultures an inter-
esting oscillator model for circadian studies, in a cancer con-
text, regardless of cell division and systemic influences from
the brain’s master clock. At times after at least 8 h of DEX
treatment, ~ 60% of cells maintained in the presence of FBS
after synchronization were arrested at the G
0
/G
1
phases and ~
30% at the G
2
/M and S phases (Supplementary Fig. 1a). The
distribution of cell populations throughout the cell cycle in the
presence of FBS medium remained constant at most times
examined up to 48 h, with an average of ~ 35% of cells in
the G
2
/M and S phases. On the contrary, at time 0 with no
DEX treatment and during the first hour after treatment, cells
exhibited the highest percentage at phase S + G
2
/M (50–65%)
and the lowest for G
o
/G
1
(30–40%) as compared with DEX-
treated cells at longer times (Supplementary Fig. 1a, b). Based
on these observations, it can be inferred that once synchro-
nized, cells from the two conditions (proliferative and
arrested) represent different proliferative situations, making
them suitable for circadian studies. In addition, when the mo-
lecular clock was perturbed in proliferating T98G cells after
transfection with PX459-Bmal1 plasmid in order to knock-
down Bmal1 expression (Supplementary Fig. 1a, c, d), there
was a higher percentage of arrested cells (77%) in the G
0
/G
1
phases than in the controls (61%).
Temporal Regulation of Clock Gene Expression
in T98G Cells Under Different Proliferation Conditions
T98G cells synchronized with a b rief DEX sh ock a nd
maintained arrested in an FBS-free medium displayed a
significant temporal variation in mRNA levels for the
clock genes Per1 and Rev-erbα along the 48 h examined,
with markedly different profiles of expression
(Supplementary Fig. 1b) and a periodicity according to a
COSINOR/RAIN analysis ranging from 24 to 28 h
(Table 1). It is noteworthy that Bmal1 was highly
expressed at time 0 and 24 h after synchronization and
lowest levels were seen at 16 h (data not shown), whereas
Per1 and Rev-erbα transcripts exhibited patterns of expres-
sion in total antiphase with Bmal1 mRNA (Supplementary
Fig. 1b). In fact, highest levels of mRNA for Per1 and Rev-
erbα were observed around 16 h after synchronization.
The statistical analysis revealed a significant time effect
for the two transcripts assessed: mRNA le vels f or Per1
and Rev-erbα at 16 h were higher than those at 0–8and
24–30 h.
Although clock gene expression in proliferating T98G
cells after DEX synchronization also showed temporal
fluctuations in the three clock genes investigated
(Supplementary Fig. 1b), the circadian rhythmicity was
lost for Rev-erb α mRNAs and for Per1 the period was
shortened to 16 h according to the periodic analysis
(Table 1). Nevertheless, the temporal variations observed
display different amplitudes and profiles o f expression
with respect to patterns described in arrested cells
(Supplementary Fig. 1b). The amplitude for Per1 tran-
scripts varied 7-fold over time in arrested cells and only
1.5-fold under proliferation. Rev-erbα reached 5- and 4-
fold incr eases in a rrested and proliferative cells, respec-
tively. In addition, synchronized proliferating cells
displayed a significant daily rhythm in levels of PER1-
Mol Neurobiol
like protein by immunocytochemistry (ICC) (p <0.005 by
analysis of variance (ANOVA)) (Fig. 1a–c), with highest
levels at 18 and 24 h after synchronization, differing from
thoseat6and12h.InproliferatingT98Gcellsaftertrans-
fection with PX459-Bmal1 plasmid (T98G E1), levels of
Per1 mRNA and protein were substantially reduced and no
time related differences were seen in PER1-like protein
along the 24 h examined (Fig. 1b, c).
Temporal Regulation of GPL Labeling in T98G Cells
In order to investigate the circadian regulation of cellular
metabolism in T98G cells, we examined GPL labeling
with
3
H-glycerol or
32
P-orthophosphate for 90 min in tu-
mor cells previously synchronized with DEX and main-
tained in an arrested state or under proliferation for 48 h,
at different times after synchronization.
Table 1 Periodic analysis by COSINOR/RAIN correlation of gene expression and metabolic parameters under both growth conditions (arrest and
proliferation)
Growth condition
Partial arrest Proliferative
p value Acrophase (h) Period (h) p value Acrophase (h) Period (h)
1- Molecular level
a. mRNA expression
Per1 0.0012
b
20 28 0.0256
a,c
416
Rev-erb α 0.0461
a
16 24 ns
b
ChoK α 0.0045
b
16–20, 48 32 ns
b
Pcyt-2 0.0329
a, c
12, 28 16 ns
b
2- Metabolic level
a. Total GPLs 0.0159
a
0200.036
a
20 24
b. Individual GPLs cpm pix
PC ns
a
0.03
a
24 32
PE 0.006
b
8320.036
a, c
12 32
PC/PE 0.008
a
4, 32 28 0.025
a, c
28 32
c. Enzymatic activities
Total PAP 0.0308
a
24 30 0.0277
a, c
024
LPAAT ns
a
ns
a
LPCAT ns
b
ns
a
LPEAT ns
a
ns
a
d. Redox state
WT 0.001
a
0120.0071
a, c
012
E1 Non-determined 0.0185
a
618
e. PRX expression SO
3
0.038
b
6240.005
b
30 30
f. Endogenous GPLs content
PC Non-determined ns
a
PE 0.0264
b
624
PC/PE 0.0323
a, c
12 24
g. BOR Non-determined 0.0109
a, c
0–630
Statistical analysis was performed with results from 3/5 independent samples for each time in T98G cell grown under arrest or proliferative conditions. To
test the time effect in the relative RNA expression or metabolic aspects, a periodic adjustment model—COSINOR—was applied for each data set
examined as described in “Materials and Methods.” When the model assumptions were infringed, the RAIN analysis was used. The COSINOR analysis
includes r
2
, acrophase, and period. Acrophase denotes the time at which the variable reaches the maximum observed value; the analysis considered a
period from 12 to 30 h and significance at p < 0.05 (values in bold). Significant periodic effects of time by RAIN or COSINOR periodic model better than
linear model according AIC are in bold. Significant periodic effects of time by COSINOR, but where the whole model does not perform better than null
linear model (according AIC) are set in italics
ns, non-significant
a
COSINOR
b
RAIN
c
Log-transformed variable
Mol Neurobiol
T98G cells totally failed to uptake
3
H-glycerol, irrespective
of the proliferative condition examined or time of labeling as
compared with positive controls (NIH3T3 fibroblasts) [3].
However, when
32
P-phosphate was used as a general precur-
sor for GPL synthesis, we found a significant temporal varia-
tion in the labeling of total GPLs of arrested cells collected at
different times, with highest levels at 0, 16–20, and 32–44 h
after DEX synchronization and lowest levels of labeling at 8,
24, and 28 h with a period of 20 h (p <0.016) (Fig. 2a(a);
Table 1). When cells were maintained in FBS medium after
DEX synchronization, a significant temporal variation was
also observed in GPL labeling with highest levels at 0–4and
16–24 h and lowest levels at 8 and 28–32 h (Fig. 2a (b)).
Under this condition, the oscillation in metabolic GPL label-
ing displayed a circadian period ~ 24 h (p <0.036)(Table1).
When PC and PE labeling and the PC/PE ratio w ere
assessed separately, a si gnificant temporal variation was
observed in arrested cells (Fig. 2b) with higher levels of
Fig. 1 Immunocytochemistry of PER1-like protein for proliferating
T98G WT and E1 cell cultures after synchronization. a, b WT (a)or
transfected cells with PX459-Bmal1 plasmid (T98GE1 cells) (b)keptin
the presence of serum were synchronized with a shock of 100 nM DEX
for 20 min and collected at different times from 0 to 24 h. Cultures were
immunolabeled with specific primary antibody for PER1-like protein
(red) and with DAPI (blue) for nuclear localization and visualized by
confocal microscopy as described in “Materials and Methods.” Scale
bar = 10 μm. c Histograms indicating relative levels of PER1-like
protein in T98G WT cells (black) or T98G E1 cells (gray) at different
times after synchronization. Data are mean ± SE. Results revealed a
significant temporal variation in levels of PER1-like prot ein by
immunocy tochem istry ( p < 0.005 by ANOVA) in WT cells, with
highest levels at 18 and 24 h after synchronization, differing from those
at 6 and 12 h. By contrast, T98GE1 cells under the same conditions
exhibited lower immunoreactivity associated with PER1-like protein
over time and no significant temporal effect at times examined. d RT-
PCR for Bmal1 (a)andPer1 (b) (a downstream target gene for Bmal1)
and the housekeeping gene tbp in T98GWT and T98GE1 cells. Results
showed a significant decrease in both transcripts in T98G E1 glial cells.
See text for further details
Mol Neurobiol
