A Step Toward Optimization of Cancer Therapeutics [Chronobiological Investigations]
ABSTRACT An integrative physiology model has been designed, which takes into account the cell proliferation at the level of a population of cells by age-structured partial differential equations (PDEs), its control by cell cycle proteins, and the control of these molecular mechanisms by the circadian system, designed as a network of coupled oscillators also described by ODEs. Cancer growth and response to therapy by anticancer drugs have been shown to be dependent on circadian clock inputs. This multiscale modeling framework will provide clinicians with a theoretical tool to bridge the gap between the pharmaceutical clinical control level and the molecular pharmacological hidden level of drug action.
- SourceAvailable from: inria.fr[Show abstract] [Hide abstract]
ABSTRACT: The circadian timing system is composed of molecular clocks, which drive 24-h changes in xenobiotic metabolism and detoxification, cell cycle events, DNA repair, apoptosis, and angiogenesis. The cellular circadian clocks are coordinated by endogenous physiological rhythms, so that they tick in synchrony in the host tissues that can be damaged by anticancer agents. As a result, circadian timing can modify 2- to 10-fold the tolerability of anticancer medications in experimental models and in cancer patients. Improved efficacy is also seen when drugs are given near their respective times of best tolerability, due to (a) inherently poor circadian entrainment of tumors and (b) persistent circadian entrainment of healthy tissues. Conversely, host clocks are disrupted whenever anticancer drugs are administered at their most toxic time. On the other hand, circadian disruption accelerates experimental and clinical cancer processes. Gender, circadian physiology, clock genes, and cell cycle critically affect outcome on cancer chronotherapeutics. Mathematical and systems biology approaches currently develop and integrate theoretical, experimental, and technological tools in order to further optimize and personalize the circadian administration of cancer treatments.Annual Review of Pharmacology 01/2010; 50:377-421. · 21.54 Impact Factor
- Mathematical Modelling of Natural Phenomena 01/2009; 4(3):12-67. · 0.56 Impact Factor
Conference Paper: Proliferation in cell population models with age structure[Show abstract] [Hide abstract]
ABSTRACT: We study proliferation in tissues from the point of view of physiologically structured partial differential models, focusing on age synchronisation in the cell division cycle in cell populations and its control at phase transition checkpoints. We show how a recent fluorescence-based technique (FUCCI) performed at the single cell level in proliferating cell populations allows identifying model parameters and how it may be applied to investigate healthy and cancer cell populations. We show how this modelling approach allows designing original optimisation methods for cancer chronotherapeutics, by controlling eigenvalues of differential operators underlying proliferation dynamics, in tumour and in healthy cell populations.Proceedings of ICNAAM 2011, Kallithea Chalkidis (Greece); 01/2011
A Step Toward Optimization
of Cancer Therapeutics
mic clock – and this is most likely related to cell cycle
molecular circadian clock and cell cycle timings? How should
anticancer therapeutics be designed so as to induce efficient
recovering of such control mechanisms? To answer these ques-
tions, an integrative physiology model has been designed,
which takes into account the cell proliferation at the level of a
population of cells by age-structured partial differential equa-
nary differential equations [ODEs]), and the control of these
molecular mechanisms by the circadian system, designed as a
ancer growth and response to therapy byanticancer
clock inputs , . Indeed, tumor growth is
enhanced by perturbations of the central hypothala-
Cell Proliferation: Populations of Cells
In tissues subject to renewal, in particular in tumors, but also in
fast renewing healthy tissues such as gut, skin, and bone mar-
row, the individual cell is the fundamental level of description
for the interacting molecular mechanisms at stake in cell cycle
progression and circadian clock timing systems. But cancer
growth and healthy tissue-controlled proliferation are a matter
of cell population dynamics. This leads us to consider age-
structured cell populations, which are described, for each phase
of the cell division cycle, classically divided in phases G1,
S?G2, and M, by evolution equations for cell densities depend-
ent on time and age spent in each phase. Cells in renewing tis-
sues are either in the quiescent, G0, phase (where no growth
occurs),orintheproliferating phasesG1,S (for DNAsynthesis,
i.e., genome duplication), G2, and M (for mitosis, or actual cell
division). Exchanges of cells between phases G0and G1may
point has not been reached; afterwards, a proliferating cell is
is stopped, e.g., by anticancer drugs, in its progression at cell
occur at the G1=S and G2=M transitions. The transition rates
from one phase to the following one (hereafter noted Ki!iþ1in
the model) are the main targets of the circadian (and pharmaco-
An Age-Structured PDE Model of the Cell Cycle
The model chosen – is linear and of the Von Foerster-
McKendrick type. For each phase i (1 ? i ? I) of thecell divi-
sion cycle, ni¼ ni(t,a) denotes the density of cells of age a in
phase i at time t. Inside each phase, the variation of the density
of cells is either due to a spontaneous death rate di¼ di(t,a) or
to a transition rate Ki!iþ1¼ Ki!iþ1(t,a) from phase i to phase
i þ 1 (withthe conventionI þ 1 ¼ 1):
við0Þniðt,a ¼ 0Þ ¼
2 ? i ? I
n1ðt,a ¼ 0Þ ¼ 2
@a½viðaÞni? þ ½diþ Ki!iþ1?ni¼ 0,
Ki?1!iðt,aÞ ni?1ðt,aÞ da,
KI!1ðt,aÞ nIðt,aÞ da,
The transition rates are chosen as wi(t)1fa?aig(a): a minimum
age aimust be spent in phase i, and a dynamic control t7!wi(t)
isexertedonthetransitionfromphaseitophasei þ 1.Thefunc-
tions wiintegrate physiological control (hormonal or circadian)
as well as pharmacological influences. Another possibility to
take circadian control into account would be by an action on the
death rates di, an apparently natural representation since anti-
cancer drugs increase death rates in cell populations. But it can
be shown that at least by itself such control representation is not
compatible with observations from laboratory experiments on
tumor growth involving circadian clock disruption , . The
mainoutput of thismodelisa (positive) growthexponent,k, the
first eigenvalue of an underlying differential operator, which
determines its asymptotic behavior; its existence is granted by
a?0ni(t,a)da ¼ 1,vi(a) denotes a speed function
of age a in phase i with respect to time t, and Ki!iþ1(t,a) ¼
Tumor Cells: Unlimited Growth
is bound to show exponential growth. The first eigenvalue, or
BY JEAN CLAIRAMBAULT
Digital Object Identifier 10.1109/MEMB.2007.907363
Physiologically Based Modeling of Circadian
Control on Cell Proliferation
IEEE ENGINEERING IN MEDICINE AND BIOLOGY MAGAZINE0739-5175/08/$25.00©2008IEEE JANUARY/FEBRUARY 2008
Malthus exponent k, of the underlying differential operator is
positive and governs the asymptotic behavior (in time t) of the
solutions ni. This means that for all i, the normalized solutions
t7!e?ktni(t) are bounded, asymptotically constant if w is con-
stant, and asymptotically periodic if the wiare periodic, with
the same common period. This Malthus exponent k may be
assessed in the case of solid tumors by in vivo tissue growth
measurements (only approximately, if one neglects the pres-
ence of necrotic material) by k ¼ (ln2)=Td, if Tdis the appa-
Healthy Cells: Tissue Homeostasis
Inthecase ofhealthy cells, therecanbe noexponentialgrowth
and tissue homeostasis must be ensured, in the sense that for
healthy renewing tissues cell loss must be made up for, and
with no excess, by influx from newly formed cells. This can
be modeled by the introduction of a G0(or quiescent) phase
exchanging cells with the G1phase, taking into account in the
control of these exchange mechanisms a limitation by cell
density dependent inhibition. This has been done in an
extended version of this model including cyclin concentration
as a structure variable . In this case, the Malthus exponent
ulation of cells is evolution to stationarity or periodicity, and
there is no exponential growth, except when the total cell pop-
ulation is small, but on the contrary convergence toward a
steady state when thetotal cellpopulation grows.
An interest in simultaneously representing a tumor and a
healthy cell population is, in the perspective of therapeutic
optimization, to control unwanted toxicity on healthy tissues,
a constant side effect in anticancer therapies. These unwanted
effects on the healthy cell division cycle are also dependent on
the control of cell cycle phase transitions by the circadian
clock, but possibly with phase differences, by comparison
with tumors: a phase delay of 12 h between circadian peaks
of anti-tumor therapeutic efficacy and circadian peaks of
unwanted toxicity to healthy tissues is usually observed, what-
ever the drug . The therapeutic control objective may be
seen as to obtain a negative Malthus exponent for tumor cells
and zero growth forhealthy cells.
Cell Proliferation Control: Cyclins
and Cyclin-Dependent Kinases
Control of Cell Cycle Phase Transitions
by Cyclins and Cyclin-Dependent Kinases
Cyclins and their activating kinases (cyclin-dependent kinases
[CDKs]) are the proteins that are the most important determi-
nants for the processing through cell cycle phases and transi-
tions between phases. Cyclin E associated with CDK2
controls G1=S transition, and its dimerization is dependent on
protein p21, and Cyclin B associated with CDK1 is essential
for G2=M transition, and the dimerization is controlled by the
clock-controlled kinase Wee1 ,, .
Circadian Control on Cyclins and CDKs
Of these two mechanisms, the most known is the inhibition of
G2=M transition by the circadian clock-controlled kinase
Wee1, which deactivates CDK1, blocking cells in G2. It has
been modeled by Goldbeter in  with an ODE system in
which Wee1 is a parameter. This CDK system oscillates tran-
siently with an intrinsic frequency that depends on its parame-
ters, but it may be entrained at a 24-h period by Wee1 if Wee1
is an output of the molecular circadian clock (in fact, the clock
protein Bmal1 plays this role whereas the control on G1=S
through p21 is exerted by PER). The entrainment of the nor-
malized M-cell population t7!e?ktRþ1
the population, by a 24 h-entrained CDK1 as function w1in
shown inFigure 1.
This means that for t? ! þ 1 the total population in phase
i, NiðtÞ ¼Rþ1
being the sameforall phases1 ? i ? I.
n2(t,a) da, in a two-
phase model (n1¼ cells in G1? S ? G2, n2¼ cells in M) of
ni(t,a) da is of the form Ni(t) ¼ ektWi(t),
where Wiis a periodic function with the same period as wi, k
CDK1 (Lower Trace)
M Cells (Upper Trace)
Fig. 1. Entrainment of G2=M transition in a two-phase cell
cycle population model (n1: G1? S ? G2cells; n2: M cells)
by the circadian clock: CDK1 kinase (lower trace), taken as
function w1in the cell cycle model and entrained by Wee1
(here a square wave, not shown, of 4 h duration with 24 h
period) in Goldbeter’s mitotic oscillator model, and normal-
ized population of cells in M phase (upper trace) (abscissae:
tens of hours; ordinates: arbitrary units).
Cancer growth and response to therapy by
anticancer drugs have been shown to be
dependent on circadian clock inputs.
IEEE ENGINEERING IN MEDICINE AND BIOLOGY MAGAZINE JANUARY/FEBRUARY 2008
Pharmacological Control on Cyclins and CDKs
Cell cycle control proteins such as p53 act both at the G1=S
and G2=M checkpoints as sensors of DNA damage and effec-
tors of cell cycle arrest by inhibiting the formation of Cyc
E/CDK2 and Cyc B/CDK1 complexes. Anticancer drugs such
as alkylating agents act by damaging DNA and thus provoke
cell cycle arrestby triggering p53 and its effects on phase tran-
sitions at G1=S and G2=M (). It has been shown that p53,
on the one hand, and many cellular enzymatic drug detoxifica-
tion mechanisms (such as reduced glutathione), on the other
hand, are dependent on the cell circadian clock, showing 24-h
periodicity in their gene expression (). Hence, the molecu-
lar circadian clock exerts its control on cell cycle transitions
both physiologically and when an external pharmacological
control is applied. In this modeling frame, the target of both
controls will be on the time-dependent transition functions
t7!wiðtÞ introduced earlier.
The Circadian System: A Network of Oscillators
This circadian clock control on the cell division cycle is not
independent of individual or environmental factors. Although
this clock is endowed with an intrinsic circa 24-h (circa
diem) period given by a hypothalamic pacemaker, the supra-
chiasmatic nuclei (SCN), it is dependent on photic inputs
(entrainment by the light/dark cycle through the retinohypo-
thalamic tract) and also on the disruptive input of circulating
molecules such as cytokines, often elevated in cancers,
which may perturb its amplitude so deeply that any physio-
logical body circadian rhythm (temperature, cortisol, rest/
activity, etc.) becomes undetectable. Indeed, it has been
shown that patients with cancer showing disrupted circadian
rhythms are less responsive to chemotherapy and have
poorer prognosis than do patients with preserved circadian
rhythmicity , .
The Network and Its Constituents
The body circadian clock , , as far as its control on
peripheral cell proliferation is concerned, may be seen as an
orchestra consisting in the same basic individual oscillators:
cell molecular circadian clocks, since the molecular mecha-
nisms are the same whatever the cells, its peripheral constitu-
ents (the musicians) being slaves, not communicating
together, to the central pacemaker located in the SCN (the
conductor), located inthe hypothalamus.
Various physiological ODE models of individual circadian
clocks have been published in the last ten years. They rely on
transcriptional regulation, a mechanism possibly yielding limit
periodic behavior , , . The simplest such physio-
logical model of circadian clock is the three-dimensional ODE
model of FRQ (or PER) protein regulation in Neurospora
Crassa , and it was chosen for the individual cell clock
Knþ Zn? Vm
Kdþ PER? k1PER þ k2Z
¼ ksRNAm? Vd
dt¼ k1PER ? k2Z,
where PER and mRNA stand for the PER cytoplasmic protein
and its messenger RNA concentrations and Z for a transcrip-
tional retroinhibition factor linkedto PER synthesis.
The Conductor: The SCN
Diffusive Neuronal Coupling in the SCN
From this individual clock model,anetwork of circadianoscil-
latorsresultsby diffusivecoupling of PER. Such couplingmay
be physiologically achieved through gap junction connections
between suprachiasmatic neurons, through local release in the
intercellular space of neurotransmitters such as vasointestinal
peptide, abundant in the ventrolateral part of these nuclei,
where the main pacemaker is supposedto be located, andbind-
ing to VPAC2membrane receptors on these same neurons, or
even through electrical signaling with glial participation .
This coupling is thus not necessarily instantaneous, possibly
relying mainly on slow tissue diffusion through gap junctions
or ligand-receptor connections, but nevertheless may be con-
sidered as fast at a time scale of 24 h, relevant for cell division
Knþ ZðiÞn? VmðiÞ
Kdþ PERðiÞ? k1PERðiÞ
¼ ksmRNAðiÞ ? Vd
þ k2ZðiÞ þ Ke
½PERðjÞ ? PERðiÞ?
¼ k1PERðiÞ ? k2ZðiÞ,
where 1?i,j?N, N being the number of neurons connected
in the pacemaker network. The degradation rate Vmof the
messenger RNA is supposed to differ from one neuron to the
other (with random distribution around a central value) and
thus holds variability in individual PER period. The output of
this pacemaker network, transmitted to the periphery, is an
In the case of healthy cells,
there can be no exponential growth and
tissue homeostasis must be ensured.
IEEE ENGINEERING IN MEDICINE AND BIOLOGY MAGAZINE JANUARY/FEBRUARY 2008
higher amplitude in its periodic variations (i.e., good synchro-
nization) asthe couplingstrength Keisstronger.
i¼1PER(i) between neurons, which shows as
Synchronizing Photic Inputs
It is also known that light, through the retinohypothalamic
tract, modulates, equally for all neurons, the transcription rate
Vs, which is thus supposed to be Vs¼ Vs01 þ Lcos(2pt)=24
This modulation actually entrains the suprachiasmatic pace-
maker, initially endowed with a period dependent on the
parameters, e.g., 21 h 30 min with our parameter set, to a
forced 24-h period, provided that the entrainment by light L
is strong enough to overcome the spontaneous period result-
ing from the coupling between neurons inside the pacemaker.
A higher risk of cancer is related with circadian disruption
induced by light/dark rhythm perturbations in shiftwork (see
review in , and the present model aims at taking such
perturbations into account.
Disruptive Inputs from Cytokines and Drugs
Circulating molecules such as cytokines (interferon, interleu-
kins), either secreted by the immune system in the presence of
cancer cells or delivered by therapy, are known to have a dis-
ruptive effect on the circadian clock, most likely at the central
pacemaker level , and elevated levels of cytokines have
been shown to be correlated with fatigue , asymptom con-
stantly found in cancer, which presents close relationships
with the jet-lag of transmeridian flights. This has also been
found with anticancer drugs and is supposed to be a result of
drug toxicity on the SCN. It may be taken into account in the
model by an additive effect on variable Z, which represents
transcriptional inhibition,in the central clocks.
Transmission Pathways from the Center
to Peripheral Cells
These pathways are not completely known but the autonomic
nervous system, and neurohormonal ways using the hypothal-
amocorticosurrenal axis, is a likely candidate to this role .
The messengers may be represented by supplementary
evolution equations describing in a simple way a chain
between messengers: intercentral, hormonal, such as adreno-
corticotropic hormone (ACTH), and peripheral tissue terminal
(such as cortisol)
dt¼ k3PERðSCNÞ ? k4U
dt¼ k4U ? k5V
b þ V? cW:
The Musicians: Peripheral Cell Circadian Clocks
This terminal control is supposed, as in the case of cytokines
in the pacemaker, to be exerted at the transcriptional level
through a modification of variable Z in individual peripheral
will act itself on each peripheral circadian clock by additively
modifying to Z þ rW the feedback variable Z, where r is a
parameter. If the central pacemaker rhythm is strong enough,
i.e., constituted of enough synchronized neuronal oscillators, it
will entrain each peripheral individual clock at a 24 h-rhythm;
but there is no intercellular communication between peripheral
cells unlike between neurons in the central pacemaker. In this
setting, cell cycle phase transition functions wiin peripheral
tissues will be supposed to be tissue-averaged versions of local
PER proteins., i.e., wi¼ kiPER, where ki are constants and
PER isa time-dependent function, averaged ofPER(j) over the
cells (1 ? j ? N, each j cell having its own period, distributed
around a central value) of a peripheral tissue assembly. They
can also be CDKs, as in Figure 1, controlled by a protein such
as Wee1, itself resulting from an averaging of PER (or rather
Bmal1 in the case of Wee1, but PER and Bmal1 are in anti-
phase, i.e., phase delayed by 12 h in their circadian rhythm).
To illustrate this, Figure 2 shows variables mRNA, PER, and Z
of the model circadian clock averaged in a population of
peripheral cells subject to entrainment by the central pace-
The behavior of the cell cycle model, which is presented here
in simple form, completely linear, limits its applicability to
the early stages of tumor growth. They may be considered as
the most important because even when a tumor is discovered
at an advanced stage of its growth, the first administered anti-
cancer treatments, when they do not eradicate it, may shrink it
to cell population conditions putting it in such early stages.
Besides, an extended version of this model with nonlinear
feedback by cyclin concentration and total cell population
number ,  opens it to the representation of different types
of tumor behavior, namely, polynomial in time, as has been
experimentally observed .
The circadian system representation, with its central and
peripheral components, is meant to be able to account for
disrupting inputs to it from tumor-produced cytokines or anti-
cancer drugs, and their possible overcoming by shielding mol-
ecules such as epidermal growth factor receptor (EGFR)
antagonists. An analysis of its effects onto cell cycle phase
transitions (by functions wiin this model) remains to be done.
More detailed representations of the molecular circadian
clock, as recent ones by Leloup and Goldbeter , with
0 100200300 400500 600700
Fig. 2. Three epochs of 240 h for variables mRNA, PER and Z
of a peripheral circadian clock (average of local circadian
clocks): a) without entrainment by light (L ¼ 0); b) with
entrainment (L ¼ 1); c) without (L ¼ 0) (abscissae: hours;
ordinates: arbitrary units).
IEEE ENGINEERING IN MEDICINE AND BIOLOGY MAGAZINE JANUARY/FEBRUARY 2008
Bmal1, Clock, Per, Cry, and other genes, rather than the sim-
plifiedmodel usedhere may also be used.
This article presents a modeling frame for the circadian control
of cell and tissue proliferation. It aims at providing a rationale
for therapeutic optimization of cancer chemotherapies, that is,
maximization of tumor cell kill while shielding healthy renew-
ing tissues from unwanted toxic side effects, by using synchro-
nization of the drug delivery schedule and its cell processing
with intrinsic cell cycle timing. Such synchronization naturally
relies on the circadian system and it has been in use in clinical
,  and in theoretical ,  settings in oncology with
macroscopic modeling for drug delivery regimen and therapeu-
tic efficacy and unwanted toxicity representation. Anticancer
drugs are delivered at the whole organism level but act at the
cell and tissue level on cell cycle control mechanisms. This
multiscale modeling framework will provide clinicians with a
theoretical tool, still under construction, to bridge the gap
between the pharmaceutical clinical control level and the
molecular pharmacological hidden level of drug action; opti-
mal design of pharmacological control on the cell cycle thus
may relyon the knowledge of the natural synchronizing control
of both cell proliferation and cell drug processing mechanisms
by the circadian system, which may be routinely assessed by
noninvasive measurements in the clinic . Future modeling
developments will aim at designing practical clinical rules for
dynamic drug delivery schedule optimization based on molec-
ular pharmacokinetic-pharmacodynamic data for the drugs in
use and drug enzymatic metabolism and circadian profiles for
patients under anticancer treatment, as recently suggested .
This article is the extended version of a paper first published
under the same title, pp. 176–179 in the proceedings of the
28th Annual International IEEE-EMBS Conference, New
York, 2006. I thank Benoı ˆt Perthame and Fadia Bekkal Brikci
for their contributions to cell cycle modeling, Be ´atrice
Laroche for helping deriving the circadian system oscillator
network model, and Sonia Montanes for helpful information
on synchronization in the SCN.
This work was supported by the EU Network of Excellence
BIOSIM, contract 005137.
Jean Clairambault received a Ph.D. in
mathematics from Universite ´ Paris VII-
Denis Diderot in 1978 and an M.D. from
Universite ´ Paris VI-Pierre-et-Marie-Curie
in 1989. He is a researcher at Institut
National de Recherche en Informatique et
en Automatique (INRIA), Rocquencourt,
France. His research interests include mo-
lecular pharmacokinetics-pharmacodynamics of anticancer
drugs, the cell cycle and its control by physiological and
pharmacological inputs for cancer therapeutics, the modula-
tion of these effects by the body circadian clock, and anti-
cancer therapeutic optimization.
Address for Correspondence: Jean Clairambault, Institut
National de Recherche en Informatique et en Automatique
(INRIA), Domaine de Voluceau, BP 105, F78153 Rocquen-
court,France. E-mail: email@example.com.
 C. Basdevant, J. Clairambault, and F. Le ´vi, ‘‘Optimisation of time-scheduled
regimen for anti-cancer drug infusion,’’ Math. Model. Numer. Anal., vol. 39,
no. 6, pp. 1069–1086, 2005.
 F. Bekkal Brikci, J. Clairambault, B. Ribba, and B. Perthame, ‘‘A cell popula-
tion model with proliferation and quiescence for healthy and tumoral tissues,’’ J.
Math. Biol., to be published.
 F. Bekkal Brikci, J. Clairambault, and B. Perthame, ‘‘Analysis of a molecular
structured population model with possible polynomial growth for the cell division
cycle,’’ J. Math. Comput. Model., to be published.
 G. A. Bjarnason, R. C. K. Jordan, and R. B. Sothern, ‘‘Circadian variation in
the expression of cell-cycle proteins in the human oral epithelium,’’ Am. J.
Pathol., vol. 154, no. 2, pp. 613–622, 1999.
 A. Bru ´, S. Albertos, J. L. Subiza, J. L. Gareia-Asenjo, and I. Bru ´, ‘‘The univer-
sal dynamics of tumor growth,’’ Biophys. J., vol. 85, no. 5, pp. 2948–2961, 2003.
 J. Clairambault, B. Laroche, S. Mischler, and B. Perthame, ‘‘A mathematical
model of the cell cycle and its control,’’ INRIA Internal Research Rep. 4892,
2003. Available: http://hal.inria.fr/docs/00/07/167/90/PDF/RR-4892.pdf
 J. Clairambault, P. Michel, and B. Perthame, ‘‘A model of the cell cycle and
its circadian control,’’ In Proc. 6th Int. Conf. of the European Society for Mathe-
matical and Theoretical Biology (ESMTB), Dresden, July, 2005.
 J. Clairambault, P. Michel, and B. Perthame, ‘‘Circadian rhythm and tumour
growth,’’ C.R. Acad. Sci. Paris Ser. I Math., vol. 342, no. 1, pp. 17–22, 2006.
 J. Clairambault, ‘‘Modeling oxaliplatin drug delivery to circadian rhythms in
drug metabolism and host tolerance,’’Adv. Drug Deliv. Rev., vol. 59, no. 9/10,
pp. 1054–1068, 2007.
 R. Dautray and J.-L. Lions, Mathematical Analysis and Numerical Methods
for Sciences and Technology. Springer, 1990, Ch. VIII, pp. 187–199.
 E. Filipski, V. M. King, X. M. Li, T. G. Granda, M.-C. Mormont, X. H. Liu,
B. Claustrat, M. H. Hastings, and F. Le ´vi, ‘‘Host circadian clock as a control
point in tumor progression,’’ J. Natl. Cancer Inst., vol. 94, no. 9, pp. 690–697,
 E. Filipski, F. Delaunay, V. M. King, M. W. Wu, B. Claustrat, A. Gre ´chez-
Chassiau, C. Guettier, M. H. Hastings, and F. Le ´vi, ‘‘Effects of chronic Jet Lag
on tumor progression in mice,’’ Cancer Res., vol. 64, no. 21, pp. 7879–7885,
 E. Filipski, P. Innominato, M. X. Wu, X. M. Li, S. Iacobelli, L. J. Xian, and
F. Le ´vi, ‘‘Effects of light and food schedules on liver and tumor molecular clocks
in mice,’’ J. Natl. Cancer Inst., vol. 97, no. 7, pp. 507–517, 2005.
 L. N. Fu, H. Pelicano, J. Liu, P. Huang, and C. C. Lee, ‘‘The circadian gene
Per2 plays an important role in tumor suppression and DNA damage response in
vivo,’’ Cell, vol. 111, no. 1, pp. 41–50, 2002.
 L. N. Fu and C. C. Lee, ‘‘Thecircadianclock:Pacemakerandtumorsuppressor,’’
 A. Goldbeter, Biochemical Oscillations and Cellular Rhythms. Cambridge,
UK: Cambridge Univ. Press, 1996.
 D. Gonze, S. Bernard, C. Wassermann, A. Kramer, and H. Herzel, ‘‘Sponta-
neous synchronization of coupled circadian oscillators,’’ Biophys. J., vol. 89,
no. 1, pp. 120–129, 2005.
 M. H. Hastings, A. B. Reddy, and E. S. Maywood, ‘‘A clockwork web: Cir-
cadian timing in brain and periphery, in health and disease,’’ Nat. Rev. Neurosc.,
vol. 4, no. 8, pp. 649–661, 2003.
 A. Kramer, F. C. Yang, P. Snodgrass, X. Li, T. E. Scammell, F. C. Davis,
and C. J. Weitz, ‘‘Regulation of daily locomotor activity and sleep by hypothala-
mic EGF receptor signaling,’’ Science, vol. 294, no. 5551, pp. 2511–2515, 2001.
 J.-C. Leloup and A. Goldbeter, ‘‘Toward a detailed computational model for
the mammalian circadian clock,’’ Proc. Natl. Acad. Sci. U.S.A., vol. 100, no. 12,
pp. 7051–7056, 2003.
 J.-C. Leloup, D. Gonze, and A. Goldbeter, ‘‘Limit cycle models for circadian
rhythms based on transcriptional regulation in Drosophila and Neurospora,’’ J.
Biol. Rhythms, vol. 4, no. 6, pp. 433–448, 1999.
 F. Le ´vi, ‘‘Circadian chronotherapy for human cancers,’’ Lancet Oncol.,
vol. 2, no. 5, pp. 307–315, 2001.
 F. Le ´vi and U. Schibler, ‘‘Circadian rhythms: Mechanisms and therapeutic
implications,’’ Ann. Rev. Pharmacol. Toxicol., vol. 47, pp. 593–628, 2007.
 T. Matsuo, S. Yamaguchi, S. Mitsui, A. Emi, F. Shimoda, and H. Okamura,
‘‘Control mechanism of the circadian clock for timing of cell division in vivo,’’
Science, vol. 302, no. 5643, pp. 255–259, 2003.
 M. Moser and R. G. Stevens, Eds., ‘‘Cancer and Rhythm,’’ Cancer Causes
Control, vol. 17, no. 4, pp. 483–622, 2006.
 S. Montanes, ‘‘Study of the biological clock within the suprachiasmatic
nucleus (SCN) in mammals’ brain,’’ Internal report, Benjamin Franklin Klinikum
fu ¨r medizinische Informatik, Charite ´-Universita ¨tsmedizin, Berlin, 2006.
 M.-C. Mormont and F. Le ´vi, ‘‘Cancer chronotherapy: Principles, applications
and perspectives,’’ Cancer, vol. 97, no. 1, pp. 155–169, 2003.
 T. Rich, P. F. Innominato, J. Boerner, M.-C. Mormont, S. Iacobelli,
B. Baron, C. Jasmin, and F. Le ´vi, ‘‘Elevated serum cytokines correlated with
altered behavior, serum cortisol rhythm, and dampened 24-hour rest-activity pat-
terns in patients with metastatic colorectal cancer,’’ Clin. Cancer Res., vol. 11,
no. 5, pp. 1757–1764, 2005.
 S. You, P. A. Wood, Y. Xiong, M. Kobayashi, J. Du-Quiton, and W. J.
M. Hrushesky, ‘‘Daily coordination of cancer growth and circadian clock gene
expression,’’ Breast Cancer Res. Treat., vol. 91, no. 1, pp. 47–60, 2005.
IEEE ENGINEERING IN MEDICINE AND BIOLOGY MAGAZINE JANUARY/FEBRUARY 2008