Mechanisms of resistance to intermittent androgen deprivation in

Author Manuscript Published OnlineFirst on May 22, 2014; DOI: 10.1158/0008-5472.CAN-13-3162
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Identifying CRPC Mechanisms in Individual Patients
Mechanisms of resistance to intermittent androgen
deprivation in prostate cancer patients identified by a novel
computational method
Jason D. Morken1,2, Aaron Packer1, Rebecca A. Everett1, John D. Nagy1,3, and Yang Kuang1
1. School of Mathematical and Statistical Sciences, Arizona State University, Tempe, Arizona
85287, USA
2. Department of Chemistry and Biochemistry, Arizona State University, Tempe, Arizona 85287,
USA
3. Department of Biology, Scottsdale Community College, Scottsdale, Arizona 85256, USA
Running title: Identifying CRPC Mechanisms in Individual Patients
Keywords: Diagnosis, hormone, therapy, androgen, modeling
Jason D. Morken (Corresponding author)
Mailing address: 2045 East Howe Avenue Tempe, AZ 85281
Phone number: 480-809-9187
Email: [email protected]
This manuscript contains approximately 5,485 words (excluding references), 3 colored figures
and 2 additional supplementary figures, 3 tables, 1 additional supplementary table, and a
mathematical derivation in Supplementary Material.
Abstract
For progressive prostate cancer, intermittent androgen deprivation (IAD) is one of the most
common and effective treatments. Although this treatment is usually initially effective at
1
Downloaded from cancerres.aacrjournals.org on February 2, 2015. © 2014 American Association for Cancer
Research.
Author Manuscript Published OnlineFirst on May 22, 2014; DOI: 10.1158/0008-5472.CAN-13-3162
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Identifying CRPC Mechanisms in Individual Patients
regressing tumors, most patients eventually develop castration-resistant prostate cancer (CRPC),
for which there is no effective treatment and is generally fatal. Although several biological
mechanisms leading to CRPC development and their relative frequencies have been identified, it
is difficult to determine which mechanisms of resistance are developing in a given patient.
Personalized therapy that identifies and targets specific mechanisms of resistance developing in
individual patients is likely one of the most promising methods of future cancer therapy.
Prostate-specific antigen (PSA) is a biomarker for monitoring tumor progression. We
incorporated a cell-death rate function into a previous dynamical PSA model that was highly
accurate at fitting clinical PSA data for 7 patients. The mechanism of action of IAD is largely
induction of apoptosis, and each mechanism of resistance varies in its cell-death rate dynamics.
Thus, we analyze the cell-death rate levels and their time-dependent oscillations to identify
mechanisms of resistance to IAD developing in individual patients.
Major Findings. Here, we introduce a novel computational method called cell-death rate
analysis that uses a mathematical model validated by clinical data to identify the major
mechanism of treatment resistance developing in a given patient. Although cell-death rate
analysis may be applicable to other cancers, we use it here to predictively diagnose the major
mechanism of castration resistance developing in individual prostate cancer patients after a
period of IAD. Our results are consistent with biological literature, and this mathematical method
using cell-death rate analysis may significantly improve personalized treatment of CRPC and
other cancers.
Quick Guide to Equations and Assumptions
2
Downloaded from cancerres.aacrjournals.org on February 2, 2015. © 2014 American Association for Cancer
Research.
Author Manuscript Published OnlineFirst on May 22, 2014; DOI: 10.1158/0008-5472.CAN-13-3162
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Identifying CRPC Mechanisms in Individual Patients
This study builds off of the previous work by Jackson et al., Ideta et al. and especially Portz et
al. (19, 22, 23). As these authors did, we model two populations of prostate cancer cells within a
single patient. One population has a castration-sensitive (CS) phenotype and the other is
castration resistant (CR). For simplicity we will refer to these different populations as “strains.”
The masses of each strain at time t are represented by
and
. Dynamics of these
populations are modeled by the following system of ordinary differential equations:
1
 q 
dX 2
= μm 1 − 2  X 2 − D2 ( Q2 ) X 2 − λ2 ( Q2 ) X 2 + λ1 ( Q1 ) X 1 ,
  

dt
Q2 
 

Cell Death
CR to CS Switch
CS to CR Switch


CR
2
Proliferation
where
represents the mass of the th strain at time . (See Table I for a comprehensive list
of parameter interpretations and values.) Cell population dynamics are driven by intracellular
concentration, or cell quota, of bound androgen receptors (ARs), which we denote
. Here,
1,2 refers to the strain identity. Androgen is viewed as an ecological resource on which
both proliferation and cell mortality depend. As did Portz et al., we represent androgendependent proliferation using an established formalism from mathematical ecology called the
Droop model (28). In this context, the Droop form for per capita proliferation rate of strain
becomes
1
, where positive constants
and
represent maximum proliferation rate
and minimum cell quota required for proliferation, respectively. Since strains 1 and 2 are CS and
CR, respectively, and since CR cells are less dependent on androgens and can proliferate at
aberrantly low levels of serum androgen (2-4), we assume that
.
3
Downloaded from cancerres.aacrjournals.org on February 2, 2015. © 2014 American Association for Cancer
Research.
Author Manuscript Published OnlineFirst on May 22, 2014; DOI: 10.1158/0008-5472.CAN-13-3162
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Identifying CRPC Mechanisms in Individual Patients
Portz et al. assume a constant per capita mortality rate for both strains. However, androgen
deprivation therapy works not only by inhibiting growth and proliferation, but also by inducing
apoptosis in prostate cancer cells (2, 4, 8, 9). To capture this reality, we represent per capita
mortality in the th strain as the following cell-death rate (CDR) function:
3
The first term on the right-hand side expresses androgen-dependent apoptosis, where positive
constants
and
are shape parameters playing similar roles to the half-saturation constant and
Hill coefficients in Hill functions; together they describe how sensitively apoptosis responds to
changes in cell quota. Positive constants
are the maximal apoptosis rates induced by complete
absence of bound ARs. Finally, positive constants
represent androgen-independent cell
mortality.
The switching rates between cell phenotypes are represented by:
K1n
Q2n
Q
c
,
λ
=
.
(
)
2
2
2
Q1n + K1n
Q2n + K 2n
 
λ1 ( Q1 ) = c1
CS to CR
Shape parameters
,
quota, and parameters
and
and
4, 5
CR to CS
deterimine the sensitivity of the switch rates to variations in cell
are maximal switch rates when cell quota approaches 0 or
infinity, respectively. All are positive constants.
The cell quota,
, is modeled by:
4
Downloaded from cancerres.aacrjournals.org on February 2, 2015. © 2014 American Association for Cancer
Research.
Author Manuscript Published OnlineFirst on May 22, 2014; DOI: 10.1158/0008-5472.CAN-13-3162
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Identifying CRPC Mechanisms in Individual Patients
dQi
q − Qi A
=ν m m
− μm ( Qi − qi ) − bQi .

dt
qm − qi A +ν h  Degredation


Growth Dilution
Cell Quota
6
Androgen-AR Binding
Here we interpret the meaning of
slightly differently than did Portz et al. Since
is the
(average) bound AR concentration in the th strain and ARs are intracellular, we view the first
term in the cell quota equation as including two processes: androgen-AR binding inside the cell
and cellular regulation of AR concentration. With this interpretation,
is the concentration of
intraprostatic androgen, which essentially freely diffuses between interstitial and cytosolic
compartments, and is interpolated using the same method as Portz et al. Androgen-AR binding is
/
assumed to be a Michaelis-Menten process with rate
/
saturation constant and
, where
. Positive constant
is the halfis the product
formation rate (rate at which enzyme-substrate complexes dissociate to release product in the
classic Michaelis-Menten formulation), whereas the quotient
/
, in which
is the maximum cell quota, is the total (bound and unbound) AR concentration regulated such
that
for all and ; therefore, the proliferation terms in equations 1 and 2 cannot become
negative. Intracellular androgen degrades at the rate , and the term
accounts for
growth dilution.
Serum prostate specific antigen (PSA) is assumed to be produced by each strain at a
constant baseline rate,
(invariant across phenotype), plus an androgen-dependent rate
governed by a Hill function with Hill parameter
parameter
and maximal PSA production rate
(also invariant across phenotype), shape
, with the latter two being phenotype specific.
Serum PSA degrades at a constant per capita rate, . These assumptions lead to the following
model of serum PSA dynamics:
5
Downloaded from cancerres.aacrjournals.org on February 2, 2015. © 2014 American Association for Cancer
Research.
Author Manuscript Published OnlineFirst on May 22, 2014; DOI: 10.1158/0008-5472.CAN-13-3162
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Identifying CRPC Mechanisms in Individual Patients
dP
Qm
Qm
P .
= σ 0 ( X 1 + X 2 ) + σ 1 X 1 m 1 m + σ 2 X 2 m 2 m − ε


dt 
Q1 + ρ1
Q2 + ρ 2 Degredation


Baseline Production
PSA
7
Androgen-Dependent Production
Since PSA is only produced in epithelial cells, our model only considers epithelial cells and thus
we only model carcinomas. The full model comprises equations 1-2, and 6-7.
Since IAD induces cell death only during on-treatment periods, CDRs oscillate in populations
that depend on androgens for survival. The CDR oscillation amplitude (the difference between
global maximum and global minimum CDR values throughout the entire period of intermittent
treatment) reflects the degree of this dependence. Since each mechanism varies in its dependence
on androgens for survival, the major mechanism of treatment resistance is identified primarily by
the amplitude of CDR oscillation,
, which we define as:
max
min
,
,
,
where
,
min
and
,
,
8
,
. (See Supplementary Material for a detailed
max
derivation of equation 8.)
Introduction
Prostate cancer and treatment
In this work we limit our attention to prostatic adenocarcinomas of non-neuroendocrine origin as
this entity has by far the greatest clinical significance of all prostate tumors. Healthy prostatic
epithelial cells and most malignant prostatic epithelial cells exhibit some degree of dependence
on androgens, including testosterone and dihydrotestosterone (DHT), for proliferation and
survival (1-5). The contribution androgens provide to cell proliferation and survival is mediated
6
Downloaded from cancerres.aacrjournals.org on February 2, 2015. © 2014 American Association for Cancer
Research.
Author Manuscript Published OnlineFirst on May 22, 2014; DOI: 10.1158/0008-5472.CAN-13-3162
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Identifying CRPC Mechanisms in Individual Patients
by the androgen-receptor (AR) axis. It is well known that AR signaling regulates expression of
genes involved in proliferation and survival and is generally accepted to be the primary driver of
tumor progression in adenocarcinomas (1, 2, 4-7). AR signaling inhibition therefore regresses
tumors (2, 4, 8, 9).
The most common and effective treatment for progressive prostate cancer is androgen
deprivation therapy (5, 6). The mechanism of action of this therapy is largely induction of
apoptosis, contingent on prostate cancer cells being dependent on androgens, via the AR axis, for
survival (2-5). Androgen deprivation can be accomplished by surgical castration (orchiectomy)
or chemical castration (with drugs). The latter method is commonly carried out by “total
androgen blockade” which combines LHRH analogues and anti-androgens (10). Chemical
castration is almost universally effective at regressing tumors initially; however, almost all
patients eventually develop resistance to treatment and advance to a more aggressive stage for
which there is no effective treatment (5, 6, 10, 11). This advanced stage has traditionally been
referred to as androgen-independent prostate cancer (AIPC), the rationale for this terminology
being that if prostate cancer cells have no androgen access, but continue to survive and
proliferate, then they must have bypassed the need for androgens. However, although chemical
castration depletes blood serum testosterone levels by >90%, intraprostatic androgen
concentrations remain at 20-50% (2, 4, 5, 12). These residual intratumoral androgen sources
allow prostate cancer cells to survive since these cells are still dependent on androgens for
survival (5). The recent clinical efficacy of abiraterone (which inhibits intraprostatic androgen
synthesis) and enzalutamide (a 2nd-generation AR antagonist), has further confirmed that these
cells are not androgen-independent (5, 13, 14). Therefore, this stage of cancer is now generally
referred to as castration-resistant prostate cancer (CRPC) (2, 4, 5, 12, 15).
7
Downloaded from cancerres.aacrjournals.org on February 2, 2015. © 2014 American Association for Cancer
Research.
Author Manuscript Published OnlineFirst on May 22, 2014; DOI: 10.1158/0008-5472.CAN-13-3162
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Identifying CRPC Mechanisms in Individual Patients
Since continuous androgen deprivation (CAD) provides a prolonged stimulus selecting for
castration-resistant clones, intermittent androgen deprivation (IAD) was proposed to delay the
onset of CRPC (10). Although this basis for IAD remains controversial (16), IAD is cheaper and
minimizes side effects relative to CAD (10). For detailed clinical examples of IAD protocol, see
(9, 17, 18).
The gene coding for prostate-specific antigen (PSA) is exclusively activated by AR signaling in
prostatic epithelial cells (1, 6). PSA is secreted by these cells into the blood serum (1). Since AR
signaling is considered the primary driver of tumor progression, before and after androgen
deprivation failure, PSA is a commonly used predictive biomarker for monitoring androgendeprivation efficacy (1, 9). Increasing PSA levels are characteristic of tumor progression while
decreasing PSA levels are characteristic of tumor regression. For IAD, patients are usually put on
treatment until their PSA levels fall below some threshold value, usually <4 ng/mL, and taken
off treatment until PSA rises above some threshold level, usually >10 ng/mL (10).
Several mathematical models have been developed with the goal of fitting model output to
clinical PSA data throughout the course of IAD (19-23). Here we address how a mechanistic
model of prostate cancer that takes an individual patient’s serum androgen data as input, and uses
data fitting to clinical PSA data to estimate parameters, might be used to inform clinicians of the
inherent mechanism(s) of treatment resistance in individual patients.
Major mechanisms of treatment resistance
Although most prostate cancer cells exhibit some degree of dependence on androgens for
survival, as tumors continue to advance to a more progressive state, the degree of dependence
tends to decrease. Several pathways to CRPC development have been identified with varying
8
Downloaded from cancerres.aacrjournals.org on February 2, 2015. © 2014 American Association for Cancer
Research.
Author Manuscript Published OnlineFirst on May 22, 2014; DOI: 10.1158/0008-5472.CAN-13-3162
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Identifying CRPC Mechanisms in Individual Patients
degrees of androgen dependence (see ref. 1 for detailed review of these mechanisms). Later, we
attempt to quantify this variation in dependence and use it to identify which mechanism is
developing in individual patients. Clinical identification of these mechanisms in individual
tumors still suffers from a lack of maturity since CRPC tissue is difficult to obtain and study
(11). Hence, mathematical models may be a potentially powerful clinical tool for analyzing
mechanisms of CRPC development. Here, we highlight mechanisms that lead to CRPC which
we propose can be identified through our model.
Hypersensitivity pathway: Cells using this as their major mechanism have a lower androgen
threshold for proliferation and survival but remain completely dependent on androgens. We
consider three specific pathways to hypersensitivity: amplification and overexpression of AR,
increased AR stability, and increased local androgen production. Overexpression of AR is
characteristic of most CRPC cases (7, 15, 24). It has been demonstrated that overexpression of
wild-type AR is adequate to fully transform a CS tumor into a CR tumor (24). AR amplification
is one of the most common genetic aberrations to occur in CRPC (15). It has been suggested that
80% of CRPC cases exhibit elevated AR gene copy number while 30% exhibit high-level
amplification (15). It has also been suggested that increased AR stability may be a mechanism
leading to CRPC (1). Increased local androgen production can result from constitutive activation
of 5α-reductase, which converts testosterone to DHT in prostate cells. For example, the V89L
mutation in the gene coding for 5α-reductase leads to an increase in its activity (1). Furthermore,
some CRPC cells may be able to synthesize the androgens themselves from cholesterol or other
androgenic precursors (1, 5, 6).
Promiscuous receptor pathway: Promiscuous ARs generally arise from missense mutations in
the ligand-binding domain that decrease the specificity of the AR for a particular ligand. These
9
Downloaded from cancerres.aacrjournals.org on February 2, 2015. © 2014 American Association for Cancer
Research.
Author Manuscript Published OnlineFirst on May 22, 2014; DOI: 10.1158/0008-5472.CAN-13-3162
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Identifying CRPC Mechanisms in Individual Patients
mutations broaden the group of ligands that activate the AR, such that the AR obtains the ability
to bind to and be activated by other non-androgen agonists in addition to androgens (1, 4).
Potential non-androgen agonists include estrogen, progesterone, cortisol, anti-androgens, or
adrenal androgens such as DHEA (4). For example, the most common promiscuous AR mutation
is T877A, which is endogenously expressed in androgen-sensitive LNCaP cell lines where it was
first discovered (4, 12). Taplin et al. found the T877A mutation to be present in 30% of prostate
cancer bone metastases (25). The T877A mutant AR still binds to, and is thus activated by,
testosterone and DHT. However, alanine replaces the threonine residue at position 877 and
allows ligands other than testosterone and DHT to also fit in the binding pocket and activate the
receptor (12). Promiscuous ARs with complete loss of responsiveness to androgens are
extremely rare (4). While mutations have been identified that reduce affinity for DHT relative to
the wild-type AR, nearly all identified promiscuous ARs are still activated by DHT (4). We
therefore assume that all CR cells with promiscuous ARs retain responsiveness to androgens;
however, they are less dependent on these androgens than cells with hypersensitivity.
Outlaw receptor pathway: Outlaw receptors are steroid-hormone receptors that are activated by
ligand-independent mechanisms. No mutations in the AR gene coding region have been
identified such that the AR is activated by ligand-independent mechanisms (1). Rather, this
pathway apparently involves the wild-type receptor or AR splice variants, since the AR gene
sequence is not changed, and may still exhibit responsiveness to androgens. There have been
several AR splice variants identified such that the ligand-binding domain is absent and these ARs
remain ligand-independently constitutively active (6, 26). However, exclusive expression of AR
splice variants is very rare. The vast majority of cells expressing AR splice variants also express
wild-type AR (7). Therefore, most tumors developing resistance through outlaw AR splice
10
Downloaded from cancerres.aacrjournals.org on February 2, 2015. © 2014 American Association for Cancer
Research.
Author Manuscript Published OnlineFirst on May 22, 2014; DOI: 10.1158/0008-5472.CAN-13-3162
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Identifying CRPC Mechanisms in Individual Patients
variant expression should still exhibit responsiveness to androgens while maintaining active AR
signaling in the absence of androgens. It has been shown the wild-type AR can be activated by
growth factors, receptor-tyrosine kinases, MAPK, AKT, and other activators in a ligandindependent fashion (1). If no ligands are available other than androgens in a cell with
promiscuous ARs, the cell then depends on androgens. If no ligands are available in a cell with
outlaw AR, AR signaling remains active. Thus, this pathway is less dependent on androgens than
the promiscuous receptor pathway since it is completely ligand independent.
The bypass pathway: This mechanism results from constitutive activation of pathways that
directly inhibit apoptosis in the absence of androgens (1). One mechanism may be the
upregulation of BCL2, BCLX, or MCL1, genes known to directly inhibit apoptosis (1, 4, 10).
Prostatic secretory epithelial cells normally do not produce BCL2, but many CRPC cells do (1,
10). However, besides prostatic sarcomas and neuroendocrine carcinomas, which our model does
not consider, the complete bypass of AR signaling does not provide a selective advantage (4, 6).
This pathway is the least dependent on androgens for survival.
Materials and Methods
Formulation of the mathematical model
We extend the prostate cancer IAD treatment model by Portz et al. (19) who consider tumors as
two cell populations: cells that are castration sensitive (CS) and those that are castration resistant
(CR). The mathematical model portrayed in the Quick Guide describes two populations of
luminal secretory epithelial cells that secrete PSA (1). The CR population accounts for prostate
cancer cells that have acquired at least one mechanism of treatment resistance. The static
parameter values, free parameter value ranges, and their interpretations are listed in Table I.
11
Downloaded from cancerres.aacrjournals.org on February 2, 2015. © 2014 American Association for Cancer
Research.
Author Manuscript Published OnlineFirst on May 22, 2014; DOI: 10.1158/0008-5472.CAN-13-3162
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Identifying CRPC Mechanisms in Individual Patients
Detailed model formulation and explanations can be found in (19) and are summarized in the
Quick Guide above. What follows is a narrative summary of the equations and assumptions.
Here we follow Portz et al. in assuming that proliferation of prostate carcinoma cells is more-orless androgen dependent, this dependence can vary among cells in a given tumor, and cells can
switch reversibly between CS and CR states. Our major extension of the Portz et al. model is the
addition of an androgen-dependent cell-death rate (CDR) function,
(equation 3). We
retain the Droop formalism (28), adapted from chemostat and ecological models, to govern
androgen-dependent proliferation, as well as the forms used by Portz et al. for androgendependent switching between CS and CR states—i.e.,
and
(equations 1-2, and 4-5).
In this model, androgen enters cells where it can bind to and dissociate from ARs in the
intracellular compartment. Androgen may also be metabolized or transported out of the system.
The dynamics of these processes are governed by the cell quota (equation 6). Bound ARs alter
cell proliferation, death and state-switching rates. In general, proliferation rates increase and
mortality rates decrease with increasing concentration of bound ARs. The total AR
concentration, both bound and unbound, is regulated in the cell quota equation such that the
proliferation terms in equations 1 and 2 can never become negative. In low androgen
environments, state switching by prostate cells is biased in favor of the CR state. This bias is
reversed in high androgen environments. Both cell strains secrete PSA to the serum at the same
basic rate. However, strains may vary in how their PSA secretion rate depends on bound AR (see
equation 7).
It is an open question whether castration resistance evolves from selection acting on a small
population of CR cells that exist in the tumor prior to treatment (the selection hypothesis), or
12
Downloaded from cancerres.aacrjournals.org on February 2, 2015. © 2014 American Association for Cancer
Research.
Author Manuscript Published OnlineFirst on May 22, 2014; DOI: 10.1158/0008-5472.CAN-13-3162
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Identifying CRPC Mechanisms in Individual Patients
whether CR strains arise during treatment (the adaptation hypothesis), or both (29, 30, 31). The
role accommodative changes play is also not well understood. Here, by accommodative changes,
we mean reversible changes in cell behavior not associated with any genetic or genomic
alterations; i.e., plasticity in the cells’ behavioral repertoire available to accommodate
fluctuations in their environment. The possibility of accommodative switching between relative
CS and CR states in a single clonal lineage was initially investigated in a model by Hirata et al.
(20, 21). Portz et al. (19) assumed reversible switching between CS and CR states at rates
and
, which both depend on the bound AR cell quota (see equations 1-2, and 4-5). Although Portz
et al. interpreted these as forward and backward mutations between CS and CR states, the form
is much more general. Any mechanism of castration resistance, including mutation in coding and
control regions of a variety of genes and epigenetic changes, can be accommodated by the
model’s form. Regardless of which process the mathematics truly reflects, the incorporation of
this switch resulted in Portz et al.’s model fitting data much more accurately than previous
models from which it was derived. However, we interpret the switch as incorporating both
adaptation and accommodation to cover all possible switching mechanisms. The adaptation label
is supported by the fact that AR point mutations are rarely found in primary prostate cancer but
are found in approximately 20% of CRPCs (11). Further, one study showed that only 2 out of
205 untreated prostate cancer cases had detectable AR amplification (32), and another study
showed that 38-63% of metastatic CRPC cases exhibited high-level AR amplification (33, 34).
This evidence strongly suggests that an oscillating androgen environment creates selective
pressure inducing an adaptive switch. Although this evidence only shows evolution towards
CRPC, it is possible that CR tumors could evolve back to CS if selective pressures change (e.g.,
androgen is reintroduced); although this return to castration sensitivity would be unlikely to be
13
Downloaded from cancerres.aacrjournals.org on February 2, 2015. © 2014 American Association for Cancer
Research.
Author Manuscript Published OnlineFirst on May 22, 2014; DOI: 10.1158/0008-5472.CAN-13-3162
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Identifying CRPC Mechanisms in Individual Patients
due to direct back mutation, we recognize that the phenotype can be regenerated by novel
alterations of coding or control sequences to a variety of genes in a given resistance pathway. It
is necessary to include accommodation into the interpretation since epigenetic modifications,
which can lead to phenotypic switching, take place at a much higher rate than genetic
modifications (35) and several epigenetic aberrations play a role in prostate cancer that can
and
facilitate the phenotypic switch represented by
(36).
Cell-death rate analysis
The CDR-analysis method we propose here consists of analyzing the CDR levels and oscillations
in response to intermittent treatment to determine which molecular basis is conferring treatment
resistance in a given patient. The rationale behind CDR analysis is as follows:
1. The mechanism of action of IAD is largely induction of apoptosis, since all CS cells and
most CR cells exhibit some degree of dependence on androgens for survival.
2. Since the treatment is intermittent, CDRs increase during on-treatment and decrease
during off-treatment giving rise to CDR oscillations.
3. The amplitude of CDR oscillation in a given population reflects the degree to which that
population depends on androgens for survival.
4. Each mechanism of treatment resistance varies in its dependence on androgens for
survival.
5. In addition, CDR levels tell us whether or not there is a constitutively active antiapoptotic pathway inherent in the population (i.e. the bypass pathway). An exceedingly
low CDR indicates there is.
14
Downloaded from cancerres.aacrjournals.org on February 2, 2015. © 2014 American Association for Cancer
Research.
Author Manuscript Published OnlineFirst on May 22, 2014; DOI: 10.1158/0008-5472.CAN-13-3162
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Identifying CRPC Mechanisms in Individual Patients
6. We identify which major mechanism of treatment resistance is developing by analyzing
the CDR levels and oscillations.
Data
The data from Akakura et al. were obtained from 7 patients with progressive prostate cancer
undergoing IAD by total androgen blockade (9). Serum PSA and testosterone were assayed
monthly to monitor therapy. Treatment was initiated and subsequently interrupted after serum
PSA had been suppressed to adequately low levels for roughly 6 months. When PSA recovered
past 15-20 ng/mL, treatment was reinitiated for a new cycle.
Treatment induced rapid decreases in serum testosterone, with levels reaching approximately
0.5-1.0 nmol/L after about 1 or 2 months. These low levels were sustained until interruption of
treatment, after which testosterone recovered relatively quickly, with some variation depending
on the patient. The coupling of these changes to those of PSA varied across not only patients but
also cycles of individual patients. Akakura and his colleagues suggested that these dynamics
were due both to changes in androgen sensitivity and to the “dual effect” of androgen on PSA;
namely, the cellular level (PSA synthesis and secretion) and the population level (proliferation
and apoptosis) (9). The model captures these effects via the dependence of
the serum androgen concentration
(equation 6) on
. CDR analysis quantifies this behavior across cycles and
patients in a way that is both simple and biologically meaningful.
For each patient, CDR values were calculated by fitting the model to the PSA data. To
implement the serum androgen function,
, testosterone data was interpolated using the same
approach found in Portz et al. (19). The Nelder-Mead simplex algorithm (37) was then used to
minimize the mean-square error (MSE) of the PSA data with the model solution
by varying
15
Downloaded from cancerres.aacrjournals.org on February 2, 2015. © 2014 American Association for Cancer
Research.
Author Manuscript Published OnlineFirst on May 22, 2014; DOI: 10.1158/0008-5472.CAN-13-3162
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Identifying CRPC Mechanisms in Individual Patients
the parameters μ ,
,
,
,
,
, and
,
1,2. The remaining model parameters were held
fixed across patients (see Table I for ranges and values).
Simulations and Diagnosis
In order to identify the major mechanism of resistance for each patient, we establish criteria
based on
, the total CDR oscillation amplitude defined in equation 8. Since the amplitude of
CDR oscillation reflects the degree to which a population depends on androgens for survival, and
since each mechanism varies in its degree of dependence on androgens for survival, we propose
that there exist threshold CDR oscillation amplitudes which distinguish one mechanism from
another. Specifically, we define two threshold values,
. For 0
and
, the CDR is
said to be relatively constant and the population is thus relatively independent of androgens for
survival. For
0, which occurs when
0 or
0(
,
and
,
are never equal),
the CDR is constant and the population is completely independent of androgens for survival. For
, the CDR is said to slightly oscillate with respect to oscillation in androgen
concentration and the population thus retains some dependence on androgens for survival. For
, we say that there is relatively high oscillation in the CDR with respect to oscillation in
androgen concentration and the population is thus completely dependent on androgens for
survival. We also define a threshold
such that for max
, the CDR is said to be
exceedingly low. An exceedingly low CDR would be characteristic of a constitutively active
anti-apoptotic pathway such that apoptosis is directly inhibited in the population, i.e. the bypass
pathway.
As a novel approach for computational molecular diagnosis, our threshold values for
,
, and
are estimates based on existing hypotheses and data relevant to this work. Model parameters
16
Downloaded from cancerres.aacrjournals.org on February 2, 2015. © 2014 American Association for Cancer
Research.
Author Manuscript Published OnlineFirst on May 22, 2014; DOI: 10.1158/0008-5472.CAN-13-3162
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Identifying CRPC Mechanisms in Individual Patients
were constrained to biologically relevant ranges when applicable. When estimating these values
we considered consistency with patient-specific clinical observations and with the relative
frequency of each mechanism reported in the literature. The threshold values were estimated to
0.013
be
0.0003
,
, and
0.0013
. Mutations in the AR
gene that confer hypersensitivity are the most frequent mutations found in CRPC and there is
strong evidence that AR overexpression is the main mechanism of CRPC development (15).
Therefore, these threshold values were estimated such that 4 out of 7 patients would be
diagnosed with hypersensitivity. Thus
oscillation amplitude value,
was estimated to be slightly below patient 7’s CDR
0.0014
. Since patient 4 was the only patient proposed
was
to have early onset of CRPC and had an exceedingly low CDR relative to others,
estimated to be slightly above max
0.0110
, the maximum CDR of patient 4.
was estimated to be slightly above patient 2’s CDR oscillation amplitude value,
0.0001
, since patient 2 exhibited a more aggressive treatment resistance compared to
others.
Although a given patient may develop CRPC due to a combination of mechanisms, we assume
that it is primarily due to one major mechanism. If more than one mechanism is inherent in the
population, the one that is the least dependent on androgens for survival is the major mechanism.
Since the hypersensitivity pathway maintains the strongest dependence on androgens, we propose
that patients developing resistance primarily through this pathway should exhibit a high
oscillation amplitude
max
with a maximum CDR that is not exceedingly low; that is,
and
. The promiscuous receptor pathway, which is less dependent on androgens
than the hypersensitivity pathway, should exhibit a smaller oscillation amplitude
while
maintaining a maximum CDR that is not exceedingly low; that is,
17
Downloaded from cancerres.aacrjournals.org on February 2, 2015. © 2014 American Association for Cancer
Research.
and
Author Manuscript Published OnlineFirst on May 22, 2014; DOI: 10.1158/0008-5472.CAN-13-3162
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Identifying CRPC Mechanisms in Individual Patients
max
. We propose that the outlaw receptor pathway is less dependent on androgens
than the promiscuous receptor pathway since it is completely ligand independent and therefore
exhibits an oscillation amplitude that is zero or relatively constant and the maximum CDR
should remain not exceedingly low; that is, 0
and max
. The bypass
pathway exhibits an exceedingly low CDR and is completely androgen independent for survival
and thus oscillates minimally; that is, 0
and max
. We propose no
mechanism for exceedingly low and oscillating CDRs as they are biologically unlikely to exist
since the apoptotic program would be directly shut down. These criteria are outlined in Table II.
Results and Discussion
For each patient we computed the amplitudes of CDR oscillation
given in equation 8 and the
maximum CDRs for cell-death rate analysis. We then diagnosed each patient using the outlined
criteria (see Table II). Figure 1 shows the PSA, cell-quota, CDR, and proliferation-rate plots for
patient 1. Figure 2 shows the PSA plots for patients 2-7. Figure 3 shows the CDR plots for
patients 2-7. The proliferation-rate and cell-quota plots for patients 2-7 are found in the
Supplementary Material, as are the MSE values of the fitted model solutions and PSA data.
Table III contains the predictive diagnosis for each patient based on our criteria for CDR
analysis. As expected, none of the patients have cell populations with exceedingly low and
oscillating CDRs.
As expected, our results suggest that most CR populations retain at least some degree of
dependence on androgens for survival. Patient 4 is the only patient whose PSA level never
dropped below 4 ng/mL within the first 8 months of treatment (see Figure 2). Akakura et al. (9)
concluded that IAD is likely ineffective for patients whose PSA levels do not fall below this
18
Downloaded from cancerres.aacrjournals.org on February 2, 2015. © 2014 American Association for Cancer
Research.
Author Manuscript Published OnlineFirst on May 22, 2014; DOI: 10.1158/0008-5472.CAN-13-3162
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Identifying CRPC Mechanisms in Individual Patients
stable level within the first 8 months of treatment. They suggest that failure for PSA levels to
normalize within this time period is a sign of early onset of CRPC (9). Our results show that
patient 4 is the only patient that did not have a highly oscillating CS CDR, meaning that the CS
population for patient 4 was not fully dependent on androgens for survival, consistent with the
notion of early onset of CRPC. Patients whose CS CDR oscillations are inside the range of
, or
, may thus be developing early onset of CRPC. Further, patient 4 was diagnosed with the
most aggressive and least androgen-dependent pathway according to our criteria, which is also
consistent with early onset of CRPC.
In future studies, it will be important to assess how much data is required for CDR analysis. In
this study, diagnoses were based on the entire data sets from the clinical trial, which varied from
1.5 to 3.5 cycles per patient. At this stage of its development, we believe at least one full cycle of
treatment is needed. This includes data from start until shortly after the second on-treatment has
begun and PSA has regressed to physiological range. Ideally, physicians will continually test
patients’ levels and update their diagnosis profile with subsequent targeted treatment.
Although the threshold values ,
, and
remain to be validated clinically, even once these
values are empirically determined it is possible they will still be subject to variation among data
sets, and among patients. As with most models used for clinical prediction, it is likely to require
repeated updating and validation with new data sets (38). Further, we expect such updates not
only to involve these thresholds, but also to include modifications to the model itself. We
recognize this study as one that is preliminary but that establishes the principles of CDR
analysis—a novel computational method for molecular diagnosis—on which others may build.
19
Downloaded from cancerres.aacrjournals.org on February 2, 2015. © 2014 American Association for Cancer
Research.
Author Manuscript Published OnlineFirst on May 22, 2014; DOI: 10.1158/0008-5472.CAN-13-3162
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Identifying CRPC Mechanisms in Individual Patients
Although we demonstrated how cell-death rate analysis works using CRPC development as an
example, this methodology may be applied to other types of cancer. In particular, other steroidhormone-driven malignancies, such as estrogen-receptor-positive (ER+) breast cancers, share
many similarities in terms of their progression. For example, certain ER+ breast cancers can be
treated intermittently with LHRH analogues to suppress serum estrogen levels in combination
with anti-estrogens that directly inhibit ER signaling (39, 40), which is similar to the
combination of LHRH analogues and anti-androgens in the total androgen blockade used for
prostate cancer IAD.
Personalized therapy on the basis of identifying and targeting specific mechanisms of treatment
resistance developing in individual patients is likely one of the most promising methods of future
cancer therapy. CDR analysis is a method for identifying the mechanism, which physicians can
then conceivably target, thus reestablishing treatment sensitivity in these patients. Sequential
identification and targeting of resistance mechanisms may potentially control or eradicate
cancers.
References
1. Feldman BJ, Feldman D. The development of androgen-independent prostate cancer.
Nature 2001;1:34-45.
2. Scher HI, Buchanan G, Gerald W, Butler LM, Tilley WD. Targeting the androgen
receptor: improving outcomes for castration-resistant prostate cancer. Endocr-Relat
Cancer 2004;11:459-76.
3. Denmeade SR, Lin XS, Isaacs JT. Role of programmed (apoptotic) cell death during the
progression and therapy for prostate cancer. The Prostate 1996;28:251-65.
20
Downloaded from cancerres.aacrjournals.org on February 2, 2015. © 2014 American Association for Cancer
Research.
Author Manuscript Published OnlineFirst on May 22, 2014; DOI: 10.1158/0008-5472.CAN-13-3162
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Identifying CRPC Mechanisms in Individual Patients
4. Heinlein CA, Chang C. Androgen Receptor in Prostate Cancer. Endocr Rev 2004;25:
276-308.
5. Mitsiades N. A road map to comprehensive androgen receptor axis targeting for
castration-resistant prostate cancer. Cancer Res 2013;73:4599-605.
6. Nelson PS. Molecular states underlying androgen receptor activation: a framework for
therapeutics targeting androgen signaling in prostate cancer. J Clin Oncol 2012;30:644-6.
7. Sun S, Sprenger CCT, Vessella RL, Haugk K, Soriano K, Mostaghel EA, et al. Castration
resistance in human prostate cancer is conferred by a frequently occurring androgen
receptor splice variant. Journal Clin Invest 2010;120:2715-30.
8. Holzbeierlein J, Lal P, LaTulippe E, Smith A, Satagopan J, Zhang L, et al. Gene
expression analysis of human prostate carcinoma during hormonal therapy identifies
androgen-responsive genes and mechanisms of therapy resistance. Am J Pathol
2004;164:217-27.
9. Akakura K, Bruchovsky N, Goldenberg SL, Rennie PS, Buckley AR, Sullivan LD.
Effects of intermittent androgen suppression on androgen-dependent tumors. Apoptosis
and serum prostate-specific antigen. Cancer 1993;71:2782-90.
10. Rashid MH, Chaudhary UB. Intermittent androgen deprivation therapy for prostate
cancer. Oncologist 2004;9:295-301.
11. Huang J, Wang JK, Sun Y. Molecular pathology of prostate cancer revealed by nextgeneration sequencing: opportunities for genome-based personalized therapy. Curr Opin
Urol 2013;23:189-93.
12. Brooke GN, Bevan CL. The role of androgen receptor mutations in prostate cancer
progression. Curr Genomics 2009;10:18-25.
21
Downloaded from cancerres.aacrjournals.org on February 2, 2015. © 2014 American Association for Cancer
Research.
Author Manuscript Published OnlineFirst on May 22, 2014; DOI: 10.1158/0008-5472.CAN-13-3162
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Identifying CRPC Mechanisms in Individual Patients
13. de Bono JS, Logothetis CJ, Molina A, Fizazi K, North S, Chu L, et al. Abiraterone and
increased survival in metastatic prostate cancer. N Engl J Med 2011;364:1995-2005.
14. Scher HI, Fizazi K, Saad F, Taplin ME, Sternberg CN, Miller K, et al. Increased survival
with enzalutamide in prostate cancer after chemotherapy. N Engl J Med 2012;367:118797.
15. Waltering KK, Urbanucci A, Visakorpi T. Androgen receptor (AR) aberrations in
castration-resistant prostate cancer. Mol Cell Endocrinol 2012;360:38-43.
16. Hussain M, Tangen CM, Berry DL, Higano CS, Crawford ED, Liu G, et al. Intermittent
versus continuous androgen deprivation in prostate cancer. N Engl J Med 2013;368:131225.
17. Prapotnich D, Cathelineau X, Rozet F, Barret E, Mombet A, Cathala N, et al. A 16-year
clinical experience with intermittent androgen deprivation for prostate cancer:
oncological results. World J Urol 2009;27:627-35.
18. Bruchovsky N, Klotz L, Crook J, Phillips N, Abersbach J, Goldenberg SL. Quality of
life, morbidity, and mortality results of a prospective phase II study of intermittent
androgen suppression for men with evidence of prostate-specific antigen relapse after
radiation therapy for locally advanced prostate cancer. Clin Genitourin Cancer
2008;6:46-52.
19. Portz T, Kuang Y, Nagy JD. A clinical validated mathematical model of prostate cancer
growth under intermittent androgen suppression therapy. AIP Adv 2012;2:011002.
20. Hirata, Y, Bruchovsky N, Aihara K. Development of a mathematical model that predicts
the outcome of hormone therapy for prostate cancer. J Theor Biol 2010;264:517-27.
22
Downloaded from cancerres.aacrjournals.org on February 2, 2015. © 2014 American Association for Cancer
Research.
Author Manuscript Published OnlineFirst on May 22, 2014; DOI: 10.1158/0008-5472.CAN-13-3162
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Identifying CRPC Mechanisms in Individual Patients
21. Hirata Y, Akakura K, Higano CS, Bruchovsky N, Aihara K. Quantitative mathematical
modeling of PSA dynamics of prostate cancer patients treated with intermittent androgen
suppression. J Mol Cell Biol 2012;4:127-32.
22. Jackson TL. A mathematical investigation of the multiple pathways to recurrent prostate
cancer: comparison with experimental data. Neoplasia 2004;6:697-704.
23. Ideta AM, Tanaka G, Takeuchi T, Aihara K. A mathematical model of intermittent
androgen suppression for prostate cancer. J Nonlinear Sci 2008;18:593-614.
24. Chen CD, Welsbie DS, Tran C, Baek SH, Chen R, Vessella R, et al. Molecular
determinants of resistance to antiandrogen therapy. Nat Med 2004;10:33-9.
25. Taplin M, Bubley GJ, Shuster TD, Frantz ME, Spooner AE, Ogata GK, et al. Mutation of
the androgen-receptor gene in metastatic androgen-independent prostate cancer. N Engl J
Med, 1995;332:1394-8.
26. Dehm SM, Tindall DJ. Alternatively spliced androgen receptor variants. Endocr-Relat
Cancer 2011;18:R183-R196.
27. Berges RR, Vukanovic J, Epstein JI, CarMichel M, Cisek L, Johnson DE, et al.
Implication of cell kinetic changes during the progression of human prostatic cancer. Clin
Cancer Res 1995;1:473-80.
28. Droop, MR. Some thoughts on nutrient limitation in algae. J Phycol, 1973;9:264–72.
29. Dawson NA. Intermittent Androgen Deprivation. Curr Oncol Rep 2000;2:409-16.
30. Ahmed M, Li LC. Adaptation and clonal selection of castration-resistant prostate cancer:
Current perspective. Int J Urol 2013;20:362-71.
31. Zong Y, Goldstein AS. Adaptation or selection—mechanisms of castration-resistant
prostate cancer. Nat Rev Urol 2013;10:90-8.
23
Downloaded from cancerres.aacrjournals.org on February 2, 2015. © 2014 American Association for Cancer
Research.
Author Manuscript Published OnlineFirst on May 22, 2014; DOI: 10.1158/0008-5472.CAN-13-3162
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Identifying CRPC Mechanisms in Individual Patients
32. Bubendorf L, Kononen J, Koivisto P, Schraml P, Moch H, Gasser TC, et al. Survey of
gene amplifications during prostate cancer progression by high-throughput fluorescence
in situ hybridization on tissue microarrays. Cancer Res 1999;59:803-6.
33. Attard G, Swennenhuis JF, Olmos D, Reid AHM, Vickers E, A’Hern R, et al.
Characterization of ERG, AR, and PTEN gene status in circulating tumor cells from
patients with castration-resistant prostate cancer. Cancer Res 2009;69:2912-8.
34. Leversha MA, Han J, Asgari Z, Danila DC, Lin O, Gonzalez-Espinoza R, et al.
Fluorescence in situ hybridization analysis of circulating tumor cells in metastatic
prostate cancer. Clin Cancer Res 2009;15:2091-7.
35. Greaves M, Maley CC. Clonal evolution in cancer. Nature 2012;481:306-13.
36. Albany C, Alva AS, Aparicio AM, Singal R, Yellapragada S, Sonpavde G, et al.
Epigenetics in Prostate Cancer. Prostate Cancer 2011;21:1-12.
37. Lagarias JC, Reeds JA, Wright MH, Wright PE. Convergence Properties of the Nelder-Mead Simplex Method in Low Dimensions. SIAM J Optim 1998;9:112-47.
38. Toll DB, Janssen KJM, Vergouwe Y, Moons KGM. Validation, updating and impact of
clinical prediction rules: a review. J Clin Epidemiol 2008;61:1085-94.
39. Goel S, Sharma R, Hamilton A, Beith J. LHRH agonists for adjuvant therapy of early
breast cancer in premenopausal women. Cochrane Database Syst Rev 2009;7.
40. Wilcken N, Stockler M. Use of luteinising-hormone-releasing hormone agonists as
adjuvant treatment in premenopausal patients with hormone-receptor-positive breast
cancer: a meta-analysis of individual patient data from randomised adjuvant trials.
Commentary. Lancet 2007;369:1711-23.
Table Captions
24
Downloaded from cancerres.aacrjournals.org on February 2, 2015. © 2014 American Association for Cancer
Research.
Author Manuscript Published OnlineFirst on May 22, 2014; DOI: 10.1158/0008-5472.CAN-13-3162
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Identifying CRPC Mechanisms in Individual Patients
Table I. Static parameter values, free parameter ranges, and parameter interpretations.
“Mutation” parameters in Portz et al. (19) are called “switching” parameters here to reflect
– adaptation and accommodation. *These parameter ranges were
the new interpretation of
set such that min
0.0076 and max
0.06 (see ref. 27).
Table II. Criteria for predictive diagnosis by CDR analysis. The major mechanism of
treatment resistance inherent in a given patient after a period of IAD is identified on the basis
of
(the amplitude of CR CDR oscillation), max
their relationships with 3 threshold values set as
and
0.0013
(the maximum CR CDR), and
0.013
,
0.0003
,
. See Simulations and Diagnosis for criteria elucidation. We
propose no mechanism for exceedingly low and oscillating CDRs.
Table III. Predictive diagnosis for each of 7 patients. These diagnoses were carried out using
criteria for predictive diagnosis by CDR analysis outlined in Table II on the basis of
amplitude of CR CDR oscillation), max
values
0.013
,
(the
(the maximum CR CDR), and threshold
0.0003
, and
0.0013
.
Figure Captions
Figure 1. (A)-(D) show serum-PSA (ng/mL), serum-testosterone (nM), CDR (days-1), and
proliferation-rate (days-1) plots for patient 1, respectively. (A) also displays CS and CR
population dynamics. X-axis is time in days.
Figure 2. (A)-(F) display serum-PSA plots (ng/mL) and CS and CR population dynamics for
patients 2-7, respectively. X-axis is time in days. (D) was scaled for easier viewing; the first
PSA datum is 113 ng/mL.
25
Downloaded from cancerres.aacrjournals.org on February 2, 2015. © 2014 American Association for Cancer
Research.
Author Manuscript Published OnlineFirst on May 22, 2014; DOI: 10.1158/0008-5472.CAN-13-3162
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Identifying CRPC Mechanisms in Individual Patients
Figure 3. (A)-(F) display CDR plots (days-1) for patients 2-7, respectively. X-axis is time in
days.
TABLE I:
26
Downloaded from cancerres.aacrjournals.org on February 2, 2015. © 2014 American Association for Cancer
Research.
Author Manuscript Published OnlineFirst on May 22, 2014; DOI: 10.1158/0008-5472.CAN-13-3162
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Identifying CRPC Mechanisms in Individual Patients
TABLE II:
TABLE III:
27
Downloaded from cancerres.aacrjournals.org on February 2, 2015. © 2014 American Association for Cancer
Research.
Author Manuscript Published OnlineFirst on May 22, 2014; DOI: 10.1158/0008-5472.CAN-13-3162
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Downloaded from cancerres.aacrjournals.org on February 2, 2015. © 2014 American Association for Cancer
Research.
Author Manuscript Published OnlineFirst on May 22, 2014; DOI: 10.1158/0008-5472.CAN-13-3162
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Downloaded from cancerres.aacrjournals.org on February 2, 2015. © 2014 American Association for Cancer
Research.
Author Manuscript Published OnlineFirst on May 22, 2014; DOI: 10.1158/0008-5472.CAN-13-3162
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Downloaded from cancerres.aacrjournals.org on February 2, 2015. © 2014 American Association for Cancer
Research.
Author Manuscript Published OnlineFirst on May 22, 2014; DOI: 10.1158/0008-5472.CAN-13-3162
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
Mechanisms of resistance to intermittent androgen deprivation
in prostate cancer patients identified by a novel computational
method
Jason D Morken, Aaron Packer, Rebecca A Everett, et al.
Cancer Res Published OnlineFirst May 22, 2014.
Updated version
Supplementary
Material
Author
Manuscript
E-mail alerts
Reprints and
Subscriptions
Permissions
Access the most recent version of this article at:
doi:10.1158/0008-5472.CAN-13-3162
Access the most recent supplemental material at:
http://cancerres.aacrjournals.org/content/suppl/2014/06/03/0008-5472.CAN-13-3162.DC1.html
Author manuscripts have been peer reviewed and accepted for publication but have not yet been
edited.
Sign up to receive free email-alerts related to this article or journal.
To order reprints of this article or to subscribe to the journal, contact the AACR Publications
Department at [email protected]
To request permission to re-use all or part of this article, contact the AACR Publications
Department at [email protected]
Downloaded from cancerres.aacrjournals.org on February 2, 2015. © 2014 American Association for Cancer
Research.