6 JANUARY 2017 • VOL 355 ISSUE 6320 29 SCIENCE sciencemag.org
imposing some of these constraints in the
development of functionals often produces
better electron densities (1, 9). As Medvedev
et al. show, this strategy improves accuracy
of both the energies and densities of atomic
systems upon ascending Jacob’s ladder, suggesting a reasonable path for systematic improvement toward higher accuracy (see the
figure). This strategy is more theoretically rigorous and more likely to approach the exact
universal functional. The resulting functionals will also be more reliable for properties
that depend on the electron density.
However, in the near future, this strategy
may not produce functionals that are as
widely applicable for modeling molecular
and condensed phase systems. The challenges are to identify which of the theoretically derived constraints are essential and
to develop functionals that satisfy them
while also producing accurate energies and
geometries. A compromise strategy is to develop empirically parameterized functionals but incorporate the electron density or
related properties, such as dipole moments,
into the databases used in the parameterization procedure.
Given the diversity of the DFT community, which includes theorists, developers,
and practitioners from many fields, all these
strategies will most likely be pursued in parallel. Some strategies will produce functionals for short-term use, whereas others will
focus on long-term solutions. Medvedev et al.
performed a statistical analysis of atomic systems, but an analogous analysis of molecular
systems is also warranted to determine the
generalizability of the conclusions. These issues merit further investigation because, despite the philosophical conundrum discussed
above, most scientists would prefer to obtain
the correct answer for the correct reason. j
REFERENCES AND NOTES
1. M. G. Medvedev, I. S. Bushmarinov, J. Sun, J. P. Perdew, K. A.
Lyssenko, Science 355, 49 (2017).
2. P. Hohenberg, W. Kohn, Phys.Rev. 136, B864 (1964).
3. W.Kohn, L.J.Sham, Phys. Rev. 140,A1133(1965).
4. A.J.Cohen, P.Mori-Sánchez, W. Yang, Chem. Rev. 112,289
5. J.P.Perdew, K.Schmidt,in AIP Conference Proceedings,V.
Van Doren, Ed. (AIP Publishing, 2001), vol. 577, pp. 1– 20.
6. H.S. Yu,X.He,D.G. Truhlar, J.Chem. Theory Comput. 12,
7. N. Mardirossian, M. Head-Gordon, J. Chem. Phys. 144,
8. C. W. Anson et al. , J. Am. Chem. Soc. 138, 4186 (2016).
9. J. Tao et al., Phys. Rev. Lett. 91, 146401 (2003).
10. The figure sho ws the relative density differences with
respect to coupled cluster singles and doubles (CCSD) for
the beryllium atom, as generated with the GAMESS quantum chemistry program ( 11), for four different density
11. M. W. Schmidt et al., J. Comput. Chem. 14, 1347 (1993).
ACKNO WLEDGMEN TS
I am grateful for helpful discussions with Y. Yang, M. Pak, K.
Brorsen, and T. Culpitt. This work was supported by the National
Science Foundation under CHE-13-61293.
Reprogramming to resist
Prostate cancer cells alter identity to resist therapy
By Kathleen Kelly1 and Steven P. Balk2
One means by which cancer cells evade therapies involves their abil- ity to reprogram to a cell type that no longer depends on the cellular pathway being targeted by the treat- ments. Hormone deprivation therapies that suppress androgen receptor (AR)
signaling are the mainstay of treatment for
metastatic prostate cancer. However, prostate cancers can become resistant to this
approach by losing dependence on androgen hormones. On pages 84 and 78 of this
issue, Mu et al. (1) and Ku et al. (2), respectively, contribute to our mechanistic understanding of this remarkable plasticity in
cell identity, which allows cancers to thrive.
Androgens stimulate prostate cancer
cell growth. The main androgens are tes-
tosterone and dihydrotestosterone, which
are synthesized primarily in the testes. De-
creasing androgen production or prevent-
ing the hormones from acting on prostate
cancer cells often makes the tumors shrink
or grow more slowly. However, prostate
cancer can adapt to androgen deprivation
through alterations that restore AR signal-
ing and maintain their luminal epithelial
adenocarcinoma phenotype, even when
androgen production is low [referred to as
castration-resistant prostate cancer-adeno
(CRPC-adeno)] ( 3). With the development
of more effective AR-targeting drugs such
as abiraterone and enzalutamide, addi-
tional resistance mechanisms are aris-
ing. About a quarter of these resistant
tumors undergo cellular reprogramming
and acquire a continuum of neuroendo-
crine characteristics (CRPC-NE) ( 4, 5). Ge-
nomic analyses have shown that CRPC-NE
evolves from CRPC-adeno. Most CRPC-NE
express one or more NE-lineage markers
[such as synaptophysin (SYP)], and there
are a range of morphological variants,
perhaps reflecting variable differentiation
states. The increased expression of AR and
AR-regulated genes is generally reduced
in CRPC-NE compared to CRPC-adeno, al-
though there is a range of overlap that may
reflect ongoing selection as well as differ-
ences in genomics ( 6).
In addition to NE-lineage markers, the
messenger RNA profiles (transcriptomes)
of CRPC-NE patient samples and prostate cancer models have shown increased
expression of genes involved in neuronal
development, such as sex determining region Y box 2 (SOX2), and genes encoding
epigenetic regulators, such as enhancer of
zeste homology 2 (EZH2) and DNA methyl-transferase 1 (DNMT1) ( 6, 7). DNMT1 may
contribute to epigenetic characteristics,
such as DNA methylation patterns, that are
markedly different between CRPC-adeno
and CRPC-NE ( 6). In the most comprehensive genomic analysis of CRPC-NE to
date ( 6), the co-occurrence of alterations
in cell signaling pathways involving the
tumor suppressor proteins retinoblastoma
1 (RB1) and tumor protein 53 (TP53) was
highly enriched in CRPC-NE (~50%) relative to CRPC-adeno (~15%), suggesting the
involvement of these pathways in the selection of CRPC-NE. Mu et al. and Ku et al.
connect the loss of the RB1 and TP53 genes
to lineage plasticity and epigenetic regulation in prostate cancer resistance to androgen deprivation therapy.
Mu et al. addressed the phenotypic consequences of RB1 and TP53 silencing in a
human cell line that overexpresses AR (
LNCaP-AR cells), a model of CRPC-adeno that
is sensitive to the AR antagonist enzalutamide. Silencing both RB1 and TP53,
but neither alone, caused marked enzalutamide resistance in these cells, although
AR activity persisted and remained responsive to enzalutamide. Notably, cells lacking
TP53 and RB1 displayed lineage plasticity,
as indicated by decreased expression of luminal epithelial cell markers and increased
expression of basal epithelial cell and neuroendocrine markers. Moreover, these gene
expression changes occurred within 48
hours of induced depletion of TP53 and RB1
and could be rapidly reversed, indicating a
direct effect rather than selection for cells
1Laboratory of Genitourinary Cancer Pathogenesis, Center
for Cancer Research, National Cancer Institute, Bethesda,
MD 20892, USA. 2Hematology-Oncology Division and Cancer
Center, Beth Israel Deaconess Medical Center and Harvard
Medical School, Boston, MA 02215, USA.
“…this remarkable plasticity
in cell identity…allows
cancers to thrive.”