308 | Nature | Vol 578 | 13 February 2020
Article
prominently downregulated the DHT signature, ABBV-075 induced
a broader distribution of expression alterations than ABBV-744 and
affected hallmarks that were not affected by ABBV-744 (Fig. 2f and
Extended Data Fig. 3d). Collectively, these results suggest that ABBV-
744 significantly inhibited AR-dependent transcription in LNCaP cells
while having a lower impact on global transcription than ABBV-075.
DbBi has been shown to downregulate AR protein expression in
some but not all experimental settings, probably owing to subtle dif-
ferences in cell lines and exact experimental conditions. In our hands,
neither ABBV-075 nor ABBV-744 reduced AR protein levels in LNCaP
cells (Extended Data Fig. 3b). Given the lack of a direct effect on the
AR protein, genome-wide AR and BRD4 occupancy was determined
to understand the sensitivity of AR-dependent transcription to ABBV-
744 in LNCaP cells. ABBV-075 but not ABBV-744 caused AR peak loss
similar to JQ1 treatment^20 (Extended Data Fig. 4a). Dependency profiles
from the DepMap portal (https://depmap.org/portal/) indicated that
prostate cancer cell lines are significantly more dependent on BRD4
than BRD2 or BRD3, and higher BRD4 dependency is associated with
higher sensitivity to ABBV-744 (Extended Data Fig. 4b), collectively
suggesting that BRD4 may be the primary BET family driver of prostate
cancer cell line viability and an important target of ABBV-744. ABBV-744
displayed a globally weaker but otherwise similar pattern of BRD4 peak
displacement relative to ABBV-075 and JQ1, and preferentially down-
regulated genes associated with super-enhancers similar to DbBi^20 ,^25 ,^26
(Fig. 3a–c and Extended Data Fig. 4c). A subset of BRD4 peaks over-
lapped with AR peaks, and notably 43% of the BRD4/AR co-occupied
sites were in super-enhancers (Extended Data Fig. 4d). Interestingly,
BRD4 was highly bound at AR-occupied super-enhancers relative to
non-AR super-enhancers, and ABBV-744 and ABBV-075 both effectively
displaced BRD4 from the AR-containing super-enhancers, suggesting
an increased dependence of BRD4–AR co-occupied super-enhancers on
BD2 (Fig. 3d). Further integrating the BRD4-binding profile with gene
regulation by AR, ingenuity pathway analysis and motif analysis revealed
enrichment of DHT pathway and androgen-response elements within
super-enhancers from which BRD4 was displaced by both ABBV-744 and
ABBV-075 (Extended Data Fig. 5a, b). For example, ABBV-744 displaced
BRD4 from BRD4–AR co-occupied super-enhancers that are closely
associated with AR-dependent genes and inhibited KLK2 expression
(Fig. 3e–g and Extended Data Fig. 3b). Similarly, ABBV-744 significantly
affected BRD4 occupancy on super-enhancers associated with BD2-
sensitive ACPP but not BD2-insensitive ZG16B (Extended Data Fig. 5c).
To understand the sensitivity of BRD4–AR co-occupied super-
enhancers to ABBV-744, we tested BD2 dependency of the reported
BRD4–AR interaction^20. A small but reproducibly detectable fraction of
BRD4 was found in complex with AR. This DHT and acetylation-depend-
ent interaction was disrupted by ABBV-744 and ABBV-075. By contrast,
the reported interactions of BRD4 with CDK9, GATA2 or CDK9/cyclin T1
with HEXIM1 were not BD2 dependent^7 ,^27 ,^28 (Extended Data Fig. 6a–c).ac e f–6.0 6.0dABBV-075 ABBV-7442
0
–2
–4
log–6FC in 2gene regulationTreatment 2 FC up2 FC down
ABBV-075 700 1,40 1
ABBV-74 420221b Vehicle ENZ ABBV-075 ABBV-7440.3 μM0.06 μM0.06 μMVehicle DMSOENZABBV-075ABBV-744
ABBV-075 (48 nM)ABBV-744 (48 nM)
ABBV-744 (6.0 μM)(^0) MYC ACPP SPDEFZG16B HEXIM1
0.5
1.0
1.5
2.0 8
6
4
2
0
Relative expression
(^00) 0.33.00.030.060.090.030.090.2
20
40
60
80
100
Senescent cells (%)
ENZ ABBV-075ABBV-744
μM
–4
–2
0
2
4
log^2 –6
(ABBV-075 versus DMSO)
–3 –2 –1 01
log 2 (ABBV-744 versus DMSO)
ABBV-075 only
ABBV-744 only+ABBV-075 and ABBV-744Not signi cant
G1S G2/MG1S G2/MG1S G2/M
Vehicle ABBV-075ABBV-744
0
20
40
60
80
100
Cell cycle distribution (%)
0
200
400
600
800
1,000
ABBV-075 ABBV-744
AML PC
IC
(nM) 50
All
DHT
Fig. 2 | ABBV-744 exhibits potent antiproliferative activity against AR-
positive prostate cancer cells and inhibits AR-dependent transcription. a,
The antiproliferative IC 50 values across cancer cell lines after treatment with
ABBV-075 or ABBV-744 for 5 days. b, ABBV-744 induced cell cycle arrest (left,
72 h 60 nM ABBV-075 or 90 nM ABBV-744; concentrations that elicited similar
degrees of inhibition of BD2 of BRD4) and senescence (right, 12 days). ENZ,
enzalutamide. Data are mean ± s.d. (n = 3 biologically independent samples)
and are representative of n = 3 independent experiments. Representative
images of β-galactosidase staining of cells at 100× magnification are shown in
the top right. c, Number of significantly regulated genes (fold change in
expression > 2-fold, P < 0.01, n = 2, statistical analysis by DESeq2 algorithm) and
scatter plot of log 2 -transformed fold change in expression after 24 h treatment
compared with DHT stimulation alone in phenol red-free, charcoal stripped
serum (vehicle, 5 nM DHT, 5 nM DHT and 60 nM ABBV-075, or 5 nM DHT and
90 nM ABBV-744). Genes significantly regulated by both ABBV-075 and
ABBV-744 or by individual compounds were labelled as ABBV-075 and
ABBV-744, ABBV-075 only, or ABBV-744 only. d, Expression of BD2-sensitive and
-insensitive genes quantified using the branched DNA (bDNA) assay after
treatment for 24 h with ABBV-075 or ABBV-744 in the presence of 5 nM DHT.
Data are mean ± s.d. (n = 3 biological replicates) and are representative of n = 2
independent experiments. e, Heat map of DHT-induced gene expression
alterations (DHT signature, fold change in expression of >2, P < 0.01 for DHT
versus vehicle, n = 2) and the response of these DHT signature genes to
treatment with enzalutamide, ABBV-075 or ABBV-744 in DHT-stimulated cells.
f, Genes significantly (q < 0.01) regulated by ABBV-075 or ABBV-744 in DHT-
stimulated cells were classified as DHT-regulated genes (overlapping with the
DHT signature) or non-DHT regulated genes (outside of the DHT signature).
The distribution of log 2 -transformed fold changes in expression is shown as a
split violin plot. The long solid line represents the mean fold change. The small
lines represent individual data points. The dotted line represents the overall
average. Statistical significance between all and DHT was determined by
two-tailed unpaired Student’s t-test, P values were calculated by DESeq2.