1. Cancer Biology
  2. Chromosomes and Gene Expression

CREB5 reprograms FOXA1 nuclear interactions to promote resistance to androgen receptor-targeting therapies

  1. Justin H Hwang  Is a corresponding author
  2. Rand Arafeh
  3. Ji-Heui Seo
  4. Sylvan C Baca
  5. Megan Ludwig
  6. Taylor E Arnoff
  7. Lydia Sawyer
  8. Camden Richter
  9. Sydney Tape
  10. Hannah E Bergom
  11. Sean McSweeney
  12. Jonathan P Rennhack
  13. Sarah A Klingenberg
  14. Alexander TM Cheung
  15. Jason Kwon
  16. Jonathan So
  17. Steven Kregel
  18. Eliezer M Van Allen
  19. Justin M Drake
  20. Matthew L Freedman
  21. William C Hahn  Is a corresponding author
  1. Masonic Cancer Center, University of Minnesota-Twin Cities, United States
  2. Department of Medicine, University of Minnesota, United States
  3. Department of Medical Oncology, Dana-Farber Cancer Institute, United States
  4. Broad Institute of MIT and Harvard, Cambridge, United States
  5. Harvard Medical School, United States
  6. Department of Pharmacology, University of Minnesota-Twin Cities, United States
  7. Warren Alpert Medical School of Brown University, United States
  8. Grossman School of Medicine, New York University, United States
  9. Department of Cancer Biology, Loyola University Chicago, United States
  10. Department of Pharmacology and Urology, University of Minnesota, United States
https://doi.org/10.7554/eLife.73223
Share this article
Cite this article
  1. Justin H Hwang
  2. Rand Arafeh
  3. Ji-Heui Seo
  4. Sylvan C Baca
  5. Megan Ludwig
  6. Taylor E Arnoff
  7. Lydia Sawyer
  8. Camden Richter
  9. Sydney Tape
  10. Hannah E Bergom
  11. Sean McSweeney
  12. Jonathan P Rennhack
  13. Sarah A Klingenberg
  14. Alexander TM Cheung
  15. Jason Kwon
  16. Jonathan So
  17. Steven Kregel
  18. Eliezer M Van Allen
  19. Justin M Drake
  20. Matthew L Freedman
  21. William C Hahn
(2022)
CREB5 reprograms FOXA1 nuclear interactions to promote resistance to androgen receptor-targeting therapies
eLife 11:e73223.
https://doi.org/10.7554/eLife.73223
  2. Download BibTeX
  3. Download .RIS

Abstract

Metastatic castration-resistant prostate cancers (mCRPCs) are treated with therapies that antagonize the androgen receptor (AR). Nearly all patients develop resistance to AR-targeted therapies (ARTs). Our previous work identified CREB5 as an upregulated target gene in human mCRPC that promoted resistance to all clinically approved ART. The mechanisms by which CREB5 promotes progression of mCRPC or other cancers remains elusive. Integrating ChIP-seq and rapid immunoprecipitation and mass spectroscopy of endogenous proteins, we report that cells overexpressing CREB5 demonstrate extensive reprogramming of nuclear protein–protein interactions in response to the ART agent enzalutamide. Specifically, CREB5 physically interacts with AR, the pioneering actor FOXA1, and other known co-factors of AR and FOXA1 at transcription regulatory elements recently found to be active in mCRPC patients. We identified a subset of CREB5/FOXA1 co-interacting nuclear factors that have critical functions for AR transcription (GRHL2, HOXB13) while others (TBX3, NFIC) regulated cell viability and ART resistance and were amplified or overexpressed in mCRPC. Upon examining the nuclear protein interactions and the impact of CREB5 expression on the mCRPC patient transcriptome, we found that CREB5 was associated with Wnt signaling and epithelial to mesenchymal transitions, implicating these pathways in CREB5/FOXA1-mediated ART resistance. Overall, these observations define the molecular interactions among CREB5, FOXA1, and pathways that promote ART resistance.

Editor's evaluation

Building on your earlier work implicating CREB5 in resistance to androgen receptor (AR) inhibition, you have now defined the CREB5 interactome in this setting, revealing physical interaction with AR, with the pioneer transcription factor FOXA1, and with other known co-interacting nuclear factors such as TBX3 and NFIC. Collectively, this work strengthens our understanding of how dysregulated epigenomic and transcriptomic processes drive disease pathogenesis and progression.

https://doi.org/10.7554/eLife.73223.sa0

Introduction

The androgen receptor (AR) plays a fundamental role in the development and function of the prostate. AR transcriptional activity is required for the initiation and progression of prostate cancer and remains critical in metastatic castration-resistant prostate cancer (mCRPC) (Abida et al., 2019; Armenia et al., 2018; Grasso et al., 2012; He et al., 2021; Robinson et al., 2015). The standard of care for patients with mCRPC is androgen deprivation therapy (ADT) and AR-targeted therapies (ARTs). While ADT and ART initially induce responses and lengthen the survival of mCRPC patients, intrinsic or acquired resistance continues to be a substantial clinical obstacle. With limited treatment options, better mechanistic understanding of the targets that drive resistant disease is urgently needed.

Several lines of evidence indicate that mCRPC may acquire resistance to ART through multiple mechanisms. Recent studies of mCRPC patients resistant to second-generation ART and ADT have identified that a subset of the resistant mCRPC harbor AR genomic alterations through mutations, copy number gain, enhancer amplification, as well as increased resistant transcript variants (ARV7) (Bubley and Balk, 2017; Henzler et al., 2016; Takeda et al., 2018; Viswanathan et al., 2018). Other clinical and functional studies together demonstrate that resistance also occurs through mechanisms, including upregulation of cell cycle genes (Comstock et al., 2013; Han et al., 2017) and/or the proto-oncogene MYC (Abida et al., 2019; Armenia et al., 2018; Bernard et al., 2003; Grasso et al., 2012; Robinson et al., 2015; Sharma et al., 2013). In addition, selective activation of pathways such as epithelial to mesenchymal transition (EMT), TGFβ and Wnt has been recently identified from profiling mCRPC patient become resistant to second-generation ART or ADT (Alumkal et al., 2020; He et al., 2021).

Recent mechanistic studies have begun to elucidate how dysregulated epigenomic and transcriptomic processes drive disease pathogenesis and progression. AR activity in prostate tissue is regulated by AR co-factors such as HOXB13 and FOXA1 at AR binding sites (ARBs) (Pomerantz et al., 2015; Pomerantz et al., 2020). ChIP-seq experiments have demonstrated that FOXA1, a co-factor with both pioneering and transcription functions, binds to distinct sites and has different protein interactions that is dependent on the disease stage and pathological phenotype (Baca et al., 2021; Pomerantz et al., 2015; Pomerantz et al., 2020). Recent studies indicate that genomic alterations of FOXA1 promote ART resistance (Adams et al., 2019; Baca et al., 2021; Parolia et al., 2019; Shah and Brown, 2019). We have shown that FOXA1 function remains functionally relevant even as mCRPC differentiate into aggressive variants that no longer require AR signaling, such as neuroendocrine prostate cancer (Baca et al., 2021). We have previously demonstrated that the transcriptional factor cAMP-responsive element binding protein 5 (CREB5) promoted resistance to FDA-approved ART, enzalutamide, darolutamide and apalutamide (Hwang et al., 2019). CREB5 has also been associated with progression of cancers, including ovarian, colorectal, and breast (Bhardwaj et al., 2017; He et al., 2017; Molnár et al., 2018; Qi and Ding, 2014). However, the molecular mechanisms by which CREB5 promotes resistance in mCRPC or general tumorigenesis remains unclear.

Here, we utilized ChIP-seq and rapid immunoprecipitation and mass spectrometry of endogenous proteins (RIME) experiments to resolve the CREB5-associated transcriptional targets and molecular interactions. These CREB5 transcriptional co-factors are potential therapeutic targets to perturb CREB5 signaling in cancers that upregulate its activity.

Results

CREB5 drives a resistance response to enzalutamide and androgen deprivation

We previously performed a large-scale screen to identify genes involved in ADT and ART resistance through overexpression of 17,255 open reading frames (ORFs) in LNCaP cells, an AR-dependent prostate cancer cell line (Hwang et al., 2019). We and others functionally validated several genes that drive ADT/ART resistance (FGFs, CDK4/6, MDM4, CREB5) in cell lines or mCRPC patients (Bluemn et al., 2017; Comstock et al., 2013; Elmarakeby et al., 2020; Han et al., 2017; Hwang et al., 2019). Here, we sought to interrogate the molecular mechanisms specifically associated with resistance to ADT and the clinically used ART, enzalutamide. Specifically, we identified ORFs that when overexpressed only promoted cell survival in the presence of ADT/ART and not in standard cultures. Upon re-examining the data, we found that unlike in ADT/ART, overexpression of CREB5 reduced viability (Z = –1.3) in standard cultures (Figure 1A). The Z-scores, which represent relative proliferation effects compared to all 17,255 ORFs, exhibited robust differentials when comparing the treated arm in comparison to the control arm (Z = +14.5 and –1.3). This observation contrasted to other genes that mediate resistance to ART, such as CDK4 or CDK6, which promoted cell fitness regardless of treatment conditions. At genome scale, many ORFs had preferential fitness effects when considering the differential Z-score of the treated and standard conditions. Among ORFs, CREB5 had the greatest differential viability effect after low androgen and enzalutamide treatment (Supplementary file 1, Table 1). Overall, these observations prompted us to pursue a deeper interrogation of binding properties of CREB5 to understand specific molecular interactions that promote resistance to ART.

Figure 1 with 1 supplement see all
Download asset Open asset
CREB5 overexpression and nuclear interactions that are reprogrammed upon androgen receptor-targeted therapy (ART) treatments.

(A) Analysis of enzalutamide resistance genes in LNCaP cells based on a genome-scale screen, including 17,255 open reading frames (ORFs). Z-scores are displayed for the experimental arms conducted in either standard culture (FCS, x-axis) or treatment (enzalutamide +CSS, y-axis) conditions. CREB5 and other enzalutamide -specific hits (Z > 3) and their proliferation scores are highlighted in red. (B) A model that depicts changes in chromatin -associated interactions of CREB5 that occur post enzalutamide treatment. Bottom,: CREB5 ChIP-seq data is presented in accordance to three categories of CREB5 binding behavior. Categories are grouped by significant changes by enzalutamide treatments. (C) GIGGLE analyses predicts transcription factors that are CREB5-bound based on the ChIP-seq experiments as categorized in B. (D) Rapid immunoprecipitation and mass spectroscopy of endogenous proteins (RIME) experiments were performed to identify CREB5 interaction profiles in control or enzalutamide -treated cultures. The common proteins identified by both RIME and GIGGLE are highlighted for the retained and gained groups.

Dynamic CREB5 nuclear interactions are associated with the ART resistance response

We next determined the CREB5 cistrome as well as other regulatory proteins critical to the ART-resistant phenotype by identifying features that were either ‘retained’ or ‘gained’ upon ART treatments. In LNCaP cells overexpressing V5-tagged CREB5 or luciferase and cultured in either standard media or media containing enzalutamide, we analyzed differential interactions of CREB5 at regulatory elements through ChIP-seq or with other proteins through RIME. We first examined CREB5 binding sites pre- and post-enzalutamide treatments by performing CREB5 ChIP-seq. CREB5 overexpression induced a robust differential phenotype (Figure 1B), in which lost (n = 5392), retained (n = 12,432), or gained (n = 12,144) CREB5 binding sites were tallied in the enzalutamide condition compared to the pretreatment condition. To nominate other candidate trans-acting factors that bind within the three defined categories, we used the GIGGLE, an analytical approach that compares collection of sequences from ChIP-seq experiments with over 10,000 experiments from the ENCODE ChIP-seq database and nominates the transcription factors with highly similar binding profiles (Layer et al., 2018). We observed that the elements in which CREB5 retained or gained sites were statistically significantly enriched of other nuclear proteins in which ChIP-seq had been performed in prostate cancer cell lines. This included binding elements (AR, FOXA1, HOXB13) that we previously demonstrated through ChIP-seq experiments (Figure 1C). In addition, we also identify that retained or gained sequences nominated well-studied regulators or co-factors of AR such as GRHL2 (Paltoglou et al., 2017), EP300 (Yu et al., 2020), and SMARCA4 (Launonen et al., 2021; Marshall et al., 2003). In confirmation of previous findings, the genetic ablation of AR or FOXA1 reduced viability of CREB5-overexpressing cells (Figure 1—figure supplement 1; Hwang et al., 2019). These observations suggested that CREB5 bound critical regulators of prostate cancer biology to drive ART resistance.

RIME has been used as a tool to study transcription co-factors interactions in hormone-regulated cancer cells (Glont et al., 2019; Mohammed et al., 2016). To build upon the predictions of CREB5 binding by GIGGLE, we directly examined CREB5 nuclear interactions via RIME. As part of the RIME analysis, we included only the unique peptides that bound CREB5 after subtraction of all proteins precipitated by the V5 antibody or luciferase in cell lysates (Supplementary file 1, Table 2). When we compared the analyzed RIME binding profiles of the control to the enzalutamide-treated conditions, analogous to the ChIP-seq experiments, we found the CREB5-bound unique peptides also segregated into lost (n = 77), retained (n = 222), and gained groups (n = 207) (Figure 1D). To consider key factors in the retained or gained groups that associate with resistance, we integrated the results derived from GIGGLE and RIME. From this, we found both approaches nominated known AR interactions (EP300, FOXA1, and GRHL2). Overall, these parallel approaches demonstrate that CREB5 binding dynamically responds to ART treatment. Moreover, the retained or gained RIME interactions indicate CREB5 may interact with distinct sets of co-factors, some of which are AR associated, to promote ART resistance.

CREB5-FOXA1 interactions converge in ART resistance

In LNCaP cells, we comprehensively compared CREB5 and FOXA1 protein interactions using RIME. In prior work in LNCaP cells, we found that overexpressing CREB3, a related CREB family member, conferred significantly weaker resistance to enzalutamide, and therefore serves as a useful control to understand CREB5 functions in resistance (Hwang et al., 2019). We used RIME to target overexpressed V5-tagged CREB5, luciferase, or CREB3 upon treating cells with enzalutamide. We optimized experimental conditions to consistently observe unique peptides representing CREB5, CREB3, and FOXA1. On average, we detected 8, 23, and 14 unique peptides that respectively mapped to CREB5, CREB3, and FOXA1 (Supplementary file 1, Table 3). RIME interaction profiles were subsequently constructed for each targeted protein based on the visualization of the counts of all detected unique peptides that were bound. The overall RIME profiles of CREB5 and FOXA1 were compared and exhibited a positive correlation (R = 0.394), while those of CREB3 with either CREB5 or FOXA1 lacked correlation (R = 0.0332, –0.0498) (Figure 2A). At the peptide level, we found that CREB5 and FOXA1 shared a total of 504 protein interactions at chromatin, and of these 504, 335 did not interact with CREB3. While almost three times the number of CREB3 peptides were detected relative to CREB5, CREB3 and FOXA1 shared only 83 unique interactions (Figure 2B). These observations nominated CREB5/FOXA1-specific protein–protein interactions.

Figure 2
Download asset Open asset
CREB5 and FOXA1 share chromatin-associated functions in metastatic castration-resistant prostate cancer (mCRPC) based on binding sequences and rapid immunoprecipitation and mass spectroscopy of endogenous proteins (RIME) interaction profiles.

(A) RIME analysis depicting the interaction profiles of FOXA1 (greay), CREB5 (red), and CREB3 (blue). Proteins that interact with FOXA1 and CREB5 are also shown. The Pearson correlation coefficients (R) are shownn. (B) Venn diagram depicting unique peptide interactions that are either independent or shared between CREB5 (red), CREB3 (blue), and FOXA1 (greay). Peptides identified to be induced by enzalutamide are highlighted as Retained/Gained. (C) ChIP-seq experiments were used to examine CREB5 and FOXA1 interactions in LNCaP cells with or without enzalutamide treatments. The Venn diagram depicts total binding sites in each condition and the overlapping sites and percentage of shared transcription regulatory elements. (D) CREB5 -bound sites are analyzed and represented as AR binding sites (ARBS) observed in clinical samples. This includes ARBS exclusive in normal (NARBS), tumor (TARBS), mCRPC (metARBS), all tissues (UARBS), as well as all non -ARBS (OTHER). E. CREB5 -bound ARBs are further classified and depicted as % of FOXA1 sites observed in mCRPC (y-axis). The colors represent the overall percentage of FOXA1 sites while the black represents non -FOXA1 sites.

As an orthogonal approach to RIME, we sought to examine ChIP-seq interactions of CREB5-FOXA1. We first examined the overlap of CREB5 and FOXA1 at transcription regulatory elements with and without ART. By analyzing ChIP-seq experiments that targeted either CREB5 or FOXA1, we found that regardless of ART, CREB5 and FOXA1 shared a strong degree of interactions as more than 90% of CREB5-bound sites were FOXA1 bound (Figure 2C). Pomerantz et al. have recently demonstrated that AR binding sites are reprogrammed during tumorigenesis and progression (Pomerantz et al., 2015; Pomerantz et al., 2020). Since CREB5 promoted ART-resistant activity, we anticipated CREB5 interactions at ARBs would be enriched in tumor-specific binding sites. When considering the subset of ARBs specific to progression, CREB5 indeed bound ARBs in prostate cancer (12.7%) or mCRPC (8.6%) tissue at higher rates as compared to normal prostate ARBs (0.2%) (Figure 2D). Unlike ARBs, FOXA1 binding was less dynamic and the binding sites were relatively ubiquitous in prostate tissue. Almost all CREB5-bound sites were also observed in mCRPCs and were FOXA1 bound (Figure 2E). Taken together these observations demonstrate that CREB5 and FOXA1 engage a similar subset of proteins in cells with ART resistance and these interactions are enriched at ARBs or FOXA1 binding sites in mCRPC.

CREB5-interacting co-factors are associated with ART resistance

To define which CREB5-specific interactions are required to drive ART resistance, we used CREB5 mutants to interrogate these interactions. Based on sequence alignment, the B-Zip and L-Zip domains are highly homologous in CREB and ATF family members (Figure 3A) and regulate binding to DNA- and CREB co-factors (Dwarki et al., 1990; Luo et al., 2012). Within the B-Zip and L-Zip domains, several leucine residues regulate transcriptional activity and heterodimerization with the transcription factors JUN and FOS in vitro (Fuchs and Ronai, 1999; Nomura et al., 1993). We engineered CREB5 point mutants that would emulate these structural perturbations by disrupting binding at chromatin (R396E), CRTCs (K405A and K406A), and JUN/FOS (L431P and L434P). We expressed these CREB5 variants in LNCaP cells (Figure 3B) and found that despite robust expression of L431P and L434P CREB5, cells expressing these mutants proliferated similarly to cells expressing luciferase controls in cell doubling assays performed in low androgen media containing enzalutamide (Figure 3B), indicating L431P and L434P CREB5 lacked interactions critical for ART resistance.

Figure 3
Download asset Open asset
A loss of resistant CREB5 mutant was identified and determines transcription co-regulators associated with androgen receptor-targeted therapy (ART)-resistant proliferation.

(A) Alignment of CREB5 sequence with ATF2, ATF7, and CREB3, highlighting the DNA binding domains (blue), CRCT2 binding domains (red), and JUN/ATF binding domains (yellow). (B) Population doubling (y-axis) of LNCaP cells overexpressing wild-type CREB5 variants (red), CREB5 JUN/FOS-binding mutants (purple), and a luciferase negative control (green) in 10 μM enzalutamide. V5 expression represents V5-tagged CREB5 protein levels. Actin is a loading control. (C) Rapid immunoprecipitation and mass spectroscopy of endogenous proteins (RIME) analysis depicting the interaction profiles of wild-type CREB5 (red), CREB5 L434P (orange), and CREB3 (blue). CREB5 interactions that were reduced, retained, or gained upon enzalutamide treatments are depicted. The Pearson correlation coefficients (R) are shown. (D) A heatmap depicts the RIME interactions of luciferase control, wild-type CREB5, L434P CREB5, and CREB3. Several canonical AR co-factors (AR, FOXA1, HOXB13) interact with both CREB5 and CREB5 L434P and are shown.

Figure 3—source data 1

Immunoblots were used to detect expression of V5-tagged CREB5 or luciferase in the indicated samples for Figure 3B.

https://cdn.elifesciences.org/articles/73223/elife-73223-fig3-data1-v2.tif
Figure 3—source data 2

The area highlighted was used to develop the figure for V5-tagged CREB5 or luciferase in the indicated samples for Figure 3B.

https://cdn.elifesciences.org/articles/73223/elife-73223-fig3-data2-v2.tif
Figure 3—source data 3

Immunoblots were used to detect expression of actin in the indicated samples for Figure 3B.

https://cdn.elifesciences.org/articles/73223/elife-73223-fig3-data3-v2.tif
Figure 3—source data 4

The area highlighted was used to develop the figure for actin in the indicated samples for Figure 3B.

https://cdn.elifesciences.org/articles/73223/elife-73223-fig3-data4-v2.tif

To identify functional interactions specific to wild-type CREB5, we performed RIME in LNCaP cells overexpressing V5-tagged CREB5, CREB5 L434P, or luciferase (Supplementary file 1, Table 4). At approximately the same average peptide counts of CREB5 (8) and L434 CREB5 (7.5), we found a striking correlation between the interaction profiles of CREB5 and CREB5 L434P (R = 0.834). This observation indicated that this CREB5 mutant binds the same proteins as the L434P mutant. In parallel, we failed to find a correlation between CREB3 and wild-type CREB5 (R = −0.0581) (Figure 3C). The limited changes in the L434P CREB5 RIME profile highlighted a subset of differential protein interactions associated with the ART-resistant phenotype. When examining the interactions of wild-type and L434P CREB5, we found smaller differences between AR, FOXA1, and HOXB13 (Figure 3D). Outside of AR co-factors, we detected unique peptide signals at comparable levels with AR, FOXA1, and HOXB13, including NFIC, TBX3 (Figure 3D). We detected peptides from these proteins preferentially in cells expressing wild-type CREB5 as compared to L434P or CREB3. We note that we also found peptides from LBD1 and DNMT1, but further evaluation showed that these proteins did not exhibit as strong a difference in AR-expressing prostate cancer cells, and we therefore concentrated on NFIC and TBX3, which also interacted with FOXA1 in our other RIME experiments (Supplementary file 1 Table 3). These observations identified candidate CREB5 interactors, specifically NFIC and TBX3, that may be essential in ART resistance.

TBX3 and NFIC are critical for AR-positive prostate cancer cells and ART resistance

To examine the relative contribution of TBX3 and NFIC to cell viability, we analyzed genome-scale RNAi screens performed in eight prostate cancer cell lines as a part of Project Achilles 2.20.1 (Cowley et al., 2014; Shalem et al., 2014; Tsherniak et al., 2017). We found that in AR-positive cell lines (Figure 4A, left) TBX3 and NFIC exhibited a pattern of cell line dependencies as observed for FOXA1. When examining AR-negative cell lines (Figure 4A, right), we found that TBX3 exhibited no clear dependency, while NFIC exhibited modest dependency as compared to FOXA1. This close correlation of TBX3 and NFIC dependency with that observed for FOXA1 supports the conclusion that TBX3 and NFIC regulate prostate cancer cell viability in the AR setting. Furthermore, upon computing the average dependency scores in DEMETER, we found that NFIC and TBX3 ranked among the strongest dependencies in these prostate cancer cell lines while exhibiting limited overall dependency in the other 495 cell lines (Figure 4B). The relative dependencies of NFIC and TBX3 were comparable to the strongest gene dependencies, such as FOXA1 and HOXB13, found in previous studies (Pomerantz et al., 2015).

Figure 4
Download asset Open asset
TBX3 and NFIC are key regulators in prostate cancer cells including those that are enzalutamide resistant.

(A) Analysis of genome-scale RNAi screening data ranking the average dependency of 16,869 genes (x-axis) in androgen receptor (AR)-positive (Lleft) and AR-negative (Rrightt.) prostate cancer cell lines (Project Achilles 2.20.1). Average DEMETER score (y-axis) indicates the dependency correlations of FOXA1 and CREB5-interacting proteins. A negative DEMTER score indicates gene dependency in these specific PC cell lines. (B) Average ranks and percentiles based on DEMETER dependency scores are shown for selected genes in AR-positive, AR-negative, and non-PC cell lines. (C). shRNA was utilized to deplete experimental (NFIC, TBX3), negative (GFP) or positive controls (FOXA1) genes in LNCaP cells overexpressing CREB5. The overall cell numbers are depicted post -perturbation. A representative immunoblots depicts depletion of proteins from the proliferation experiments. Tubulin was used as a loading control. The relative depletion is quantified based on the average of all experiments after normalizing to tubulin. (D) CRISPR-Cas9 was utilized to deplete experimental (NFIC, TBX3), negative (GFP) or positive controls (FOXA1) genes in LNCaP cells that spontaneously developed resistance to enzalutamide. The overall cell numbers are depicted post -perturbation. A representative immunoblots depicts depletion of proteins from (C, upper panel) in proliferation experiments. Tubulin was used as a loading control. The relative depletion is quantified based on the average of all experiments after normalizing to tubulin. (E) ChIP-seq data from NFIC and TBX3 was analyzed to predict interaction with CREB5 or FOXA1 motifs. Enriched motifs, the targeted cell lines, and significance levels are depicted.

Figure 4—source data 1

Immunoblots were used to detect expression of TBX3 and FOXA1 in the indicated samples for Figure 4C.

https://cdn.elifesciences.org/articles/73223/elife-73223-fig4-data1-v2.zip
Figure 4—source data 2

The area highlighted was used to develop the figure for TBX3 and FOXA1 in the indicated samples for Figure 4C.

https://cdn.elifesciences.org/articles/73223/elife-73223-fig4-data2-v2.zip
Figure 4—source data 3

Immunoblots were used to detect expression of tubulin in the indicated samples for Figure 4C.

https://cdn.elifesciences.org/articles/73223/elife-73223-fig4-data3-v2.zip
Figure 4—source data 4

The area highlighted was used to develop the figure for tubulin in the indicated samples for Figure 4C.

https://cdn.elifesciences.org/articles/73223/elife-73223-fig4-data4-v2.zip
Figure 4—source data 5

Immunoblots were used to detect expression of NFIC in the indicated sample for Figure 4C.

https://cdn.elifesciences.org/articles/73223/elife-73223-fig4-data5-v2.tif
Figure 4—source data 6

The area highlighted was used to develop the figure for NFIC in the indicated samples for Figure 4C.

https://cdn.elifesciences.org/articles/73223/elife-73223-fig4-data6-v2.tif
Figure 4—source data 7

Immunoblots were used to detect expression of V5-tagged CREB5 or luciferase in the indicated samples for Figure 4C.

https://cdn.elifesciences.org/articles/73223/elife-73223-fig4-data7-v2.tif
Figure 4—source data 8

The area highlighted was used to develop the figure for V5-tagged CREB5 or luciferase in the indicated samples for Figure 4C.

https://cdn.elifesciences.org/articles/73223/elife-73223-fig4-data8-v2.tif
Figure 4—source data 9

Immunoblots were used to detect expression of FOXA1 and TBX3 in the indicated samples for Figure 4D.

https://cdn.elifesciences.org/articles/73223/elife-73223-fig4-data9-v2.zip
Figure 4—source data 10

The area highlighted was used to develop the figure for FOXA1 and TBX3 in the indicated samples for Figure 4D.

https://cdn.elifesciences.org/articles/73223/elife-73223-fig4-data10-v2.zip
Figure 4—source data 11

Immunoblots were used to detect expression of tubulin in the indicated samples for Figure 4D.

https://cdn.elifesciences.org/articles/73223/elife-73223-fig4-data11-v2.zip
Figure 4—source data 12

The area highlighted was used to develop the figure for tubulin in the indicated samples for Figure 4D.

https://cdn.elifesciences.org/articles/73223/elife-73223-fig4-data12-v2.zip
Figure 4—source data 13

Immunoblots were used to detect expression of NFIC in the indicated samples for Figure 4D.

https://cdn.elifesciences.org/articles/73223/elife-73223-fig4-data13-v2.zip
Figure 4—source data 14

The area highlighted was used to develop the figure for NFIC in the indicated samples for Figure 4D.

https://cdn.elifesciences.org/articles/73223/elife-73223-fig4-data14-v2.zip

To further examine if TBX3 and NFIC were critical in models that exhibit ART resistance, we first performed RNAi-mediated gene depletion experiments in CREB5-overexpressing LNCaP cells in cultures exposed to enzalutamide (Figure 4C). We found that suppression of TBX3 or NFIC using two shRNAs, as compared to a negative control (GFP), reduced the viability of CREB5-overexpressing LNCaP cells to the same extent as two positive control shRNAs that targeted FOXA1. This observation indicates that TBX3 and NFIC, like FOXA1, are required for optimal viability of CREB5 cells in the presence of enzalutamide. We also performed CRISPR-Cas9-mediated gene depletion experiments in a cell line model that spontaneously acquired enzalutamide resistance (Kregel et al., 2016). In this cell line, we expressed three distinct sgRNAs that depleted FOXA1 or two sgRNAs that ablated TBX3 or NFIC, and we found that the sgRNAs against FOXA1, TBX3, or NFIC all reduced protein levels and decreased viability (Figure 4D). We next examined if TBX3 or NFIC interacted with the key CREB5 co-factor FOXA1 in ChIP-seq experiments performed as part of the Encode project (Consortium, 2012). Based on motif enrichment analyses of the TBX3 or NFIC binding sites, we observed statistically significant enrichment of FOXA1 binding motifs identified in experiments on breast or prostate cancer cell lines (Figure 4E). Like CREB5, NFIC also bound B-Zip motifs.

Given their role in regulating FOXA1 functions, we further examined the landscape of TBX3 and NFIC dysregulation in prostate cancer studies using whole-exome sequencing and whole transcriptome sequencing data from cBioPortal (Cerami et al., 2012; Gao et al., 2013) based on the 209 mCRPC samples collected by Stand Up 2 Cancer/Prostate Cancer Foundation (SU2C/PCF) (Abida et al., 2019). We found that TBX3 and NFIC genomic amplifications together represented up to 13.3% of prostate cancers, in which notably higher amplification rates were observed in mCRPC studies as opposed to those sampling primary prostate cancer (Figure 5A). Upon examining expression and amplification data mCRPC samples (Abida et al., 2019), we found that TBX3, NFIC, and FOXA1 were amplified or overexpressed in 7, 3, and 7% of mCRPC samples, respectively (Figure 5B). Of the 209 samples with expression data, we found a robust positive correlation between FOXA1 with TBX3 (Spearman’s correlation: 0.416; q-value: 1.48e-7) or NFIC (Spearman’s correlation: 0.536; q-value: 4.34e-15) (Figure 5C). Together, we found TBX3 and NFIC were nuclear proteins associated with CREB5, FOXA1, and functionally impacted prostate cancer cell viability and ART resistance even absent of CREB5. In addition, they interacted specifically at FOXA1 motifs in cell lines, are amplified or overexpressed in prostate cancer patients, and their expression is associated with that of FOXA1 in mCRPC.

Figure 5
Download asset Open asset
TBX3 and NFIC are amplified in prostate cancer cells.

(A) The genomic amplification rates of TBX3 and NFIC are examined in various prostate cancer studies. (B) In one metastatic castration-resistant prostate cancer (mCRPC) study, the rates of TBX3, NFIC, and FOXA1 gains are depicted. (C and D). The expression of TBX and NFIC are compared in one mCRPC study in which the regression line, Spearman’s correlation coefficients, and q-values are depicted.

CREB5 regulates ART-resistant pathways in cell lines and mCRPC patients

To determine signaling functions of CREB5, we determined transcriptome expression patterns that were statistically associated with CREB5 expression in both cell lines and in mCRPC patients (Figure 6A). Of the proteins that commonly interacted with CREB5, we utilized Enrichr (Chen et al., 2013; Kuleshov et al., 2016) to perform pathway enrichment analysis on the 335 proteins that bound both CREB5 and FOXA1 in our RIME profiling analysis. Based on the top 10 statistically significant signatures, we found that these 335 interactions associated with AR, cell cycle, as well as Notch, Wnt, and SMAD/TGFβ pathways (Figure 6B).

Figure 6
Download asset Open asset
Integrative analysis of CREB5 activity.

(A) A workflow of the informatics analysis of CREB5 using in vitro and metastatic castration-resistant prostate cancer (mCRPC) data. (B) Spectrum of shared CREB5 and FOXA1 protein interactions identified by rapid immunoprecipitation and mass spectroscopy of endogenous proteins (RIME) are analyzed. The enriched pathways and statistical significance are presented for specific pathways. (C) Gene Set Enrichment Analysis (GSEA) analysis of RNA-seq data from CREB5 or luciferase overexpressing androgen receptor (AR)-positive (LNCaP and LAPC-4) and AR-negative (PC3, DU145) prostate cancer cells. (D) Based on RNA-seq from clinical mCRPC, Spearman’s correlation coefficients compare CREB5 expression with EMT, betaβ-catenin, and G2/M signaling. Correlation coefficient values (Rho, σ, x-axis) for CREB5 against each gene, as represented by a single dot, and the statistical significance (negative log of p-value, y-axis) areis displayed. Pp-vValue is marked (red dotted line).

To evaluate these findings, we collected mRNA from luciferase or CREB5-overexpressing prostate cancer cells, including the AR-dependent LNCaP and LAPC-4 cells, as well as the AR-negative PC3 and DU145 cells. We then used GSEA to identify pathways associated with CREB5 these cell lines (Figure 6C). We found that CREB5 overexpression was significantly associated with enrichment of signaling pathways EMT and β-catenin, the Wnt transcription effector in the AR-positive LNCaP and LAPC4 cells, but not in the AR-negative cell lines PC3 and DU145. In clinical mCRPC samples, we analyzed RNA-sequencing datasets from the SU2C/PCF mCRPC cohort (n = 209) (Abida et al., 2019). We computed the Spearman’s correlation coefficient for CREB5 expression against each transcriptional target gene in the EMT, β-catenin, and cell cycle signatures. In mCRPC, we found that CREB5 expression was correlated with the expression of similar signaling programs we found in vitro, including EMT (Alumkal et al., 2020; He et al., 2021) and β-catenin (Isaacsson Velho et al., 2020; Lee et al., 2015; Murillo-Garzón and Kypta, 2017; Figure 6D). Together, the CREB5-associated nuclear protein interactions in cells and transcripts in mCRPC provide insights into the specific ART-resistant pathways that are activated by this transcription factor, showing the role CREB5 plays in regulating EMT and β-catenin signaling genes.

Discussion

We and others have demonstrated that transcription regulators, such as FOXA1, promote prostate cancer progression and resistance to ART and ADT. Our work identified other molecular events associated with transition towards resistance to second-generation ART and ADT. We did so by identifying reprogramming events associated with overexpressed CREB5, which exhibited shifts in binding at transcription regulatory elements as well as interaction with co-factors when cells were challenged with enzalutamide. In promoting proliferation in ART, CREB5 exhibited a strong degree of interaction with known AR transcription machinery, including FOXA1, HOXB13, and GRHL2, as well as novel prostate cancer regulators TBX3 and NFIC. The robust convergence of CREB5-FOXA1 function was observed through binding of transcription regulatory elements and interactions among nuclear proteins. The factors TBX3 and NFIC interacted with CREB5 but required an intact B/L-Zip domain. TBX3 and NFIC, two nuclear factors that are amplified or overexpressed in mCRPC, were vulnerabilities in other prostate cancer models, including ones that were ART resistant, and bound to FOXA1 transcription regulatory elements. Informatics modeling of the CREB5 activity through protein interactions and mCRPC transcription patterns indicated that CREB5 is associated with pathways found in patients resistant to enzalutamide (Alumkal et al., 2020; He et al., 2021). Altogether, our study indicates that the dynamic binding properties of CREB5 mediate assembly of essential factors to AR and FOXA1 to promote resistant transcripts (Figure 7).

Figure 7
Download asset Open asset
A molecular model of the CREB5 complex and transcription promoting androgen receptor-targeted therapy (ART) resistance.

FOXA1 functions as an oncogenic pioneering and transcription factor in cancers, including prostate and breast (Gerhardt et al., 2012; Nakshatri and Badve, 2009; Shah and Brown, 2019). FOXA1 is a critical dependency in prostate cancer cell line models (Pomerantz et al., 2015), including those that transdifferentiate into AR-agnostic, neuroendocrine-like features (Baca et al., 2021). Prior studies have broadly examined the binding patterns of FOXA1 with transcription regulatory elements in normal prostate tissue, primary prostate cancer, mCRPC, and neuroendocrine prostate cancers (Baca et al., 2021; Pomerantz et al., 2015; Pomerantz et al., 2020), demonstrating transcriptional programs associated with prostate in development, tumor progression, and drug resistance. Although EMT is associated with neuroendocrine prostate cancers (Beltran et al., 2016), AR-expressing mCRPCs that become enzalutamide resistance also show expression of EMT markers (Alumkal et al., 2020; He et al., 2021). Together, the work presented herein demonstrates that the interactions of CREB5 and FOXA1 act as one mechanism that promotes EMT signaling in AR-positive-resistant cells.

We also found that TBX3 and NFIC interacted and correlated with FOXA1 in the setting of ART resistance based on RIME experiments and gene expression profile analyses of mCRPC samples. Independent of CREB5 and FOXA1, TBX3 expression has been shown to be required for viability of breast cancer cells (Amir et al., 2016; Krstic et al., 2016). In addition, NFI factors, NFIA/B/C/X, have previously been shown to interact with FOXA1 to promote the transcription of AR target genes (Grabowska et al., 2014). Future studies that examine FOXA1 interactions in parallel through ChIP-seq and RIME are necessary to elucidate context-specific functions. In particular, it would be useful to characterize the universal subset of FOXA1 interactions in prostate cancer tumorigenesis as well as molecular changes associated with mutated forms of FOXA1 (Adams et al., 2019; Parolia et al., 2019).

Studies have identified increases in CREB5 as a marker for metastasis in ovarian, breast, and colorectal cancers (Bhardwaj et al., 2017; He et al., 2017; Molnár et al., 2018; Qi and Ding, 2014). Functionally, Bardwaj et al. have identified CREB5 transcripts as a repressed target of miRNA-29c, a tumor-suppressive miRNA lost in the triple-negative subtype of breast cancer (Bhardwaj et al., 2017). Forced overexpression of CREB5 promoted cell cycle and colony formation in this study. Altogether, these studies implicate pro-tumor CREB5 functions in cancers. While other CREB family members (Welti et al., 2021) have been associated with therapy resistance in advanced prostate cancer, the differential RIME interaction profiles displayed by these two family member proteins, CREB5 and CREB3, exhibited dichotomous behavior with respect to binding of nuclear proteins in cancer cells. This suggests that the oncogenic roles of CREB5 and other CREB family members may be distinct and mediated through their structurally different N-terminal domains. As another key observation, we also find that experimental conditions, including cell culture and drug treatment, dramatically influenced CREB5 molecular interactions.

Our findings also support that transcription regulators may act as effective therapeutic targets in mCRPC. As examples, TBX3 and NFCI have been previously detected in large-scale proteomic approaches that interrogated prostate cancer tissue (Sinha et al., 2019). This study demonstrates that they have key regulatory roles in prostate cancer cell viability and ART resistance. Antagonizing nuclear or transcription factors has been efficacious in recent examples, as inhibitors against EP300, an AR interacting protein, were efficacious in prostate cancer models (Jin et al., 2017; Lasko et al., 2017; Welti et al., 2021). In regulating a pro-tumorigenic role in mCRPC, CREB5 require additional factors (Supplementary file 1, Tables 2–4), including JUN/ATF and SWI/SNF complex. While we have previously discussed that CREB5 functions differ from other CREB or ATF family members (Hwang et al., 2019), the context of how CREB5 interacts with CREB or ATF factors, such as potential hetero-dimerization at open regions of chromatin, still requires further resolution through biochemical approaches. Of bound SWI/SNF family members, we have previously discussed SMARCB1 as recurrent mutation in mCRPC (Armenia et al., 2018), and its loss as a biomarker in malignant pediatric tumors (Hong et al., 2019; Howard et al., 2019). How CREB5 interacts with other chromatin remodeling complexes in ART resistance is an additional research direction that could improve our understanding of transcription processes that could act as targets in therapy-resistant mCRPC.

In summary, our observations implicate CREB5 as a driver of mCRPC. At the molecular level, our findings depict a complex model of therapy resistance that occurs in the nucleus of tumor cells that permits the activation of oncogenic signaling pathways. Furthering the understanding of these underlying changes may inform of additional research avenues and precision strategies for advanced cancer patients that depend on CREB5.

Materials and methods

Genome-scale ORF screen analysis

Request a detailed protocol

We analyzed a published genome-scale ORF screen performed in LNCaP cells (Hwang et al., 2019). Specifically, we compared the experimental arms conducted in control media (FCS) with androgens and androgen stripped media (CSS) containing enzalutamide. Z-scores represent the relative effects of each ORF on cell proliferation after 25 days in culture.

Rapid immunoprecipitation mass spectrometry of endogenous proteins (RIME)

Request a detailed protocol

Upon preparation of cells in each experimental arm, two replicates of 50 million cells each were fixed in 1% formaldehyde for 10 min. Reactions were terminated in 0.125 M glycine for 5 min. Cells were subsequently collected and lysed at 4°C in RIPA buffer (Cell Signaling Technology, 13202S) containing protease inhibitor cocktail (Roche, 11836145001). Cells were then sonicated with a Covaris sonicator to yield DNA fragments averaging around 3000 nucleotides. To target V5-tagged CREB5, CREB5 L434P, or CREB5, 20 μL of V5 antibody (Cell Signaling Technology, 58613) was added to the supernatant. To target FOXA1, each sample was incubated with 20 μL of two FOXA1 antibodies against distinct epitopes (Abcam, ab23738, and Cell Signaling Technology, 58613). Samples were incubated with gentle mixing at 4°C overnight. The following morning, RIPA buffer was used to wash Protein A magnetic beads (Life Technologies, 88846) five times, and the beads were then resuspended into the original volume. 50 μL of the bead mixture was added to each sample and incubated for 2 hr at 4°C. Each sample was then washed five times with 300 μL of RIPA buffer, followed by five times with 300 μL of ammonium bicarbonate (50 μM), and finally resuspended in 50 μL of ammonium bicarbonate (50 μM). These samples on the beads were then sent for proteomic analysis at the Taplin Mass Spectrometry Facility at Harvard Medical School. Upon obtaining mapped reads, only unique peptides of proteins were considered for subsequent analysis. Beads were subjected to trypsin digestion procedure (Shevchenko et al., 1996), then washed and dehydrated with acetonitrile for 10 min followed by removal of acetonitrile. The beads were then completely dried in a speed-vac. Rehydration of the beads was with 50 mM ammonium bicarbonate solution containing 12.5 ng/µL modified sequencing-grade trypsin (Promega, Madison, WI) at 4°C. After 45 min, the excess trypsin solution was removed and replaced with 50 mM ammonium bicarbonate solution to just cover the gel pieces. Samples were then placed in a 37°C room overnight. Peptides were later extracted by removing the ammonium bicarbonate solution, followed by one wash with a solution containing 50% acetonitrile and 1% formic acid. The extracts were then dried in a speed-vac (~1 hr). The samples were then stored at 4°C until analysis. On the day of analysis, the samples were reconstituted in 5–10 µL of HPLC solvent A (2.5% acetonitrile, 0.1% formic acid). A nano-scale reverse-phase HPLC capillary column was created by packing 2.6 µm C18 spherical silica beads into a fused silica capillary (100µm inner diameter × ~30cm length) with a flame-drawn tip (Peng and Gygi, 2001). After equilibrating the column, each sample was loaded via a Famos auto sampler (LC Packings, San Francisco, CA) onto the column. A gradient was formed and peptides were eluted with increasing concentrations of solvent B (97.5% acetonitrile, 0.1% formic acid). As peptides eluted, they were subjected to electrospray ionization and then entered into an LTQ Orbitrap Velos Pro ion-trap mass spectrometer (Thermo Fisher Scientific, Waltham, MA). Peptides were detected, isolated, and fragmented to produce a tandem mass spectrum of specific fragment ions for each peptide. Peptide sequences (and hence protein identity) were determined by matching protein databases with the acquired fragmentation pattern by the software program, Sequest (Thermo Fisher Scientific) (Eng et al., 1994). All databases include a reversed version of all the sequences, and the data was filtered to between a 1 and 2% peptide false discovery rate.

Population doubling

Request a detailed protocol

100~200k cells were plated in 12-well plates in either control media (FCS) or low androgen media (CSS) with 10 μM of enzalutamide. Cell counts and relative cell viability were determined after 7 days using a Vi-Cell. Original cell counts were subtracted before doubling was computed.

Generation of CREB5 point mutants

Request a detailed protocol

A pDNR221 CREB5 plasmid was used as a backbone for point mutagenesis. Per the five point mutants, forward and reverse primers were designed to mutagenize nucleotides of CREB5 to reflect the corresponding changes in protein sequence. Primers were designed with primerX (http://www.bioinformatics.org/primerx). Detailed reaction conditions were followed according to a QuikChange II Site-Directed Mutagenesis Kit (Agilent, 2005230). The mutant clones were each sequenced to confirm their identities before subsequent use in recombination reactions. LR Clonase II (Invitrogen, 11791-020) was used to catalyze the recombination reactions to insert each mutated CREB5 ORF into a puromycin-resistant lentiviral pLX307 vector. Positive clones were then sequenced to confirm the identity of the resulting mutant vectors used for further experimentation.

CRISPR-Cas9 experiments

Request a detailed protocol

To generates sgRNAs, oligos were cloned into a pXPR_003 vector as previously cited (Hwang et al., 2019). Blasticidin-resistant Cas9-positive LNCaP enzalutamide-resistant cells were cultured in 10 μg/mL blasticidin (Thermo Fisher Scientific, NC9016621) for 72 hr to select for cells with optimal Cas9 activity. One million cells were seeded in parallel in 6-well plates and infected with lentiviruses expressing puromycin-resistant sgRNAs targeting FOXA1, TBX3, NFIC, or GFP control. After 48 hr, cells were counted and seeded, using a Vi-Cell, in FCS media with 20 μM enzalutamide at a density of 20,000 cells per well in 6-well plate for a proliferation assay. After 24 hr, cells were subjected to puromycin selection for 3 days. 7 days later, cells were counted again with a Vi-Cell to assess viability, representing a total of 12 days. The target sequences against GFP were AGCTGGACGGCGACGTAAA (sgGFP1) and GCCACAAGTTCAGC GTGTCG (sgGFP2). The target sequences against FOXA1 were GTTGGACGGC GCGTACGCCA (sgFOXA1-1), GTAGTAGCTGTTCCAGTCGC (sgFOXA1-2), and ACTGCGCCCCCCATA AGCTC (sgFOXA1-4). The target sequences against TBX3 were GAAAAGGTGAGCCTTGACCG (sgTBX3-1) and GCTCTTACAATGTGGAACCG (sgTBX3-2). The target sequences against NFIC were ACGGCCACGCCAATGTGGTG (sgNFIC-1) and GCTGAGCATCACCGGCAAGA (sgNFIC-2).

shRNA experiments

Request a detailed protocol

Lentiviruses expressing shRNAs for AR, FOXA1, TBX3, NFIC, or GFP were used to infect LNCaP cells grown in FCS media. Between 48 and 72 hr post infection, protein lysates were collected to determine the extent of suppression. After confirming suppression, respective cells were counted and directly seeded for proliferation experiments. All shRNA constructs were acquired from the Broad Institute Genetic Perturbation Platform (https://portals.broadinstitute.org/gpp/public/). The target sequence against GFP was ACAACAGCCACAACGTCTATA (sgGFP). The target sequences against AR were GAGCGTGGACTTTCCGGAAAT (shAR1), GATGTCTTCTGCCTGTTATAA (shAR2), and CGCGACTACTACAACTTTCCA (shAR3). The target sequences against FOXA1 were TCTAGTTTGTGGAGGGTTATT (shFOXA1-1) and GCGTACTACCAAGGTGTGTAT (shFOXA1-2). The target sequences against TBX3 were GCATACCAGAATGATAAGATA (shTBX3-2) and GCTGCTGATGACTGTCGTTAT (shTBX3-3). The target sequences against NFIC were GATGGACAAGTCACCATTCAA (shNFIC-2) and CCCGGTGAAGAAGACAGAGAT (shNFIC-4).

RNA-seq experiments

Request a detailed protocol

For RNA-seq experiments, AR-positive (LNCaP, LAPC-4) cells and AR-negative (PC3, DU145) cells expressing either luciferase or CREB5 were cultured in FCS media. For RNA-seq experiments, library preparations, quality control, and sequencing on a HiSeq2500 (Illumina) were performed and analyzed by the Dana-Farber Molecular Biology core facility based on prior studies (Hwang et al., 2019).

Immunoblotting

Request a detailed protocol

Cells were lysed using 2× sample buffer (62.5 mM Tris pH 6.8, 2% SDS, 10% glycerol, Coomassie dye) and freshly added 4% β-mercaptoethanol. Lysed cells were scraped, transferred into a 1.5 mL microcentrifuge tube, sonicated for 15 s, and boiled at 95°C for 10 min. Proteins were resolved in NuPAGE 4–12% Bis-Tris Protein gels (Thermo Fisher Scientific) and run with NuPAGE MOPS SDS Running Buffer (Thermo Fisher Scientific, NP0001). Proteins were transferred to nitrocellulose membranes using an iBlot apparatus (Thermo Fisher Scientific). Membranes were blocked in Odyssey Blocking Buffer (LI-COR Biosciences, 927-70010) for 1 hr at room temperature, and membranes were then cut and incubated in primary antibodies diluted in Odyssey Blocking Buffer at 4°C overnight. The following morning, membranes were washed with phosphate-buffered saline, 0.1% Tween (PBST) and incubated with fluorescent anti-rabbit or anti-mouse secondary antibodies at a dilution of 1:5000 (Thermo Fisher Scientific, NC9401842 [rabbit] and NC0046410 [mouse]) for 1 hr at room temperature. Membranes were again washed with PBST and then imaged using an Odyssey Imaging System (LI-COR Biosciences). Primary antibodies used include V5 (Cell Signaling Technology, 13202S), β-actin (Cell Signaling Technology, 8457L), FOXA1 (Cell Signaling, 58613), TBX3 (Life Technologies, 424800), NFIC (Abcam, ab245597), and tubulin (Cell Signaling, 3873S).

Overlap analysis of CREB5 binding sites

Request a detailed protocol

Bed files containing peak summit locations determined by MACS2 from CREB5 ChIP-seq data were intersected with the indicated datasets using BEDtools v2.27.1 (Quinlan and Hall, 2010). To assess overlap with FOXA1 binding sites, FOXA1 ChIP-seq datasets from 23 prostate adenocarcinoma patient-derived xenografts (Nguyen et al., 2017) and tissue (Pomerantz et al., 2015; Pomerantz et al., 2020) were merged to create a FOXA1 union peak set. To assess overlap of CREB5 binding and various sets of AR binding sites, we first created a union set of CREB5 peaks that were present with or without enzalutamide treatment. This CREB5 peak set was intersected with the indicated AR peaks sets (Figure 2B) from Pomerantz et al., 2015 and Pomerantz et al., 2020. The percentages of each class of AR+CREB5+peaks were assessed for overlap with the union set of FOXA1 peaks in Figure 2C.

Project Achilles 2.20.1 analysis

Request a detailed protocol

Of the total of 503 cell lines, we analyzed a published genome-scale RNAi screen of eight prostate cancer cell lines (Cowley et al., 2014; Shalem et al., 2014; Tsherniak et al., 2017) whereby we averaged the dependency for each gene. Cell lines included NCIH660 (NEPC-like), PC3 and DU145 (AR negative), 22RV1 (expressing an AR V7 splice variant), LNCaP, VCaP, and MDAPCA2B (AR positive and dependent), and PRECLH (normal immortalized prostate epithelium).

Motif enrichment analysis

Request a detailed protocol

Known motifs enriched in TBX3 and NFIC ChIP-seq data from HepG2 cells (GEO: GSM2825557, GSM2902642) compared to a whole-genome background were identified with Homer version 4.17 (Heinz et al., 2010). Selected examples from the most significantly enriched known motifs are shown (Figure 4D).

Expression association analysis of CREB5 in mCRPC

Request a detailed protocol

We analyzed RNA-sequencing data from an updated combined cohort of men with mCRPC from multiple institutions comprising the SU2C/PCF Prostate Cancer Dream Team (Abida et al., 2019). RNA-seq data, normalized in units of transcripts per million (TPM), was available from 208 patients. Expression data was previously examined and adjusted for batch effects using ComBat (Johnson et al., 2007) via the R Bioconductor package ‘sva’ (Leek et al., 2012), V3.22.0. The Spearman correlations were determined for CREB5 against all detectable transcripts in these samples. This profile of association was further examined with focus on CREB5 and its association with genes in the indicated pathways using predefined MSigDB signatures.

Cell lines and authentication

Request a detailed protocol

The cell lines used in this study were directly ordered from American Type Culture Collection (ATCC) or identities have been confirmed through their STR profiling analyses. None of the cell lines we have used are frequently misidentified by standards of the International Cell Line Authentication Committee. Micoplasma contamination was routinely tested using MycoAlert (LT07-118, Lonza).

Data availability

RIME data has been shared through supplementary tables in Supplementary File 1.

The following previously published data sets were used
    1. Ji-Heui S
    2. Xintao Q
    (2019) NCBI Gene Expression Omnibus
    ID GSE137775. CREB5 promotes resistance to androgen-receptor antagonists and androgen deprivation in prostate cancer.
    https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE137775

References

    1. Nomura N
    2. Zu YL
    3. Maekawa T
    4. Tabata S
    5. Akiyama T
    6. Ishii S
    (1993)
    Isolation and characterization of a novel member of the gene family encoding the cAMP response element-binding protein CRE-BP1
    The Journal of Biological Chemistry 268:4259–4266.

Decision letter

  1. Charles L Sawyers
    Reviewing Editor; Memorial Sloan Kettering Cancer Center, United States
  2. Kathryn Song Eng Cheah
    Senior Editor; University of Hong Kong, Hong Kong

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Decision letter after peer review:

Thank you for sending your article entitled "CREB5 reprograms nuclear interactions to promote resistance to androgen receptor targeting therapies" for peer review at eLife. Your article is being evaluated by 2 peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation is being overseen by Kathryn Cheah as the Senior Editor.

While both reviewers appreciate the work that went into defining the CREB5 cistrome and interactome, you will see two major concerns have been raised.

1. Reviewer 1 wants to understand whether CREB5 confers ART resistance by working through AR, by restoring/reshaping its function, versus a lineage plasticity model as suggested by enrichment of EMT/TGF-β signatures. This would require further experiments in CREB5 wild-type vs overexpression models and ideally other prostate models beyond LNCaP to address questions of generalizability.

2. Reviewer 2 raises concerns about the underlying premise of CREB5/AR biology based on literature data showing that CREB5 is an AR-repressed gene.

Reviewer #1:

Previously, these authors identified CREB5 overexpression as a mechanism that drives enzalutamide resistance, which they reported in eLife. In this paper, the authors use ChIP-seq and RIME to further characterize CREB5 in this context. The main findings are co-association/overlap of motifs and protein interactions with AR and FOXA1. Through mutagenesis of the known FOS/JUN interaction motif in CREB5 followed by addition RIME, the authors identify NFIC and TBX3 as additional interaction partners associated with enzalutamide resistance. Analysis of Achilles data and an enzalutamide-resistant LNCaP subline suggest dependencies in prostate cancer lines, similar to those seen with FOXA1. This work further develops the previously published CREB5 story and provides useful datasets on CREB5 binding sites and protein interaction partners, but the overall novelty and impact are limited.

Specific comments

1. The enrichment for signatures of EMT and TGF-β (Figure 5) is puzzling since lineage shifts of this type (away from luminal) is generally associated with loss of AR dependency. Yet the ChIP-seq and interactome data point to AR, FOXA1, etc as the primary drivers of CREB5 biology. How do the authors reconcile these findings? Does enzalutamide resistance mediated by CREB5 still require AR? (i.e., what happens with genetic ablation of AR?)

2. Figure 4B shows that both TBX3 and NFIC are required for cell viability in the LNCaP cells that spontaneously developed resistance to enzalutamide. This is the major novelty in the work but is only addressed in this one experiment. At a minimum, the authors should test the function of TBX3 and NFIC in a CREB5 overexpressing cell line because that is the biological context in which the screens were done. As with their earlier paper, it would also be more persuasive to conduct these experiments in vivo (e.g. using xenograft models) to show that TBX3 and NFIC are important regulators for castration + enzalutamide resistance. The in vivo experiment would also be a better context to address the potential lineage changes such as EMT (see comment #1).

3. Figure 4A plots the dependency results in prostate cancer cell lines from Project Achilles, noting that TBX3 and NFIC both score. Are all of these lines AR-positive? What about widely used AR-negative lines like PC3 and DU145? Also, how do TBX3 and NFIC score in a pan cancer analysis? And what are the results from DepMap?

4. In figure 3D, LDB1 and DNMT1 are another two proteins that have reduced binding with CREB5 mutants comparing with wildtype CREB5. The authors ignored these two proteins and focused on NFIC and TBX3. Why not examine all four? Do LDB1 and DNMT1 also play a role in ART resistance?

5. For the CREB5 binding proteins identified using RIME, especially the ART reprogrammed protein-protein interactions, the results would be more convincing if the authors could provide a few validations using co-immunoprecipitation (co-IP).

6. In my opinion, the title "CREB5 reprograms nuclear interactions to promote resistance to androgen receptor targeting therapies" seems a bit misleading. The experiments and analysis focus on the effects of CREB5 in the presence of ART, which is already known to have a major impact on protein interactions and chromatin landscape.

The work needs to clarify whether the effects of CREB5 overexpression and the dependencies on TBX3 and NIFC are restricted to AR-positive/AR-dependent models.

Reviewer #2:

This reviewer acknowledges the tremendous effort that went into producing high-quality OMIC data. The authors were likely unaware that CREB5 was an AR target gene and if they confirm it to be so in their model system they will have to reformulate their hypothesis and completely rewrite the paper. Further, there are no data presented in this paper that shows that CREB5 is causally involved in processes that confer resistance to enzalutamide. They need to validate he results of the OMICs approaches with such studies.

1. The apparent goal of the ORF screen (described in figure 1) was to identify factors whose involvement in PCa pathobiology was dependent on ENZ. However, a quick review of several published RNAseq datasets indicates that CREB5 is highly downregulated in PCa cells (normal and malignant) treated with R1881, Enz or Bicalutamide (its also downregulated by agonist activated progesterone receptor in several systems). Have the authors considered that they may have identified a gene (CREB5) that is AR repressed that is important for AR action and that the protein is not specifically related to ENZ/ADT resistance but is required for AR action? The potential involvement of CREB in AR action more broadly needs to be probed, but this reviewer suspects that when explored further the authors will find that the protein is not involved in resistance-dependent reprogramming of the AR cistrome per se.

2. Given that CREB5 is highly downregulated in cells treated with R1881, Enz or Bicalutamide one interpretation of the authors findings is that overexpression of CREB5 bypasses this regulatory pathway but that should impact the activity of both agonists and antagonists. Thus, the authors need to consider that they may have identified a gene (CREB5) that is AR repressed that its reduced expression is important for AR action and that dysregulation of the expression of this protein is not specifically related to ENZ/ADT resistance? To support their specific hypothesis the authors would have to show that CREB5 overexpression has no effect on R1881 dependent transcription. Otherwise the focus of the paper must change completely to consider a more "physiological" role for CREB5 in AR action.

3. The appearance of FOXA1 GRHL2, FOXA1 et at CREB5 binding sites is interesting but the requirement for AR in these studies is not demonstrated.

4. In designing the screen the assumption is made that CSS is just FBS "without androgens" which is not the case. FBS +/- enz would seem to be a more relevant model to look for important mediators of resistance (enz is not very active in this scenario but that is another story!). Further, if androgens suppress CREB5 (as has been demonstrated) and this is required for proliferation then overexpression of CREB5 would bypass this regulation and thus would explain the results observed in FBS (inhibition of proliferation).

5. If CREB5 downregulation is required for normal AR function then it is hard to explain the dependencies highlighted in DEPMAP. Unless its overexpression prevents the repression/downregulation of proliferation that occurs in PCa cancer cells as androgen levels rise.

https://doi.org/10.7554/eLife.73223.sa1

Author response

Reviewer #1:

Previously, these authors identified CREB5 overexpression as a mechanism that drives enzalutamide resistance, which they reported in eLife. In this paper, the authors use ChIP-seq and RIME to further characterize CREB5 in this context. The main findings are co-association/overlap of motifs and protein interactions with AR and FOXA1. Through mutagenesis of the known FOS/JUN interaction motif in CREB5 followed by addition RIME, the authors identify NFIC and TBX3 as additional interaction partners associated with enzalutamide resistance. Analysis of Achilles data and an enzalutamide-resistant LNCaP subline suggest dependencies in prostate cancer lines, similar to those seen with FOXA1. This work further develops the previously published CREB5 story and provides useful datasets on CREB5 binding sites and protein interaction partners, but the overall novelty and impact are limited.

Specific comments

1. The enrichment for signatures of EMT and TGF-β (Figure 5) is puzzling since lineage shifts of this type (away from luminal) is generally associated with loss of AR dependency. Yet the ChIP-seq and interactome data point to AR, FOXA1, etc as the primary drivers of CREB5 biology. How do the authors reconcile these findings? Does enzalutamide resistance mediated by CREB5 still require AR? (i.e., what happens with genetic ablation of AR?)

We agree with the Reviewer that initial studies suggested that EMT signaling or other lineage plastic pathways occurred after adenocarcinomas differentiated into AR independent states, including neuroendocrine prostate cancer (NEPC) (Beltran et al., Nat Med, 2016). However, genes implicated as part of the EMT pathway are also associated with numerous cancer processes other than plasticity observed in NEPC. Recent single-cell RNA sequencing studies (He et al., Nat Med, 2021) reported that increased EMT signaling co-occurs with increased levels of AR or various splice variants in biopsy-matched treatment-resistant adenocarcinoma samples. Upon further analysis of data from this study, we present the supporting evidence as Author response image 1. These AR or AR spice variant expressing tumor cells did not exhibit a NEPC histology. In another independent study, several mCRPCs that developed enzalutamide resistance exhibited increased EMT signaling (Alumkal et al., Proc Natl Acad Sci U S A, 2020). These findings indicate that enzalutamide-resistant adenocarcinomas harbor a dependency on AR functions while also co-expressing markers consistent with EMT signaling, suggesting that enzalutamide resistance, loss of AR expression and EMT do not occur in a linear manner.

Author response image 1
Download asset Open asset
A.

Figure adapted from He et al. (He et al., Nat Med, 2021). Single cells from a mCRPC patient pre- and post-treatment were examined based on single cell RNA-seq approach and the most enriched pathway was EMT after enzalutamide treatment. B. AR expression levels were examined in the sample tumor cells, which increased with statistical significance (Student’s t-test).

In examining CREB5 and its regulation of EMT signaling, we have previously conducted RNA-seq on LNCaP cells that overexpress control (luciferase) or CREB5. As part of this revision, we performed additional experiments to overexpress CREB5 in several additional cell models including the AR-dependent LAPC-4 cells and AR-negative PC3 and DU145 (discussed in Results, p. 7). We previously showed that CREB5 overexpression promoted enzalutamide resistance in the AR-positive LNCaP and LAPC-4 cells (Hwang et al., Cell Rep, 2019). Here we conducted RNA-seq on LAPC-4, PC3 and DU145 cells that overexpress a control vector or CREB5. We then took an unbiased GSEA approach to identify pathways that are associated with CREB5 overexpression in the two AR-positive cell lines and separately in 209 mCRPC samples. We found that CREB5 overexpression was significantly associated with enrichment of EMT signaling in the AR-positive LNCaP and LAPC4 cells, but not in the AR-negative cell lines PC3 and DU145 (updated Figure 6C, discussed in Results, p. 7). Furthermore, CREB5 expression correlated with EMT in the mCRPC samples (Figure 6D, discussed in Results, p. 7). In these analyses, TGF-β expression was not statistically significant in the AR-positive cell lines or mCRPC samples. Based on these new findings we have removed the references to TGF-Β signaling from this manuscript (Figure 6C, 6D). Collectively, these findings indicate that EMT occurs even in AR-dependent prostate tumors and that some enzalutamide resistant cancers remain dependent on AR signaling even in the midst of changes in differentiation state.

We found in our prior manuscript (Hwang et al., Cell Rep, 2019) that the genetic ablation of AR or FOXA1 reduced viability of CREB5 overexpressing cells. We also reported that CREB5 regulated AR binding in a subset of AR driven genes. As part of this revision, we have performed a new experiment in which we suppressed AR in CREB5 expressing cells, which confirmed that these cells require AR expression for cell fitness (now Figure 1 —figure supplement 1). Together, these observations implicate AR function as necessary for CREB5 driven prostate cancer cell survival.

The observations in our current manuscript indicate that CREB5 is one mechanism that may regulate a subset of co-factors and transcription targets of EMT while also retaining AR-dependency.

Author response image 1 is provided but not included in the manuscript since it is analysis of published data. We have elaborated these points in the Discussion and the added Supplementary Figure 1 of the revised manuscript (p. 8).

2. Figure 4B shows that both TBX3 and NFIC are required for cell viability in the LNCaP cells that spontaneously developed resistance to enzalutamide. This is the major novelty in the work but is only addressed in this one experiment. At a minimum, the authors should test the function of TBX3 and NFIC in a CREB5 overexpressing cell line because that is the biological context in which the screens were done. As with their earlier paper, it would also be more persuasive to conduct these experiments in vivo (e.g. using xenograft models) to show that TBX3 and NFIC are important regulators for castration + enzalutamide resistance. The in vivo experiment would also be a better context to address the potential lineage changes such as EMT (see comment #1).

We thank the reviewer for this suggestion. We have now performed experiments in which we found that suppression of TBX3 or NFIC using 2 shRNAs, as compared to a negative control (GFP), reduced the viability of CREB5 overexpressing LNCaP cells in enzalutamide cultures to the same level as 2 positive control shRNAs that targeted FOXA1, the CREB5 co-factor. These observations indicate that TBX3 and NFIC, like FOXA1 are required for optimal viability of CREB5 cells in the presence of enzalutamide (now Figure 4C, discussed in Results, p. 6). In our original submission, we demonstrated the role of TBX3 and NFIC as a necessary regulator for cell fitness in one LNCaP cell line that spontaneously acquired resistance to enzalutamide after long term cultures in enzalutamide (Kregel et al., Oncotarget, 2016) (now Figure 4D, discussed in Results, p. 6, p. 7).

In addition, as part of this revision, we have partitioned results based on AR status of the prostate cancer cell lines (also see response to Reviewer 1, comment 3). When examining the relative dependencies of TBX3 and NFIC in the AR-positive cell lines, we found that relative to all genes, TBX3 and NFIC exhibited similar dependencies as FOXA1, a pan prostate cancer cell dependency (Pomerantz et al., Nat Genet, 2015). This observation supports the conclusion that TBX3 and NFIC regulate viability in the setting of AR (Figure 4A, and 4B, discussed in Results, p. 6).

As part of this revision, we also determined that CREB5 expression is associated with increased expression of genes related to EMT in AR-positive cell lines and the subset of adenocarcinoma mCRPC (Figure 6C, 6D, Reviewer 1 comment 1, discussed in Results, p. 7 and Discussion, p. 8). We considered performing the xenograft experiments; however, these experiments would require 8-12 months to perform, we believe that these experiments will not change the interpretation of the revised manuscript.

These new experiments together support the role of TBX3 and NFIC in CREB5-driven models and indicate that CREB5 plays a role in regulating EMT genes in AR-positive cells.

3. Figure 4A plots the dependency results in prostate cancer cell lines from Project Achilles, noting that TBX3 and NFIC both score. Are all of these lines AR-positive? What about widely used AR-negative lines like PC3 and DU145? Also, how do TBX3 and NFIC score in a pan cancer analysis? And what are the results from DepMap?

We agree that the dependency data from Depmap was not partitioned to exhibit the effects of TBX3 and NFIC in each cell line or on AR-positive (LNCaP, 22Rv1, MDAPC2B, VCAP – now Figure 4A, left) or AR-negative (PC3, DU145, NCIH660 – now Figure 4A, right, discussed in Results, p. 6) cell models. In the revised manuscript, we have now examined the overall average dependency of TBX3 and NFIC in all the AR-positive and –negative cell lines as compared to AR and FOXA1, the pan prostate cancer cell dependency (Pomerantz et al., Nat Genet, 2015) (Figure 4A). We also present ranking and percentile of these average dependencies with respect to all genes that were screened (now Figure 4B, discussed in Results, p. 6). We found that in AR-positive cell lines, TBX3 and NFIC exhibited a similar level of dependency as observed for FOXA1. When examining AR-negative cell lines in a similar fashion, we found that TBX3 was not required for cell fitness in these cell lines, while NFIC exhibited modest dependency as compared to FOXA1. These observations support the conclusion that TBX3 and NFIC regulate prostate cancer cell viability in the AR-setting (discussed in the response to Reviewer 1, comment 2).

When we compared the dependency profile of TBX3 and NFIC in 495 non-prostate cancer cell lines, we found that TBX3 and NFIC exhibited a pattern and degree of dependency similar to what we observed for AR and FOXA1 (now Figure 4B, discussed in Results, p. 6). These findings support the notion that TBX3 and NFIC are necessary for the fitness of AR-expressing prostate cancer models.

4. In figure 3D, LDB1 and DNMT1 are another two proteins that have reduced binding with CREB5 mutants comparing with wildtype CREB5. The authors ignored these two proteins and focused on NFIC and TBX3. Why not examine all four? Do LDB1 and DNMT1 also play a role in ART resistance?

In the updated manuscript, we indeed presented genomic, epigenomic, functional dependency, and clinical analyses on TBX3 and NFIC with respect to FOXA1 (Figure 4E, Figure 5A, B, C, discussed in Results, p. 6, p. 7).

As part of this revision, we compared DNMT1 and LDB1 to TBX3 and NFIC (Author response table 1). In RIME experiments targeting CREB5 or FOXA1 in enzalutamide cultures, we found TBX3 and NFIC interacted with CREB5 and FOXA1 strongly while we found less robust interactions with DNMT1 and LDB1 based on unique peptide counts (First two columns of Revision Table 1, also in Supplementary File 1, Table 3). In addition, TBX3 and NFIC expression also showed a stronger correlation with FOXA1 in 209 mCRPC samples (3rd column in Author response table 1). In the functional dependency data (4th to 6th columns in Author response table 1), the average DEMETER scores of TBX3 and NFIC ranked greater than the 99th percentile in AR-positive PC cell lines. In the 495 non-PC cell lines, TBX3 and NFIC were less than the 1st percentile. In contrast, we found that DNMT1 and LDB1 did not exhibit this same pattern of dependency. For these reasons, while DNMT1 and LDB1 also interacted with FOXA1 and CREB5, our subsequent analyses and experiments focused on TBX3 and NFIC. We clarified these findings in the text of the revised manuscript (Results, p. 6, p. 7).

Author response table 1
Results from RIME, transcript association analyses, Dependency analyses are presented for TBX3, NFIC, DNMT1 and LDB1.
Unique Peptide Counts, CREBS RIMEUnique Peptide Counts, FOXA1 RIMEmCRPC Transcript Correlation with FOXA1Dependency in AR-positive PC, PercentileDependency in AR-negative PC, PercentileOverall Dependency, Percentile
TBX33.006.500.5499.7823.270.19
NFIC4.004.000.4296.3885.220.82
DNMT12.000.50-0.0928.2560.2163.38
LBD12.001.500.0714.4265.5685.23

Author response table 1 is not included in the manuscript; however, we have clarified the rationale for studying TBX3 and NFIC in the revised manuscript (Results, p. 6, p. 7).

5. For the CREB5 binding proteins identified using RIME, especially the ART reprogrammed protein-protein interactions, the results would be more convincing if the authors could provide a few validations using co-immunoprecipitation (co-IP).

We agree with the Reviewer. We have attempted co-immunoprecipitation experiments with CREB5 to examine interaction with TBX3 and NFIC. The initial results are promising, in that we could detect a faint band that migrated at the predicted molecular weight for TBX3 and NFIC in the V5-CREB5 immune complexes as compared to an IgG control (Author response image 2, left). However, we concluded that the current commercially available TBX3 and NFIC antibodies are not sufficient to perform co-IPs of sufficient quality to conclude that these interactions are robust.

Author response image 2
Download asset Open asset
Left.

V5-tagged CREB5 was targeted for immunoprecipitation in cell lysates in vehicle control (DMSO) or enzalutamide treated (ENZ) LNCaP cells. TBX3 and NFIC were detected in the total lysate (input) and precipitates (IP) using immunoblots. IgG was used as a negative precipitation control. Right. FOXA1 was targeted in enzalutamide treated cells in a Co-IP experiment. CREB5 and FOXA1 were detected in the total lysate (input) and precipitates (IP) using immunoblots. IgG was used as a negative precipitation control.

To test the other central hypothesis that CREB5 and FOXA1 are co-factors, we conducted co-IP experiments and found that CREB5 interacted with FOXA1 (Author response image 2, right). This finding supports the observation in that CREB5 and FOXA1 interactions with additional co-factors were similar in RIME, and they bound to similar transcription regulatory sites (Figure 2).

We have presented Author response image 2 for the Reviewer’s use but given the question about the quality of the antibodies have not included this in the revised manuscripts.

6. In my opinion, the title "CREB5 reprograms nuclear interactions to promote resistance to androgen receptor targeting therapies" seems a bit misleading. The experiments and analysis focus on the effects of CREB5 in the presence of ART, which is already known to have a major impact on protein interactions and chromatin landscape.

In agreement with the reviewer, we recognize that several paths can lead to reprogramming. To increase the specificity in the title, we highlighted the CREB5 and FOXA1 interactions with the following proposed title for the revised manuscript (in Title, p.1):“

CREB5 reprograms FOXA1 nuclear interactions to promote resistance to androgen receptor targeting therapies"

The work needs to clarify whether the effects of CREB5 overexpression and the dependencies on TBX3 and NIFC are restricted to AR-positive/AR-dependent models.

We have examined this based on experiments proposed in the response to Reviewer 1 comment 2 and comment 3 (Public Reviews).

Reviewer #2:

This reviewer acknowledges the tremendous effort that went into producing high-quality OMIC data. The authors were likely unaware that CREB5 was an AR target gene and if they confirm it to be so in their model system they will have to reformulate their hypothesis and completely rewrite the paper. Further, there are no data presented in this paper that shows that CREB5 is causally involved in processes that confer resistance to enzalutamide. They need to validate he results of the OMICs approaches with such studies.

We thank the reviewer for their review of our manuscript.

We agree that showing that CREB5 drives enzalutamide resistance is critical. Specifically, we previously demonstrated (Hwang et al., Cell Rep, 2019) that CREB5 overexpression directly promoted resistance to enzalutamide in LNCaP cells. We also showed that CREB5 was overexpressed in an enzalutamide resistant, mCRPC patient-derived organoid and that suppression of CREB5 expression induced cell death in a prostate-patient derived organoid model. These findings are in consonance with the clinical observations in which CREB5 is amplified or overexpressed in a subset of mCRPC. Together, these experiments provide strong evidence that CREB5 indeed drives resistance to enzalutamide.

1. The apparent goal of the ORF screen (described in figure 1) was to identify factors whose involvement in PCa pathobiology was dependent on ENZ. However, a quick review of several published RNAseq datasets indicates that CREB5 is highly downregulated in PCa cells (normal and malignant) treated with R1881, Enz or Bicalutamide (its also downregulated by agonist activated progesterone receptor in several systems). Have the authors considered that they may have identified a gene (CREB5) that is AR repressed that is important for AR action and that the protein is not specifically related to ENZ/ADT resistance but is required for AR action? The potential involvement of CREB in AR action more broadly needs to be probed, but this reviewer suspects that when explored further the authors will find that the protein is not involved in resistance-dependent reprogramming of the AR cistrome per se.

We appreciate that the reviewer used different datasets than those that were used in this manuscript to conclude that CREB5 expression may be regulated by AR. We were surprised by this comment and have taken further analyses to further explore our original observations. As detailed below, we have confirmed that AR and CREB5 expression are independent in primary prostate and mCRPC.

Specifically, we previously queried the SU2C/PCF dataset (Abida et al., Proc Natl Acad Sci U S A, 2019) and found that CREB5 expression was independent of AR expression (Pearson correlation = 0.03) in mCRPC (Hwang et al., Cell Rep, 2019). In the same analyses, the AR target genes KLK2 and KLK3 are positively correlated with one another (R=0.82). We have now further evaluated CREB5 expression as a function of AR in both primary prostate cancer and mCRPC in data publicly available on cBioPortal. As summarized in Author response table 2, we found that CREB5 and AR expression are independent of each other in each of these datasets. In comparison, we also analyzed the correlation of AR and FOXA1 expression, which exhibited a strong correlation as expected (Pomerantz et al., Nat Genet, 2015, Pomerantz et al., Nat Genet, 2020). Together, these analyses indicate that CREB5 expression is not primarily regulated by AR in these widely used tumor datasets. However, as Reviewer 2 indicated, we surmise that it remains possible that in specific models and conditions, AR may suppress CREB5 transcripts.

Author response table 2
Transcription data from three independent studies were examined.

Pearson correlations were performed to examine the associations between CREB5 and AR as well as AR and FOXA1. The cohort features are described along with the correlation value and p-value. The p-values were are marked as significant (**) or not significant (n.s.) based on type I error levels at less than 0.05.

CREB5, ARAR, FOXA1
Rp-valRp-val
PCF/SU2C (n=266, mCRPC only) (Abida et al., Proc Natl Acad Sci USA, 2019)-0.110.0612 (n.s.)0.413.70E-12**
TCGA PRAD (n=488, Primary only)0.080.0707 (n.s.)0.372.29E-17**
MSK (n=128, primary and mCRPC) (Taylor et al., Cancer Cell, 2010)0.080.34 (n.s.)0.42.92E-06**

We have provided Author response table 2 for the editors and reviewers to use but have not included this Table in the revised manuscript.

2. Given that CREB5 is highly downregulated in cells treated with R1881, Enz or Bicalutamide one interpretation of the authors findings is that overexpression of CREB5 bypasses this regulatory pathway but that should impact the activity of both agonists and antagonists. Thus, the authors need to consider that they may have identified a gene (CREB5) that is AR repressed that its reduced expression is important for AR action and that dysregulation of the expression of this protein is not specifically related to ENZ/ADT resistance? To support their specific hypothesis the authors would have to show that CREB5 overexpression has no effect on R1881 dependent transcription. Otherwise the focus of the paper must change completely to consider a more "physiological" role for CREB5 in AR action.

We refer the Reviewer to consider the correlation of AR activity and CREB5 expression in mCRPC in the response to Reviewer 2, comment 1. We found that the relationship between CREB5 and AR expression was independent in prostate tumors.

The Reviewer suggests that CREB5 may play a role in responses to both AR agonists and antagonists. In our prior work, we found that CREB5 regulated general AR transcription targets, which would also be regulated by R1881 based on a qRT-PCR panel of AR target genes (Hwang et al., Cell Rep, 2019). However, in this manuscript we showed that CREB5 expression had a modest anti-proliferative effect in ORF screens conducted in cultures with full serum, which represents cultures in which androgens are in excess (Figure 1, discussed in Results, p. 7). In contrast, we found that CREB5 was the top candidate out of 17,255 ORFs when cells were treated with enzalutamide and androgen deprivation. As R1881 is an androgen analog, we predict that CREB5 would not have significant pro-proliferative phenotypes.

Due to the significant phenotype we observed with androgen receptor inhibitors, we focused on the mechanistic characterization of CREB5 in cells and tumor samples that have received some form of AR signaling inhibition. We have also updated the title of this study to reflect this emphasis (Reviewer 1, comment 6).

We also note that we previously reported that CREB5 is genomically amplified with limited deletion events in mCRPC (only one observed), in which samples are derived from patients treated with ART. We re-examined the data as part of this revision and presented the updates here (Author response image 3). These results support that prostate tumors, which are AR-positive, select for increases in CREB5 expression.

Author response image 3 is based on re-analyses of published studies and thus not included in the revised manuscript.

Author response image 3
Download asset Open asset
CREB5 amplifications (red) and deletions (blue) are examined and displayed across prostate cancer cohorts as of 12.2.2021.

In total, there were variable amplification rates and we only detected one homozygous deletion in a primary prostate tumor in the 2018 study that included 680 primary prostate cancer samples and 333 mCRPC. Adapted from cBioportal.

3. The appearance of FOXA1 GRHL2, FOXA1 et at CREB5 binding sites is interesting but the requirement for AR in these studies is not demonstrated.

We have previously reported that suppressing FOXA1 or AR expression by RNAi in CREB5 overexpressing LNCaP cell lines induced cell death (Hwang et al., Cell Rep, 2019). This observation supports the conclusion that CREB5 requires AR and the key co-factor FOXA1 in the CREB5-mediated enzalutamide resistant phenotype. We have also repeated this observation as part of this response (Supplementary Figure 1). To make the title more specific, we have changed the title of the revised manuscript to highlight CREB5 as a co-factor to FOXA1 (Reviewer 1, comment 6, updated in Title, p. 1), which better represents this work.

While GRHL2 is also a key AR co-factor, we believe that the RNAi experiments targeting AR or FOXA1 demonstrate that AR and the key AR co-factor FOXA1 are necessary in CREB5 overexpressing cells.

4. In designing the screen the assumption is made that CSS is just FBS "without androgens" which is not the case. FBS +/- enz would seem to be a more relevant model to look for important mediators of resistance (enz is not very active in this scenario but that is another story!). Further, if androgens suppress CREB5 (as has been demonstrated) and this is required for proliferation then overexpression of CREB5 would bypass this regulation and thus would explain the results observed in FBS (inhibition of proliferation).

We agree with this comment and have clarified how experiments after the ORF screen were conducted with enzalutamide in FBS (discussed in Materials and methods, p. 9).

Materials and methods, Page 9. “We analyzed a published genome-scale ORF screen performed in LNCaP cells (Hwang et al., 2019). Specifically, we compared the experimental arms conducted in control media (FCS) with androgens and androgen stripped media (CSS) containing enzalutamide. Z-scores represent the relative effects of each ORF on cell proliferation after 25 days in culture.”

While the screen was conducted in CSS and enzalutamide, in which we describe as experimental conditions of ADT/ART (Figure 1A, discussed in Results, p. 3. p. 4), all following –omic experiments in this study were performed in FBS and enzalutamide, as we have not indicated otherwise. In the control experiments in which no enzalutamide was added, we used FBS and not CSS. We thank the Reviewer for this suggestion.

5. If CREB5 downregulation is required for normal AR function then it is hard to explain the dependencies highlighted in DEPMAP. Unless its overexpression prevents the repression/downregulation of proliferation that occurs in PCa cancer cells as androgen levels rise.

See the response to Reviewer 2, comment 1 and 2 in which we observed that human tumors independently expressed CREB5 and AR.

References

Abida, W., J. Cyrta, G. Heller, D. Prandi, J. Armenia, I. Coleman, M. Cieslik, M. Benelli, D. Robinson, E. M. Van Allen, A. Sboner, T. Fedrizzi, J. M. Mosquera, B. D. Robinson, N. De Sarkar, L. P. Kunju, S. Tomlins, Y. M. Wu, D. Nava Rodrigues, M. Loda, A. Gopalan, V. E. Reuter, C. C. Pritchard, J. Mateo, D. Bianchini, S. Miranda, S. Carreira, P. Rescigno, J. Filipenko, J. Vinson, R. B. Montgomery, H. Beltran, E. I. Heath, H. I. Scher, P. W. Kantoff, M. E. Taplin, N. Schultz, J. S. deBono, F. Demichelis, P. S. Nelson, M. A. Rubin, A. M. Chinnaiyan and C. L. Sawyers (2019). "Genomic correlates of clinical outcome in advanced prostate cancer." Proc Natl Acad Sci U S A 116(23): 11428-11436.

Alumkal, J. J., D. Sun, E. Lu, T. M. Beer, G. V. Thomas, E. Latour, R. Aggarwal, J. Cetnar, C. J. Ryan, S. Tabatabaei, S. Bailey, C. B. Turina, D. A. Quigley, X. Guan, A. Foye, J. F. Youngren, J. Urrutia, J. Huang, A. S. Weinstein, V. Friedl, M. Rettig, R. E. Reiter, D. E. Spratt, M. Gleave, C. P. Evans, J. M. Stuart, Y. Chen, F. Y. Feng, E. J. Small, O. N. Witte and Z. Xia (2020). "Transcriptional profiling identifies an androgen receptor activity-low, stemness program associated with enzalutamide resistance." Proc Natl Acad Sci U S A 117(22): 12315-12323.

Beltran, H., D. Prandi, J. M. Mosquera, M. Benelli, L. Puca, J. Cyrta, C. Marotz, E. Giannopoulou, B. V. Chakravarthi, S. Varambally, S. A. Tomlins, D. M. Nanus, S. T. Tagawa, E. M. Van Allen, O. Elemento, A. Sboner, L. A. Garraway, M. A. Rubin and F. Demichelis (2016). "Divergent clonal evolution of castration-resistant neuroendocrine prostate cancer." Nat Med 22(3): 298-305.

He, M. X., M. S. Cuoco, J. Crowdis, A. Bosma-Moody, Z. Zhang, K. Bi, A. Kanodia, M. J. Su, S. Y. Ku, M. M. Garcia, A. R. Sweet, C. Rodman, L. DelloStritto, R. Silver, J. Steinharter, P. Shah, B. Izar, N. C. Walk, K. P. Burke, Z. Bakouny, A. K. Tewari, D. Liu, S. Y. Camp, N. I. Vokes, K. Salari, J. Park, S. Vigneau, L. Fong, J. W. Russo, X. Yuan, S. P. Balk, H. Beltran, O. Rozenblatt-Rosen, A. Regev, A. Rotem, M. E. Taplin and E. M. Van Allen (2021). "Transcriptional mediators of treatment resistance in lethal prostate cancer." Nat Med 27(3): 426-433.

Hwang, J. H., J. H. Seo, M. L. Beshiri, S. Wankowicz, D. Liu, A. Cheung, J. Li, X. Qiu, A. L. Hong, G. Botta, L. Golumb, C. Richter, J. So, G. J. Sandoval, A. O. Giacomelli, S. H. Ly, C. Han, C. Dai, H. Pakula, A. Sheahan, F. Piccioni, O. Gjoerup, M. Loda, A. G. Sowalsky, L. Ellis, H. Long, D. E. Root, K. Kelly, E. M. Van Allen, M. L. Freedman, A. D. Choudhury and W. C. Hahn (2019). "CREB5 Promotes Resistance to Androgen-Receptor Antagonists and Androgen Deprivation in Prostate Cancer." Cell Rep 29(8): 2355-2370 e2356.

Kregel, S., J. L. Chen, W. Tom, V. Krishnan, J. Kach, H. Brechka, T. B. Fessenden, M. Isikbay, G. P. Paner, R. Z. Szmulewitz and D. J. Vander Griend (2016). "Acquired resistance to the second-generation androgen receptor antagonist enzalutamide in castration-resistant prostate cancer." Oncotarget 7(18): 26259-26274.

Pomerantz, M. M., F. Li, D. Y. Takeda, R. Lenci, A. Chonkar, M. Chabot, P. Cejas, F. Vazquez, J. Cook, R. A. Shivdasani, M. Bowden, R. Lis, W. C. Hahn, P. W. Kantoff, M. Brown, M. Loda, H. W. Long and M. L. Freedman (2015). "The androgen receptor cistrome is extensively reprogrammed in human prostate tumorigenesis." Nat Genet 47(11): 1346-1351.

Pomerantz, M. M., X. Qiu, Y. Zhu, D. Y. Takeda, W. Pan, S. C. Baca, A. Gusev, K. D. Korthauer, T. M. Severson, G. Ha, S. R. Viswanathan, J. H. Seo, H. M. Nguyen, B. Zhang, B. Pasaniuc, C. Giambartolomei, S. A. Alaiwi, C. A. Bell, E. P. O'Connor, M. S. Chabot, D. R. Stillman, R. Lis, A. Font-Tello, L. Li, P. Cejas, A. M. Bergman, J. Sanders, H. G. van der Poel, S. A. Gayther, K. Lawrenson, M. A. S. Fonseca, J. Reddy, R. I. Corona, G. Martovetsky, B. Egan, T. Choueiri, L. Ellis, I. P. Garraway, G. M. Lee, E. Corey, H. W. Long, W. Zwart and M. L. Freedman (2020). "Prostate cancer reactivates developmental epigenomic programs during metastatic progression." Nat Genet 52(8): 790-799.

Taylor, B. S., N. Schultz, H. Hieronymus, A. Gopalan, Y. Xiao, B. S. Carver, V. K. Arora, P. Kaushik, E. Cerami, B. Reva, Y. Antipin, N. Mitsiades, T. Landers, I. Dolgalev, J. E. Major, M. Wilson, N. D. Socci, A. E. Lash, A. Heguy, J. A. Eastham, H. I. Scher, V. E. Reuter, P. T. Scardino, C. Sander, C. L. Sawyers and W. L. Gerald (2010). "Integrative genomic profiling of human prostate cancer." Cancer Cell 18(1): 11-22.

https://doi.org/10.7554/eLife.73223.sa2

Article and author information

Author details

  1. Justin H Hwang

    1. Masonic Cancer Center, University of Minnesota-Twin Cities, Minneapolis, United States
    2. Department of Medicine, University of Minnesota, Minneapolis, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Validation, Visualization, Writing – original draft, Writing – review and editing
    Contributed equally with
    Rand Arafeh and Ji-Heui Seo
    For correspondence
    jhwang@umn.edu
    Competing interests
    is a consultant for Astrin Biosciences, Principal Investigator for Caris Life Sciences Genitourinary disease working group
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-1686-7103

  2. Rand Arafeh

    1. Department of Medical Oncology, Dana-Farber Cancer Institute, Boston, United States
    2. Broad Institute of MIT and Harvard, Cambridge, Cambridge, United States
    3. Harvard Medical School, Boston, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Methodology, Validation, Writing – original draft, Writing – review and editing
    Contributed equally with
    Justin H Hwang and Ji-Heui Seo
    Competing interests
    No competing interests declared

  3. Ji-Heui Seo

    1. Department of Medical Oncology, Dana-Farber Cancer Institute, Boston, United States
    2. Broad Institute of MIT and Harvard, Cambridge, Cambridge, United States
    3. Harvard Medical School, Boston, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Validation, Writing – original draft
    Contributed equally with
    Justin H Hwang and Rand Arafeh
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-7280-3334

  4. Sylvan C Baca

    1. Department of Medical Oncology, Dana-Farber Cancer Institute, Boston, United States
    2. Broad Institute of MIT and Harvard, Cambridge, Cambridge, United States
    3. Harvard Medical School, Boston, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Visualization, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared

  5. Megan Ludwig

    Department of Pharmacology, University of Minnesota-Twin Cities, Minneapolis, United States
    Contribution
    Conceptualization, Investigation, Methodology, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared

  6. Taylor E Arnoff

    Warren Alpert Medical School of Brown University, Providence, United States
    Contribution
    Formal analysis, Investigation, Methodology, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared

  7. Lydia Sawyer

    Department of Medical Oncology, Dana-Farber Cancer Institute, Boston, United States
    Contribution
    Formal analysis, Investigation, Methodology, Validation, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared

  8. Camden Richter

    1. Department of Medical Oncology, Dana-Farber Cancer Institute, Boston, United States
    2. Broad Institute of MIT and Harvard, Cambridge, Cambridge, United States
    3. Harvard Medical School, Boston, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Visualization, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared

  9. Sydney Tape

    Department of Medicine, University of Minnesota, Minneapolis, United States
    Contribution
    Formal analysis, Investigation, Methodology, Validation, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared

  10. Hannah E Bergom

    Department of Medicine, University of Minnesota, Minneapolis, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Visualization, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared

  11. Sean McSweeney

    Department of Medicine, University of Minnesota, Minneapolis, United States
    Contribution
    Investigation, Methodology, Resources, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-7682-2073

  12. Jonathan P Rennhack

    1. Department of Medical Oncology, Dana-Farber Cancer Institute, Boston, United States
    2. Broad Institute of MIT and Harvard, Cambridge, Cambridge, United States
    3. Harvard Medical School, Boston, United States
    Contribution
    Conceptualization, Formal analysis, Investigation, Methodology, Visualization, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared

  13. Sarah A Klingenberg

    Department of Medicine, University of Minnesota, Minneapolis, United States
    Contribution
    Conceptualization, Methodology, Resources, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared

  14. Alexander TM Cheung

    1. Broad Institute of MIT and Harvard, Cambridge, Cambridge, United States
    2. Grossman School of Medicine, New York University, New York, United States
    Contribution
    Conceptualization, Formal analysis, Investigation, Methodology, Visualization, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared

  15. Jason Kwon

    1. Department of Medical Oncology, Dana-Farber Cancer Institute, Boston, United States
    2. Broad Institute of MIT and Harvard, Cambridge, Cambridge, United States
    3. Harvard Medical School, Boston, United States
    Contribution
    Conceptualization, Methodology, Resources, Supervision, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared

  16. Jonathan So

    1. Department of Medical Oncology, Dana-Farber Cancer Institute, Boston, United States
    2. Broad Institute of MIT and Harvard, Cambridge, Cambridge, United States
    3. Harvard Medical School, Boston, United States
    Contribution
    Conceptualization, Investigation, Methodology, Resources, Supervision, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared

  17. Steven Kregel

    Department of Cancer Biology, Loyola University Chicago, Maywood, United States
    Contribution
    Conceptualization, Methodology, Resources, Supervision, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared

  18. Eliezer M Van Allen

    1. Department of Medical Oncology, Dana-Farber Cancer Institute, Boston, United States
    2. Broad Institute of MIT and Harvard, Cambridge, Cambridge, United States
    3. Harvard Medical School, Boston, United States
    Contribution
    Conceptualization, Investigation, Resources, Supervision, Validation, Writing – original draft, Writing – review and editing
    Competing interests
    serves as Advisory/Consulting for Tango Therapeutics, Genome Medical, Invitae, Enara Bio, Janssen, Manifold Bio, Monte Rosa, received research support from Novartis, BMS, has equity with Tango Therapeutics, Genome Medical, Syapse, Enara Bio, Manifold Bio, Microsoft, Monte Rosa, receives travel reimbursement from Roche/Genentech, and holds patents including Institutional patents filed on chromatin mutations and immunotherapy response, and methods for clinical interpretation

  19. Justin M Drake

    1. Masonic Cancer Center, University of Minnesota-Twin Cities, Minneapolis, United States
    2. Department of Pharmacology and Urology, University of Minnesota, Minneapolis, United States
    Contribution
    Conceptualization, Investigation, Methodology, Resources, Supervision, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared

  20. Matthew L Freedman

    1. Department of Medical Oncology, Dana-Farber Cancer Institute, Boston, United States
    2. Broad Institute of MIT and Harvard, Cambridge, Cambridge, United States
    3. Harvard Medical School, Boston, United States
    Contribution
    Conceptualization, Formal analysis, Methodology, Resources, Supervision, Validation, Writing – review and editing
    Competing interests
    No competing interests declared

  21. William C Hahn

    1. Department of Medical Oncology, Dana-Farber Cancer Institute, Boston, United States
    2. Broad Institute of MIT and Harvard, Cambridge, Cambridge, United States
    3. Harvard Medical School, Boston, United States
    Contribution
    Conceptualization, Investigation, Methodology, Project administration, Supervision, Writing – original draft, Writing – review and editing
    For correspondence
    william_hahn@dfci.harvard.edu
    Competing interests
    Reviewing editor, eLife
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-2840-9791

Funding

University of Minnesota (Start up funds)

  • Justin H Hwang

National Cancer Institute (U01 CA176058)

  • William C Hahn

National Cancer Institute (U01 CA233100)

  • Eliezer M Van Allen

Kureit Cancer Research Foundation

  • Sylvan C Baca

Weizmann Institute of Science

  • Rand Arafeh

Ray of Light

  • Justin H Hwang
  • Hannah E Bergom

University of Minnesota. Targets of Cancer Training program grant T32 CA009138T32 (CA009138)

  • Megan Ludwig

American Cancer Society-AstraZeneca (PF-16-142-01-TBE)

  • Justin H Hwang

Young Investigator Award from the American Society of Clinical Oncologists (ASCO)

  • Sylvan C Baca

PhRMA Foundation

  • Sylvan C Baca

Mark Foundation Emerging Leader Award

  • Eliezer M Van Allen

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

This work was supported in part by the Weizmann Institute of Science – National Postdoctoral Award Program for Advancing Women in Science (to RA), Targets of Cancer Training program grant T32 CA009138 (to ML), Ray of Light Foundation (to HEB), NIH/NCI (K00 CA212221) (to JPR), American Cancer Society-AstraZeneca (PF-16-142-01-TBE) (to JH), Young Investigator Award from the American Society of Clinical Oncologists (ASCO) and by the PhRMA Foundation and Kure It Cancer Research Foundation (to SCB), U01 CA233100 (EMV), Mark Foundation Emerging Leader Award (EMV), U.S. National Institutes of Health/National Cancer Institute: U01 CA176058 (to WCH). We acknowledge Joshua Pan from Dana Farber Institute and Broad Institute of MIT and Harvard for designing approaches for co-dependency analysis.

Graphical figures were created with https://biorender.com/.

Senior Editor

  1. Kathryn Song Eng Cheah, University of Hong Kong, Hong Kong

Reviewing Editor

  1. Charles L Sawyers, Memorial Sloan Kettering Cancer Center, United States

Version history

  1. Preprint posted: August 18, 2021 (view preprint)
  2. Received: August 20, 2021
  3. Accepted: May 11, 2022
  4. Accepted Manuscript published: May 12, 2022 (version 1)
  5. Version of Record published: May 26, 2022 (version 2)

Copyright

© 2022, Hwang, Arafeh, Seo et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,558
    Page views
  • 325
    Downloads
  • 8
    Citations

Article citation count generated by polling the highest count across the following sources: PubMed Central, Crossref, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Justin H Hwang
  2. Rand Arafeh
  3. Ji-Heui Seo
  4. Sylvan C Baca
  5. Megan Ludwig
  6. Taylor E Arnoff
  7. Lydia Sawyer
  8. Camden Richter
  9. Sydney Tape
  10. Hannah E Bergom
  11. Sean McSweeney
  12. Jonathan P Rennhack
  13. Sarah A Klingenberg
  14. Alexander TM Cheung
  15. Jason Kwon
  16. Jonathan So
  17. Steven Kregel
  18. Eliezer M Van Allen
  19. Justin M Drake
  20. Matthew L Freedman
  21. William C Hahn
(2022)
CREB5 reprograms FOXA1 nuclear interactions to promote resistance to androgen receptor-targeting therapies
eLife 11:e73223.
https://doi.org/10.7554/eLife.73223

Categories and tags

Research organism

Further reading

    1. Cancer Biology
    2. Computational and Systems Biology

    Pancreatic cancer symptom trajectories from Danish registry data and free text in electronic health records

    Jessica Xin Hjaltelin, Sif Ingibergsdóttir Novitski ... Søren Brunak
    Research Article

    Pancreatic cancer is one of the deadliest cancer types with poor treatment options. Better detection of early symptoms and relevant disease correlations could improve pancreatic cancer prognosis. In this retrospective study, we used symptom and disease codes (ICD-10) from the Danish National Patient Registry (NPR) encompassing 6.9 million patients from 1994 to 2018,, of whom 23,592 were diagnosed with pancreatic cancer. The Danish cancer registry included 18,523 of these patients. To complement and compare the registry diagnosis codes with deeper clinical data, we used a text mining approach to extract symptoms from free text clinical notes in electronic health records (3078 pancreatic cancer patients and 30,780 controls). We used both data sources to generate and compare symptom disease trajectories to uncover temporal patterns of symptoms prior to pancreatic cancer diagnosis for the same patients. We show that the text mining of the clinical notes was able to complement the registry-based symptoms by capturing more symptoms prior to pancreatic cancer diagnosis. For example, ‘Blood pressure reading without diagnosis’, ‘Abnormalities of heartbeat’, and ‘Intestinal obstruction’ were not found for the registry-based analysis. Chaining symptoms together in trajectories identified two groups of patients with lower median survival (<90 days) following the trajectories ‘Cough→Jaundice→Intestinal obstruction’ and ‘Pain→Jaundice→Abnormal results of function studies’. These results provide a comprehensive comparison of the two types of pancreatic cancer symptom trajectories, which in combination can leverage the full potential of the health data and ultimately provide a fuller picture for detection of early risk factors for pancreatic cancer.

    1. Cancer Biology

    Disseminating cells in human oral tumours possess an EMT cancer stem cell marker profile that is predictive of metastasis in image-based machine learning

    Gehad Youssef, Luke Gammon ... Adrian Biddle
    Research Article

    Cancer stem cells (CSCs) undergo epithelial-mesenchymal transition (EMT) to drive metastatic dissemination in experimental cancer models. However, tumour cells undergoing EMT have not been observed disseminating into the tissue surrounding human tumour specimens, leaving the relevance to human cancer uncertain. We have previously identified both EpCAM and CD24 as CSC markers that, alongside the mesenchymal marker Vimentin, identify EMT CSCs in human oral cancer cell lines. This afforded the opportunity to investigate whether the combination of these three markers can identify disseminating EMT CSCs in actual human tumours. Examining disseminating tumour cells in over 12,000 imaging fields from 74 human oral tumours, we see a significant enrichment of EpCAM, CD24 and Vimentin co-stained cells disseminating beyond the tumour body in metastatic specimens. Through training an artificial neural network, these predict metastasis with high accuracy (cross-validated accuracy of 87-89%). In this study, we have observed single disseminating EMT CSCs in human oral cancer specimens, and these are highly predictive of metastatic disease.

    1. Cancer Biology
    2. Medicine

    Tissue-derived exosome proteomics identifies promising diagnostic biomarkers for esophageal cancer

    Dingyu Rao, Hua Lu ... Defa Huang
    Research Article

    Esophageal cancer (EC) is a fatal digestive disease with a poor prognosis and frequent lymphatic metastases. Nevertheless, reliable biomarkers for EC diagnosis are currently unavailable. Accordingly, we have performed a comparative proteomics analysis on cancer and paracancer tissue-derived exosomes from eight pairs of EC patients using label-free quantification proteomics profiling and have analyzed the differentially expressed proteins through bioinformatics. Furthermore, nano-flow cytometry (NanoFCM) was used to validate the candidate proteins from plasma-derived exosomes in 122 EC patients. Of the 803 differentially expressed proteins discovered in cancer and paracancer tissue-derived exosomes, 686 were up-regulated and 117 were down-regulated. Intercellular adhesion molecule-1 (CD54) was identified as an up-regulated candidate for further investigation, and its high expression in cancer tissues of EC patients was validated using immunohistochemistry, real-time quantitative PCR (RT-qPCR), and western blot analyses. In addition, plasma-derived exosome NanoFCM data from 122 EC patients concurred with our proteomic analysis. The receiver operating characteristic (ROC) analysis demonstrated that the AUC, sensitivity, and specificity values for CD54 were 0.702, 66.13%, and 71.31%, respectively, for EC diagnosis. Small interference (si)RNA was employed to silence the CD54 gene in EC cells. A series of assays, including cell counting kit-8, adhesion, wound healing, and Matrigel invasion, were performed to investigate EC viability, adhesive, migratory, and invasive abilities, respectively. The results showed that CD54 promoted EC proliferation, migration, and invasion. Collectively, tissue-derived exosomal proteomics strongly demonstrates that CD54 is a promising biomarker for EC diagnosis and a key molecule for EC development.