Neuroendocrine differentiation of prostate cancer leads to PSMA suppression

Abstract

Prostate-specific membrane antigen (PSMA) is overexpressed in most prostate adenocarcinoma (AdPC) cells and acts as a target for molecular imaging. However, some case reports indicate that PSMA-targeted imaging could be ineffectual for delineation of neuroendocrine (NE) prostate cancer (NEPC) lesions due to the suppression of the PSMA gene (FOLH1). These same reports suggest that targeting somatostatin receptor type 2 (SSTR2) could be an alternative diagnostic target for NEPC patients. This study evaluates the correlation between expression of FOLH1, NEPC marker genes and SSTR2. We evaluated the transcript abundance for FOLH1 and SSTR2 genes as well as NE markers across 956 tumors. A significant suppression of FOLH1 in NEPC patient samples and AdPC samples with high expression of NE marker genes was observed. We also investigated protein alterations of PSMA and SSTR2 in an NE-induced cell line derived by hormone depletion and lineage plasticity by loss of p53. PSMA is suppressed following NE induction and cellular plasticity in p53-deficient NEPC model. The PSMA-suppressed cells have more colony formation ability and resistance to enzalutamide treatment. Conversely, SSTR2 was only elevated following hormone depletion. In 18 NEPC patient-derived xenograft (PDX) models we find a significant suppression of FOLH1 and amplification of SSTR2 expression. Due to the observed FOLH1-supressed signature of NEPC, this study cautions on the reliability of using PMSA as a target for molecular imaging of NEPC. The observed elevation of SSTR2 in NEPC supports the possible ability of SSTR2-targeted imaging for follow-up imaging of low-PSMA patients and monitoring for NEPC development.

Full text

AUTHOR COPY ONLY
https://doi.org/10.1530/ERC-18-0226
https://erc.bioscientica.com © 2019 Society for Endocrinology
Printed in Great Britain
Published by Bioscientica Ltd.
26:2
Endocrine-Related
Cancer
M K Bakht etal. NE dierentiation suppresses
PSMA
131–146
-18-0226
RESEARCH
Neuroendocrine dierentiation of prostate
cancer leads to PSMA suppression
MartinKBakht1,2,3, IulianDerecichei3, YinanLi4, Rosa-MariaFerraiuolo3, MarkDunning5, SoWonOh1,
AbdulkadirHussein6, HyewonYoun1,2,7,8, KeithFStringer3,9, ChangWookJeong10, GiJeongCheon1,2, CheolKwak10,
KeonWookKang1,2, AlastairDLamb5,11, YuzhuoWang4,12, XuesenDong4 and LisaA Porter3
1Department of Nuclear Medicine, Seoul National University College of Medicine, Seoul, Korea
2Laboratory of Molecular Imaging and Therapy, Cancer Research Institute, Seoul National University College of Medicine, Seoul, Korea
3Department of Biological Sciences, University of Windsor, Windsor, Ontario, Canada
4Vancouver Prostate Centre, Department of Urologic Sciences, University of British Columbia, Vancouver, British Columbia, Canada
5Cancer Research UK Cambridge Institute, University of Cambridge, Cambridge, UK
6Department of Mathematics and Statistics, University of Windsor, Windsor, Ontario, Canada
7Tumor Microenvironment Global Core Research Center, Seoul National University, Seoul, Korea
8Cancer Imaging Center, Seoul National University Hospital, Seoul, Korea
9Department of Pathology, Cincinnati Children’s Hospital Medical Center, Cincinnati, Ohio, USA
10Department of Urology, Seoul National University College of Medicine, Seoul, Korea
11Nueld Department of Surgical Sciences, University of Oxford, Oxford, UK
12Department of Experimental Therapeutics, BC Cancer Research Centre, Vancouver, British Columbia, Canada
Correspondence should be addressed to G J Cheon or L A Porter: larrycheon@snu.ac.kr or lporter@uwindsor.ca
Abstract
Prostate-specic membrane antigen (PSMA) is overexpressed in most prostate
adenocarcinoma (AdPC) cells and acts as a target for molecular imaging. However, some
case reports indicate that PSMA-targeted imaging could be ineectual for delineation
of neuroendocrine (NE) prostate cancer (NEPC) lesions due to the suppression of the
PSMA gene (FOLH1). These same reports suggest that targeting somatostatin receptor
type 2 (SSTR2) could be an alternative diagnostic target for NEPC patients. This study
evaluates the correlation between expression of FOLH1, NEPC marker genes and SSTR2.
We evaluated the transcript abundance for FOLH1 and SSTR2 genes as well as NE markers
across 909 tumors. A signicant suppression of FOLH1 in NEPC patient samples and AdPC
samples with high expression of NE marker genes was observed. We also investigated
protein alterations of PSMA and SSTR2 in an NE-induced cell line derived by hormone
depletion and lineage plasticity by loss of p53. PSMA is suppressed following NE induction
and cellular plasticity in p53-decient NEPC model. The PSMA-suppressed cells have more
colony formation ability and resistance to enzalutamide treatment. Conversely, SSTR2
was only elevated following hormone depletion. In 18 NEPC patient-derived xenograft
(PDX) models we nd a signicant suppression of FOLH1 and amplication of SSTR2
expression. Due to the observed FOLH1-supressed signature of NEPC, this study cautions
on the reliability of using PMSA as a target for molecular imaging of NEPC. The observed
elevation of SSTR2 in NEPC supports the possible ability of SSTR2-targeted imaging for
follow-up imaging of low PSMA patients and monitoring for NEPC development.
2
Key Words
fprostate cancer
fPSMA
fneuroendocrine
dierentiation
fnuclear medicine
fsomatostatin receptor
26
Endocrine-Related Cancer
(2019) 26, 131–146

AUTHOR COPY ONLY
Printed in Great Britain
Published by Bioscientica Ltd.
https://doi.org/10.1530/ERC-18-0226
https://erc.bioscientica.com © 2019 Society for Endocrinology
132
M K Bakht etal. NE dierentiation suppresses
PSMA
26:2
Endocrine-Related
Cancer
Introduction
The main treatment protocol for patients suffering from
castration-resistant prostate cancer (CRPC) is androgen
receptor pathway inhibition (ARPI). Selection pressure
and lineage plasticity of ARPI can lead to neuroendocrine
(NE) differentiation of prostate adenocarcinoma (AdPC),
promoting the more prevalent subtype of CRPC,
which is termed treatment-induced neuroendocrine
prostate cancer (NEPC) (Davies et al. 2018). It has been
speculated that mutations in lineage regulators such as
retinoblastoma 1 (Rb1) and tumor protein 53 (p53) could
also pave the way of ARPI to confer AdPC lineage plasticity
and development of NEPC (Chen etal. 2018).
Due to the presumed androgen receptor (AR)
negativity of NEPC tumors, treatment options are
restricted to platinum- and cisplatin-based combinations
and median survival of NEPC patients is much lower than
patients with AdPC (Vlachostergios & Papandreou 2015).
Early identification of NEPC and novel targeting options
could be beneficial. The transmembrane protein prostate-
specific membrane antigen (PSMA) is an appealing
target for molecular imaging and therapy of AdPC since
it is overexpressed in a majority of AdPC tumors and
metastatic lesions (Rai et al. 2016). Targeting PSMA is
feasible by a wide variety of radioligands (Kopka et al.
2017, Rowe etal. 2017, Jadvar & Ballas 2018). The PSMA
ligands can also be labeled with therapeutic radionuclides
which can irradiate PSMA-expressing cells (Boegemann
etal. 2017). Despite the positive implications of PSMA for
many forms of advanced AdPC there are clinical reports
supporting that PSMA-targeted imaging is not able to
delineate NEPC tumors (Parimi et al. 2014, Chakraborty
etal. 2015, Sheikhbahaei etal. 2017, Tosoian etal. 2017,
Usmani etal. 2017). In four specific cases, NEPC patients
did not show substantial PSMA-radioligand uptake, in one
case, this was described due to a downregulation of PSMA
(Chakraborty etal. 2015, Tosoian etal. 2017, Usmani etal.
2017, Parida etal. 2018). To date, the relevance of these
clinical reports has not been investigated.
NEPC tumors express common markers such as
neuron-specific enolase (NSE; gene ENO2), chromogranin
A (CHGA) and synaptophysin (SYP) (Parimi et al.
2014). NEPC is associated with the loss of RE1-silencing
transcription factor (REST) due to alternate splicing
by the RNA splicing factor serine/arginine repetitive
matrix 4 (SRRM4). We revealed that SRRM4 plays a key
role in transdifferentiation of AdPC to NEPC under
ARPI (Li et al. 2017). The somatostatin receptor subtype
2 (SSTR-2) is prevalent in the majority NE tumors and
DOTATOC and DOTATATE radioligands with affinity
to SSTR are radiotracers for delineation of NE tumors
(Poeppel etal. 2011). Recently, Gofrit etal. suggested that
68Ga-DOTATATE might be a potential option for NEPC
imaging (Gofrit et al. 2017). Interestingly, Usmani et al.
compared 68Ga-PSMA vs 68Ga-DOTA in a NEPC patient
and reported that the NEPC tumor and metastatic lesions
have avid uptake of 68Ga-DOTA while the 68Ga-PSMA
uptake was faint and inconclusive (Usmani etal. 2017).
In this study, we use bioinformatic datasets, cell lines
and patient-derived xenograft (PDX) models to study the
correlation between expression of the PSMA gene, FOLH1,
NE biomarkers and SSTR2. This work supports that NEPC
tumors have a distinct PSMA-suppressed signature and
demonstrate the possibility that SSTR2-targeted imaging
could be an alternative diagnostic target for this aggressive
form of prostate cancer.
Materials and methods
Cell lines and cell culture
The LNCaP and DU-145 cell lines were purchased from
ATCC. The high passage LNCaP cell line (LNCaP-HP)
was a generous gift from Dr Dora Cavallo-Medved of
University of Windsor. LNCaP and LNCaP-HP cells were
grown in RPMI-1640 in the presence of 10% fetal bovine
serum (FBS). DU-145 cells were grown in Eagle’s minimum
essential medium (E-MEM) supplemented with 10% FBS.
For induction of NE transdifferentiation, LNCaP cells
were cultured in RPMI-1640 medium with 10% charcoal-
stripped serum (CSS) over ten passages during 4 weeks.
LNCaP cells were cultured in RPMI medium supplemented
with 10% CSS when treated with 10 µM enzalutamide
(ENZ) from Selleckchem (Houston, TX, USA).
Plasmids and infection
Small hairpin RNA (shRNA)-based p53 knockdown was
attained by transducing LNCaP cells with p53 shRNA
lentiviral particles and transduction control cells were
produced by using control shRNA lentiviral particles.
pLKO1 shp53-targeting shRNA (Addgene, #19119)
(MOI = 8) and pLKO1-control (Addgene, #8453) as an
empty backbone were gifts from Dr Bob Weinberg of
Massachusetts Institute of Technology. The lentiviral
production was previously described (Al Sorkhy etal. 2012).
Overall, 10,000 cells were seeded in fully supplemented
growth media in 24-well plates for 24 h. Cells were starved
by removing serum from the media, followed by the use

AUTHOR COPY ONLY
https://erc.bioscientica.com © 2019 Society for Endocrinology
Printed in Great Britain
Published by Bioscientica Ltd.
https://doi.org/10.1530/ERC-18-0226
133
M K Bakht etal. NE dierentiation suppresses
PSMA
26:2
Endocrine-Related
Cancer
of 1 mg/mL polybrene (Santa Cruz, sc-134220) and MOI 3
of the specific vector used. Infected media was changed to
fully supplemented media 24 h after infection. Cells were
incubated with 1 mg/mL puromycin (Sigma, P8833) 48 h
after infection for 72 h to allow for puromycin selection.
Media is thereafter changed every 48 h with puromycin
included. The isolated single-cell clones were cultured in
RPMI medium supplemented with 10% CSS.
Immunoblotting and immunocytochemistry
Immunoblotting and immunocytochemistry techniques
were described previously (da Silva et al. 2016, Li et al.
2017). In brief, cells were lysed in TNE buffer (50 mM
Tris, 150 mM NaCl, 5 mM EDTA) with protease inhibitors
(leupeptin 2 µg/mL, aprotinin 5 µg/mL, PMSF 100 µg/mL).
Protein concentrations were assessed using the Bradford
assay and equal amounts of protein were analyzed
using SDS-PAGE and transferred to PVDF membranes.
Membranes were blocked for 1 h at room temperature in
1% BSA and incubated in primary antibody overnight at
4°C, followed by secondary at a concentration of 1:10,000
for 1 h at room temperature. Visualization was conducted
using chemiluminescent peroxidase substrate (Pierce) as
per manufacturer’s instructions. Images were captured on
Alpha Innotech HD 2 using AlphaEase FC software.
For immunocytochemistry, primary antibodies were
diluted in 3% BSA-0.1% Tween-20 in 1× PBS and used
at a concentration of 1:200. Secondary antibodies were
used at a concentration of 1:750. Slides were imaged using
the LEICA DMI6000 inverted microscope with LAS 3.6
software.
Actin antibody was purchased from Chemicon-
Millipore (MAB150 1R). PSMA (D4S1F) was purchased
from Cell Signaling (12702). The secondary rabbit and
mouse antibodies were purchased from Sigma. SSTR2
(sc-365502), AR (sc-518), NSE (sc-271384) and p53
(sc-53394) were obtained from Santa Cruz Biotechnology.
Cell proliferation assay
Cell proliferation assay was conducted by plating 20,000
LNCaP cells per well of a 24-well cell culture plate in 500 µL
of RPMI medium supplemented with 10% CSS and treated
with vehicle control (DMSO) or enzalutamide (10 µM) for
6 days. Cell numbers were counted using trypan blue
exclusion and using a hemocytometer; counts were also
verified using a TC10 automated cell counter (Biorad).
Colony formation assay
To evaluate the anchorage-independent ability of LNCaP
cells with different levels of PSMA, the colony formation
assay was used. LNCaP cells were seeded in six-well
plates at a concentration of 2500 cells per well. They
were then cultured for 1 week in RPMI supplemented
with CSS, followed by treatment for one more week with
vehicle control (DMSO) or enzalutamide (10 µM). The
cell colonies were fixed with 4% paraformaldehyde and
stained with crystal violet (Sigma-Aldrich). Colonies were
photographed and scored via NIH ImageJ software.
Neurite length measurement and statistical analysis
Neurite length for each separate cell was measured
by manual tracing and determined using NIH ImageJ
software as previously described (Ding et al. 2015). The
neurites were defined as a process with lengths equivalent
to one diameters of a cell body. The percentage of neurite-
bearing cells was calculated from the total number of
counted cells (n = 3, ~1000 cells measured).
Quantitative real-time PCR analysis
RNA was isolated using Qiagen RNeasy Plus Mini Kit as
per manufacturer’s instructions. cDNA was synthesized
using Superscript II (Invitrogen) as per manufacturer’s
instructions. SYBR Green detection (Applied Biosystems)
was used for real-time PCR and was performed and analyzed
using Viia7 Real-Time PCR System (Life Technologies) and
software. The primers used are listed on Supplementary
Table1 (see section on supplementary data given at the
end of this article).
In silico dataset
Using Human Protein Atlas (www.proteinatlas.org) (Thul
et al. 2017, Uhlen et al. 2017), cBioPortal web-portal
(http://cbioportal.org) (Gao etal. 2013) and the web-portal
UALCAN (Chandrashekar et al. 2017), we evaluated the
transcript abundance for PSMA gene (FOLH1), prostate-
specific antigen (PSA) gene (KLK3), four well-known NE
biomarker genes including NSE gene (ENO2), CD56 gene
(NCAM1), synaptophysin gene (SYP), chromogranin A
gene (CHGA).
Cambridge Carcinoma of the Prostate App (http://
bioinformatics.cruk.cam.ac.uk/apps/camcAPP) as a

AUTHOR COPY ONLY
Printed in Great Britain
Published by Bioscientica Ltd.
https://doi.org/10.1530/ERC-18-0226
https://erc.bioscientica.com © 2019 Society for Endocrinology
134
M K Bakht etal. NE dierentiation suppresses
PSMA
26:2
Endocrine-Related
Cancer
multiple AdPC genomic datasets was used in this
study (Dunning et al. 2017). The web-portal UALCAN
(Chandrashekar et al. 2017) was also used which focuses
on TCGA level 3 RNA-seq and clinical data from 31 cancer
types. We analyzed relative expression of genes across
AdPC tumors and normal samples, based on Gleason
grading system using this resource. In addition, we used
the Beltran dataset (Beltran etal. 2016) including 34 AdPC
samples from 33 patients and 15 NEPC samples from 10
patients. Table1 provides an overview of the datasets used
in analysis.
The expression of PSMA protein was examined in a
variety of organs using the Human Protein Atlas (www.
proteinatlas.org). Two anti-PSMA antibodies including
HPA010593 (Sigma-Aldrich) and CAB001451 (Novocastra)
were used for visualization and measurement of PSMA in
all major tissues and organs (n = 45) in the human body,
supplemented with RNA-sequencing data for 31 tissues.
The same antibodies used for immunohistochemistry
images of AdPC tissues in different stages and each
annotated by pathologists (Thul etal. 2017, Uhlen etal.
2017).
The survival data and pairwise-correlations of
gene expression
The specialized web interface, Cambridge Carcinoma of
the Prostate App (http://bioinformatics.cruk.cam.ac.uk/
apps/camcAPP/) as a source for multiple AdPC genomic
datasets was used in this study (Dunning et al. 2017).
Table1 provides an overview of the datasets used in this
analysis (Taylor etal. 2010, Ross-Adams et al. 2015). The
Cambridge (Ross-Adams etal. 2015) cohort involved 358
fresh frozen samples from 156 patients comprising; 125
primary prostate cancer from radical prostatectomy with
matched benign tissue, 64 matched germline genomic
DNA (gDNA), 19 CRPC from channel transurethral
resection of the prostate, 13 with matched germline gDNA,
and 12 independent benign samples from holmium laser
enucleation of the prostate. The camcAPP was used for
recursive partitioning-based survival analysis and Kaplan–
Meier plots, pairwise-correlations of gene expression and
heatmaps of gene expression data.
Survival data and heatmaps were generated using
camcAPP, which is implemented in R code as a Shiny
application (Dunning et al. 2017). Kaplan–Meier
biochemical relapse-free survival plots were generated
using a recursive partitioning analysis named unbiased
recursive partitioning (Hothorn et al. 2006). This
conditional inference framework was used to determine
if the samples could be split into groups based on the
expression data from each of the genes of interest. The
algorithm tailored for our case can be described as follows:
a statistical test of independence was run between gene
expression levels and the survival times. When the
P value of initial test of independence (ITI) was found to
be less than 0.05, an optimal cutoff point was determined
in the expression data such that a weighted log-rank
statistic (a loss function based on log-rank) comparing the
two groups resulting from dividing the sample of patients
by the cutoff point would be maximized. Afterward,
the samples were split at the optimal cutoff point on
the expression scale and represented as a log-rank
comparing those two groups. When the algorithm did
not confirm dependence between survival times and gene
expression, we split samples into two groups based on
median expression level of the gene. In the cases with ITI
P values more than 0.05, we are unable to state a definitive
relationship between the expression of the gene and
survival. The value quoted on the Kaplan–Meier plots of
this manuscript indicate where the recursive partitioning
algorithm has found two distinct groups of samples
as high and low expression levels and the log-rank test
was employed to identify statistical difference between
Table 1 An overview of the prostate cancer datasets used in this study.
Dataset name Source
Platform: gene expression
Platform: copy number
Sample size and
tumor type
Cambridge 2015 Ross-Adams etal. (2015) Illumina HT12 Illumina Omni 2.5 125 Primary
19 Advanced
MSKCC 2010 Taylor etal. (2010) Aymetrix Human 1.0 ST Agilent 244k 109 Primary
19 Advanced
Michigan 2012 Grasso etal. (2012) Agilent Whole Human 44k Agilent 105k/244k 59 Primary
32 Advanced
TCGA for prostate
cancer
UALCAN (Chandrashekar etal. 2017) Illumina Illumina HiSeq 2000N 497 Primary
Beltran 2016 Beltran etal. (2016) Agilent 2100 Illumina HiSeq 34 AdPC
15 NEPC

AUTHOR COPY ONLY
https://erc.bioscientica.com © 2019 Society for Endocrinology
Printed in Great Britain
Published by Bioscientica Ltd.
https://doi.org/10.1530/ERC-18-0226
135
M K Bakht etal. NE dierentiation suppresses
PSMA
26:2
Endocrine-Related
Cancer
the high and low expressing groups. Pearson correlation
was used for pairwise-correlations of the studied gene
expression analysis.
Animals and PDX models
Fresh AdPC or NEPC tissues from patients were grafted
under the kidney capsules of non-obese diabetic/severe
combined immunodeficient (NOD/SCID) mice. This study
followed the ethical guidelines stated in the Declaration of
Helsinki, specimens were obtained from patients with their
informed written consent approved by the Institutional
Review Board of the University of British Columbia (UBC).
Animal care and experimental procedures were carried
out in accordance with the guidelines of the Canadian
Council on Animal Care (CCAC) under the approval of
the Animal Care Committee of the UBC.
The expression of the PSMA gene in five AdPC PDX
models (LTL-313-B, LTL-313-B-R, LTL-418, LTL-418-R
and LTL-331-7) and two NEPC PDX models (LTL-331-7-R
and LTL-352) was studied by real-time PCR analysis.
Transcription of FOLH1 and SSTR2 genes in 18 PDX
models including three NEPC models were analyzed.
Transcriptomic analysis for all PDX models, with the
exception of the LTL331-331R castration time-series
samples, was achieved by GE 8 × 60 K microarray and
transcriptomic analysis of the LTL331-331R time-series
performed using RNA-sequencing data (Akamatsu et al.
2015, Ci et al. 2018). We previously characterized and
validated these models as having AdPC and NEPC mRNA
and protein signatures (Li et al. 2017, Lin et al. 2014,
Nabavi etal. 2017).
Statistical analysis
All of the in vitro experiments were performed in triplicates
and repeated three times. All in vitro and in vivo results are
expressed as the mean ± standard error of the mean (s.e.m.).
In the box whisker plots, the line inside each box is the
median, upper box border represents the 75th quartile,
lower box border represents the 25th quartile and whiskers
represent the range. Statistical analysis was done using
the GraphPad Prism 5.01 software (GraphPad Software).
Differences between the two groups were compared by
unpaired Student’s t-test. One-way ANOVA followed by a
Benjamini–Hochberg, Tukey or Newman–Keuls multiple
comparison test was used to compare differences among
multiple groups. The false discovery rate (FDR) less than
0.05 considered as significant in Benjamini–Hochberg
adjustment. The levels of significance were set at P < 0.05
as *, P < 0.01 as ** and P < 0.001 as ***.
For quantification of FOLH1 heatmaps, −0.5 < Z-score
and +0.5 > Z-score are considered as suppression and
amplification thresholds. The percent of patients
with suppression (Z-score ≤ −0.5), no alteration
(−0.5 < Z-score < +0.5) and amplification (Z-score ≥ +0.5)
of FOLH1 in each group of samples were calculated.
Total number of patients with amplification of the
gene of interests were counted and set as the maximum
value. Of the maximum value patients, those who
were co-upregulated with FOLH1 were categorized as
‘Amplification’, while those whose FOLH1 status did not
change or was downregulated were categorized as ‘No
Alteration’ and ‘Suppressed’ respectively. The values for
all three categories were then divided by the maximum
value and graphed under their respective categories.
Results
High-grade CRPC has inconsistent expression of the
PSMA gene (FOLH1)
FOLH1 expression was highly upregulated in prostate tissue
and relatively upregulated in both the hippocampus and
salivary gland (Supplementary Fig.1A). Similarly, protein
level of PSMA was higher in prostate tissue than all other
organs and tissues. Only the kidney, small intestine and
duodenum reached medium levels of protein expression
(Supplementary Fig. 1B). As compared to a spectrum of
other cancer types (Gao et al. 2013), FOLH1 is highly
amplified in AdPC (Supplementary Fig. 2A). AdPC
patients with amplification in FOLH1 have a significant
poorer survival rate in MSKCC (Taylor etal. 2010) dataset
(Supplementary Fig.2B). Similarly, the Cambridge (Ross-
Adams etal. 2015) dataset displayed differential outcomes
for men with low vs high FOLH1 expression (log-rank
P = 0.047); however, the initial test of independence shows
no significance and hence conclusions cannot be drawn
from this dataset (ITI P = 0.31).
With the aim of exploring PSMA gene levels during
progression of AdPC to clinically relevant CRPC, the
Michigan and Cambridge datasets were used and gene
expression levels studied (Fig.1A and B). The Michigan
dataset (Grasso et al. 2012) showed that there is an
expected rise from normal to AdPC tissue in expression
of FOLH1, but there was no statistically significant
difference between hormone-responsive AdPC and
CRPC tumors (Fig. 1A). A cluster of samples with very
low expression of the PSMA gene is observable in CRPC

AUTHOR COPY ONLY
Printed in Great Britain
Published by Bioscientica Ltd.
https://doi.org/10.1530/ERC-18-0226
https://erc.bioscientica.com © 2019 Society for Endocrinology
136
M K Bakht etal. NE dierentiation suppresses
PSMA
26:2
Endocrine-Related
Cancer
samples. Similarly, Fig. 1B illustrates an increase from
benign to hormone-responsive tumor in the expression of
FOLH1 in the Cambridge dataset and, interestingly, there
was a significant drop in expression when transitioning
to CRPC. To further explore this inconsistency, we
evaluated the expression of FOLH1 by Gleason grading
score (Fig. 1C) in TCGA dataset (Chandrashekar et al.
2017). From benign tissue (normal) to Gleason score
(GS) 8 FOLH1 levels steadily increase and then they
remain unchanged between the transition from 8 to 9.
Samples with GS of ten showed no significant elevation in
comparison with the normal samples. While the sample
size (n = 4) prevents definitive conclusions for GS 10, the
observed fall in FOLH1 expression supports the possibility
of a suppression in high-grad AdPC. Collectively, this led
us to conclude that expression of FOLH1 in high-grade
CRPC is variable and that there is a need to question
whether FOLH1 expression and PSMA protein levels
are effective in determining progression to the highest
grades of AdPC.
An inverse correlation between the expression of
FOLH1 and common NE biomarker genes
Figure 1D shows the mean expression levels of FOLH1,
PSA gene (KLK3) and NE marker genes over 144 patients.
As we expected, KLK3 and FOLH1 expression had a direct
correlation where 49% of patients with amplified KLK3 also
showed an amplification of FOLH1 (Fig.1E). In contrast,
an inverse correlation between FOLH1 and NE biomarker
gene expression was identified. Particularly, suppression
of FOLH1 was observed in 65 and 53% of patients who
were overexpressing the NE genes ENO2 and NCAM1,
respectively. The numbers of patients with amplification of
genes of interest based on level of FOLH1 gene expression are
presented on Supplementary Table2. In Fig.1F, G, H, I and J,
Figure1
Expression of PSMA at varying grades of CRPC. (A and B) Box-whisker plots showing the expression of FOLH1 gene in three dierent classes of samples
from (A) Michigan (Grasso etal. 2012) and (B) Cambridge (Ross-Adams etal. 2015) datasets. (C) The expression of FOLH1 during progression of AdPC
based on Gleason score from TCGA dataset generated by web-portal UALCAN (Chandrashekar etal. 2017). One-way ANOVA followed by unpaired t-tests
were performed with Benjamini–Hochberg adjustment for multiple test correction; **P < 0.01 and ***P < 0.001, n.s.: no signicant. (D) Heatmap plot of
the mean expression levels of FOLH1, PSA gene (KLK3) and four major clinically signicant NE marker genes including among patients of Cambridge
(Ross-Adams etal. 2015) datasets. (E) Percent of patients with suppression (Z-score ≤ +0.5), no alteration (−0.5 < Z-score < +0.5) and amplication
(Z-score ≥ +0.5) of FOLH1 in each group of samples. (F, G, H, I and J) Pairwise correlations of the studied gene expression and Pearson correlation analysis
from Cambridge (Ross-Adams etal. 2015) datasets. A full-colour version of this gure is available at https://doi.org/10.1530/ERC-18-0226.

AUTHOR COPY ONLY
https://erc.bioscientica.com © 2019 Society for Endocrinology
Printed in Great Britain
Published by Bioscientica Ltd.
https://doi.org/10.1530/ERC-18-0226
137
M K Bakht etal. NE dierentiation suppresses
PSMA
26:2
Endocrine-Related
Cancer
this relationship was further characterized through the
use of a Pearson correlation. On Fig.1G and H, a strong
inverse correlation between FOLH1 and ENO2 and NCAM1
is prevalent (r = −0.46 and −0.33), while correlation between
FOLH1 and the other NE markers was not significant.
Figure1F shows there was a significant direct correlation
between FOLH1 and KLK3 genes (r = 0.41). Supplementar y
Table3 provides confidence interval (CI) parameters of
these gene correlations. Supplementary Figure3 shows the
mean expression levels of FOLH1, NE genes, AR and AR
target genes. Similar to the PSA gene (KLK3), AR and some
other AR target genes such as KLK2, FKBP5, CAMKK2 and
NKX3-1 are also inversely correlated with FOLH1
expression. The overall regression analysis led us to
conclude that PSMA and NE biomarkers are inversely
correlated.
Treatment-induced NEPC correlates with
PSMA suppression
Previously we showed SRRM4 can induce NEPC in patients
treated by ARPI through compromising the function
of genes such as REST (Li et al. 2017). The elevation of
SRRM4 and the loss of REST are indicators of treatment-
induced NEPC. Figure2A and B show that when SRRM4
is amplified FOLH1 is highly suppressed and, conversely,
when REST is amplified FOLH1 is amplified. Specifically,
the suppression of FOLH1 was observed in 57% of patients
who were overexpressing SRRM4, while only 15% of
patients who were overexpressing REST had suppression
of FOLH1.
As summarized on Table1, Beltran dataset (Beltran etal.
2016) patient samples are histologically characterized as
AdPC and NEPC. Figure2C shows NEPC has significantly
lower expression of FOLH1 (P < 0.001) as compared to
AdPC. Following this we analyzed the association between
FOLH1 and SRRM4 expression in NEPC samples and
found that a significant (P = 0.011) inverse correlation was
present between the two biomarkers (r = −0.358) (Fig.2D).
Also, Fig. 2E shows FOLH1 directly correlated to REST
expression (r = 0.561). The relationship of SRRM4 and REST
to FOLH1 collectively supports that treatment-induced
NE transdifferentiation correlates with suppressed PSMA
gene expression.
An inverse correlation between the expression of
FOLH1 and SSTR2
The heatmap plot in Fig. 3A displays NE marker genes
clustered together at the top which showed less difference
Figure2
Analysis of FOLH1, SRRM4 and REST in tumor
datasets. (A) The heatmap plot of the mean
expression levels of FOLH1, SRRM4 and REST genes
among patients of Cambridge dataset (Ross-
Adams etal. 2015). (B) The percent of patients
with suppression (Z-score ≤ +0.5), no alteration
(−0.5 < Z-score < +0.5) and amplication
(Z-score ≥ +0.5) of FOLH1 in each group of
samples. (C) The comparison of FOLH1, SRRM4 and
REST expressions between AdPC and NEPC
samples of Beltran dataset (Beltran etal. 2016).
Error bars reect s.e.m. and Student’s t-test was
performed. (D, E) The relationship between FOLH1
and SRRM4 levels in NEPC samples in Beltran
dataset (Beltran etal. 2016) by Pearson
correlation analysis. A full-colour version of this
gure is available at https://doi.org/10.1530/
ERC-18-0226.

AUTHOR COPY ONLY
Printed in Great Britain
Published by Bioscientica Ltd.
https://doi.org/10.1530/ERC-18-0226
https://erc.bioscientica.com © 2019 Society for Endocrinology
138
M K Bakht etal. NE dierentiation suppresses
PSMA
26:2
Endocrine-Related
Cancer
while SSTR2 and FOLH1 represented the largest difference;
consequently, Fig.3B demonstrates that more than 61%
of patients with amplified SSTR2 are FOLH1 suppressed.
Figure 3C and Supplementary Table 3 show a modest
inverse correlation between SSRT2 and FOLH1 (r = −0.50).
Figure3D shows that from benign tissue (normal) to GS
of 9, FOLH1 is suppressed while there is an evidence of
a two-fold increase in the expression of SSTR2 at GS of
10 (P < 0.05). SSRT2 level in NEPC sample showed no
suppression (Fig. 3E). This led us to conclude that the
observed suppression of FOLH1 could be accompanied by
SSTR2 gene overexpression at high-grade CRPC.
NEPC-like patients have signicantly worse survival
rates than non-NEPC-like patients
Kaplan–Meier survival curves studying high and low
expression levels of the PSA gene, KLK3, fail to reveal
any consistent correlation with patient survival over two
different datasets (Fig. 4A and Supplementary Fig. 4A).
SSTR2 and NE genes including ENO2 and NCAM1
did not correlate with a difference in BCR (Fig. 4B, D,
H and Supplementary Fig. 4B, D, H). Figure 4E and
Supplementary Fig.4E show that high expression of SYP
was associated with decreased relapse-free survival (poor
outcome) in both of studied datasets (log-rank P < 0.05).
High expression of CHGA was also correlated with poor
outcome in MSKCC (Taylor etal. 2010) dataset (Fig.4C).
Similarly, Kaplan–Meier analysis of CHGA in Cambridge
(Ross-Adams etal. 2015) dataset showed clear partitioning
for men with low vs high CHGA expression (log-rank
P = 0.025); however, the initial test of independence was
not significant and hence no conclusion can be drawn
from this dataset (ITI P = 0.28).
The evaluation of treatment-induced NEPC, SRRM4,
showed the SRRM4 high group had a significantly (log-
rank P = 0.046) poorer prognosis than SRRM4 low group
(Fig.4F). This trend continued with its downstream target,
REST; where REST high group had a significantly (log-rank
P = 0.0056) better prognosis than REST low group (Fig.4G).
The relationships observed in MSKCC (Taylor et al.
2010) dataset were not consistently observed in the
Figure3
Correlative analysis of FOLH1 with SSRT2 and NE genes. (A) The heatmap plot of the mean expression levels of FOLH1, NE genes and somatostatin
receptor-2 gene (SSTR2) expression among patients of Cambridge dataset (Ross-Adams etal. 2015) (method to calculate distances is euclidean). (B) The
percent of patients with suppression (Z-score ≤ +0.5), no alteration (−0.5 < Z-score < +0.5) and amplication (Z-score ≥ +0.5) of FOLH1 in each group of
samples. (C) Pairwise correlation of treatment-induced gene expressions and Pearson correlation analysis from Cambridge dataset (Ross-Adams etal.
2015). (D) The expression of SSTR2 during progression of AdPC based on Gleason score from TCGA dataset generated by web-portal UALCAN
(Chandrashekar etal. 2017). One-way ANOVA followed by a t-test was performed with Benjamini–Hochberg adjustment for multiple test correction;
**P < 0.01 and ***P < 0.001, n.s.: no signicant. (E) The comparison of SSTR2 expressions between AdPC and NEPC samples of Beltran dataset (Beltran
etal. 2016) Error bars reect s.e.m. and Student’s t-test was performed. A full-colour version of this gure is available at https://doi.org/10.1530/
ERC-18-0226.

AUTHOR COPY ONLY
https://erc.bioscientica.com © 2019 Society for Endocrinology
Printed in Great Britain
Published by Bioscientica Ltd.
https://doi.org/10.1530/ERC-18-0226
139
M K Bakht etal. NE dierentiation suppresses
PSMA
26:2
Endocrine-Related
Cancer
Cambridge (Ross-Adams etal. 2015) dataset. Results found
non-significant in the Cambridge dataset do not prove
that there is no relationship with survival for those genes,
but this inconsistency provides evidence that the effects
of several genes on survival are not clear-cut.
In all, the trends in Kaplan–Meier survival curves
in Fig. 4 and Supplementary Fig. 2B, C imply that NE
markers, especially treatment-induced NEPC, could be
clinically significant in determining survival rating and
correlate with poorer prognosis. Further examination into
FOLH1 and KLK3 is required before a definitive correlation
can be made for these genes.
PSMA suppression and SSTR2 overexpression in
NE-induced AdPC cell line
The levels of PSMA, SSTR2, AR and NSE were measured
in three different cell line models representing androgen-
sensitive AdPC, CRPC and NEPC (Fig.5A and B). LNCaP
cell line as an AR-positive cell line has wild‐type p53, and
it is considered a typical model for androgen-sensitive
AdPC (Chen etal. 2018). In addition, LNCaP-HP as a high
passage LNCaP cell line could be a model of CRPC due to its
androgen insensitivity while it has positivity for AR (Unni
etal. 2004). AR-negative DU145 cells, suggested to be used
as a NEPC model, contain two different point mutations
in the TP53 gene (Phe223Leu and Val274Phe), one on
either allele, producing nonfunctional protein product
(Chappell et al. 2012, Li et al. 2016). Figure 5B shows
the level of NSE as a NE marker has continuous elevation
from AdPC to NEPC. AdPC and CRPC in vitro models
are AR and PSMA positive while NEPC models are PSMA
and AR negative. SSTR2 has significantly higher level in
DU145 model.
To examine the impact of NE transdifferentiation of
AdPC on PSMA levels we used the LNCaP cell line which
mimics the phenotype of NEPC cells when maintained
overtime in steroid-reduced conditions (Zelivianski etal.
2001). LNCaP cells maintained in 10% CSS for 1month
have an altered phenotype compared to control cells
maintained in 10% FBS (Fig. 5C). Almost 75% of the
CSS-treated LNCaP cells (LNCaP-CSS) extended neurites,
whereas less than 10% of control or LNCaP-FBS cells bore
neurites (Fig. 5D). Additionally, both the mean length
of the longest neurite and the total neurite length were
significantly increased in LNCaP-CSS cells compared
with the LNCaP-FBS (Fig.5E and F). These observations
indicate that LNCaP growth conditions can play the role
of in vitro models for AdPC and NEPC. The western blot
assay (Fig.5G and H) demonstrated that SSTR2 and NSE
as a NE marker protein are expressed at a higher level in
the LNCaP-CSS cell line. In contrast, PSMA, AR and p53
proteins have a significant decline. Immunofluorescence
microscopy of LNCaP cell lines supports western blot data
(Fig.5I, J, K and L). Collectively, these data demonstrate
that LNCaP-CSS are differentiating toward a NE phenotype
and that this occurs coincidentally with a decrease in
PSMA and AR levels.
Figure4
The probability of freedom from biochemical recurrence (BCR) of prostate cancer patients grouped according to the gene expression levels. Kaplan
Meyer survival curves for high and low expression levels of (A) KLK3, (B) ENO2, (C) CHGA, (D) NCAM1, (E) SYP, (F) SRRM4, (G) REST, (H) SSTR2 genes generated
by MSKCC (Taylor etal. 2010). A full-colour version of this gure is available at https://doi.org/10.1530/ERC-18-0226.

AUTHOR COPY ONLY
Printed in Great Britain
Published by Bioscientica Ltd.
https://doi.org/10.1530/ERC-18-0226
https://erc.bioscientica.com © 2019 Society for Endocrinology
140
M K Bakht etal. NE dierentiation suppresses
PSMA
26:2
Endocrine-Related
Cancer
Development of ENZ resistance following a
p53-dependent suppression of PSMA
Treatment of LNCaP cell line with ENZ (10 µM) resulted
in an increase in PSMA protein levels in medium
containing either FBS or CSS (Fig. 6A and B). AR levels
show no significant alteration under these conditions.
Treatment with ENZ in medium supplemented with
CSS had an increase in levels of NSE. This experiment,
and previously reported data, supports that short-term
exposure of LNCaP cell lines to either ENZ treatment
or serum removal cannot be a viable approach to make
a clinically relevant PSMA-suppressed in vitro model of
AdPC with NE features. Recently, it had been reported
that p53 knockdown could lead to suppression of luminal
markers and overexpression of basal and NE markers (Li
etal. 2017, Mu etal. 2017). Figure6C and D shows p53
knockdown of LNCaP cells in a medium supplemented in
Figure5
Analysis of PSMA and SSTR2 in a NEPC induced cell line. (A and B) Western blot analyses of protein level of PSMA, SSTR2, AR, NSE and p53 in 3 dierent
prostate cancer cell line models. (A) Immunoblotting (B) diagram showing the relative density of protein levels. (C) Representative photos of control (left)
and CSS-treated (right) LNCaP cells stained with Hoechst. Scale bar: 50 μm. (D, E and F) Neurites were studied under an inverted microscope: (D) % of cells
with neurites counted over 3 elds of view over 3 separate experiments. (E) Neurites were measured using ImageJ software and longest neurite
calculated. (F) Average neurite. (G, H, I, J, K and L) LNCaP cells are treated with either FBS or CSS as indicated and level of PSMA, SSTR2, AR, NSE and p53
were detected by (G and H) immunoblotting and (I, J and K) immunocytochemistry. (L) Data are quantied using ImageJ software. Stat: Error bars reect
s.e.m. between three separate experiments. The data were analyzed by either Student’s t-test or one-way ANOVA followed by a Tukey’s multiple
comparison tests; **P < 0.01 and ***P < 0.001.

AUTHOR COPY ONLY
https://erc.bioscientica.com © 2019 Society for Endocrinology
Printed in Great Britain
Published by Bioscientica Ltd.
https://doi.org/10.1530/ERC-18-0226
141
M K Bakht etal. NE dierentiation suppresses
PSMA
26:2
Endocrine-Related
Cancer
CSS for 6days causes suppression AR and PSMA. DU145 as
an AR- and PSMA-negative cell line has a significant higher
level of NSE (P < 0.01). Due to overexpression of NSE and
suppression of PSMA, we used the p53-knockdown cell
line as a model of low-PSMA cancer with NE features for
the next steps of our experiments. The LNCaP cell line
transduced with shControl is also referred to as high-PSMA
cell line. Figure6E shows that LNCaP cells with low level
of PSMA have slightly higher proliferation during 6-day
treatment with the control vehicle (DMSO), but it was
not statistically significant (P = 0.65). During ENZ (10 µM)
treatment, low-PSMA cells demonstrate a higher growth
rate (P = 0.0048) and on the sixth day of cell counting
low-PSMA cell numbers were 2.8 times higher than high-
PSMA cells. It can be inferred that low-PSMA cells are less
sensitive to ENZ (10 µM) than high-PSMA cells over this
6-day experiment. In addition, we tested sensitivity of
LNCaP cells with different levels of PSMA using colony
formation assay (Fig.6F) and find that low-PSMA cells have
more colonies formed under ENZ (10 µM) treatment than
treatment of cells with high PSMA. Therefore, low-PSMA
cells were less sensitive to ENZ treatment, supporting the
hypothesis that suppression of PSMA, as found in NEPC,
correlates with more aggressive disease.
Figure 6H shows a schematic of the associations in
changing gene expression found using in vitro models.
Hormone depletion, which is modeled by serum
deprivation, suppresses PSMA, AR and p53 levels and
elevates levels of SSTR2. ARPI which is modeled by
treating cells with ENZ showed no alteration of AR or
SSTR2 but increased levels of PSMA. A p53 knockdown
in vitro model supports that suppression of PSMA and AR
are associated with p53-dependent cellular plasticity and
that this is independent of SSTR2 levels.
NEPC has a distinct FOLH1-suppressed signature in
PDX models
Following pathological investigation, patient tumors were
stratified into either AdPC or NEPC phenotypes put into
PDX mouse models and studied before or after castration
(schema Fig. 7A, B and C). Significant suppression of
FOLH1 was observed in the LTL-331R-G7 and LTL-352
mice, which were found to progress to NEPC, as compared
to those mice lacking NE marker expression (P < 0.05)
(Fig.7D). There was a wide spectrum of FOLH1 expression
in different AdPC models. In the LTL-313-B model, after
castration of mice, which can mimic treatment-induced
Figure6
Analysis of treatment response to ENZ following a p53-dependent suppression of PSMA. (A and B) Western blot analyses of protein level of PSMA, SSTR2,
AR and NSE in LNCaP cell line treated with vehicle control (DMSO) or ENZ (10 µM) supplemented with either FBS or CSS for 6days (A) representative
immunoblot (B) the relative density of protein levels. (C and D) Western blot analyses of protein level of PSMA, SSTR2, AR, NSE and p53 in LNCaP cell line
transduced with annotated shRNA supplemented with CSS for 6days. (C) Representative immunoblot (D) the relative density of protein levels. (E) Growth
curve of LNCaP cell lines with dierent levels of PSMA following treatment with vehicle control (DMSO) or ENZ (10 µM) in supplemented with CSS. (F and
G) The colony-forming ability of high-PSMA and low-PSMA seeded in 10% CSS for 1week and treated with either ENZ (10 µM) or DMSO for one more
week. (F) Representative wells (G) quantication of the number of the colonies using CellProler software. (H) Schematic of the impact of ARPI, hormonal
deletion and loss of p53 on PSMA, AR and SSTR2 based on the obtained data in Figs5 and 6. Error bars reect s.e.m. between three separate experiments.
The data were analyzed by either Student’s t-test or one-way ANOVA followed by a Tukey’s multiple comparison tests; **P < 0.01 and ***P < 0.001.
A full-colour version of this gure is available at https://doi.org/10.1530/ERC-18-0226.

AUTHOR COPY ONLY
Printed in Great Britain
Published by Bioscientica Ltd.
https://doi.org/10.1530/ERC-18-0226
https://erc.bioscientica.com © 2019 Society for Endocrinology
142
M K Bakht etal. NE dierentiation suppresses
PSMA
26:2
Endocrine-Related
Cancer
relapse in the form of CRPC, we observed a 2.4-fold
elevation in FOLH1. However, in the LTL-418 model,
there was no significant alteration in FOLH1 expression.
For simulation of treatment-induced NEPC, after inducing
NE to LTL-331-7 model as a CRPC model, a 2.1 times
suppression of FOLH1 was observed. Similarly, the PDX
model obtained by direct engraftment of tissue from a
NEPC metastases sample had significantly lower FOLH1
expression in comparison with all AdPC models.
The transcription of FOLH1 and SSTR2 in 18 PDX
models including three NEPCs were investigated (Fig.8A
and B). Among these PDX models, LTL331 and LTL331R
are the first-in-field PDX model of AdPC-to-NEPC
transdifferentiation. Figure 8C illustrates upon LTL331
model castration, the primary AdPC initially regresses
but relapses within 24–32 months as typical CRPC and
the whole transdifferentiation process is predictive of
disease progression and is fully recapitulated in the donor
patient (Akamatsu et al. 2015, Ci et al. 2018). Figure 8B
shows a five-fold suppression of FOLH1 gene expression in
NEPC models subsequent with a two fold increase in the
expression of SSTR2. Figure 8D shows of the expression
levels of FOLH1 and SSTR2 through the progression to
NEPC. In the AdPC model, FOLH1 is found at a maximum
level and SSTR2 at a minimum level. Following castration
up to 8weeks there is low fluctuations in the transcription
of both of the genes. However, on 8th week, a sharp
decline in FOLH1 and a profound elevation of SSTR2
starts. The terminally differentiated NEPC model has a
minimal transcriptional level of FOLH1 and maximal
transcription of SSTR2.
PSMA levels fail to adequately predict NE
transdierentiation of high-grade AdPC
Similar to its RNA, PSMA protein level could have a variable
series of staining levels of low grade (Supplementary
Fig. 5A and D), medium grade (Supplementary Fig. 5B
and E) and high grade (Supplementary Fig. 5C and F)
AdPC. These staining patterns imply a series of possible
progression paths for AdPC progression and eventual
NEPC. On Fig.8E, we proposed five possible models for
alteration of PSMA during progression of AdPC. Model 1 is
the classical model of PSMA expression in which the level
of PSMA correlates positively with the stage of AdPC, our
data support that this model may not adequately represent
all stages of disease. Model 5 is a representation of a de
novo occurrence of NEPC, which shows poor-onset PSMA
staining. However, models 2 and 3 are relevant as AdPC
progression pathways that lead to NE transdifferentiation
and subsequent loss of PSMA. In addition, model 4 is
representation of a constant poor PSMA staining, which
against its low abundance is still capable of detecting
metastatic sites. The HPA dataset PSMA antibody staining
0
2
4
6
8
10
Relative mRNA
level of
FOLH1
p<0.05
AdPC
Models
LTL-313-B
LTL-313-B-R
LTL-418
LTL-418-R
LTL-331-7
LTL-331R-G7
LTL-352
NEPC
Models
*LTL-313-B
**LTL-418
AdPC Model (PSA+/NE-)
*Biopsy of primary AdPC
**Radical prostatectomy of AdPC LTL-313-B-R
LTL-418-R
CRPC Model (PSA+/NE-)
Mice castration
A
LTL-331
PCa Model (PSA+/NE-)
Biopsy of primary AdPC LTL-331-7
CRPC Model (PSA+/NE-)
Passaging
Mice castration
LTL-331R-G7
NE Model (PSA-/NE+)
B
Surgical Removal of
NEPC metastases
LTL-352
NE Model (PSA-/NE+)
C
D
AdPC Models
NEPC Models
Figure7
Establishment and analysis of AdPC and NEPC PDX mouse models. (A, B, C and D) Schematic of the established PDX mice models of AdPC and NEPC
(adapted, with permission, from Lin etal. (2014)). (D) The levels of FOLH1 in dierent PDX models. One-way ANOVA followed by a Newman–Keuls multiple
comparison test was used (n = 3). Some elements of this gure were produced using Servier Medical Art image bank (www.servier.com) under the terms
of a Creative Commons Attribution 3.0 Unported licence. A full-colour version of this gure is available at https://doi.org/10.1530/ERC-18-0226.

AUTHOR COPY ONLY
https://erc.bioscientica.com © 2019 Society for Endocrinology
Printed in Great Britain
Published by Bioscientica Ltd.
https://doi.org/10.1530/ERC-18-0226
143
M K Bakht etal. NE dierentiation suppresses
PSMA
26:2
Endocrine-Related
Cancer
patterns for AdPC progression led us to conclude that
model 2 is the most likely progression pattern for the
treatment-induced NEPC.
Discussion
The development of NE phenotype following hormone
depletion in LNCaP cell line is a well-studied observation
(Shen etal. 1997, Yuan etal. 2006). Liu etal. (2012) reported
that hormone depletion suppresses AR and PSMA levels in
LNCaP cells; however, they did not evaluate the expression
of NE makers. Suppression of AR and overexpression of
NE makers in LNCaP following p53 knockdown in LNCaP
cell has also been described recently (Li et al. 2017, Mu
etal. 2017). In this manuscript, we report the suppression
of PSMA and overexpression of NSE and SSTR2 in LNCaP
cell line following hormone depletion and loss of p53.
Mannweiler et al. reported that a direct correlation
between histological parameters and PSMA could not be
established; specifically, in AdPC patients with distant
metastasis (Mannweiler et al. 2009). In this study, we
report the possibility of PSMA suppression in high grades
of CRPC. We further show that expression of the PSMA
gene, FOLH1, inversely correlates with markers of NE
Time after castration of LTL-331 model
0
3
6
9
12
15
18
FOLH1SSTR2
Normalized RNA-seqread counts
AdPC PDX models
NEPC PDX models
B
p<0.00001
P=0.047
Pathology
FOLH1
SSTR2
Sample name
NEPC AdPC
PDX models
A
Row Z-score
-3 -1 0 +1 +3
LTL-331
AdPC Model (PSA+/NE-)
ARPI sensitive
Biopsy of primary AdPC LTL-331R
NEPC Model (PSA-/NE+)
ARPI insensitive
Mice castration Terminally differentiated NEPC
RNA-seqanalysis
Day 1-7Week 2-12
Week 24-32
(Relapse)
0
2
4
6
8
10
12
14
SSTR2FOLH1
ANR dezilamroN-seqread counts
D
PSMA level
High
Medium
Low
Not
detectable
Prostate Adenocarcinoma grade
LowMedium High
Model 1
Model 2
Model 3
Model 4
Model 5
C
E
Figure8
NEPC represents a distinctive FOLH1-supressed signature in a series of PDX model. (A and B) Transcriptomic proles from the PDX models (15
adenocarcinomas vs 3 NEPCs), (A) heatmap showing the clustering among all PDX samples (B) the average level of FOLH1 and SSTR2 suggests a unique
downregulation of FOLH1 in NEPC PDX tumors while SSTR2 levels are slightly increased. The data were analyzed by Student’s t-test. (C) Schematic of
development of LTL-331R as terminally dierentiated NEPC PDX model following castration of hormone-sensitive LTL-331 PDX model. The time points at
which tumors were collected along progression to NEPC are illustrated by blue color arrows. Some elements of this gure were produced using Servier
Medical Art image bank (www.servier.com) under the terms of a Creative Commons Attribution 3.0 Unported licence. (D) Transcription of FOLH1 and
SSTR2 during NE transdierentiation in the LTL331 system highlighting the suppression of FOLH1 and amplication of SSTR2 during development of NEPC
as a result of hormone depletion. (E) Possible models of alteration of PSMA level during progression of AdPC. Schema shows possible kinetic changes in
PSMA level (Y axis) during progression from low-, medium- and high-grade AdPC. A full-colour version of this gure is available at https://doi.org/10.1530/
ERC-18-0226.

AUTHOR COPY ONLY
Printed in Great Britain
Published by Bioscientica Ltd.
https://doi.org/10.1530/ERC-18-0226
https://erc.bioscientica.com © 2019 Society for Endocrinology
144
M K Bakht etal. NE dierentiation suppresses
PSMA
26:2
Endocrine-Related
Cancer
differentiation. These data support previous clinical
reports showing a suppression of PSMA-targeted nuclear
scans for NEPC patients (Chakraborty etal. 2015, Tosoian
etal. 2017, Usmani etal. 2017). This mounts the possibility
that portions of AdPC patients undergoing ARPI will
transdifferentiate to NEPC and will subsequently fail to
adequately uptake PSMA-targeted radiopharmaceuticals.
Clinical reports show the possibility of a false-positive
uptake of PSMA-targeted radioligand after ARPI (Hope
etal. 2017) and a false-negative uptake of this radioligand
in NEPC (Chakraborty et al. 2015, Tosoian et al. 2017,
Usmani etal. 2017). We previously reviewed the available
preclinical evidence justifying molecular backgrounds
of a false-positive uptake of PSMA-targeted radioligand
after ARPI (Bakht etal. 2017). However, there is a lack of
preclinical evidence for the reported false-negative uptake
of PSMA-targeted radioligand. The observed significant
suppression of FOLH1 expression in PDX models shown
in this work are the first preclinical evidence validating
false-negative uptake of PSMA-targeted radioligand in
NEPC. The suggested second model of alteration of PSMA
on Fig.8E best describes the progression of AdPC patients
undergoing ARPI who end up with NEPC.
There is literature to support that SSTR-targeted
radioligands such as 68Ga-DOTATOC or 68Ga-DOTATATE
could be valuable alternatives for PSMA-negative and
non-hormone-naive patients. Usmani et al. compared
68Ga-PSMA vs 68Ga-DOTA NOC PET-CT in a 62-year-old
NEPC patient and reported that the NEPC tumor and
its metastatic lesions have avid uptake of 68Ga-DOTA
NOC while the PSMA uptake was faint and inconclusive
(Usmani et al. 2017). Additionally, in a study on 12
patients with CRPC, 68Ga-DOTATATE PET/CT is suggested
for NEPC imaging and early detection of metastatic lymph
node and blastic or lytic bone lesions (Gofrit etal. 2017,
Sheikhbahaei et al. 2017). In this study, we report an
overexpression of SSTR2 in NEPC that could potentially
lead to overexpression of SSTR2 at the protein level and
positivity in the uptake of SSTR radioligand. Interestingly,
Morichetti etal. reported a strong specific staining for SSTR2
in 20 cases of NEPC (Morichetti etal. 2010). We speculate
that SSTR radioligand might be a feasible tool to delineate
PSMA-negative NEPC from the normal tissue while PSMA-
negative NEPC is not detectable by PSA measurements
or by PSMA radiotracers. Future work to explore the role
of SSTR2-targeted radionuclide therapy for the detection
and/or management of NEPC is of high priority.
Two possible scenarios for a patient with a suppressed
PSMA radioligand uptake after ARPI are illustrated on
Supplementary Fig. 6. In an ideal condition, the PSMA
radioligand should be able to delineate AdPC tumor and all
metastatic lesions (Supplementary Fig.6A); consequently,
after therapy faint or a no PSMA radioligand uptake
should be observed since ARPI suppresses the malignancy
(Supplementary Fig. 6B). However, as a second possible
scenario, we might face a high-grade NEPC with suppressed
PSMA expression level (Supplementary Fig. 6C). In this
case, SSTR2 radioligand might be able to delineate NEPC
tumor and its metastatic lesions.
Conclusion
The use of potent antiandrogens and lineage plasticity
may contribute to the increasing prevalence of NEPC,
an aggressive and hormone-resistant form of AdPC.
While PSMA targeting is a promising approach for the
nuclear imaging and therapy of many forms of aggressive
AdPC, our data based on transcriptome analysis of tumor
samples, cell line models and PDX mice models supports
that PSMA (FOLH1 gene) levels are not consistent with
all forms of high-grade CRPC. Our in vitro data showed
a significant suppression of PSMA as a result of hormone
depletion, loss of p53 and lineage plasticity. A PSMA-
suppressed NEPC cell line model showed higher colony
formation and resistance to ARPI by ENZ. Our data further
show that induced NE transdifferentiation correlates
with an elevated level of the protein SSTR2. SSTR2 levels
appear to elevate due to hormone depletion but are not
affected by altered cellular plasticity. PDX data support
these conclusions, showing an inverse correlation
between the expression of the PSMA gene and SSTR2
gene. Specifically we find declining PSMA and increasing
SSTR2 gene expression occurring during the development
of CRPC and becoming more pronounced in terminally
differentiated NEPC. Collectively, this study cautions on
the reliability of using PMSA levels as a diagnostic target
for molecular imaging in advanced treatment-induced
NEPC. Our work supports the recommendation that
SSTR2-targeted imaging approaches may permit more
accurate monitoring of PSMA-suppressed patients due to
SSTR2-positivity of NEPC tumors.
Supplementary data
This is linked to the online version of the paper at https://doi.org/10.1530/
ERC-18-0226.
Declaration of interest
The authors declare that there is no conict of interest that could be
perceived as prejudicing the impartiality of the research reported.

AUTHOR COPY ONLY
https://erc.bioscientica.com © 2019 Society for Endocrinology
Printed in Great Britain
Published by Bioscientica Ltd.
https://doi.org/10.1530/ERC-18-0226
145
M K Bakht etal. NE dierentiation suppresses
PSMA
26:2
Endocrine-Related
Cancer
Funding
This research was supported by a grant of the Korea Health Technology
R&D Project through the Korea Health Industry Development Institute
(KHIDI), funded by the Ministry of Health & Welfare, Republic of Korea
(grant number: HI14C1072). The nancial support of Prostate Cancer Fight
Foundation (Motorcycle Ride for Dad Charity, ON, Canada) and Ontario
Trillium Scholarship program (Government of Ontario, ON, Canada) is
greatly appreciated.
Author contribution statement
The conception and design of the study was by M K B, G J C, A D L, K F S,
K W K, Y W, X D and L A P. M K B, I D, Y L and Y W did the experiments and
obtained the data. M K B, I D, Y L, A H and A D L analyzed the data. M K B, I D,
R M F, S W O, H Y and L A P drafted the manuscript. M K B, R F, H Y, K F S,
C W J, G J C, C K, M D, K W K, A D L, X D and L A P revised the manuscript.
References
AkamatsuS, WyattAW, LinD, LysakowskiS, ZhangF, KimS, TseC,
WangK, MoF & HaegertA 2015 The placental gene PEG10
promotes progression of neuroendocrine prostate cancer. Cell Reports
12 922–936. (https://doi.org/10.1016/j.celrep.2015.07.012)
Al SorkhyM, FerraiuoloR-M, JaliliE, MalysaA, FratiloiuAR, SloaneBF
& PorterLA 2012 The cyclin-like protein Spy1/RINGO promotes
mammary transformation and is elevated in human breast cancer.
BMC Cancer 12 45. (https://doi.org/10.1186/1471-2407-12-45)
BakhtMK, OhSW, YounH, Cheon GJ, KwakC & Kang KW 2017
Influence of androgen deprivation therapy on the uptake of PSMA-
targeted agents: emerging opportunities and challenges. Nuclear
Medicine and Molecular Imaging 51 202–211. (https://doi.org/10.1007/
s13139-016-0439-4)
BeltranH, PrandiD, MosqueraJM, BenelliM, PucaL, CyrtaJ, MarotzC,
GiannopoulouE, ChakravarthiBV & VaramballyS 2016 Divergent
clonal evolution of castration-resistant neuroendocrine prostate
cancer. Nature Medicine 22 298–305. (https://doi.org/10.1038/
nm.4045)
BoegemannM, SchraderAJ & RahbarK 2017 [177Lu-PSMA therapy:
current evidence for use in the treatment of patients with metastatic
prostate cancer]. Urologe Ausgabe A 56 1440–1444. (https://doi.
org/10.1007/s00120-017-0510-5)
ChakrabortyPS, TripathiM, AgarwalKK, KumarR, VijayMK & Bal C
2015 Metastatic poorly differentiated prostatic carcinoma with
neuroendocrine differentiation: negative on 68Ga-PSMA PET/CT.
Clinical Nuclear Medicine 40 e163–e166. (https://doi.org/10.1097/
RLU.0000000000000594)
ChandrashekarDS, BashelB, BalasubramanyaSAH, CreightonCJ,
Ponce-RodriguezI, ChakravarthiBVSK & VaramballyS 2017
UALCAN: a portal for facilitating tumor subgroup gene expression
and survival analyses. Neoplasia 19 649–658. (https://doi.
org/10.1016/j.neo.2017.05.002)
ChappellWH, LehmannBD, TerrianDM, AbramsSL, SteelmanLS &
McCubreyJA 2012 p53 expression controls prostate cancer
sensitivity to chemotherapy and the MDM2 inhibitor Nutlin-3. Cell
Cycle 11 4579–4588. (https://doi.org/10.4161/cc.22852)
ChenR, DongX & GleaveM 2018 Molecular model for neuroendocrine
prostate cancer progression. BJU International 122 560–570. (https://
doi.org/10.1111/bju.14207)
CiX, HaoJ, DongX, ChoiSY, XueH, WuR, QuS, GoutPW, ZhangF,
HaegertAM, etal. 2018 Heterochromatin protein 1alpha mediates
development and aggressiveness of neuroendocrine prostate cancer.
Cancer Research 78 2691–2704. (https://doi.org/10.1158/0008-5472.
CAN-17-3677)
da SilvaEF, BotsfordS & PorterLA 2016 Derivation of a novel G2
reporter system. Cytotechnology 68 19–24. (https://doi.org/10.1007/
s10616-014-9757-5)
DaviesAH, BeltranH & ZoubeidiA 2018 Cellular plasticity and the
neuroendocrine phenotype in prostate cancer. Nature Reviews Urology
15 271–286. (https://doi.org/10.1038/nrurol.2018.22)
DingY, LiY, LuL, ZhangR, ZengL, WangL & ZhangX 2015 Inhibition
of nischarin expression promotes neurite outgrowth through
regulation of PAK activity. PLoS ONE 10 e0144948. (https://doi.
org/10.1371/journal.pone.0144948)
DunningMJ, VowlerSL, LalondeE, Ross-AdamsH, BoutrosP, MillsIG,
LynchAG & LambAD 2017 Mining human prostate cancer datasets:
the “camcAPP” shiny app. EBioMedicine 17 5–6. (https://doi.
org/10.1016/j.ebiom.2017.02.022)
GaoJ, AksoyBA, DogrusozU, DresdnerG, GrossB, SumerSO, SunY,
JacobsenA, SinhaR & LarssonE 2013 Integrative analysis of
complex cancer genomics and clinical profiles using the cBioPortal.
Science Signaling 6 pl1. (https://doi.org/10.1126/scisignal.2004088)
GofritON, FrankS, MeirovitzA, NechushtanH & OreviM 2017 PET/CT
with 68Ga-DOTA-TATE for diagnosis of neuroendocrine:
differentiation in patients with castrate-resistant prostate cancer.
Clinical Nuclear Medicine 42 1–6. (https://doi.org/10.1097/
RLU.0000000000001424)
GrassoCS, WuY-M, RobinsonDR, CaoX, DhanasekaranSM, KhanAP,
QuistMJ, JingX, LonigroRJ, BrennerJC, etal. 2012 The mutational
landscape of lethal castration-resistant prostate cancer. Nature 487
239–243. (https://doi.org/10.1038/nature11125)
HopeTA, TruilletC, EhmanEC, Afshar-OromiehA, AggarwalR,
RyanCJ, CarrollPR, SmallEJ & EvansMJ 2017 68Ga-PSMA-11 PET
imaging of response to androgen receptor inhibition: first human
experience. Journal of Nuclear Medicine 58 81–84. (https://doi.
org/10.2967/jnumed.116.181800)
HothornT, HornikK & ZeileisA 2006 Unbiased recursive partitioning:
a conditional inference framework. Journal of Computational and
Graphical Statistics 15 651–674. (https://doi.org/10.1198/1061860
06X133933)
JadvarH & BallasLK 2018 PSMA PET: transformational change in
prostate cancer management? Journal of Nuclear Medicine 59
228–229. (https://doi.org/10.2967/jnumed.117.203604)
KopkaK, BenešováM, BařinkaC, HaberkornU & BabichJ 2017 Glu-
ureido–based inhibitors of prostate-specific membrane antigen:
lessons learned during the development of a novel class of low-
molecular-weight theranostic radiotracers. Journal of Nuclear Medicine
58 17S–26S. (https://doi.org/10.2967/jnumed.116.186775)
LiW, CohenA, SunY, SquiresJ, BraasD, GraeberTG, DuL, LiG, LiZ,
XuX, etal. 2016 The role of CD44 in glucose metabolism in
prostatic small cell neuroendocrine carcinoma. Molecular Cancer
Research 14 344–353. (https://doi.org/10.1158/1541-7786.MCR-15-
0466)
LiY, DonmezN, SahinalpC, XieN, WangY, XueH, MoF, BeltranH,
GleaveM, WangY, etal. 2017 SRRM4 drives neuroendocrine
transdifferentiation of prostate adenocarcinoma under androgen
receptor pathway inhibition. European Urology 71 68–78. (https://doi.
org/10.1016/j.eururo.2016.04.028)
LinD, WyattAW, XueH, WangY, DongX, HaegertA, WuR,
BrahmbhattS, MoF & JongL 2014 High fidelity patient-derived
xenografts for accelerating prostate cancer discovery and drug
development. Cancer Research 74 1272–1283. (https://doi.
org/10.1158/0008-5472.CAN-13-2921-T)
LiuT, WuLY, FultonMD, Johnson JM & BerkmanCE 2012 Prolonged
androgen deprivation leads to downregulation of androgen receptor
and prostate-specific membrane antigen in prostate cancer cells.
International Journal of Oncology 41 2087–2092. (https://doi.
org/10.3892/ijo.2012.1649)
MannweilerS, AmersdorferP, TrajanoskiS, TerrettJA, KingD &
MehesG 2009 Heterogeneity of prostate-specific membrane antigen

AUTHOR COPY ONLY
Printed in Great Britain
Published by Bioscientica Ltd.
https://doi.org/10.1530/ERC-18-0226
https://erc.bioscientica.com © 2019 Society for Endocrinology
146
M K Bakht etal. NE dierentiation suppresses
PSMA
26:2
Endocrine-Related
Cancer
(PSMA) expression in prostate carcinoma with distant metastasis.
Pathology Oncology Research 15 167–172. (https://doi.org/10.1007/
s12253-008-9104-2)
MorichettiD, MazzucchelliR, SantinelliA, StramazzottiD, Lopez-
BeltranA, ScarpelliM, BonoA, ChengL & MontironiR 2010
Immunohistochemical expression and localization of somatostatin
receptor subtypes in prostate cancer with neuroendocrine
differentiation. International Journal of Immunopathology and
Pharmacology 23 511–522. (https://doi.
org/10.1177/039463201002300213)
MuP, ZhangZ, BenelliM, KarthausWR, Hoover E, ChenCC,
WongvipatJ, KuSY, GaoD, CaoZ, etal. 2017 SOX2 promotes
lineage plasticity and antiandrogen resistance in TP53- and RB1-
deficient prostate cancer. Science 355 84–88. (https://doi.org/10.1126/
science.aah4307)
NabaviN, SaidyNRN, VenalainenE, HaegertA, ParoliaA, XueH,
WangY, WuR, DongX & CollinsC 2017 miR-100-5p inhibition
induces apoptosis in dormant prostate cancer cells and prevents the
emergence of castration-resistant prostate cancer. Scientific Reports 7
4079. (https://doi.org/10.1038/s41598-017-03731-8)
ParidaGK, TripathyS, Datta GuptaS, SinghalA, KumarR, BalC &
ShamimSA 2018 Adenocarcinoma prostate with neuroendocrine
differentiation: potential utility of 18F-FDG PET/CT and
68Ga-DOTANOC PET/CT over 68Ga-PSMA PET/CT. Clinical Nuclear
Medicine 43 248–249. (https://doi.org/10.1097/RLU.0000000000002013)
ParimiV, GoyalR, PoropatichK & YangXJ 2014 Neuroendocrine
differentiation of prostate cancer: a review. American Journal of
Clinical and Experimental Urology 2 273–285.
PoeppelTD, BinseI, PetersennS, LahnerH, SchottM, AntochG,
BrandauW, BockischA & BoyC 2011 68Ga-DOTATOC versus
68Ga-DOTATATE PET/CT in functional imaging of neuroendocrine
tumors. Journal of Nuclear Medicine 52 1864–1870. (https://doi.
org/10.2967/jnumed.111.091165)
RaiBP, BaumR, PatelA, HughesR, Alonzi R, LaneT, AdsheadJ &
VasdevN 2016 The role PET with 68 gallium (Ga)-labelled prostate-
specific membrane antigen (PSMA) in the Management of Patient
with organ confined and locally advanced prostate cancer prior to
radical treatment and after radical prostatectomy. Urology 95 11–15.
(https://doi.org/10.1016/j.urology.2015.12.048)
Ross-AdamsH, LambAD, DunningMJ, HalimS, LindbergJ, MassieCM,
EgevadLA, RussellR, Ramos-MontoyaA, VowlerSL, etal. 2015
Integration of copy number and transcriptomics provides risk
stratification in prostate cancer: a discovery and validation cohort
study. EBioMedicine 2 1133–1144. (https://doi.org/10.1016/j.
ebiom.2015.07.017)
RoweSP, GorinMA & PomperMG 2017 Imaging of prostate-specific
membrane antigen using [18F] DCFPyL. PET Clinics 12 289–296.
(https://doi.org/10.1016/j.cpet.2017.02.006)
SheikhbahaeiS, Afshar-OromiehA, EiberM, SolnesLB, JavadiMS,
RossAE, PientaKJ, AllafME, HaberkornU, PomperMG, etal. 2017
Pearls and pitfalls in clinical interpretation of prostate-specific
membrane antigen (PSMA)-targeted PET imaging. European Journal of
Nuclear Medicine and Molecular Imaging 44 2117–2136. (https://doi.
org/10.1007/s00259-017-3780-7)
ShenR, DoraiT, SzabolesM, KatzAE, OlssonCA & ButtyanR 1997
Transdifferentiation of cultured human prostate cancer cells to a
neuroendocrine cell phenotype in a hormone-depleted medium.
Urologic Oncology: Seminars and Original Investigations 3 67–75.
(https://doi.org/10.1016/S1078-1439(97)00039-2)
TaylorBS, SchultzN, HieronymusH, Gopalan A, XiaoY, CarverBS,
AroraVK, KaushikP, CeramiE, RevaB, etal. 2010 Integrative
genomic profiling of human prostate cancer. Cancer Cell 18 11–22.
(https://doi.org/10.1016/j.ccr.2010.05.026)
ThulPJ, ÅkessonL, WikingM, MahdessianD, GeladakiA, Ait BlalH,
AlmT, AsplundA, BjörkL, BreckelsLM, etal. 2017 A subcellular
map of the human proteome. Science 356 eaal3321. (https://doi.
org/10.1126/science.aal3321)
TosoianJJ, GorinMA, RoweSP, AndreasD, SzaboZ, PientaKJ,
PomperMG, LotanTL & RossAE 2017 Correlation of PSMA-targeted
(18)F-DCFPyL PET/CT findings with immunohistochemical and
genomic data in a patient with metastatic neuroendocrine prostate
cancer. Clinical Genitourinary Cancer 15 e65–e68. (https://doi.
org/10.1016/j.clgc.2016.09.002)
UhlenM, ZhangC, LeeS, SjöstedtE, FagerbergL, BidkhoriG,
BenfeitasR, ArifM, LiuZ, EdforsF, etal. 2017 A pathology atlas of
the human cancer transcriptome. Science 357 eaan2507. (https://
doi.org/10.1126/science.aan2507)
UnniE, SunS, NanB, McPhaulMJ, CheskisB, ManciniMA &
MarcelliM 2004 Changes in androgen receptor nongenotropic
signaling correlate with transition of LNCaP cells to androgen
independence. Cancer Research 64 7156–7168. (https://doi.
org/10.1158/0008-5472.CAN-04-1121)
UsmaniS, AhmedN, MarafiF, RasheedR, AmangunoHG & Al
KandariF 2017 Molecular imaging in neuroendocrine differentiation
of prostate cancer: 68Ga-PSMA versus 68Ga-DOTA NOC PET-CT.
Clinical Nuclear Medicine 42 410–413. (https://doi.org/10.1097/
RLU.0000000000001618)
VlachostergiosPJ & PapandreouCN 2015 Targeting neuroendocrine
prostate cancer: molecular and clinical perspectives. Frontiers in
Oncology 5 6. (https://doi.org/10.3389/fonc.2015.00006)
YuanTC, VeeramaniS, LinFF, KondrikouD, ZelivianskiS, IgawaT,
KaranD, BatraSK & LinMF 2006 Androgen deprivation induces
human prostate epithelial neuroendocrine differentiation of
androgen-sensitive LNCaP cells. Endocrine-Related Cancer 13 151–167.
(https://doi.org/10.1677/erc.1.01043)
ZelivianskiS, VerniM, MooreC, KondrikovD, TaylorR & LinM-F 2001
Multipathways for transdifferentiation of human prostate cancer
cells into neuroendocrine-like phenotype. Biochimica et Biophysica
Acta (BBA) – Molecular Cell Research 1539 28–43. (https://doi.
org/10.1016/S0167-4889(01)00087-8)
Received in final form 24 September 2018
Accepted 11 October 2018
Accepted Preprint published online 11 October 2018