The HDL receptor SR-BI is associated with human prostate cancer progression and plays a possible role in establishing androgen independence

Evaluation of SR-BI and LDLR expression as markers for prostate cancer progression

We analyzed SR-BI and LDLR mRNA expression in primary human prostate cancers from
the Expression Project for Oncology and from a hallmark study by Lapointe et al.16]. When comparing prostate cancer samples with high Gleason scores (equal to or higher
than 7) and samples with low Gleason scores (equal to or lower than 6), SR-BI was
more highly expressed in GSE2109 (n?=?56, P?=?0.039, Fig. 1a) and GSE3933 (n?=?58, P?=?0.016, Fig. 1c). LDLR was not increased in high Gleason score samples in GSE2109 (n?=?56, P?=?0.844, Fig. 1b) and in GSE3933 (n?=?60, P?=?0.219, Fig. 1d).

Fig. 1. Expression of SR-BI and LDLR in clinical prostate samples. Differential expression
of SR-BI and LDLR according to Gleason scoring (a–d). All samples were grouped according to Gleason score (GS)???6 and???7. Differential
expression of SR-BI and LDLR in metastasizing and non-metastasizing prostate tissues
(e-j). All samples were grouped into primary tumors (primary site) and metastasizing tumors
(metastasis). P-values are presented within each graph. The arithmetic mean is given as a line within
the dots. Data displayed as log 2 median-centered intensity

The most important parameter used to decide upon patient survival is the occurrence
of metastasis. Therefore, we next determined the expression of SR-BI and LDLR in clinical
prostate samples derived either from non-metastatic or metastatic prostate cancer.
For this analysis, we investigated the dataset GSE35988, which contains benign and
localized prostate cancer from radical prostatectomy as well as metastatic, castration-resistant
prostate cancer obtained from rapid autopsy 17]. Furthermore, we included the datasets GSE6919 and GSE3933, which contain samples
from the primary tumor site as well as metastasized prostate cancer samples from the
liver, lung, kidney, adrenal gland or lymph nodes 16], 18], 19]. We identified an increased expression of SR-BI in metastatic prostate samples compared
to non-metastatic prostate samples in GSE35988, GSE3933 and GSE6919 (P??0.001, P?=?0.009 and P?=?0.017, respectively) (Fig. 1e, g, i,). Contrary to SR-BI, LDLR expression was not increased in metastatic prostate samples
compared to non-metastatic prostate samples in GSE35988, GSE3933 and GSE6919 (P?=?0.341, P?=?0.139 and P?=?0.856, respectively) (Fig. 1f, h, j).

Association of clinicopathological parameters with SR-BI expression

To confirm findings from mRNA expression studies, we assessed SR-BI protein expression
in normal prostate tissue and prostate cancer samples derived from patients with known
TNM status. A total of 106 biopsy cores were independently assigned to Gleason scoring.
Cores were subjected to immunohistochemical staining for SR-BI and afterwards analyzed
for their staining intensity. Table 1 shows the clinicopathological characteristics of the cohort studied. Samples were
classified into non-cancer and cancer samples. Cancer samples were further characterized
by Gleason score, pathologically classified tumor stage (pT) and metastasis. Nearly
half of all cancer samples showed a Gleason score equal to or above 7. Furthermore,
about a third of all cancer samples showed advanced tumor stages (pT3/4), and nearly
half of them were positive for metastasis. Representative cores for different staining
intensities of SR-BI and their respective scores are shown in Fig. 2 (a–d). We tested the specificity of our antibodies by staining human liver sections (Fig. 2e and f). The distribution of the results for the whole tissue collective is shown in Fig. 2g. On the basis of a binary classification system for low (score 0 and score 1) and
high (score 2 and score 3) SR-BI staining intensities, we evaluated associations between
SR-BI and the presence or absence of clinicopathological parameters. Notably, out
of 23 normal prostate samples, none showed high staining results (score 2 and score
3), and among all cancer samples, approximately 53.6 % showed high staining for SR-BI.

Table 1. Frequencies of clinicopathological characteristics

Fig. 2. Immunohistochemical staining of prostate tissue and control tissue for SR-BI expression.
Each panel shows representative prostate samples scored for staining intensity as
follows: 0 for negative (a), 1 for low (b), 2 for moderate (c) and 3 for high (d). Liver sections were used as positive controls (pos. ctrl.) in standard magnification
(e) and in high magnification (high mag.), demonstrating SR-BI localization in the outer
cell membrane of hepatocytes (f). The overall scoring distribution for SR-BI staining intensity in different clinicopathological
groups (g). Scores of 0 and 1 represented “low” expression and are shown in green colors, whereas
scores of 2 and 3 represented “high” expression and are shown in blue colors. Clinicopathological
groups are plotted on the x-axis: non-cancer and cancer, Gleason score (GS)???6 and???7,
pathologic tumor stage 2 (pT2) and pathologic tumor stage 3/4 (pT3/4), metastasis
negative (met. neg.) and metastasis positive (met. pos.)

As shown in Table 2, we identified high SR-BI expression to be associated with the presence of prostate
cancer when compared to non-cancer prostate tissue (risk ratio?=?2.154, P??0.0001). Furthermore, we identified an association of high SR-BI score with a Gleason
score equal to or higher than 7 (risk ratio?=?2.907, P??0.0001).

Table 2. Evaluation of the prognostic significance of SR-BI staining intensity

Because SR-BI expression showed an association with prostate cancer differentiation,
we also tested for LDLR expression on selected sections with either low or high SR-BI
staining intensity (Fig. 3a–d). Our case study showed that LDLR was constitutively expressed in prostate tissue,
with lower expression levels in high-grade cancer samples. Interestingly, we also
observed cases of low-grade prostate cancer, which displayed high SR-BI expression
in a subpopulation of cells showing signs of tissue invasion (Fig. 3e and f). Cancer cells of this subpopulation either grew detached from the primary tumor,
floating in the remaining glands (Fig. 3e), or separated from the solid tumor mass, infiltrating the surrounding tissue (Fig. 3f).

Fig. 3. Lipoprotein receptor expression patterns in prostate cancer (high grade and low grade).
Histologic staining for LDLR and SR-BI in selected patients. Tissue derived from a
27-year-old patient diagnosed with prostate hyperplasia was stained for LDLR (a); a consecutive area of the same tumor was stained for SR-BI (b). Tissue derived from a 75-year-old patient diagnosed with prostate cancer T2N1M1c
and Gleason score 5?+?4 was stained for LDLR (c); a consecutive area of the same tumor was stained for SR-BI (d). Tumor biopsies from a 72-year-old patient diagnosed with prostate cancer T2N0M0
and Gleason score 3?+?3 were analyzed for SR-BI expression (e and f). Black arrows indicate clusters of cells strongly positive for SR-BI. Kaplan Meier
analysis of LDLR and SR-BI expression in GSE40272 (g–h). High LDLR expression had no effect on disease-free survival time (g). High SR-BI expression was associated with decreased disease-free survival time
(h). Green?=?high expression (high), blue?=?low expression (low), DFS?=?disease-free
survival time. P-values of the log-rank test are presented within each graph

To assess whether SR-BI and LDLR had any influence on the clinical outcome of patients,
we chose to evaluate the disease-free survival time in relation to SR-BI and LDLR
expression. Therefore, we performed Kaplan-Meier analyses of the dataset GSE40272,
which contained mRNA expression data on prostate tissue samples from men who underwent
radical prostatectomy 20]. The disease-free survival time was defined as the time between surgery and the recurrence
of disease (serum PSA??0.1 ng/ml on two consecutive measurements after surgery).
We identified samples with low SR-BI expression to have a significantly better survival
outcome compared to samples with high SR-BI expression (p?=?0.02, Fig. 3h). By contrast, there was no significant difference in disease-free survival time
between samples with low LDLR expression and samples with high LDLR expression (Fig. 3g).

Correlation of SR-BI with androgen-synthesizing enzymes and the mTOR pathway

Because SR-BI mediates the selective uptake of cholesterol, which can be used for
steroidogenesis, we analyzed enzymes that participate in androgen synthesis. We identified
the ?-hydroxysteroid-dehydrogenases HSD17B1 and HSD17B3 to be significantly up-regulated
in metastatic compared to non-metastatic prostate cancer; they also correlated with
the intensity of SR-BI expression (Fig. 4a–h).

Fig. 4. Correlation of SR-BI with androgen-synthesizing enzymes and the mTOR pathway. Differential
expression of HSD17B1 and HSD3B1 in primary tumors (primary site) and metastasizing
tumors (metastasis) (a, c, e, g). The arithmetic mean is given as a line within the dots and the P-values of t-test analysis is given within each graph. Scatter plots correlating SR-BI expression
with HSD17B1 and HSD3B1 in prostate cancer (b, d, f, h). P-values of Pearson correlation analyses and regression lines are presented within
each graph. Red dots represent metastatic and blue dots represent non-metastatic prostate
cancer. Analysis of the co-occurrence of SR-BI and ribosomal protein S6 phosphorylation
(i–l). A representative high grade prostate cancer is shown with staining for SR-BI (i) and for S6 phosphorylation at serine 240 and 244 in a consecutive area of the same
tumor (j). A representative low grade prostate cancer is shown with staining for SR-BI (k) and for S6 phosphorylation at serine 240 and 244 in a consecutive area of the same
tumor (l). pS6?=?ribosomal protein S6 phosphorylation at serine 240 and 244

Mechanistic studies have shown that mTOR signaling can mediate androgen independence
21]. Therefore, we further assessed the association of SR-BI expression with serine phosphorylation
of ribosomal protein S6 at position 240 and 244. A total of 22 biopsy cores were subjected
to immunohistochemical staining for pS6, and adjacent sections from the same patients
were simultaneously subjected to SR-BI staining. Representative histologic staining
for high grade and low grade prostate carcinoma samples is shown in Fig. 4 (i–l). After the pS6 and SR-BI staining, the samples were analyzed for their staining
intensity. Representative cores for different staining intensities of pS6 and their
respective scores are shown in Additional file 1: Figure S1 (A–D). Spearman correlation analysis revealed a significant positive correlation
of SR-BI and pS6 (R?=?0.828, p??0.001).