Introduction

Alternative mRNA isoforms have important roles in normal development and physiology. Almost every human gene produces more than one mRNA isoform, vastly expanding the information content of the human genome (Djebali et al. 2012). Alternative and aberrant pre-mRNA splice isoforms also play an important role in cancer. In fact, altered splicing patterns have been suggested as a new “hallmark” of cancer cells, in addition to other well-established hallmarks of cancer such as evasion of cell death and metastasis (Hanahan and Weinberg 2011; Ladomery 2013; Oltean and Bates 2014). Recent data indicate a key role for splicing pattern changes in the pathology of prostate cancer. Alternative splicing programmes in prostate cancer have been the topic of excellent reviews (Hagen and Ladomery 2012; Lu et al. 2015; Nakazawa et al. 2014; Rajan et al. 2009b; Sette 2013). Here we particularly concentrate on developments in the last 3 years.

Alternative splicing patterns include whole exons being either spliced in or left out (exon skip**) and alternative utilisation of both 5′ and 3′ splice sites to insert exons of different sizes. Alternative mRNA isoforms can also be generated by the selection of different promoters and different polyadenylation sites. Each of these pathways produces mRNA splice variants that can impact on prostate cancer development or response to treatment (summarised in Fig. 1).

Fig. 1
figure 1

Different kinds of splicing pattern and their effect on prostate cancer cell biology. The most common form of alternative splicing in human cells is shown, with key examples from prostate cancer (not shown, whole introns can also be left in the mRNA)

Full size image

Prostate cancer is the second most frequent male cancer in the UK, with 47,300 cases being diagnosed in the UK in 2014 (corresponding to 130 new cases diagnosed per day, http://www.cancerresearchuk.org/health-professional/cancer-statistics/statistics-by-cancer-type/prostate-cancer#heading-Zero). The prostate is a small gland located just below the bladder that produces seminal fluid components. Prostate cancers typically do not cause symptoms, but in some more advanced stages can block urine flow from the bladder, invade the adjacent seminal vesicles and metastasise more distantly to bone. Primary prostate cancer leads to a breakdown in the normal glandular structure of the prostate gland. Prostate cancer is classified histologically by morphological features using the Gleason scoring system (Gleason and Mellinger 1974) (for example, see Fig. 2).

Fig. 2
figure 2

Prostate tissue visualised using tissue biopsies. a, b. Histological sections made from benign prostatic hyperplasia (BPH, with normal glandular structure embedded in stroma). Prostate cancer development is clinically described as a series of Gleason grades (1–5, with 1 corresponding to well-differentiated tissue containing a glandular structure and 5 being the most advanced with only few glands still visible) (Gleason and Mellinger 1974; Mellinger et al. 1967). c, d Histological sections made from prostate cancer (Gleason grade 5, notice breakdown of glandular structure). Left panels, sections processed using H&E staining. Right panels, sections processed by staining with haematoxylin and counterstaining with the RNA-binding protein Sam68 (brown stain).

Figure adapted from (Rajan et al. 2008) with permission from the Journal of Pathology

Full size image

Alternative splicing of genes under androgen control in prostate cancer

Clinical progression of prostate cancer is fuelled by a group of steroid hormones called androgens (Livermore 2016; Mills 2014). Androgens are small hydrophobic molecules that can cross the cell membrane, and include the male hormone testosterone. Once within cells, androgens bind to a nuclear hormone receptor protein called the androgen receptor (AR). The AR typically has a default cytoplasmic location in the absence of androgens, but translocates into the nucleus after binding to androgens via its ligand-binding domain (LBD). Once inside the nucleus, the AR binds to DNA target sequences via its DNA-binding domain (DBD) and controls patterns of downstream gene transcription via its N-terminal TF domain (Fig. 3a). The AR transcriptionally controls in the order of 700 genes within prostate cancer cells (Munkley et al. 2016).

Fig. 3
figure 3

Transcriptional control by a the full-length androgen receptor and b constitutively active AR isoforms made by splice variants. In (a), testosterone enters the prostate cancer cell and becomes modified to dihydroxytestosterone (DHT) by 5α-reductase. DHT binds to the androgen receptor (AR), displacing heat shock protein 90 (HSP) and resulting in AR translocation into the nucleus. Once inside the nucleus, the AR binds to consensus DNA sequence elements called androgen response elements (AREs) to control target gene expression. In (b), an androgen receptor variant protein (AR-V) lacking the ligand-binding domain is able to directly translocate into the nucleus without binding to DHT, resulting in androgen-independent control of gene expression

Full size image

Androgen hormones can affect splicing patterns as well as transcription. Many splicing decisions are made on nascent RNAs while their transcription is still in progress (Kornblihtt 2006). Increased transcriptional speeds from androgen-regulated promoters in the presence of androgens could potentially provide the spliceosome with a choice of exons to include into the mRNA. Consistent with this, transcripts from a CD44 minigene driven from a steroid-responsive promoter show increased exon skip** in response to the joint presence of the AR and androgens (Rajan et al. 2008). The AR also recruits some RNA-binding proteins as cofactors that can affect splicing. These include the RNA-binding protein Sam68 that decreases skip** of CD44 variable exons, and the RNA helicase p68 that increases skip** of CD44 variable exons (Fig. 1) (Clark et al. 2008; Rajan et al. 2008).

The above experiments used artificial model minigenes to investigate splicing control of AR target genes. Initial searches to find endogenous androgen-dependent splice isoforms used exon microarrays to probe the entire transcriptome (Rajan et al. 2011). These initial experiments identified two endogenous androgen-dependent cassette exons, one in the ZNF121 gene that encodes a zinc finger-containing protein (this ZNF121 exon was activated by androgens) and one in the NDUFV3 gene that encodes a mitochondrial respiratory protein (this NDUFV3 exon was repressed by androgens). Although these ZNF121 and NDUFV3 exons changed splicing in response to androgens, their clinical importance is not known, nor if these genes are direct targets for the AR. However, this same study also identified a number of androgen-dependent mRNAs made from alternative promoters, including an alternative mRNA isoform of the normally tumour suppressor TSC2 (Tuberous Sclerosis 2) gene that is transcribed from an internal promoter (so only contains downstream TSC2 exons). The well-characterised full-length TSC2 protein represses cell growth via the mTOR pathway. The androgen-driven alternative TSC2 mRNA isoform encodes a shorter (C-terminal only) TSC2 protein that promotes rather than represses cell growth (Munkley et al. 2014).

More recent RNAseq analysis of prostate cancer cell transcriptomes identified an androgen-regulated and prostate-specific ST6GALNAC1 mRNA splice isoform. ST6GALNAC1 encodes the important ST6GalNac1 enzyme that synthesises the cancer-associated sialyl-Tn (sTn) antigen (Munkley et al. 2015). In prostate cancer cells, an alternative spliced ST6GALNAC1 mRNA isoform that encodes a shorter ST6GalNac1 protein isoform is made. This shorter ST6GalNac1 protein isoform is actually made from a longer mRNA, since an extra exon is included within the 5′ untranslated region (UTR) of the ST6GALNAC1 mRNA in the prostate. The presence of this extra exon causes an alternative start codon to be utilised, resulting in the shorter version of the ST6GalNac1 protein (Munkley et al. 2015). This shorter ST6GalNac1 protein is produced at higher levels in prostate cancer cells than the previously reported full-length protein; yet it is able to synthesise the sTn antigen which is linked to patient survival and metastasis, and controls cell adhesion (Munkley et al. 2015). The changed 5′ UTR structure of the prostate-specific mRNA isoform of ST6GalNac1 might even enhance its translation, resulting in increased ST6GalNAc1 enzyme levels and more synthesis of sTn antigen (Munkley 2016). Both the short and long isoforms of ST6GalNac1 use the same androgen-driven promoter, but the 5′ UTR exon-skipped isoform has only been reported thus far in the prostate.

Alternative splicing patterns of an androgen-regulated oncogenic fusion gene called ETS-related gene (ERG) are associated with more advanced forms of prostate cancer progression. ERG is a proto-oncogene that plays a key role in the pathology of prostate cancer. ERG encodes a transcription factor that controls the expression of many genes during normal development (Adamo and Ladomery 2016). ERG is not normally transcriptionally controlled by androgens, but gene fusions can place ERG under transcriptional control by the androgen-regulated TMPRSS2 gene on the transition between prostatic epithelial neoplasia to prostate carcinoma. Such genetic rearrangements frequently link exon 1 or exons 1 and 2 of the TMPRSS2 gene, and the upstream TMPRSS2 gene promoter, to exon 4 and downstream regions of the ERG gene, by removing intervening regions of chromosome 21.

The TMPRSS2–ERG fusion gene is one of the most frequently over-expressed genes in prostate cancer. The encoded TMPRSS2–ERG fusion protein controls many important properties of prostate cancer cells, including cytoskeletal organisation, cell proliferation, expression of prostate-specific antigen (abbreviated PSA; this is a key serum biomarker for detecting prostate cancer) and epithelial–mesenchymal transitions (EMT) crucial for prostate cancer cell metastasis (Adamo and Ladomery 2016). Reverse transcription PCR (RT-PCR) analyses of tumour mRNA from patients show that more clinically advanced prostate cancers with histological evidence for seminal gland invasion have decreased skip** of two cassette exons within the ERG gene, compared with more localised prostate cancers and benign prostate tissue (Hagen et al. 2014). Increased splicing inclusion of these two ERG exons might contribute to the production of more oncogenic isoforms of the ERG fusion protein. One of these alternatively spliced ERG exons (of 72 nucleotides in length) encodes an in-frame peptide sequence. Inclusion of this 72-nucleotide exon might affect the interactions of the encoded ERG protein with transcription factors and other nuclear proteins, since it is immediately adjacent to the region of the ETS gene that encodes a protein–protein interaction domain called the sterile alpha motif (SAM)/pointed domain (http://pfam.xfam.org/family/SAM_PNT).

Changes in AR mRNA splicing patterns enable prostate cancer cells to become hormone resistant

The primary therapeutic strategy for advanced prostate cancer treatment is to block androgen signalling through androgen deprivation therapy or AR blockade, thereby halting tumour progression. Abiraterone is a drug that inhibits androgen biosynthesis and so reduces the levels of androgens within prostate cancer cells. Enzalutamide is a drug that antagonises the interaction of androgens with AR protein. Although prostate tumours initially respond to androgen deprivation therapy, later stages of the disease can develop into a treatment-resistant form of the disease called castration-resistant prostate cancer. Mechanisms of prostate cancer cell resistance cross over, so a prostate cancer cell develo** resistance to enzalutamide will also show a reduced (~20%) response to abiraterone, and vice versa (Liu et al. 2016).

Changing patterns of AR pre-mRNA splicing play a critical role in enabling prostate cancer cells to develop castration resistance. Changed splicing patterns generate variant AR protein isoforms (abbreviated AR-V) that lack ligand-binding domains, frequently as a result of splicing inclusion of cryptic exons (abbreviated CE), making them independent of androgen control. Some AR-V protein variants translocate into the nucleus even in the presence of enzalutamide (Fig. 3). AR pre-mRNA splicing changes thus enable prostate cancer cells to proliferate during androgen deprivation therapy in reduced circulating androgen concentrations.

Around 20 AR variant splice isoforms have been implicated in the development of hormone refractory prostate cancer [reviewed by Lu and Luo (2013)]. The clinically most frequent AR splice variant, called ARv7, is produced by splicing inclusion of a cryptic exon called CE3 located within intron 3 (Fig. 4). AR CE3 is a terminal exon, meaning that splicing inclusion is linked to the selection of a new poly(A) site, thus creating a truncated AR mRNA that lacks coding information for the Ligand-Binding Domain (abbreviated LBD). As a result, ARv7 encodes a short yet constitutively active isoform of the AR (active in the absence of androgens). While full-length AR protein is dependent on androgen binding via its LBD to translocate into the nucleus and control transcriptional activity, ARv7 protein is constitutively present in the nucleus even in the absence of androgens and so can provide AR activity in androgen-depleted prostate cancer cells (Cao et al. 2014). Expression of ARv7 is important for prostate cancer cell growth: siRNA-mediated depletion of ARv7 (using siRNAs complementary to CE3) inhibits the growth of the VCaP cell line in androgen-limiting conditions (Liu et al. Full size image

Supporting ARv7 protein or RNA expression being a potentially useful biomarker for disease progression, AR variant splice isoform levels change during prostate cancer development. Expression levels of ARv7 mRNA in patients with prostate cancer predict their pharmacological response to enzalutamide and abiraterone (Antonarakis et al. 2014). Levels of nuclear Arv7 protein can also be monitored by immunohistochemistry using a monoclonal antibody and are predictive of overall survival (Welti et al. 2016). Prostate cancer cells metastasise to bone via circulating tumour cells which are released from the primary tumour. Single-cell RNAseq analysis of circulating tumour cells purified from the bloodstream identified AR splice variants in most (8 out of 11 sequenced) patients, but less frequently in primary tumours (Miyamoto et al. 2015). Consistent with ARv7 expression being an adaptive response of cells to reduced androgen levels, ARv7 splice isoform levels increase in response to androgen deprivation therapy and decrease on the reintroduction of androgens (Yu et al. 2014). Expression of ARv7 is controlled by the transcription factors Myc (Myc also controls the expression of full-length AR) and NFκβ2, both of which increase expression in prostate cancer (Nadiminty et al. 2015).

Rather than providing a like-for-like replacement with the full-length androgen receptor, ARv7 protein instead preferentially regulates the expression of genes involved in active cell division and so promotes cell division rather than differentiation (Hu et al. 2012; Nakazawa et al. 2014). ARv7 expression in the LNCaP prostate cancer cell line also changes patterns of cell metabolism, decreasing the production of citrate and increasing the breakdown of glutamine—features of the “Warburg effect” changes in cancer metabolism that are also observed in prostate tumours (Shafi et al. 2015).

ARv567es is a further cancer-associated AR splice form observed in patients and also encodes a constitutively active AR protein (Fig. 4). ARv567es mRNA is made through skip** of exons 5–7 of the AR pre-mRNA and is only expressed within prostate cancer (and not the normal prostate). The ARv567es protein isoform regulates the transcription of genes involved in cell cycle control and particularly activates the expression of the oncogene UBE2C. UBE2C is a ubiquitin-conjugating protein active during mitosis, as part of the anaphase-promoting complex (APC) that inactivates the mitotic checkpoint control. UBE2C protein is highly expressed in solid tumours and promotes cell proliferation. Transcriptional activation by the ARv567es AR isoform occurs via a DNA loo** mechanism that involves interaction with the transcription factor MED1 (part of the mediator complex involved in transcriptional initiation), and within castration-resistant but not hormone-responsive prostate cancer (Liu et al. 2015). In mice, the expression of AR variant isoforms can be sufficient themselves to induce cancer development. Transgenic mouse models expressing ARv567es and ARv7 proteins within their prostate glands develop prostate cancer (Liu et al. 2013; Sun et al. 2014).