Role of PRMTs in cancer: Could minor isoforms be leaving a mark?

Abstract

Protein arginine methyltransferases (PRMTs) catalyze the methylation of a variety of protein substrates, many of which have been linked to the development, progression and aggressiveness of different types of cancer. Moreover, aberrant expression of PRMTs has been observed in several cancer types. While the link between PRMTs and cancer is a relatively new area of interest, the functional implications documented thus far warrant further investigations into its therapeutic potential. However, the expression of these enzymes and the regulation of their activity in cancer are still significantly understudied. Currently there are nine main members of the PRMT family. Further, the existence of alternatively spliced isoforms for several of these family members provides an additional layer of complexity. Specifically, PRMT1, PRMT2, CARM1 and PRMT7 have been shown to have alternative isoforms and others may be currently unrealized. Our knowledge with respect to the relative expression and the specific functions of these isoforms is largely lacking and needs attention. Here we present a review of the current knowledge of the known alternative PRMT isoforms and provide a rationale for how they may impact on cancer and represent potentially useful targets for the development of novel therapeutic strategies.

Key Words: Protein arginine methyltransferase, Arginine methylation, Cancer, Alternative splicing, Isoforms

Core tip: This review focuses on the current knowledge regarding alternative protein arginine methyltransferases (PRMT) isoforms and evidence supporting their potential impact in cancer. Alternative PRMT isoforms have been identified for PRMT1, PRMT2, CARM1 and PRMT7 and more may exist for the other PRMT family members. The presence of these isoforms adds a layer of complexity to the functional roles PRMTs play in normal and disease contexts. These alternative isoforms have unique characteristics that may offer clarification to conflicting roles documented in the literature. Finally, understanding the specific functions of these isoforms is crucial for fully characterizing the therapeutic potential of PRMTs in cancer.

INTRODUCTION

Cancer is a leading cause of death worldwide. As we improve our understanding of the complex biologic processes behind this devastating disease we are able to develop improved treatments and increase patient survival. The biology of human tumours has been characterized as having six key hallmarks: sustained proliferative capacity, evasion of growth suppressors, resisting death, enabling replicative immortality, inducing angiogenesis and activating invasion and metastasis[1]. Each of these features is distinct, but they all cooperate to promote tumour development, growth and aggressiveness. Identifying key molecular regulators of one or more of these characteristics is essential in understanding cancer and potentially discovering new and better therapeutic strategies.

Arginine methylation is a common posttranslational modification that is known to have a role in several cellular processes, including signal transduction, DNA repair, transcription, protein subcellular localization and RNA processing[2,3]. Arginine methylation, in mammalian cells, is catalyzed by a family of enzymes called protein arginine methyltransferases (PRMTs). This family currently consists of nine characterized members in higher eukaryotes. These enzymes are subdivided into three categories based on the type of methyl mark produced on the arginine residue. These methylation reactions are depicted in Figure 1. Type I [PRMT1, 3, 4 (CARM1), 6, and 8] generate ω-N^G,N^G-asymmetric dimethylarginine. Type II (PRMT 5 and potentially PRMT9) generate ω-N^G,N’^G-symmetric dimethylarginine. Finally, Type III generate ω-N^G-monomethylarginine residues. Recently, it has been demonstrated that PRMT7 is the only bona fide type III methyltransferase[4,5]. The majority of arginine methylation is catalyzed by PRMT1 (asymmetric) and PRMT5 (symmetric), and loss of expression of either of these enzymes is not compatible with life[6,7]. Currently, there is more that 120 known arginine methylated proteins, including histone and non-histone proteins[8,9]. The list of arginine methylated protein substrates is constantly growing, and along with it the discovery of new functional roles and involvement in numerous regulatory pathways[8,10,11].

Figure 1 Arginine methylation reactions catalyzed by protein arginine methyltransferases. Type I protein arginine methyltransferases catalyze the asymmetric dimethylation of arginine residues, Type II symmetrically dimethylated arginine and Type III monomethylated arginine residues.

Accumulating evidence convincingly shows that arginine methylation may represent a driving force behind the development, progression and aggressiveness of several cancer types. While the link between arginine methylation and cancer is a relatively new area of interest, the roles that the PRMTs have been shown to play in cancer thus far demonstrate their importance. These roles and the cancer types that have been studied are highlighted in Table 1. Dysregulated PRMT expression has been observed in a number of human tumours, including lung, breast, prostate, colorectal, bladder and leukemia[12-19]. For a comprehensive review summarizing the roles of each PRMT family member in cancer see Yang and Bedford’s review article in Nature Reviews: Cancer entitled, Protein arginine methyltransferases and cancer[20]. The primary focus of this review is to specifically highlight the current knowledge regarding alternatively spliced PRMT family members and the potentially distinct roles that they play in cancer. While a survey within the Ensembl database predicts the existence of alternatively spliced isoforms for all the PRMT gene family members, only the expression of PRMT1, PRMT2, CARM1 and PRMT7 isoforms has been characterized and confirmed in mammalian cells[21-27].

Table 1 Protein arginine methyltransferases in cancer cells.

PRMT	Cancer type	Role(s) in cancer	Ref.
PRMT1	Breast cancer, Lung cancer, Colon cancer, Bladder cancer, Acute myeloid leukemia, Mixed lineage leukemia	Cell proliferation and survival, Transformation, Resistance to DNA damaging agents, Invasion	[13,15-17,19, 21,36-38]
PRMT2	Breast cancer	Cell proliferation and invasion	[22,72]
PRMT3	Breast cancer	Cell survival	[101,102]
CARM1/PRMT4	Breast cancer, Prostate cancer, Colorectal cancer	Cell proliferation	[12,14,77-79,88]
PRMT5	Lung cancer, Leukemia, Lymphoma, Melanoma, Gastric cancer, Colorectal cancer	Cell proliferation, Transformation, Invasion, Resistance to DNA damaging agents	[18,103-109]
PRMT6	Lung cancer, Bladder cancer	Cell proliferation	[17,110]
PRMT7	Breast cancer	Resistance to DNA damaging agents	[27,91,92,94]
PRMT8	ND	ND
PRMT9	ND	ND

ND: Not determined; PRMT: Protein arginine methyltransferase.

Interestingly, the majority of these alternative isoforms were found in cancer cells, suggesting they may have specific roles in cancer. Characterization of several of these alternative PRMT isoforms has shown that they are differentially expressed in various cell types and they possess distinct functional characteristics. However, the individual roles that these alternative isoforms play in cells remains poorly understood and understudied. Therefore, more attention needs to be given to their individual functions under normal biological conditions, as well as their contribution to diseases such as cancer. PRMTs are thought to be potentially useful therapeutic targets for the treatment of diseases such as cancer[28]. Moreover, these alternative PRMT isoforms must be taken into account when designing and evaluating potential candidate therapeutic strategies or compounds. This is essential so there is a clear understanding of the precise mechanism of action. Although our knowledge of the specific roles of these isoforms is limited, there is evidence in the literature strongly suggesting that they are not redundant. While they may share some similar functions, they also have clearly distinct roles.

PRMT ISOFORMS AND CANCER

PRMT1

PRMT1 is a Type I arginine methyltransferase and is responsible for generating upwards of 85% of the asymmetrically dimethylated proteins within cells[29]. PRMT1 is the most well characterized protein within this family of enzymes. While the PRMT1 protein is mainly described in the literature as a single entity, it has been identified, that at least seven distinct PRMT1 isoforms are generated by complex alternative splicing in the 5’ region of its pre-mRNA[21,30,31]. The exon structure for the identified PRMT1 isoforms is summarized in Figure 2 and detailed in Goulet et al[21] 2007. Each of these isoforms, named PRMT1v1-v7, has distinct characteristics in terms of expression. PRMT1v1 is the most abundantly expressed isoform and likely represents the isoform that is described as PRMT1 in most reports. The expression levels of PRMT1v1, v2 and v3 have all been shown to be ubiquitous across tissues[21,30,31]. Interestingly, a higher level of PRMT1v1 mRNA expression is observed in the kidney, liver, lung, skeletal muscle and spleen[21]. PRMT1v2 mRNA was found to be elevated in the kidney, liver and pancreas, while, PRMT1v3 mRNA expression was observed at similar levels in all tissues examined (brain, heart, kidney, liver, lung, pancreas, skeletal muscle and spleen), however at low levels compared to PRMT1v1 and PRMT1v2. The mRNA expression levels of PRMT1v4 to v7 showed a more tissue specific profile, with v4 being detected only in the heart, v5 mainly in the pancreas, and v7 observed in the heart and skeletal muscle. PRMT1v6 mRNA was not detected in any normal tissues examined[21]. Further studies would need to be performed to determine if this differential expression has any correlation with the development of cancer from a particular tissue of origin.

Figure 2 Protein arginine methyltransferase variant isoforms. Schematic representation of the identified variant isoforms of protein arginine methyltransferase (PRMT) 1, PRMT2, CARM1/PRMT4 and PRMT7. The PRMT1 sequence has 12 exons. Exon organization of the seven identified PRMT1 isoforms are shown. The intronic sequences (-) that have been shown to be included in several of these alternative PRMT1 isoform transcripts are due to the splicing sites[21]. PRMT2 is made up of 11 exons. The PRMT2L2 transcript is produced as a result of alternative polyadenylation[72]. This silences the 5’ splice site on exon 7 and results in a transcript retains a significant portion of intron 7 and a premature termination codon. PRMT2α has a deletion of exons 8-10 with a frame shift that produces 12 new amino acids at the C-terminus (ν). The PRMT2β isoform has a deletion of exons 7, 8, 9 resulting in a frame shift that generates 83 alternate amino acids at the C-terminus (νν). PRMT2γ has an in frame deletion of exons 7 to 10. The full-length CARM1 gene, CARM1/CARM1v1/CARM1FL, consists of 16 exons. CARM1v2 is generated through retention of the intron 15 sequence; CARM1v3 is produced through the retention of introns 15 (-) and 16 (-). CARM1v4/CARM1Δ15 results from the skipping of exon 15[23,24]. The PRMT7 sequence consists of 19 exons. In Hamster cells, these two PRMT7 isoforms (α and β) are thought to be generated by the use of distinct 5’ translation initiation codons within the primary transcript. The PRMT7β isoform sequence contains 37 extra amino acids at the N-terminus. Alternatively, at least 2 alternatively spliced PRMT7 isoforms can be produced from the human PRMT7 gene. These two isoforms have the same N- and C-terminal regions but variant 2 (PRMT7v2) has an in frame deletion of exon 5.

While tissue specific expression of PRMT1 isoforms is observed, at the cellular level there are also differences in their subcellular localization (Table 2). PRMT1v3, v4, v5 and v6 all show an equal distribution of nuclear and cytoplasmic expression[21]. In contrast, PRMT1v1, v2 and v7 display a more compartmentalized expression profile within cells. PRMT1v1 and v7 display a more intense nuclear expression, while PRMT1v2 is expressed predominantly in the cytoplasm, however this may vary depending on cell type and methylation status of substrates as it was clarified by the Fackelmayer lab[32,33]. The cytoplasmic expression of PRMT1v2 is due to the retention of exon 2 within the N-terminal coding sequence. This short exon contains a leucine-rich nuclear export sequence (NES). Careful analysis showed that this NES does in fact control the nuclear export of PRMT1v2 and that its export is dependent on the nuclear export receptor CRM1[21].

Table 2 Protein arginine methyltransferase isoform specific subcellular localization and current cancer cell types in which they have been shown to be expressed.

PRMT isoform	Molecular weight (kDa)	Subcellular localization	Cancer cell type	Ref.
PRMT1v1	40.5	Predominantly nuclear	Breast cancer cell lines and tumour samples, cervical cancers cells	[21]
PRMT1v2	42.5	Predominantly cytoplasmic	Breast cancer cell lines and tumour samples, cervical cancers cells	[21,39]
PRMT1v3	39.9	Cytoplasmic and nuclear	Breast cancer cell lines and tumour samples	[21]
PRMT1v4	40.1	Cytoplasmic and nuclear	Breast cancer cell lines	[21]
PRMT1v5	39.4	Cytoplasmic and nuclear	Breast cancer cell lines	[21]
PRMT1v6	37.7	Cytoplasmic and nuclear	Breast cancer cell lines	[21]
PRMT1v7	36.7	Predominantly nuclear	Breast cancer cell lines	[21]
PRMT2	48.5	Predominantly nuclear, excluding nucleoli	Breast cancer cell lines and tumour samples	[22,72]
PRMT2L2	32	Predominantly cytoplasmic	Breast cancer cell lines and tumour samples	[72]
PRMT2α	32.6	Predominantly nuclear, excluding nucleoli	Breast cancer cell lines and tumour samples	[22]
PRMT2β	34	Cytoplasmic and nuclear, including nucleoli	Breast cancer cell lines and tumour samples	[22]
PRMT2γ	25.8	Predominantly nuclear, excluding nucleoli	Breast cancer cell lines and tumour samples	[22]
CARM1/CARM1v1/CARM1FL	66	ND	Breast cancer cell lines	[23,24]
CARM1v2	71	ND	Breast cancer cell lines	[23,24]
CARM1v3	63	ND	Breast cancer cell lines	[23,24]
CARM1v4/CARM1Δ15	64	ND	Breast cancer cell lines	[23,24]
PRMT7α	78	Cytoplasmic and nuclear	ND	[27]
PRMT7β	82	Predominantly cytoplasmic	ND	[27]

ND: Not determined; PRMT: Protein arginine methyltransferase.

A comparison of the PRMT1 isoforms revealed they have dstinct enzymatic activity and substrate specificity profiles[21]. Additionally, stable isotope labeling by amino acids in cell culture (SILAC[34,35]) followed by immunopurification of PRMT1v1 and PRMT1v2 from cells has been used to identify their isoform-specific protein binding partners and/or substrates (Figure 3). In Figure 3 we show the full data set from this analysis comparing the SILAC ratios of PRMT1v1 and PRMT1v2 binding proteins (unpublished data). Each point represents an identified interacting protein. This clearly shows that there is a potential set of PRMT1v1-specific interacting proteins (lower right quadrant) and PRMT1v2-specific interacting proteins (upper left quadrant). Also, there are some common binding partners (upper left quadrant). This emphasizes the importance of understanding their individual functions. Conservation of these alternatively spliced isoforms of PRMT1 through evolution suggests they are likely to each have their own function(s) within cells and tissues.

Figure 3 Protein arginine methyltransferase 1v1 and protein arginine methyltransferase 1v2 have potentially different interacting protein profiles. Stable isotope labeling by amino acids in cell culture (SILAC) and mass spectrometry was used to identify protein arginine methyltransferase (PRMT) 1v1 protein binding partners and PRMT1v2 protein binding partners. Cells stably expressing GFP alone, GFP-tagged PRMT1v1 or GFP-tagged PRMT1v2 were grown independently in media containing light (L), medium (M) and heavy (H) isotopes of arginine and lysine residues, respectively. Protein lysates were collected, immunoprecipitated for GFP (isolation of PRMT1v1 and PRMT1v2 interacting protein), and subjected to mass spectrometry for peptide identification. The Log2 of the SILAC ratios for the peptides identified from this experiment are plotted on the scatter plot. The x-axis is the Log2 of the H:L SILAC ratio or PRMT1v2 interacting proteins. The y-axis is the Log2 of the M:L SILAC ratio or PRMT1v1 interacting proteins. Each data point represents a single protein that was identified in this experiment. The greater this ratio is for a protein, the higher the probability of the interaction being real. This revealed a protein interacting profile identifying PRMT1v1-specific interacting proteins (PRMT1v1 quadrant), PRMT1v2-specific interacting proteins (PRMT1v2 quadrant) and common interacting proteins (PRMT1v1/PRMT1v2 quadrant; unpublished data). These results require further validation.

Deregulated PRMT1 expression has been observed in a number of tumour types, which include those of the lung, breast, colon, bladder and leukemia[13,15-17,19,21,36-38]. The question is then, “What are the functions of these isoforms and do they have specific roles in cancer?” To date our knowledge is limited as to the specific functions of each of these PRMT1 isoforms. However, there is evidence showing potential individual roles for them in cancer. In breast cancer, both the mRNA and protein expression of several alternative PRMT1 isoforms is elevated (Table 2)[21,31]. This is observed not only in breast cancer cell lines, but also in breast tumours. Specifically, the mRNA expression of PRMT1v1, v2, v3 and v7 is elevated across several breast cancer cell lines compared to a non-transformed mammary epithelial cell line[21]. In contrast, PRMT1v5 and v6 were upregulated only in a subset of breast cancer cell lines. Furthermore, PRMT1v1, v2 and v3 mRNA expression was increased in breast cancer tumour tissue compared to normal tissue. Interestingly, while this study concluded an overall upregulation of PRMT1 alternative isoforms in breast cancer, the cytoplasmically localized PRMT1v2 isoform had the greatest increase in expression in breast cancer compared to PRMT1v1, the most abundantly expressed isoform. It is difficult to assess the protein expression of each of these individual isoforms due to the sequence similarities between them. However, in the case of PRMT1v2, exploitation of the exon 2 sequence has allowed for a more specific examination. Indeed, results have shown that PRMT1v2 protein expression is elevated in breast cancer cells[21]. A recent clinical assessment of PRMT1v1, v2 and v3 expression within breast cancer tissues has identified that high PRMT1v1 mRNA expression correlates with poor patient prognosis and a reduced disease-free survival[16]. An examination of PRMT1 protein expression within breast tumours via immunohistochemistry demonstrated a predominantly cytoplasmic expression and only in rare cases nuclear expression. We, and others have shown that PRMT1v2 is predominantly localized to the cytoplasm[21,32,39]. Therefore, one could speculate that PRMT1v2 could represent a significant proportion of the cytoplasmic PRMT1 detected in these breast tumour samples. This evidence shows that the expression of the PRMT1v2 isoform is elevated in breast tumours and it may have its own unique contributions to breast cancer progression. This also emphasizes the need to study these alternative isoform individually, in order to determine their specific functions and contribution to disease. While this has been mainly assessed in breast cancer thus far, it does not rule out that these PRMT1 isoforms may be expressed in other cancer types as well and this should be explored further.

The involvement of PRMT1 in cancer is supported by evidence showing its involvement in pivotal oncogenic processes. PRMT1 plays an active role in MLL-mediated transformation of primary myeloid progenitor cells[13]. PRMT1 has also been shown to have a significant role in cell proliferation/viability and cell cycle progression. Depletion of PRMT1 resulted in a significant decrease in the proliferation of osteosarcoma, breast, bladder and lung cancer cell lines[6,17,37]. This reduction in cell proliferation was associated with cell cycle arrest at the G₀/G₁ phase. Additionally, breast cancer cells showed a loss of cyclin D1 and increase in p21^cip1 expression, indicative of a cell cycle arrest at this phase[37]. While these studies examined PRMT1 as a whole, PRMT1 isoform-specific contributions have also been investigated. The specific depletion of the PRMT1v2 isoform using RNA interference in breast cancer cells resulted in a significant reduction in cell viability and growth[40]. This decreased cell viability was attributed, at least in part, to an induction of apoptosis occurring with the suppression of PRMT1v2 expression. Additionally, breast cancer cells overexpressing PRMT1v2 showed an increased growth rate, which was not observed upon PRMT1v1 overexpression and points to isoform specific effects. This evidence suggests that in these breast cancers cells PRMT1v2 may represent a key cell survival-promoting factor. Overall, this evidence links PRMT1 to the self-sustaining proliferative signaling acquired by cancer cells, enabling them to grow and survive.

The impact that PRMT1 has on the survival and aggressiveness of cancer cells is becoming increasingly evident with the identification of new intracellular substrates. It has been demonstrated that the asymmetric dimethylation of histone H4R3 is associated with active transcription and increased tumour grade in prostate cancer[41-43]. However, the downstream consequences of this methylation event are poorly understood in most cases[44]. Many of the recently identified PRMT1 substrates are key regulators of cancer cell growth, survival and invasion signaling. PRMT1 has been shown to influence receptor activation at the cell surface through direct methylation of the receptor or indirect methylation of a receptor associated protein. PRMT1 was shown to directly methylate the estrogen receptor α (ERα) at arginine (R) 260 and affects its downstream signaling[37,45]. This results in cytoplasmic retention of ERα and the interaction of ERα with Src, focal adhesion kinase (FAK) and the regulatory subunit of PI-3 kinase (p85). All three of which are involved in oncogenic intracellular signaling that promotes cancer cell survival and invasiveness[46-50]. Furthermore, loss of this methylation site on ERα, by point mutation, impaired downstream signaling, as evidenced by a loss of PKB/Akt phosphorylation. Recently, it was shown that PRMT1 is involved in the induction of TGFβ signaling in response to bone morphogenetic protein (BMP) binding its transforming growth factor (TGF) β receptors, RI and RII[51]. Activation of this receptor is achieved through the ligation and dimerization of the RI and RII receptors[52]. The RI receptor is held in an inactive state by its association with Smad 6. Upon BMP ligation and dimerization of RI and RII, PRMT1 methylates Smad 6, causing its dissociation from RI and activation, thereby inducing BMP signaling which has a role in cancer stem cell proliferation and cancer cell invasion[53]. PRMT1 has also been shown to interact with PRMT8[54]. PRMT8 harbours a unique property, as it is tethered to the plasma membrane via an N-terminal myristoylation motif. Additionally, PRMT8 is specifically only expressed in brain tissue. This PRMT1-PRMT8 interaction effectively localizes PRMT1 activity at the plasma membrane and could potentially be affecting a distinct set of substrates. A specific role for PRMT8 in cancer has not been examined. These functions of PRMT1 occur in the cytoplasm of cells, and the RNA interference method used in these studies targeted all PRMT1 isoforms. Therefore, it would be of interest to assess whether specific PRMT1 isoforms might differentially contribute to the above-mentioned regulatory pathways. This would offer not only more functional understanding, but therapeutic insight as well.

PRMT1 has been shown to methylate key cytoplasmic proteins that are linked to apoptotic signaling pathways. Intriguingly, there have been conflicting roles presented for PRMT1 in apoptotic signal regulation. One study demonstrated that PRMT1 methylates apoptosis signal-regulating kinase 1 (ASK1) and this inhibits its activity[55]. This methylation promotes the interaction of ASK1 with its negative regulator, thioredoxin. As a consequence breast cancer cells were shown to be more resistant to treatment with paclitaxel. In contrast, the BCL-2 antagonist of cell death (BAD) has also been identified as a PRMT1 substrate in breast cancer cells[56]. This methylation prevents PKB/Akt mediated phosphorylation of BAD, thus preventing its inactivation, resulting in enhanced BAD-induced apoptosis. These conflicting roles highlight the complex role that methylation plays within cellular signaling pathways. These observations were seen in two distinct breast cancer cells, MDA-MB-231 and MCF7 respectively. Therefore, it is unknown whether these observations are due to cell specific behaviors or more interestingly the genetic differences between these two distinct breast cancer cells. Furthermore, they may also be influenced by differential expression of alternative PRMT1 isoforms, potentially reflecting differences in function and substrate specificities within cancer cells.

A recent study identified Axin, a mainly cytoplasmic protein, as a PRMT1 substrate[57]. Importantly, it was shown that Axin could be methylated by two PRMT1 isoforms, PRMT1v1 and PRMT1v2 in vitro. However, this methylation analysis was not conducted within cells and would have been a very informative experiment, considering both Axin and PRMT1v2 share a cytoplasmic localization. Axin is a critical scaffolding protein that complexes with adenomatous polyposis coli (APC), casein kinase 1 (CK1) and glycogen synthase kinase 3β (GSK3β), forming a degradation complex. This complex negatively regulates Wnt signaling and impacts actin cytoskeletal dynamics through the degradation of β-catenin[57,58]. Methylation of Axin by PRMT1 increases Axin protein stability, resulting in decreased β-catenin protein levels. Interestingly, isoform specific overexpression of PRMT1v1 or PRMT1v2 in a weakly invasive breast cancer cell line (MCF7) resulted in an increase in cell motility[40]. However, only the overexpression of the PRMT1v2 isoform increased cell invasion through a Matrigel barrier. Additionally, specific depletion of PRMT1v2 in an invasive breast cancer cell line, MDA-MB-231, resulted in decreased invasion through a Matrigel barrier. PRMT1v2 overexpression caused a decrease in β-catenin protein expression, which was not seen with the overexpression of PRMT1v1. This loss in β-catenin protein expression was directly linked to the PRMT1v2-induced invasion observed in breast cancer cells. Furthermore, PRMT1v2 enzymatic activities as well as proper subcellular localization were required for its ability to promote invasion. Therefore, it is conceivable that within cells Axin is preferentially methylated by PRMT1v2, thereby regulating β-catenin protein levels. This evidence has shown for the first time direct functional differences between PRMT1 isoforms in cancer, and identified a specific role for PRMT1v2 in promoting breast cancer cell invasion.

PRMT1 methylates several proteins within the nucleus that are involved in transcription, telomere stability and DNA repair. Similarly to the methylation of BAD, PRMT1 methylates the forkhead box protein 1 (FOXO1) at R248 and R250 blocking PKB/Akt-mediated phosphorylation of S253[59]. This methylation results in nuclear retention of FOXO1, increased transcriptional activity and increased oxidative-stress induced cell death. This evidence again supports a role for PRMT1 promoting cell death. PRMT1 also affects telomere length and stability, which impacts the replicative capacity of cancer cells[1,60]. PRMT1 methylates the telomeric repeat binding factor 2 (TRF2), thereby regulating its association with telomeres. TRF2 is a component of the sheltering complex that binds telomeric DNA and functions to protect telomeres and maintain their length. Depletion of PRMT1 in cancer cells increased the association of TRF2 with telomeres and promoted shortening. This supports a role for PRMT1 in dysregulated cancer cell replication. Additionally, PRMT1 is linked to the DNA damage response and DNA repair pathways through the methylation of MRE11 and p53 binding protein 1 (53BP1). PRMT1 has been shown to methylate MRE11 and 53BP1 within their GAR motif[61-63]. Methylation of MRE11 regulates its DNA exonuclease activity in response to DNA damage[61]. Similarly, methylation of 53BP1 is necessary for its DNA binding activity and localization to sites of DNA damage[63]. Mutation of this methylation motif in both MRE11 and 53BP1 disrupts the functions of these two key proteins in the DNA damage pathway. Finally, PRMT1 was shown to methylate the tumour suppressor gene BRCA1[36]. Methylation of BRCA1 had a significant impact on its ability to bind to different gene promoters, adding a level of complexity to the transcriptional regulating function of PRMT1. It would be interesting to determine if these effects are isoform specific, as it has been shown that the PRMT1v1 isoform is predominantly localized to the nucleus.

These studies demonstrate that PRMT1 has a significant impact on the vital processes and signaling that are involved in the development, progression and aggressiveness of cancer cells. The majority of these studies have examined PRMT1 as one single enzyme, however the existence of the distinct PRMT1 isoforms adds a level of complexity that requires further study and clarification. This evidence suggests that PRMT1 may be a potentially valuable therapeutic target for the treatment of several cancer types, however our knowledge of this target is limited due to our lack of understanding of the precise roles of the alternative isoforms that are present.

PRMT2

PRMT2, also known as HRMT1L1, was discovered through its sequence homology with the catalytic domain of PRMT1 (approximately 50%)[30]. Interestingly, within its sequence it contains an Src homology 3 (SH3) binding domain, which potentially links it to many intracellular processes. Initially, it had no characterized methyltransferease activity. However, more recent evidence has shown that it possesses Type I arginine methyltransferase activity, albeit much lower than that of PRMT1[64]. There is limited knowledge with regards to PRMT2 methyl substrates. Evidence has shown PRMT2 is recruited by β-catenin to histone H3 where it deposits an asymmetric dimethyl mark on R8 of target gene promoters[65]. However, further experiments are required in order to generate a more complete substrate repertoire for PRMT2. Nevertheless, it has been demonstrated that PRMT2 can affect the activation of several key receptors via a co-activator function within cells. PRMT2 has been shown to interact with and enhance the transactivation of ERα, progesterone receptor (PR), androgen receptor (AR), peroxisome proliferator-activated receptor γ (PPARγ) and the retinoic acid receptor α (RARα) in a ligand independent fashion[66]. Interestingly, the activation of these receptors within cells has both distinct and in some cases opposing effects. Activation of ERα, PR and AR has been implicated in tumour cell growth and progression, while PPARγ and RARα activation results in growth arrest and apoptosis[67-71]. This suggests that the functional role PRMT2 plays within cells is quite diverse.

Recently, in two separate papers by Zhong et al[22,72], four alternatively spliced PRMT2 isoforms (PRMT2L2, PRMT2α, β, and γ) in addition to the original PRMT2 isoform were identified. The PRMT2 gene consists of 11 exons and these alternative isoforms are generated through alterations in sequence that occur from exon 7 to exon 10 (Figure 2). The first report identified a novel PRMT2L2 transcript that is produced as a result of alternative polyadenylation[72]. This polyadenylation silences the 5’ splice site on exon 7 and results in a transcript that retains a significant portion of intron 7 and a premature termination codon. Subsequently, they identified PRMT2α, β and γ and showed that these isoforms are generated through splicing events occurring in the 3’ C-terminal region of the PRMT2 pre-mRNA leading to exon exclusion[22]. PRMT2α has a deletion of exons 8-10 with a frame shift that produces 12 new amino acids at the C-terminus. The PRMT2β isoform has a deletion of exons 7, 8, 9 resulting in a frame shift that generates 83 alternate amino acids at the C-terminus, while PRMT2γ has an in frame deletion of exons 7 to 10. All of these deletions in the alternatively spliced isoforms result in the loss of conserved protein arginine methyltransferase motifs. They have each lost domain III and the THW loop. The THW loop has been shown to form part to the AdoMet-binding pocket with domains I and post I[73], therefore these variant isoforms may lack arginine methylation activity. Methylation activity of these isoforms has not yet been examined. An examination of the subcellular localization of GFP tagged PRMT2 isoforms showed that PRMT2, PRMT2α and PRMT2γ have a predominantly nuclear localization, excluding the nucleolus (Table 2)[22]. The PRMT2β isoform showed a relatively even distribution throughout the nucleus, including the nucleolus, and also localized to the cytoplasm within cells. The PRMT2L2 had a predominantly cytoplasmic localization with concentrated perinuclear staining observed[72]. It is thought that the 3’ sequence may impact the localization of these isoforms.

Characterization of these alternative isoforms showed differential expression across a panel of breast cancer cell lines (Table 2). Interestingly, mRNA and protein expression of all PRMT2 isoforms are elevated in ER, PR-positive cell lines (MCF7, T47D, BT474 and ZR-75-1) compare to double negative cell lines (MDA-MB-231, MDA-MB-453 and SK-BR-3)[22,72]. Furthermore, in breast tumour samples, the mRNA expression of all PRMT2 isoforms was shown to be significantly increased in breast tumour tissues compared to normal adjacent breast samples. Additionally, the expression of each isoform was shown to be slightly higher in ER-positive compared to ER-negative tumours. Moreover, an immunohistochemical analysis, which did not differentiate between isoforms, showed that PRMT2 protein expression is elevated in breast tumour samples compared to normal breast tissue[22]. Additionally, similar to the mRNA, PRMT2 protein expression was elevated to a greater extent in ER-positive tumours compared to ER-negative tumours.

A functional assessment of the PRMT2 isoforms showed that they are able to directly bind and enhance estrogen-mediated transactivation of ERα, and also enhance the promoter activity of the downstream target gene, snail[22,72]. Increased snail transcriptional activity is associated with an increased cancer cell invasive potential[74]. Interestingly, all the isoforms had a lower transcriptional activity compared to PRMT2. Additionally, PRMT2β also had the lowest estrogen stimulated transcriptional activity and showed the lowest interaction affinity for ERα. This demonstrates that these isoform may perform different functions within cells. This interaction with ERα occurs via the N-terminus of the PRMT2 isoforms. Each PRMT2 isoform was also shown to directly bind to the AR. Intriguingly, it was revealed that PRMT2 negatively impacts the proliferation of ERα positive breast cancer cells in response to estrogen stimulation[22]. Depletion of the PRMT2 isoforms caused an increase in estrogen-induced proliferation and an enhancement in E2F expression and downstream activity. This is consistent with results showing that PRMT2 can bind to retinoblastoma protein (RB), and this interaction causes repression of E2F transcriptional activity[75]. It should be highlighted that the increase in proliferation may be specific to the original PRMT2 isoform, as depletion of this specific isoform caused a result similar in magnitude to the depletion of all four isoforms (PRMT2, PRMT2α, PRMT2β, PRMT2γ) simultaneously. Therefore, the contribution of the PRMT2α, PRMT2β, PRMT2γ isoforms to this proliferation phenotype is unclear. Similar to PRMT1, further research is required into the specific functions of these newly identified PRMT2 isoforms in order to determine their exact contributions to cancer development and progression. Nevertheless, these results demonstrate that the expression of PRMT2 and its alternative isoforms are clearly positively correlated with ERα status in breast cancers, consistent with a regulatory role in this pathway.

PRMT4/CARM1

PRMT4, more commonly known as Co-activator-associated arginine methyltransferase 1 (CARM1), was originally identified through its binding to GRIP1, the p160 steroid receptor co-activator[76]. It is involved in the regulation of a number of cellular processes including, transcription, pre-mRNA splicing, cell cycle progression and the DNA damage response. CARM1 is a type I arginine methyltransferase. In contrast to other type I PRMTs, which generally recognize substrate GAR motifs, it has no known substrate methylation motif[8,44]. CARM1 is most well characterized for its co-activator role in transcription which it performs through its interaction and methylation of a diverse substrate repertoire, including both histone and non-histone proteins[77-81]. The activity of CARM1 has also been shown to be influenced by posttranslational modifications. Specifically, CARM1 can be phosphorylated at several sites that can inhibit both dimerization (S229) and AdoMet binding (S217)[82,83]. Alternatively, phosphorylation at another site (S448) facilitates association with the ERα and stimulates ligand-independent activation of ERα[84]. Recently, it was identified that CARM1 is also regulated by auto-methylation[85]. The auto-methylation site was mapped to R551 in exon 15 of the mouse homolog of CARM1. This site is conserved in all vertebrate CARM1 proteins. Mutation of this auto-methylation site did not affect the enzymatic activity of CARM1, however it significantly impaired both CARM1-activated ERα mediated transcription and CARM1 regulated pre-mRNA splicing. Furthermore, it has been shown that essentially 100% of CARM1 is auto-methylated at R551 in cells[24]. Therefore, the regulation of CARM1 activity appears to be complex.

The expression of CARM1 has been shown to be dysregulated in colorectal, prostate and breast cancer[12,14,15]. CARM1 was found to be overexpressed in a significant number of colorectal tumours[14]. In prostate cancer, CARM1 was found to be overexpressed not only in tumours, but also in prostatic intraepithelial neoplasia (PIN). PINs are thought to be a precursor to the development of prostate cancer[12,14]. Finally, CARM1 expression was also found to be upregulated in breast cancer[14,86]. Interestingly, in the study conducted by Kim et al[14], for both prostate and breast cancers the expression level of CARM1 was lower. In a more recent study by Cheng et al[86], CARM1 expression was observed to be increased in invasive breast cancer, correlating with high tumour grade and to a greater extent with HER2, p53 and Ki-67 expression. CARM1 expression showed a lower correlative rate with ER and PR expression. The results from these studies are surprising given the role that CARM1 plays in the association and co-activation of ERα and AR[87,88]. They suggest that CARM1 has a multifaceted contribution to the development and progression of cancers. Furthermore, it shows that CARM1 may be an informative prognostic marker for breast cancer.

Within tumour cells, CARM1 plays a role in regulating cell proliferation and survival through its interaction and cooperation with several critical cancer related proteins. CARM1 is recruited to the promoter of the cyclin E1 gene, where it acts as a transcriptional co-activator in regulating cyclin E1 protein expression. Furthermore, both CARM1 and cyclin E1 were shown to be co-overexpressed and correlated with grade 3 breast tumours[78]. CARM1 has also been shown to be necessary for estrogen-stimulated proliferation of breast cancer cells[77]. This occurs via estrogen-stimulated methylation of H3R17 by CARM1, resulting in expression of the cell cycle regulator E2F1. Moreover, CARM1 is involved in the regulation of both the stability and activity of AIB1, a transcriptional co-activator that is often overexpressed in breast tumours. Additionally, it has been recently shown that CARM1 can promote breast cancer cell migration and metastasis through the methylation of BAF155, a component of the chromatin-remodeling complex[89].

While these studies define a role for CARM1 in promoting cancer progression, a study by Al-Dhaheri et al[79] showed some conflicting effects. Overexpression of CARM1 in MCF7 breast cancer cells, an ER+ cell line, inhibited estrogen-stimulated cell growth, while overexpression or depletion of CARM1 in MDA-MB-231 (ER-) breast cancer cells had no effect on their growth. Interestingly, the inhibited cell growth observed in MCF7 cells with CARM1 overexpression was accompanied by increased expression of cell cycle inhibitors, p21^cip1 and p27^kip1 and a change in cell morphology reminiscent of a more differentiated phenotype. Additionally, CARM1 was shown to repress the expression of approximately 16% of estrogen-activated target genes. An expression analysis in a set of ER+ tumours showed that CARM1 expression positively correlates with ERα expression. However, it inversely correlated with tumour grade. It should also be noted that a recent report suggested that only small proportion of endogenous CARM1 protein expression is required in order to perform its biological functions in cells[89]. Therefore, suppression of 100% of CARM1 protein expression is required in experimentation because it is thought that only a very small amount of CARM1 protein is necessary for its normal functioning. These reports suggest that a further understanding of CARM1 regulation and function is required in order to clarify its role and potential marker/therapeutic value in cancer.

A plausible explanation for these opposing results in breast cancer cells is the existence of alternatively spliced isoforms of CARM1. In the literature there are two papers that describe the presence of distinct alternatively spliced CARM1 isoforms. The first by Ohkura et al[23] describes, that in normal rat tissue, four isoforms are transcribed from the CARM1 gene; the primary isoform CARM1 (CARM1v1) and three alternative isoforms, v2, v3 and v4 (Figure 2). All four contain the arginine methyltransferase domain and the GRIP1-binding domain. The primary CARM1 isoform, CARM1v1, consists of 16 exons. CARM1v2 is generated through retention of the intron 15 sequence, CARM1v3 is produced through the retention of introns 15 and 16 and CARM1v4 results from the skipping of exon 15[23]. Each of these enzymes showed a distinct mRNA expression profile when examined across a panel of normal rat tissues. Functionally, the CARM1v3 isoform was shown to alter the splicing pattern of both E1A and CD44 reporters. This was not observed with the other isoforms suggesting they may have different functions. The splicing activity demonstrated for CARM1v3 was shown to be independent of the CARM1v3 methylation activity. In contrast to this, Cheng et al[90] showed that CARM1 enzymatic activity is required for its effect on alternative splicing of the CD44 pre-mRNA, which is thought to occur co-transcriptionally. They also suggest that while CARM1v3 is an alternative isoform, it may represent a very rare form not playing a major role in cells. Hence the precise biological roles of these CARM1 isoforms remains unclear.

Alternatively, in the second paper, Wang et al[24] showed that in human cells and tissues, two CARM1 isoforms are present. These are designated CARM1 full length (CARM1FL) and CARM1Δ15 (Figure 2). The CARM1Δ15 is a transcript in which exon 15 is excluded by alternative splicing. This alternative isoform represents the CARM1v4 isoform described previously. The other two isoforms were not detected in human cells or tissues. Importantly, exclusion of exon 15 removes the auto-methylation site that can functionally regulate CARM1, however it does not impact the methylation activity. An examination of mRNA expression across a panel of normal human tissues revealed that the CARM1Δ15 isoform is the major isoform expressed, with the exception of the brain, heart, skeletal muscle and testis. The CARM1FL isoform is expressed highest in these tissues. Additionally, the CARM1FL isoform is predominantly auto-methylated in cells.

In breast cancer cells, the CARM1Δ15 was shown to be the predominant isoform expressed (Table 2)[24]. However, only a limited number of cancer cell lines were assessed. It would be interesting to know the expression profile in other cancer types as well. Specifically, an assessment of CARM1 isoform expression in a panel of breast cancer cell lines showed a greater percentage of the CARM1Δ15 isoform compared to the CARM1FL isoform. This is surprising due to the fact that the CARM1Δ15 isoform has impaired ERα co-activator activity and failed to stimulate ERα transcriptional activity. However, it may have distinct roles with respect to activity and functions within cells. The existence of these two isoforms may shed light on some of the conflicting reports in the literature with respect to the biological functions of CARM1 and potential roles in cancer. Further study of these isoforms is required to establish if they are responsible for the methylation of distinct substrates and their individual functions.

PRMT7

PRMT7 was originally identified from a screen of genetic suppressor elements (GSE) aimed at identifying genes conferring resistance to cytotoxic agents performed in Chinese Hamster cells[91]. This screen identified a gene that encoded two proteins, p77 and p82, that were highly homologous to the PRMT family and later designated PRMT7α and β, respectively[27,91]. In Hamster cells, these two isoforms are thought to be generated by the use of distinct 5’ translation initiation codons within the primary transcript (Figure 2). The PRMT7β isoform sequence contains an extra 37 amino acids at the N-terminus. Both isoforms were shown to be active and have slightly different methylation profiles[27], though further analysis is required to clarify these differences between the isoforms. Each isoform has a distinct subcellular localization patterns (Table 2). PRMT7α localizes to the cytoplasm and nucleus, whereas PRMT7β is exclusively cytoplasmic[27]. In human tissues, only a single PRMT7 transcript is detected (approximately 3.6 kb) and in two human cell cancer cell lines, HeLa and HuH7, one protein at 78 kDa was detected. This transcript was shown to share the greatest homology to the PRMT7α isoform[25-27]. However, the limited subset of cell lines used cannot completely rule out the existence of PRMT7β isoform expression in human cells and a more comprehensive examination of expression in cells is required. Moreover, a survey within both NCBI and Ensembl databases predicts the existence of at least 2 alternatively spliced PRMT7 isoforms that can be produced from the human PRMT7 gene (Figure 2). These two isoforms have the same N- and C-terminal regions but variant 2 (PRMT7v2) has an in frame deletion of exon 5. Importantly, this may affect methyltransferase activity because it removes the post I domain. Functionally, PRMT7 was initially characterized as a Type II methyltransferase[26], but it has recently been deemed a Type III and is thus the only PRMT enzyme known to catalyze predominantly this reaction in mammalian cells[4,5]. The generation of monomethylarginine is thought to represent a reaction intermediate for the other PRMTs.

There is limited knowledge into the precise biological functions of PRMT7, however evidence has shown it is linked to cancer. A gene expression analysis of independent data sets of more than 1200 breast tumours identified increased expression in the chromosomal region where the PRMT7 gene is located (16q22)[92]. Importantly, this was also correlated with an increased metastatic potential of breast cancer. The PRMT7 gene locus was also identified in an unbiased genome-wide study to confer resistance to etoposide-induced cytotoxicity in patients[93]. As previously mentioned, PRMT7 was originally identified by a screen for GSEs conferring resistance to cytotoxic agents (etoposide and 9-OH-E)[91]. This study showed that GSE-mediated repression of PRMT7 conferred resistance to topoisomerase II inhibitors and also cisplatin. In contrast, in this same study, repression of PRMT7 caused increased sensitivity to other DNA-damaging agents, such as the topoisomerase I inhibitor, camptothecin, as well as UV-irradiation. Increased sensitivity to camptothecin was also observed when PRMT7 was depleted from HeLa cells[94]. Intriguingly, depletion of PRMT7 from NIH 3T3 cells conferred resistance to cisplatin, mytomycin C and chlorambucil[95]. Additionally, one of its only identified interacting protein partners, CTCFL, is a proposed proto-oncogene[96,97]. Further studies are required to identify additional PRMT7 substrates to better understand its role in cells. While these results strongly suggest that PRMT7 may play a key role in several cancer related processes, the opposing functions of PRMT7 in response to cytotoxic agents requires some attention. The reason for these differential effects is unclear; perhaps PRMT7 has distinct functions in different cell types. More interestingly, they could be the result of PRMT7 isoforms specific expression and function within cells.

CONCLUSION

The importance of PRMTs in cancer is only beginning to be examined. There have been many key discoveries thus far that have demonstrated the potential impact that the PRMTs have in regulating critical effectors and pathways involved in the development and progression of cancer. In fact, the PRMTs have the potential of impacting the majority of the described hallmarks of cancer proposed by Hanahan et al[1,98]. Current research efforts aim to identify and characterize the precise mechanistic roles that these PRMTs play in cancer. Importantly, the functional contribution of PRMTs to different cancer types, as well as subtypes within the same cancer, requires further investigation. The significance of this requirement is highlighted by several of the conflicting findings describe in this review.

Here we have highlighted the existence of alternatively spliced PRMT isoforms that have been identified for PRMT1, PRMT2, CARM1 and PRMT7. While not currently realized, more PRMT isoforms for these and other PRMT family members may be present in cells. The presence of distinct PRMT alternative isoforms adds a further level of complexity to this family of enzymes. Additionally, the isoforms identified for PRMT1, PRMT2, CARM1 and PRMT7 have mainly been assessed in breast cancer cells and tissues as indicated in Table 2. A more extensive analysis of their expression in other tumour types has not been performed and could uncover more interesting results with respect to these PRMT isoforms. This fact requires more attention as it may provide possible explanations for the opposing functions identified within cells. Furthermore, while these isoforms may have overlapping functions, it is clear from the data presented here that they also possess distinct functions. Interestingly, while dysregulated PRMT expression has been observed in cancer, no genetic abnormalities have been identified, with one exception being PRMT8[99,100]. While there may be no obvious change at the genome level, a shift in the expression from one alternative PRMT isoform to another may be a crucial event that occurs in cancer cells, thereby affecting development, progression and aggressiveness. Interestingly, a particular PRMT isoform may not be expressed or is expressed at lower levels in normal tissues and as a consequence of the tumourigenic process cancer cells may preferentially upregulate a specific isoform due to its advantageous functions. Understanding both the shared and distinct functions of these alternative PRMT isoforms will not only improve our knowledge of their biological significance but also provide insight into their specific contributions to diseases, such as cancer.

The roles that the PRMTs play in cancer make them an attractive target for the development of drugs that could be used in treatment strategies. This increases the importance of gaining more knowledge about the alternative PRMT isoforms, so that there is a complete understanding of the therapeutic mechanism. This will enable the development of an optimal therapeutic strategy and an improved understanding of the resulting outcomes when targeting PRMT enzymes as a treatment in cancer.

ACKNOWLEDGMENTS

Baldwin RM is a Postdoctoral Fellow of the Canadian Institute of Health Research; Côté J holds a Canada Research Chair (Tier II) in RNA Metabolism funded through the Canadian Institutes of Health Research.