The Complex Landscape of PTEN mRNA Regulation

Abstract

Phosphatase and tensin homolog deleted on chromosome 10 (PTEN) is a key tumor suppressor in the development and progression of different tumor types. Emerging data indicate that small reductions in PTEN protein levels can promote cancer. PTEN protein levels are tightly controlled by a plethora of mechanisms beginning with epigenetic and transcriptional regulation and ending with control of protein synthesis and stability. PTEN messenger RNA (mRNA) is also subject to exquisite regulation by microRNAs, coding and long noncoding RNAs, and RNA-binding proteins. Additionally, PTEN mRNA is markedly influenced by alternative splicing and variable polyadenylation. Herein we provide a synoptic description of the current understanding of the complex regulatory landscape of PTEN mRNA regulation including several specific processes that modulate its stability and expression, in the context of PTEN loss-associated cancers.

Exquisite regulation of phosphatase and tensin homolog deleted on chromosome 10 (PTEN) protein expression is critical to suppress the development and progression of many tumor types. Cellular levels of PTEN protein are governed by a variety of mechanisms, including histone modification, DNA methylation, transcriptional regulation, translational regulation, and posttranslational modifications (Song et al. 2012; Correia et al. 2014; Lee et al. 2018). Interestingly, observations made in mouse models indicate that small changes in PTEN protein levels can have profound consequences on tumor incidence, penetrance, and aggressiveness (Alimonti et al. 2010; Carracedo et al. 2011). Unlike classical models of tumor suppression where genetic perturbations are proposed to lead to allelic losses and consequent gross losses (50% or 100%) of protein expression or activity, the “continuum model” highlights the importance of nongenomic losses of tumor suppressor expression by more subtle mechanisms, such as the regulation by microRNA (miRNA) or other posttranscriptional mechanisms regulating PTEN messenger RNA (mRNA) levels (Berger and Pandolfi 2011).

Clinical observations that brain, prostate, endometrial, and gastric tissues, among others, are sensitive to small nongenetic perturbations in PTEN levels support the notion that PTEN may comply with a “continuum model” of tumor suppression rather than the more classical “two-hit” or haploinsufficiency models (Kwabi-Addo et al. 2001; Trotman et al. 2003; Alimonti et al. 2010; Berger and Pandolfi 2011). In this review, we focus on PTEN transcript regulation. We present and discuss progress in the understanding of PTEN mRNA regulation by various mechanisms including miRNA, competing endogenous RNA (ceRNA), alternative splicing, and use of alternative polyadenylation sites on PTEN transcripts, as well as the impact of N⁶-methyladenosine (m⁶A) modification.

M. icroRNA and PTEN

miRNA are small noncoding RNA molecules of ∼19–24 nucleotides in length that mediate gene silencing and play important roles in a wide range of fundamental biological processes and a variety of diverse diseases. miRNA are an essential component of the miRNA-induced silencing complex (miRISC), a multiprotein RNA complex that contains members of the Argonaute (Ago) family of proteins (Jonas and Izaurralde 2015). The principal function of miRNA is to guide the miRISC to specific RNA targets through partial Watson–Crick base pairing, typically on the 3′ untranslated region (UTR) of protein-coding and noncoding transcripts (Chandradoss et al. 2015). miRISC binding to a 3′UTR leads to gene silencing through translation repression and mRNA decay (Grimson et al. 2007; Bartel 2009; Kehl et al. 2017). Because miRNA can repress expression of most genes without completely silencing them, they are considered an important class of factors with an ability to fine-tune nearly all cellular processes.

Interest in regulation of the PTEN mRNA transcript by miRNA is extensive, with currently over 1500 publications that describe interactions of PTEN mRNA with one or more miRNA. While the landscape of miRNA regulation is complex and thought to be context-dependent, there are numerous studies that have identified specific high confidence miRNA:PTEN interactions that also present biological relevance. The pleiotropic function of PTEN in many cancers is mirrored by the miRNA-mediated regulation, which have been found to target and silence PTEN expression in numerous cancers including non-small-cell lung cancer (NSCLC), colorectal carcinoma, ovarian cancer, triple-negative breast cancers, prostate cancer, and melanoma (for review, see Bermúdez Brito et al. 2015). In each case, silencing of PTEN by miRNA leads to more aggressive forms of the disease. For example, miR-221 and miR-222 are highly expressed in NSCLC and hepatocellular carcinoma, and promote tumorigenesis by targeting PTEN (Garofalo et al. 2009). In the same fashion, PTEN was repressed by aberrant overexpression of the miR-106b∼25 cluster in human prostate cancer, leading to more aggressive forms of prostate adenocarcinoma in mouse models (Poliseno et al. 2010a).

The first validated PTEN miRNA recognition element (MRE) was that of miR-21 (Ji et al. 2007; Meng et al. 2007), which was shown to bind to the PTEN 3′UTR (Ji et al. 2007; Meng et al. 2007) and is generally regarded as a potent bona fide PTEN targeting miRNA in several malignant and nonmalignant contexts (Feng and Tsao 2016). PTEN has been reported to interact with many other miRNA families, and it is widely accepted that expression of PTEN is heavily regulated through the function of miRNA. For instance, members of miR-9, -17, -19, -21, -23, -25, -26, -33, -103, -130, -136, -144, -154, -181, -214, -216, -217, -221, -222, and -506 gene families have all been implicated in fine-tuning PTEN expression in cancer and other contexts (detailed in Table 1).

Table 1.

The interaction of microRNA (miRNA) with phosphatase and tensin homolog deleted on chromosome 10 (PTEN) subdivided by gene family, miRNA construct identified, cell model and tissue type, and an analysis of the degrees of validation

miRNA gene family	miRNA examined	Cell model	Tissue type	Biological evidence	Luciferase reporter assay	MRE mutated	References
miR-9	miR-9	TU212, Hep-2	Laryngocarcinoma	Yes	Yes	Yes	Lu et al. 2016
miR-17	miR-17	DU145	Prostate cancer	Yes	Yes	Yes	Poliseno et al. 2010a
	miR-17-5p	NIH3T3	Mouse fibroblast	Yes	Yes	Yes	Xiao et al. 2008
	miR-18a-5p	U2OS	Osteosarcoma	Yes	NA	NA	Fei et al. 2018
	miR-20a	DU145/HEK 293T/HUVEC/HEK 293	Prostate cancer/SV40 large T antigen model/endothelial/model line	Yes/yes/yes/yes	NA/yes/yes/yes	NA/yes/yes/yes	Poliseno et al. 2010b; Zhang et al. 2015; Wang et al. 2017a; Jiang et al. 2018
	miR-93	MCF-7, MDA-MB-231	Breast cancer	Yes	Yes	Yes	Li et al. 2017a
	miR-106a	SGC7901/PC-3	Gastric cancer/prostate cancer	Yes/yes	Yes/yes	Yes/NA	Fang et al. 2013/Lu et al. 2019
	miR-106a-3p	HT29	Colorectal adenocarcinoma	Yes	Yes	Yes	Qin et al. 2018
	miR-106b	MCF-7 and MDA-MB-231	Breast cancer	Yes	Yes	Yes	Li et al. 2017a
miR-19	miR-19	DU145	Prostate cancer	Yes	Yes	Yes	Poliseno et al. 2010a
	miR-19a	NCTC1469 cells and HEP1–6/HOS	Macrophage-like and hepatocellular carcinoma/osteosarcoma	Yes/yes	Yes/yes	NA/yes	Dou et al. 2015; Zhao et al. 2017
	miR-19a-3p	U2OS	Osteosarcoma	Yes	Yes	Yes	Zhang et al. 2019
	miR-19b	DU145/H9C2/MDA-MB-231	Prostate cancer/rat cardiomyocyte/breast cancer	Yes/yes/yes	NA/NA/yes	NA/NA/yes	Poliseno et al. 2010b; Xu et al. 2016; Li et al. 2017b
miR-21	miR-21	SK-HEP-1 and SNU-182/VSMC/A549	Hepatocellular carcinoma/vascular smooth muscles/non-small lung cancer	Yes/yes/yes	Yes/NA/NA	NA/NA/NA	Meng et al. 2006, 2007; Ji et al. 2007;
miR-22	miR-22	DU145	Prostate cancer	Yes	Yes	Yes	Poliseno et al. 2010a
miR-23	miR-23a-3p	HEK 293	Model	Yes	Yes	NA	Lozano-Bartolomé et al. 2018
	miR-23b-3p	A-498	Kidney carcinoma	Yes	Yes	NA	Zaman et al. 2012
miR-25	miR-25	DU145	Prostate cancer	Yes	Yes	Yes	Poliseno et al. 2010a
	miR-92a	U2OS/U14/A549	Osteosarcoma/cervical cancer/adenocarcinoma	Yes/yes/yes	Yes/yes/yes	Yes/yes/yes	Ren et al. 2016; Li et al. 2017b; Xiao et al. 2017
miR-26	miR-26a	HEK 293	Model	Yes	Yes	Yes	Huse et al. 2009
miR-33	miR-33a	HEK 293	Model	Yes	Yes	NA	Ling et al. 2013
miR-103	miR-103	UMUC2 and 5637	Bladder cancer	Yes	Yes	Yes	Yu et al. 2019
	miR-103a	HEK 293/HEK 293	Model/model	Yes/yes	Yes/yes	Yes/yes	Geng et al. 2014; Hsu et al. 2017
	miR-107	EJ cells	Bladder cancer	Yes	Yes	Yes	Chi et al. 2018; Yu et al. 2019
miR-130	miR-130b	HEK 293T /HCCLM3 and HepG2	SV40 large T antigen model/hepatocellular carcinoma	Yes/yes	Yes/yes	Yes/yes	Chang et al. 2016; Miao et al. 2017
miR-136	miR-136	U343	Glioblastoma	Yes	Yes	Yes	Lee et al. 2010
miR-144	miR-144	U343	Glioblastoma	Yes	Yes	Yes	Lee et al. 2010
miR-154	miR-494	Huh-7/INS-1	Hepatocellular carcinoma/rat insulinoma	Yes/yes	NA/yes	NA/yes	He et al. 2017; Pollutri et al. 2018
	miR-494-3p	SMMC-7721 and HCCLM3	HPV-related endocervical adenocarcinoma	Yes	Yes	Yes	Lin et al. 2018
miR-181	miR-181-5p	HEK 293	Model	Yes	Yes	NA	Lozano-Bartolomé et al. 2018
miR-193	miR-193a	786-O cells	Renal cell carcinoma	Yes	Yes	Yes	Liu et al. 2017
miR-214	miR-214	A2780CP and HIOSE-80	Ovarian cancer	Yes	Yes	Yes	Yang et al. 2008
miR-216	miR-216a	Mouse glomerular mesangial cells	Normal kidney	Yes	Yes	Yes	Kato et al. 2009
miR-216a+miR-217		PLC/-PRF/5	Hepatocellular carcinoma	Yes	Yes	Yes	Xia et al. 2013
miR-217	miR-217	mouse glomerular mesangial cells/HEK 293T	Normal kidney/SV40 large T antigen model	Yes/yes	Yes/yes	Yes/yes	Kato et al. 2009; Nie et al. 2018
miR-221	miR-222	OVCAR-3	Ovarian cancer	Yes	Yes	Yes	Zhou et al. 2012; Gong et al. 2018
miR-302	miR-302	DU145	Prostate cancer	Yes	Yes	Yes	Poliseno et al. 2010a
miR-506	miR-510-3p	NCI-H441 and PC-9	Non-small lung cancer	Yes	Yes	Yes	Yu et al. 2019

Validation ranges from having any biological evidence such as western blots and qPCR data, to analyses of the isolated fragment of the 3′UTR in a luciferase reporter assay, which may have the miRNA recognition element (MRE) mutated to further demonstrate loss of repression.

(NA) not applicable.

An ongoing challenge in the study of miRNA biology is the correct identification of miRNA:mRNA interactions; this is especially arduous because each miRNA has numerous mRNA targets and vice versa. Prediction of an MRE within a transcript typically begins with the use of one of many prediction algorithms. The diverse biological features that govern miRNA binding to an MRE are differentially weighed within these algorithms, further to being informed by experimental data (Liu et al. 2013a,b; Peterson et al. 2014). Subsequent validation of miRNA:mRNA interactions range from further in silico associations to in-depth in vitro studies and may include (1) reporter construct assays to show a direct miRNA/transcript association, (2) validation with a mutated MRE, and (3) miRNA-inhibitor-mediated derepression of expression. Advanced techniques such as high-throughput sequencing of RNAs isolated by cross-linking immunoprecipitation (HITS-CLIP) or photoactivatable ribonucleoside-enhanced CLIP (PAR-CLIP) have experimentally identified functional miRNA:mRNA interaction maps (Chi et al. 2009; Hafner et al. 2010). Although these emerging techniques are complex and biochemically challenging, they provide great promise for high-throughput simultaneous discovery of many miRNA:mRNA interactions. Indeed, a recent study by Zarringhalam and colleagues has analyzed multiple datasets from PAR-CLIP experiments in conjunction with RNA-Seq data to evaluate PTEN miRNA networks and their putative biological functions (Zarringhalam et al. 2017).

Competing Endogenous RNA and PTEN

It has been hypothesized that RNA transcripts have the potential to regulate the expression of other RNAs in trans. Such transcripts have been termed ceRNA because of their ability to effectively bind and sequester specific miRNAs, which prevents them from binding to other target RNAs (Poliseno et al. 2010b; Cesana et al. 2011; Salmena et al. 2011). The ceRNA hypothesis posits that as the levels of one RNA transcript that contains a given MRE increases, the levels of another mRNA that contains that same MRE also increases in a dose-dependent manner (Seitz 2009; Salmena et al. 2011; Ala et al. 2013). Despite recognition of the ceRNA hypothesis, a full understanding of ceRNA is still in its infancy and its generalized function, physiological ramifications, and widespread acceptance remain contentious (Bosson et al. 2014; Denzler et al. 2014, 2016). Nevertheless, close to 50 publications collectively describe a large network of coding and noncoding ceRNA that regulate PTEN expression (Table 2). This collection of articles suggest that the PTEN “ceRNome” can influence PTEN expression and, as a consequence, perturbations in this network may have effects on PTEN function and tumorigenesis (Fig. 1).

Table 2.

Phosphatase and tensin homolog deleted on chromosome 10 (PTEN) competing endogenous RNA (ceRNA)

ceRNA	miRNA implicated	Disease/condition	References
PTENP1	miR-17, miR-19, miR-21, miR-214, and miR-26	Prostate cancer and others	Poliseno et al. 2010b
FER 14	miR-106a, miR-18a-5p	Gastric cancer, osteosarcoma	Xia et al. 2015; Yue et al. 2015
MEG3	miR19a	Glioma	Qin et al. 2017
HOTAIR	miR19a	Cardiac hypertrophy	Lai et al. 2017
SLC25A5-AS1	miR19a	Gastric cancer	Li et al. 2019
Linc-USP16	miR21	Breast cancer	Li et al. 2017a
Gas5	miR-222-3p and miR-21	Papillary thyroid carcinoma/breast cancer	Zhang et al. 2013, 2018b; Gong et al. 2018
NEAT1	miR9	Bladder cancer	Lu et al. 2016; Xie et al. 2018
TP73-AS1	miR103	Colon cancer	Jia et al. 2019; Yu et al. 2019
LINC00702	miR-510	Non-small-cell lung cancer (NSCLC)	Yu et al. 2019
ORLNC1	miR-296	Osteogenesis and adipogenesis	Kakizaki et al. 2017; Yang et al. 2018

Figure 1.

MicroRNA (miRNA) networks regulating phosphatase and tensin homolog deleted on chromosome 10 (PTEN) expression. The complex competing endogenous RNA (ceRNA) regulatory network of PTEN that emerges as a result of miRNA's ability to bind to multiple transcripts (long noncoding RNA [purple] and protein-coding RNA [blue]). miRNAs that have been demonstrated to bind to PTEN and to a second RNA transcript are highlighted and colored by their seed family (see Table 1).

The first ceRNA interaction was reported between PTEN and PTEN pseudogene 1 (PTENP1) (Poliseno et al. 2010b). The PTENP1 locus was first described in 1998 in two concurrent studies (Dahia et al. 1998; Whang et al. 1998). PTENP1, localized at 9p13.3 is a processed pseudogene that is highly homologous to PTEN containing only 18 mismatches throughout the coding sequence and a missense mutation of the initiator methionine codon that prevents translation, thus sharing several common MREs with the PTEN mRNA (Fujii et al. 1999). Like PTEN, PTENP1 was demonstrated to have tumor-suppressive features including growth suppression on loss of expression in some cancers (Poliseno et al. 2010b).

Consistent with the ceRNA hypothesis, it was observed that PTEN mRNA expression is directly proportional to PTENP1 transcript expression in diverse settings. This was posited to be a consequence of PTEN and PTENP1 transcripts simultaneously controlled by several miRNA at their shared MREs (Poliseno et al. 2010b; Ioffe et al. 2012; Kovalenko et al. 2013, 2018; Yu et al. 2014). Within highly homologous regions, MREs for miR-17, miR-19, miR-21, and miR-214 families have been found, with miR-19b and miR-20a shown to directly down-regulate PTENP1 (Poliseno et al. 2010b). Subsequent studies expanded the list of miRNAs with direct effects on both PTEN and PTENP1, to include miR-26a and miR-193a-3p (Poliseno et al. 2010b; Chen et al. 2015; Gao et al. 2017; Li et al. 2017b; Qian et al. 2017; Yu et al. 2017; Yndestad et al. 2018). By mutating the MRE sequences of these miRNA families within the PTENP1 gene, degradation of PTENP1 was impaired, the effect on PTENP1 expression was derepressed, and PTEN protein levels decreased (Poliseno et al. 2010b; Yu et al. 2014). To date, PTENP1 and PTEN interactions have been reported to influence several tumor types including NSCLC, prostate cancer, melanoma, endometrial cancer, acute myeloid leukemia, clear-cell renal carcinoma, bladder cancer, hepatocellular carcinoma, gastric cancer, oral squamous cell carcinoma, and breast cancer (Marsit et al. 2005; Poliseno et al. 2010b, 2011; Ioffe et al. 2012; Wang et al. 2012; Yu et al. 2014, 2017; Chen et al. 2015; Guo et al. 2016; Gao et al. 2017; Li et al. 2017b; Qian et al. 2017; Yndestad et al. 2017, 2018; Zhang et al. 2017; Shi et al. 2018).

Beyond the PTENP1 pseudogene, a number of additional long noncoding RNA (lncRNAs) have also been reported to regulate PTEN through the ceRNA mechanism. By exploring the association between miR-106a and PTEN, and the data demonstrating miR-106a MRE on FER1L4, it was postulated that the ceRNA network of the FER1L4 lncRNA includes PTEN (Xia et al. 2014a). This possibility was supported by data in gastric cancer where FER1L4 was down-regulated and that its level corresponded with that of PTEN mRNA. Further experimental data showing that FER1L4 down-regulation decreased the abundances of PTEN mRNA supported these findings (Xia et al. 2015; Yue et al. 2015). In a related study, miR-18a-5p was found to mediate a ceRNA interaction between FER1L4 and PTEN in osteosarcoma (Fei et al. 2018). miR-19a was shown to mediate a direct and reciprocal ceRNA interaction between PTEN and MEG3 in glioma (Qin et al. 2017); HOTAIR and PTEN in cardiac hypertrophy (Lai et al. 2017) and SLC25A5-AS1 and PTEN in gastric cancer (Lai et al. 2017; Li et al. 2019). miR-21 mediates a ceRNA interaction between Linc-USP16 and PTEN in hepatocellular carcinoma where Linc-USP16 overexpression also resulted in increased PTEN expression (Zhang et al. 2010; Yang et al. 2013; Sui et al. 2017). Finally, lncRNAs including Gas5, LINC00702, NEAT1, RP11-79H23.3 and TP73-AS1, and ORLNC1 have also been reported to regulate PTEN through the ceRNA mechanism (Xia et al. 2014a; Kakizaki et al. 2017; Lai et al. 2017; Qin et al. 2017; Sui et al. 2017; Chi et al. 2018; Xie et al. 2018; Yang et al. 2018; Jia et al. 2019; Li et al. 2019; Yu et al. 2019).

In addition to lncRNA acting as a ceRNA, several protein-coding genes have also been reported to act as a ceRNA for PTEN. For example, expression of VCAN was found to modulate the expression of PTEN and RB1 (Lee et al. 2010). Both PTEN and VCAN were down-regulated in response to miR-144 and miR-136 expression, supporting a possible interaction through a ceRNA mechanism (Lee et al. 2010). PTEN was later reported to be coregulated with the DKK1 transcript, as both are inhibited by miR-217, -33a, -33b, -103a, -93, and -106a, through a proposed ceRNA interaction (Ling et al. 2013). Similarly, both CASC2 and PTEN mRNA contain miR-21 MREs, and are posited to be coregulated through a ceRNA regulatory mechanism (Meng et al. 2007; Zhang et al. 2010; Feng et al. 2017). Finally, variations in SERINC1, VAPA, CNOT6L, and ZEB2 transcripts were reported to impact PTEN expression levels through ceRNA (Karreth et al. 2011; Tay et al. 2011).

Large numbers of other genes including TNRC6B, RB1, TP53, NRAS, KLF6, HIF1A, HIAT1, CTBP2, and TNKS2 and other very large RNA networks have been identified as PTEN ceRNA through bioinformatic approaches; however, many such computational efforts lack any extensive experimental validation (Sumazin et al. 2011; Zarringhalam et al. 2017; Chiu et al. 2018). Overall, there is an overwhelming abundance of data suggesting that both coding and noncoding transcripts can act as PTEN ceRNAs; however, many studies fail to perform experimental validation that will be necessary to identify bona fide PTEN ceRNA and ceRNA networks.

PTEN Regulation by RNA-Binding Proteins

RNA-binding proteins (RBPs) are critical regulators of transcriptional and posttranscriptional gene expression in all eukaryotic organisms. Unlike other regulatory modalities affecting PTEN expression such as miRNA, there is limited evidence on the role of RBPs in PTEN biology. The most extensively studied PTEN:RBP interaction is with the RNA-binding motif protein 38 (RBM38), an RBP and a target of the P53 family, which has broad roles in transcriptional and posttranscriptional regulation. RBM38 directly interacts with, and increases the stability of, PTEN mRNA by binding to the AU-rich elements within the 3′UTR in multiple breast cancer cell lines (Zhou et al. 2017). Subsequent work in a Rbm38 knockout mouse confirmed that by binding to and stabilizing Pten mRNA, Rbm38 was integral to Pten tumor suppression and that loss of Rbm38 synergizes with mutant p53 to promote lymphomagenesis (Zhang et al. 2018a). Related studies indicate that RBM38 may be occluding miRNA binding to the PTEN-3′UTR and thereby stabilizing PTEN mRNA, as shown previously for the TP53 transcript (Léveillé et al. 2011). RBM38 may also affect PTEN mRNA splicing, similar to its impact on red blood cell (Heinicke et al. 2013) and myoblast differentiation (Miyamoto et al. 2009). Overall, RBM38 is thought to function as a tumor-suppressor gene in part through its role in stabilizing PTEN and p53 transcripts in renal, liver, and breast cancers and acute myeloid leukemia (Wampfler et al. 2016; Huang et al. 2017; Li et al. 2017c; Wu et al. 2017a; Ye et al. 2018).

In addition to RBM38, the cellular senescence-inhibited gene (CSIG) interacts with the PTEN mRNA in the 5′UTR, impacting PTEN translational regulation (Ma et al. 2008). An interaction between IGF2 mRNA-binding protein 1 (IGF2BP1) and PTEN was shown to be important in directing tumor cell migration through phosphoinositide signaling control (Stöhr et al. 2012). Finally, RBPs including polypyrimidine tract-binding protein 1 (PTBP1), IGF2BP1, cytoplasmic polyadenylation element-binding protein (CPBP), and Staufen1, have been reported to regulate PTEN mRNA stability and translation; however, direct binding data was less than convincing in these studies (Alexandrov et al. 2012; Stöhr et al. 2012; Crawford Parks et al. 2017; Wang et al. 2018). Further work on the relationship of RBPs and PTEN is needed to fully understand their relevance to PTEN biology.

PTEN Regulation by Alternative Splicing and Polyadenylation

Alternative splicing of the PTEN transcript has long been suggested as an important aspect of PTEN regulation and biology (Sharrard and Maitland 2000; Okumura et al. 2011; Oltean and Bates 2014; Malaney et al. 2017). Early studies uncovered two different PTEN splice variants (SVs) in glioblastoma and prostate cancer cells that were termed PTEN-Δ and PTEN-B (Sharrard and Maitland 2000). Both PTEN SVs were demonstrated to generate severe carboxy-terminal truncations in the PTEN protein with little to no phosphatase activity, likely because of decreased stability and rapid degradation. Remarkably, recent work in renal cell carcinoma revealed that PTEN-Δ exerts similar tumor-suppressive effects as full-length PTEN, and that patients with high PTEN-Δ expression had longer metastasis-free and overall survival (Breuksch et al. 2018).

A proportion of individuals with Cowden syndrome (CS) or Bannayan–Riley–Ruvalcaba syndrome (BRRS), the multiple hamartoma conditions associated with germline loss of PTEN, have been found to harbor PTEN SVs, which exclude exons 3, 4, or 6, leading to decreased levels of PTEN protein (Agrawal et al. 2005; Sarquis et al. 2006). These findings were further corroborated by the discovery of PTEN intronic variants in CS and BRRS patients mapping near the splice sites and resulting in exon skipping or activation of cryptic splice sites, eventually resulting in lower expression of PTEN and activation of Akt (Chen et al. 2017). At least eight additional PTEN SVs have been identified, which involve maintaining portions of introns 3 (transcripts 3a, 3b, and 3c) and 5 (transcripts 5a, 5b, and 5c) in the final, processed transcript, as well as SVs that result in the partial deletion of exon 5 (DelE5) or the complete exon 6 (DelE6), and have been found to be differentially expressed in breast cancer (Agrawal and Eng 2006). Other studies have highlighted the inherent complexity of PTEN SVs regulation. For instance, Wang and Chang (1999) found high levels of abnormal PTEN SV transcripts in both normal and tumor-derived tissues, but could not connect them to tumorigenesis (Wang and Chang 1999), whereas Liu et al. (2010) found discrepancies in the detection and proportion of PTEN SVs between fresh and aged mononuclear cells (Liu et al. 2010).

PTEN transcript is also subject to alternative polyadenylation with the main site residing 3 kb downstream of the canonical stop codon (Gray et al. 1998) and the alternative polyadenylation sites 60 and 290 bp downstream of the stop codon, resulting in shorter PTEN transcripts (Hamilton et al. 2000; Xia et al. 2014b). In a study by Thivierge et al., it was hypothesized that shorter PTEN 3′UTRs would limit the access of certain miRNAs and lead to PTEN mRNA stabilization (Thivierge et al. 2018). Surprisingly, this was not observed as longer 3′UTRs were resistant to miRNA-mediated silencing and deemed responsible for the bulk of PTEN protein dosage and signaling functions. Therefore, longer 3′UTRs may possess other features that lead to stabilization and miRNA resistance (Thivierge et al. 2018).

Alternative splicing is a major contributor to transcriptome and proteome diversity; however, deregulated splicing of important cancer genes such as PTEN can lead to a variety of human diseases including cancer. Thus, a better understanding of the mechanisms governing PTEN alternative splicing and the consequences on gene dosage and downstream signaling will shed light on the complex roles of PTEN in tumor suppression and cancer. Importantly, cancer-associated splicing variants of PTEN are potential molecular markers that may be used in prognosis and diagnosis of disease.

PTEN Regulation by mRNA Modification

N⁶-methyladenosine (m⁶A) is the most common mRNA modification in eukaryotes (Desrosiers et al. 1974; Wei et al. 1976). An m⁶A modification entails the covalent addition of a methyl group to an mRNA by a methyltransferase complex comprised of several proteins including METTL3 (Bokar et al. 1997) and METTL14 (Liu et al. 2014). The importance of m⁶A-modified mRNAs in cancer has only recently been appreciated with the discovery that m⁶A methylation contributes to tumor suppression and that loss of m⁶A methylation correlates with malignant, stem-like phenotypes in breast and cervical cancers (Zhang et al. 2016a,b; Wang et al. 2017b). Moreover, overexpression of METTL3 promotes cancer progression and chemoresistance in other cancer types, including hepatic and pancreatic cancers (Chen et al. 2018; Taketo et al. 2018).

To date the role of m⁶A modification on PTEN regulation is still unclear. However, a single study has shown that METTL3 was able to directly regulate expression of critical downstream effectors, including, but not limited to, c-MYC, BCL-2, and PTEN (Vu et al. 2017). Specifically, loss of METTL3 expression led to reduced m⁶A methylation of the PTEN transcript and consequently reduced expression of the PTEN protein, suggesting an important role for m⁶A methylation in the modulation of PTEN expression levels (Vu et al. 2017).

CONCLUDING REMARKS

The variety of mRNA regulatory pathways that converge on PTEN mRNA and impact the levels of PTEN protein are complex and relatively unexplored. Modifications, sequestration, and miRNA-targeting of the PTEN mRNA present an important layer of PTEN regulation in normal and PTEN deregulation in cancer cells (Fig. 2). Importantly, these pathways present possible therapeutic opportunities in attempting to boost PTEN expression in germline carriers of heterozygous mutations, as well as pathologies where PTEN expression is silenced by overactive miRNA silencing. Attempts to inhibit PTEN-targeting miRNAs, but more generally methods to stabilize or restore PTEN mRNA levels (Islam et al. 2018), should be explored as possible therapeutics in cancer, thus adding to the variety of treatment options traditionally focused on interfering with the downstream consequences of PTEN loss.

Figure 2.

Regulatory pathways modulating phosphatase and tensin homolog deleted on chromosome 10 (PTEN) messenger RNA (mRNA). PTEN expression is modulated by a number of factors that directly interact with or are inherent to the transcript. Included in these pathways are alternative splice variants of PTEN that are found to be expressed in tumor tissue and may lead to tumor-associated phenotypes; these splice variants include but are not limited to inclusion of portions of intron 3 (3a, 3b, and 3c) and intron 5 (5a, 5b, and 5c), partial deletion of exon 5 (DelE5), and complete deletion of exon 6 (DelE6) (Okumura et al. 2011). The PTEN transcript also contains AU-rich binding sites in the 3′UTR for the RNA-binding protein RBM38 that occludes miRNA binding (Zhou et al. 2017) and N⁶-methyladenosine (m⁶A) sites for binding and subsequent methylation by the METTL3/METTL14 methyltransferase complex, which promotes translation (Vu et al. 2017).