**The Role of miRNA for the Treatment of MGMT Unmethylated Glioblastoma Multiforme**

### **Anna Kirstein 1,2 , Thomas E. Schmid 1,2 and Stephanie E. Combs 1,2,3,\***


Received: 7 April 2020; Accepted: 26 April 2020; Published: 28 April 2020

**Abstract:** Glioblastoma multiforme (GBM) is the most common high-grade intracranial tumor in adults. It is characterized by uncontrolled proliferation, diffuse infiltration due to high invasive and migratory capacities, as well as intense resistance to chemo- and radiotherapy. With a five-year survival of less than 3% and an average survival rate of 12 months after diagnosis, GBM has become a focus of current research to urgently develop new therapeutic approaches in order to prolong survival of GBM patients. The methylation status of the promoter region of the O<sup>6</sup> -methylguanine–DNA methyltransferase (MGMT) is nowadays routinely analyzed since a methylated promoter region is beneficial for an effective response to temozolomide-based chemotherapy. Furthermore, several miRNAs were identified regulating MGMT expression, apart from promoter methylation, by degrading MGMT mRNA before protein translation. These miRNAs could be a promising innovative treatment approach to enhance Temozolomide (TMZ) sensitivity in MGMT unmethylated patients and to increase progression-free survival as well as long-term survival. In this review, the relevant miRNAs are systematically reviewed.

**Keywords:** glioblastoma; miRNA; MGMT; survival; radiotherapy; chemotherapy; temozolomide; translational medicine

### **1. Introduction**

Cancer is one of the leading causes of death worldwide with 14 million new cases diagnosed and eight million deaths every year. With 256,000 cases per year, tumors of the central nervous system account for 2% of all diagnosed cancers and are therefore the 17th most common cancer [1]. Each year the American Cancer Society compiles a cancer statistic for the USA, estimating the annual cancer incidences and mortalities based upon mortality data from 1930 to 2017 and incidence data from 1975 to 2016 [2]. For 2020, they estimate 23,890 new brain and other nervous system cases with 18,020 deaths [2]. Although the yearly incidence rate with 4-8 cases per 100,000 worldwide is relatively low, mortality rates are significantly high, making it the 12th most frequent cause of cancer-related deaths. The most common primary malignancy of the central nervous system is Glioblastoma multiforme (GBM) [1].

Glioblastoma multiforme is one of the most common and most aggressive primary brain tumors with a five-year survival of less than 3% [3] and an increasing incidence rate [4]. In recent years, the mechanisms explaining the radio- and chemoresistance of glioblastoma have been extensively studied but are still poorly understood. Radiotherapy with concomitant or adjuvant temozolomide-based chemotherapy following surgery has become the standard treatment for GBM [5]. However, short median survival is still observed in patients with an unmethylated promoter region of the O6 -methylguanine–DNA methyltransferase (MGMT) [6]. Despite the minor benefit of additional Temozolomide (TMZ) to unmethylated patients and regardless of the treatment regimen, MGMT promoter methylation status is routinely investigated in all patients after surgery as an independent prognostic biomarker. Therefore, an unmethylated MGMT promoter is an inherent prognostic indicator for poor overall-survival, which demonstrates the urgent need for the identification of new prognostic factors, especially for these patients. For this specific patient group, tailored study concepts have been performed with intensified TMZ or with concepts omitting TMZ but adding novel potentially effective substances such as Vascular Endothelial Growth Factor (VEGF)-inhibitors, integrin-antagonists or other molecular targeted substances. To date, all of these studies were negative and have not offered additional benefit [7–10].

In recent years, circulating microRNAs have been extensively studied as tumor biomarkers to predict therapy outcome and to follow up therapy response. miRNAs are endogenous, single-stranded, non-coding small RNA molecules with a length of about 22 nucleotides [11]. The interaction of miRNAs with the target mRNA leads to the degradation or translational repression of the target mRNA, which ultimately results in the down-regulation of the designated protein. This regulatory network of miRNAs affects many different biological functions and therefore represents a great potential for clinical applications [12]. The up- or down-regulation of miRNAs in tumor cells is deterministic for either a tumor-suppressive or an oncogenic characteristic of the respective miRNA [12].

In this manuscript, the relevant literature investigating the relationship between different miRNAs and glioblastoma was systematically reviewed and the results were analyzed to evaluate the value of different miRNAs in the treatment of GBM.

#### **2. Glioblastoma Multiforme**

The 2007 World Health Organization (WHO) classification of tumors of the central nervous system [13] is mainly based on microscopic analyses of hematoxylin and eosin-stained sections, immunohistochemistry of lineage-associated proteins and characterization of ultrastructures. Important characteristics include nuclear atypia, mitotic activity, vascularization, necrosis, pleomorphism and microvascular proliferation [14]. In the updated version from 2016 [15], molecular markers are taken into account proposing a more detailed classification of glioblastoma. GBM is classified as grade IV diffuse astrocytic tumor and is characterized by uncontrolled cellular proliferation, diffuse infiltration and intense resistance to radiotherapy [3]. Due to increasing evidence towards a different origin of primary and secondary GBM [16–19], GBM is now subdivided into isocitrate dehydrogenase (IDH)-wildtype, IDH-mutant and not otherwise specified (NOS) glioblastoma. NOS glioblastoma are either primary or secondary glioblastoma, but a full evaluation of the IDH status is either inconclusive or not performed due to the patient's age [15]. IDH-wildtype or primary glioblastoma develops rapidly within 3-6 months directly from glial progenitor cells and is characterized by diffuse infiltration, extensive necrosis and a unique mutation pattern [1]. EGFR amplification [17], PTEN mutation [18,20,21] and loss of chromosome 10 [17,18,20] are particular features of primary GBM, as well as the older age of the patients. The median age at diagnosis is 62 years with a male-to-female ratio of 1.46 [14,20], and the median overall-survival is 15 months [15]. A total of 90% of all glioblastoma are primary glioblastoma [14,15]. IDH-mutant or secondary glioblastomas, in contrast, develop over several years from low-grade astrocytomas (WHO grade II) and anaplastic astrocytomas (WHO grade III) and feature a different, unique mutation pattern, which was postulated to be the result of a sequential acquisition [1,19,22,23]. This pattern includes a TP53 mutation [1,17,20], LOH on chromosomes 10q and 19q [1,17,18,20,24] as well as deletion of *p16* [20,25] and inactivation of RB [20,23,25]. Median age at diagnosis is 44 years with a median overall-survival of 31 months and a male-to-female ratio of 1.12 [14,15,19,20]. Although there is no universally accepted glioblastoma stem cell marker and there

might be several stem cell markers [26], CD133 expression is significantly higher in primary, compared to secondary glioblastoma [27]. This might explain the intense resistance to chemo- and radiotherapy of primary glioblastoma due to the presence of potential glioblastoma stem cells.

#### *2.1. Current Treatment of GBM*

Treatment of patients with GBM is always interdisciplinary. For all treatments, the strongest prognostic factors are patient's age, performance score, tumor volume as well as molecular characterization. Imaging information from magnetic resonance imaging (MRI), computer tomography (CT), positron-emission tomography (PET) as well as other functional imaging, such as 5ALA, provide a basis for solid characterization of tumor extension. After imaging diagnosis, surgical resection of the tumor mass is crucial to relieve symptoms such as headache, vision and memory problems as well as nausea [28] and should be performed following the rules of maximal-safe resection. Resection allows for pathological examinations to confirm the diagnosis and to investigate several molecular markers, such as MGMT and IDH status. The diffuse infiltrative characteristic, as well as extensive vascularization into the surrounding healthy tissue, limits the complete resection of GBM and makes recurrence highly possible [3]. Hence, complete surgical resection is almost impossible and, therefore, surgery is followed by radiotherapy, generally concomitant with chemotherapy to eliminate tumor cells in the microenvironment as well.

In the 1970s, BCNU (bis-chloroethylnitrosourea—carmustine) was discovered and since then administered as an alkylating antineoplastic agent as it was shown to penetrate the blood brain barrier (BBB) and to be effective in treating intracranial neoplasms [29]. However, the combination of BCNU and radiotherapy did not significantly enhance median survival [29].

Since 2005, administration of the oral alkylating agent temozolomide (TMZ) presents the standard agent for GBM patients, as it causes only mild side-effects and efficacy has been proven in clinical trials [5]. It is given as a daily dose of 75 mg per m<sup>2</sup> body-surface area for five consecutive days for six weeks [28]. After four weeks, the dose is increased to 150 mg per m<sup>2</sup> . Adjuvant, conventional radiotherapy is given in 30 fractions at 2 Gy to a total dose of 60 Gy over a period of six weeks [28]. Alternatively, hyperfractionated radiotherapy is given for 15 days with a total dose of 34 Gy in 3.4 Gy fractions or in 15 daily fractions to a total dose of 10 Gy in 2.6 Gy fractions [28].

After radiochemotherapy with TMZ was introduced, it has been shown that patients with an unmethylated MGMT promoter as well as older patients benefit less from TMZ [30]. However, it has also been shown that even in elderly patients treated with short course radiotherapy concomitant treatment improves outcome [31]. These inconclusive data argue for more accurate discrimination of patient subgroups. A 4-miRNA signature consisting of let-7b-5p, miR-125a-5p, miR-615-5p and let-7a-5p was proposed to assign patients into high- and low-risk groups [32]. Three of the four miRNAs—let-7b-5p, let-7a-5p and miR-125a-5p—are tumor suppressive in GBM and are higher expressed in the low-risk GBM group [32]. Only miR-615-5p does not show a tendency towards a certain expression level in either risk group [32]. This leads to the promising conclusion that this 4-miRNA signature is associated with overall survival of GBM patients. This 4-miRNA could be used to differentiate GBM patients and predict therapy outcome. Still, all possibilities should be evaluated in newly diagnosed as well as recurrent patients, including surgery, radiotherapy and chemotherapy. Again, the extent of surgical resection is crucial [33] and the benefit of radiotherapy for recurrent GBM is evident for resected as well as unresected lesions [34–37].

Recurrence or progression is almost inevitable and is postulated after a median time of 32 to 36 weeks after treatment completion and a final mortality rate close to 100% [38]. This alone describes the urgent need for treatment improvement and the discovery of alternative treatment regimes.

#### *2.2. TMZ and MGMT*

Since 2005, the standard treatment of glioblastoma involves early adjuvant chemotherapy with the administration of TMZ [5,39]. TMZ is a prodrug from an imidazotetrazine derivative, which is stable in acidic pH and rapidly hydrolyzes by passing through neutral to basic pH [40–42]. Therefore, it survives the gastric acid enabling an oral administration. Due to the lipophilic character of the prodrug, it is able to penetrate the BBB [41]. Only in the brain, where the pH is around 7, spontaneous ring-opening hydrolysis of the imidazotetrazine leads to the formation of the active alkylating metabolite 3-methyl-(triazen-1-yl) imidazole-4-carboximide (MTIC) intermediate [40]. MTIC, in turn, is unstable at pH values below 7 but stable in an alkaline environment [42]. Further hydrolysis of MTIC forms 5-amino-imidazole-4-carboxamide (AIC) and methyl diazonium ions, which react with nucleophilic sites on the DNA producing methyl adducts [41]. There are several sites for DNA methylation, such as N<sup>7</sup> (70%) and O<sup>6</sup> (5%) of the base guanine as well as the N<sup>3</sup> (9%) site of adenine [41,43]. However, only the relatively rare site of the O<sup>6</sup> position at the base guanine is of importance for the anti-cancer activity of TMZ [41,42] and this site is, therefore, speculated to be mutagenic and cytotoxic [44,45].

During DNA replication, O<sup>6</sup> -methylguanine pairs with thymine creating a wobble base pair. This mismatch is repaired by the DNA mismatch repair (MMR) pathway, which involves the recognition of the mismatch via several mismatch recognition complexes [46]. Single-stranded DNA nicks are created in close proximity to the wobble base pair allowing accessibility to the mismatched base thymine, which is digested by the 50 -30 exonuclease I [46]. Eventually, DNA polymerase δ fills the gap with a new thymine [46,47]. Continuous rounds of thymine deletion and insertion eventually lead to a depletion of deoxythymidine triphosphates (dTTP). Lack of dTTP will result in a lack of DNA synthesis and ultimately causes cell death via DNA double-strand breaks [47].

MGMT or sometimes also called the O<sup>6</sup> -alkylguanine-DNA-alkyltransferase is a nuclear protein involved in this mismatch repair pathway [48]. MGMT, therefore, protects not only normal cells from apoptosis but also tumor cells. It removes alkyl groups, preferably methyl groups, from the O6 -methylguanine to counteract the futile circles of thymine deletion and insertion [49]. The removed methyl groups are covalently transferred to a cysteine acceptor residue contained within the active site of MGMT [50]. This results in a conformational change, which leads to degradation of the MGMT protein. As the cysteine site is not regenerated, the reaction is a suicide reaction [50], which makes MGMT a protein and not an enzyme [49]. Hence, the amount of methyl groups that can be removed is limited to the amount of MGMT present in the cell, which is dependent on the MGMT promoter methylation status. So, the absence or presence of MGMT mainly contributes to the chemoresistant character of GBM [48,49].

MGMT, therefore, counteracts the therapeutic efficacy of TMZ and promotes treatment failure. Stupp et al. discovered in their studies from 2000 to 2002 that administration of TMZ starting early in the treatment course and adjuvant to radiotherapy increases median survival to 2.5 months and a resulting survival rate of 27% [5]. This constant treatment regime makes dose escalation possible as well as depletion of MGMT.

In 2005, Hegi et al. published that the promoter methylation status of MGMT is an important prognostic biomarker to predict the TMZ chemotherapy outcome [39]. Overall survival of patients with a methylated MGMT promoter who received radiotherapy plus temozolomide was significantly increased compared to patients with an unmethylated MGMT promoter [39]. MGMT promoter unmethylated patients have no or only little benefit from TMZ adjuvant to radiotherapy, which suggests that other mechanisms play a role to overcome TMZ resistance. Since then, the MGMT promoter methylation status in GBM patients is routinely investigated after surgery to predict which patients would benefit most from TMZ.

Recent studies have shown that MGMT expression does not always correlate with MGMT promoter status and that some individual patients with an unmethylated MGMT promoter show comparable long-term survival [51]. This leads to the assumption that other mechanisms are active in regulating MGMT expression, which includes miRNAs [52]. Therefore, new innovative and personalized treatment options need to be developed, especially for patients with an unmethylated MGMT promoter. Some compounds were already tested or are currently tested in clinical trials for the treatment of unmethylated patients.

#### *2.3. Current Diagnostic and Prognostic Biomarkers for GBM*

The most commonly analyzed biomarkers in GBM are currently IDH status, MGMT status, 1p/19q co-deletion and ATRX loss [53]. There are, however, several classes of molecules, proposed to aim as biomarkers for GBM detection, which are found in the blood, cerebrospinal fluid (CSF) and urine.

Proteins are detectable in all kinds of body fluids and can be easily withdrawn from the patient. GBM-specific protein markers include VEGF, angiogenesis-associated proteins, extracellular matrix proteins, matrix metalloproteinases, cell line associated proteins, macrophage migration inhibitory factor (MIF) as well as functionally-related proteins, such as CD44 [53,54]. CD44 was shown as a potential marker for survival outcome and treatment resistance [54]. All these have shown deviating amounts and compositions in patients where tumor progression was observed [53].

Another class used for biomarkers are small molecules, such as lipids and metabolites. Due to their low specificity and small size, they can only be used to verify a diagnosis after other markers were tested positive [53].

Circulating tumor cells (CTCs), which are primary tumor cells circulating in the body via the blood stream, for example, might be important in other cancers apart from GBM [53]. As GBM rarely metastasizes and is described as a cranial-restricted tumor, CTCs might not be found in GBM patients in blood samples [53].

Extracellular vesicles, secreted by the tumor and containing material characteristic of the parental cells, can be found in the serum as well as the CSF. It is known, that GBM secrete exosomes, microvesicles, apoptotic bodies and oncosomes containing the glioma-specific receptor of epidermal growth factor (EGFRvIII), miR-21 as well as mutant IDH1 mRNA [53].

Circulating miRNAs have recently gained attention in research and present promising new biomarkers [55]. They can usually be found in peripheral blood of GBM patients and plasma levels of some miRNAs were already shown to be altered [56]. Some of these circulating miRNAs seem predictive in early diagnosis and helpful during treatment monitoring [55].

#### *2.4. Innovative Treatment Options for MGMT Unmethylated Patients*

Apart from TMZ, other compounds and therapeutic candidates have also been discovered and are currently tested for the treatment of unmethylated patients. Most of these compounds aim for radiosensitization [10,57,58] affecting the DNA repair pathway or other related pathways. However, in the following, two therapy alternatives will be presented, which target MGMT for radiosensitization.

#### 2.4.1. O<sup>6</sup> -Benzylguanine

O6 -benzylguanine is a guanine analog with antineoplastic activity and has been proposed to serve as a therapeutic agent to improve efficiency of alkylating agents [59]. Since benzyl groups get displaced faster compared to methyl groups, O<sup>6</sup> -benzylguanine would serve as an effective agent to inactivate MGMT [60]. O<sup>6</sup> -benzylguanine binds to the active site of MGMT, thereby transferring the benzyl moiety to the cysteine residue blocking the active site for methyl groups [47]. Dolan et al. have shown that O<sup>6</sup> -benzylguanine enhances the cytotoxicity of alkylating agents, which specifically produce O6 -methylguanine [61]. They observed a direct correlation in vitro between increased effectiveness of methylating agents upon O<sup>6</sup> -benzylguanine addition and depletion of MGMT [61]. Furthermore, Dolan et al. have shown in vivo that already low doses of O<sup>6</sup> -benzylguanine completely deplete MGMT activity [60]. However, to achieve long-lasting efficiencies, higher doses were required, which exhibited increased acute cytotoxicities, especially to the hematopoietic system. The assumption that due to the already low levels of MGMT in the bone marrow, the toxicity in the bone marrow would not significantly increase should later be proven wrong [61].

Quinn et al. reported in a phase I trial [62] and in a phase II trial [63], where TMZ plus O6 -benzylguanine was administered to patients with recurrent, TMZ-resistant glioblastoma, that myelosuppression was most commonly identified. Patients experienced grade 4 neutropenia, grade 4 thrombocytopenia, grade 4 lymphopenia and grade 3 and 4 anemia, which required a TMZ dose reduction in several patients. Although they observed MGMT depletion after O<sup>6</sup> -benzylguanine administration in blood samples [62], they did not observe a TMZ sensitization in MGMT unmethylated patients [63]. Therefore, O<sup>6</sup> -benzylguanine was not included in the standard therapy of GBM patients.

#### 2.4.2. PARP Inhibitors

The poly(ADP-ribose) polymerase (PARP) family consists of 18 PARP enzymes mainly involved in DNA damage repair and programmed cell death. PARP-1 and PARP-2 are activated upon DNA damages caused, for example, by ionizing radiation or alkylating agents to repair the DNA damage via the base-excision repair (BER) pathway [64]. Both, PARP-1 and PARP-2, were found to increase the antitumor effects of cytotoxic agents and offer treatment options for chemo- and radiosensitization.

PARP-1 binds to the damage on the DNA and generates poly ADP-ribose (PAR) polymers using NAD+. Further polymers are transferred to histones and chromatin-associated proteins on the DNA [65]. Once the repair enzymes are recruited, PARP-1 is released from the DNA break to give way for XRCC1. XRCC1 assembles the repair enzymes and factors onto the DNA to repair the break. While the DNA is repaired, PARP-1 gets reactivated by the glycohydrolase PARG removing the PARylations [65]. Therefore, PARP-1 enhances cell survival and mediates resistance to radioand chemotherapy.

PARP inhibitors either inhibit NAD<sup>+</sup> binding and following PARylation or trap PARP, thereby, blocking the damaged site for repair enzyme assembly [65]. Both lead to replicative stress and DNA double-strand breaks [66].

Several PARP inhibitors are currently tested in phase I, II and III clinical trials, including olaparib, iniparib, pamiparib, niraparib, veliparib, and talazoparib. Dungey et al. showed that olaparib increased radiosensitivity of GBM cells in vitro due to collapsed replication forks after radiation treatment [67]. They propose that the radiosensitizing effect occurs due to the replicating cells necessitating a fractionated treatment regimen [67]. Here, the PARP inhibitor does not directly have an effect on MGMT but rather on DNA replication, making it a good example for the radiosensitizing effects of PARP inhibitors.

Veliparib, in contrast, was found an alternative treatment option for MGMT unmethylated GBM patients as a combination of veliparib with irradiation inhibited cell proliferation in MGMT unmethylated primary cell lines as well as increased survival and apoptosis and decreased cell proliferation in vivo [68]. However, a randomized phase I/II study from 2016 combining TMZ and veliparib in recurrent GBM patients did not significantly increase overall survival and progression-free survival [69]. However, the results of a more recent published phase II trial (2019) comparing standard of care to veliparib concomitant to radio- as well as chemotherapy indicate an advantage of veliparib compared to standard of care treatment with an extended six months progression-free survival [70]

PARP inhibitors present a novel, innovative and personalized treatment option for MGMT unmethylated GBM patients; however, clinical trials are currently ongoing and analyses need to be completed before adding PARP inhibitors to the standard treatment of GBM.

#### **3. miRNA**

microRNAs (miRNA) are small non-coding RNA molecules consisting of 19-22 nucleotides first described in Caenorhabditis elegans in 1993 [71]. Lee et al. discovered that the *lin-4* gene produces short RNAs that are complementary to the 30UTR of *lin-14* mRNA and further observed a down-regulation of LIN-14 protein. This led them to the assumption that the direct RNA-RNA interaction between the *lin-4* transcript and the *lin-14* 3 0UTR leads to LIN-14 protein down-regulation [71]. Further, they proposed the existence of a class of regulatory genes producing small antisense RNAs influencing gene expression later to be known as microRNAs [71].

In 2001, the word microRNA was first introduced by Lagos-Quintana et al. [72] who could show that many miRNAs are expressed in several species and are highly conserved. The main role of

miRNAs is posttranscriptional regulation by sequence-specific repression of mRNAs [72]. To date, more than 2000 miRNAs have been discovered in the human genome [73], which each regulates hundreds of targets including genetic pathways, indicating their role in gene regulation, disease development and also tumorigenesis [74].

#### *3.1. miRNA Biogenesis*

miRNAs are initially produced in the nucleus from large hairpin looped RNA precursors by the RNA polymerase II [75,76]. These precursors are termed pri-miRNAs and are processed to pre-miRNAs of varying length [75] by the RNase III enzyme Drosha [77,78] and the double-stranded RNA-binding protein Pasha [79]. Via exportin 5 [80], the pre-miRNAs get exported into the cytoplasm [75], where the RNAse III enzyme Dicer processes it to 22 nucleotides long double-stranded RNAs that form the miRNA: miRNA\*duplex. The mature miRNA is unwound and released from Dicer [75] and Argonaut protein 2 (Ago2) [81] mediates the assembly to the multiprotein RNA-induced-silencing complex (miRISC) [81]. Which strand eventually enters the miRISC depends on the internal strand stability [82]. The end of the strand with the lowest stability is likely to be the target of a helicase-like enzyme, which unwinds the duplex [82]. Here, the 50 end exhibits the lowest internal stability. Perfect complementarity between the mature miRNA and the mRNA target leads to the cleavage of the target mRNA, whereas imperfect complementarity only leads to translational repression [83].

Dysregulation of miRNAs due to gene deletions, amplifications and translocations or defects in the miRNA biogenesis machinery seem to be the mechanisms contributing to the malignant cell types eventually leading to cancer.

#### *3.2. miRNA in Cancer*

In 2002, Calin et al. were the first to discover an association between miRNA dysregulation and cancer: a deletion on chromosome 13q14 coding for the *miR15* and *miR16* genes was observed in more than half of the B-cell chronic lymphocytic leukemia (CLL) and deletions or down-regulations of miR-15 and miR-16 were observed in 68% of the B-cell chronic lymphocytic leukemia [84].

Further, in 2004, they published that miRNAs are either tumor suppressive or oncogenic depending on their location; located at regions of loss of heterozygosity suggests tumor suppressors, while located at regions of amplifications suggests oncogenes [85]. In their genome-wide examination, they discovered an association between miRNA location and cancer. miRNAs are commonly found at cancer-associated regions, in which loss of heterozygosity regions may contain tumor suppressor genes and amplifications harbor oncogenes or the other way around [85]. An example for tumor suppressive miRNAs are miR-15 and miR-16, as their absence due to deletions on chromosome 13 leads to CLL. Further, it was shown by Cimmino et al. that the deletion of miR-15 and miR-16 leads to increased expression of Bcl-2 resulting in the formation of leukemias and lymphomas [86]. From this discovery, they proposed tumor suppressive miRNAs as inhibitors of their oncogenic targets in cancer therapy. Another mechanism for dysregulation of miRNAs in cancer apart from deletions and amplifications is the control of the transcription factors. The dysregulation of transcription factors regulating, for example, cell cycle progression, apoptosis, autophagy, invasion, and neoangiogenesis is tightly linked to cancer development. A key regulator of cell cycle progression and a commonly known tumor suppressor gene is p53. Mutation of p53 is frequently found in many cancers and its interaction with miRNAs suggests tumor suppressive features. Yamakuchi and Lowenstein discovered that miR-34a expression is induced by p53, which in turn suppresses p53, negatively regulating SIRT1 to induce apoptosis [87].

Therefore, miRNAs play an important role in tumor development, progression and recurrence. However, miRNAs also represent an innovative treatment option as prognostic and diagnostic biomarkers as well as therapeutic targets in cancer therapy [11,88,89].

The most common upregulated miRNA in many cancers is miR-21. miR-21 is an oncogenic miRNA inhibiting key regulator of apoptotic genes [90]. It was first found to be significantly upregulated in

human glioblastoma and its inhibition leads to increased caspase activation followed by apoptotic cell death [90]. Therefore, miR-21 is an example of an oncogenic miRNA, in which upregulation is associated with cancerogenesis. Various bioinformatics and experimental studies have tried to identify a set of de-regulated miRNAs in glioblastoma that are responsible for this tumor.

#### *3.3. miRNA in GBM*

The dysregulation of miRNAs in several cancers was shown to contribute to cancer development and progression. These miRNAs, their targets, prognostic and diagnostic value, as well as their potential in the treatment of GBM, need to be identified. Table 1 gives an overview of some miRNAs already discovered in GBM, their targets (if known) and their prognostic value (if available). This table gives a small insight into some of the most important miRNAs in GBM and, by far, does not include all up to date identified miRNA dysregulated in GBM.

**Table 1.** miRNAs in Glioblastoma multiforme, their targets, function, and prognostic value (↓ = decreased, ↑ = increased)



**Table 1.** *Cont.*

#### *3.4. miRNAs Targeting MGMT*

Since the discovery of the importance of the MGMT promoter methylation status in GBM therapy outcome [39], it is now known that the promoter methylation is not the only deterministic factor for MGMT protein expression. In 2013, Kreth et al. discovered the presence of two MGMT transcripts, which are both expressed in GBM [52]. In normal brain tissue, only the shorter transcript with a size of 440 bp is found, which contains a canonical poly(A) signal as well as a 30UTR of 105 nt. The longer transcript of about 850 bp contains an alternative poly(A) signal of 522 nt and is found only in GBM. In patient samples, they discovered that the length of the transcript is associated with high or low MGMT expression; high MGMT expression correlated with the normal 30UTR length, whereas reduced MGMT expression levels were associated with the MGMT transcript containing the elongated 30UTR [52]. Analysis of potential miRNA binding sites revealed 29 miRNAs specific for the long 30UTR and only two for the short 30 UTR; miR-181d was found in both. This led to the conclusion that longer UTRs render transcripts more accessibility to miRNA targets.

First, an in silico analysis of miRNAs targeting MGMT using the TarBase v.8 online tool (DIANA-LAB, Biomedical Sciences Research Center Alexander Fleming, Vari, Greece) [131] was done. The 43 found miRNAs are present in Table 2 below.


**Table 2.** miRNAs targeting O<sup>6</sup> -methylguanine–DNA methyltransferase (MGMT) derived from an in silico analysis.

Table 3 gives an overview of those miRNAs regulating MGMT expression, which were identified experimentally and which exhibit significant effects in cell lines. Detailed descriptions follow in the sections below.

**Table 3.** miRNAs involved in MGMT regulation in Glioblastoma multiforme.


#### 3.4.1. miR-142-3p

Lee et al. determined an inverse correlation between MGMT and miR-142-3p expression levels in GBM cell lines: high MGMT expressing cell lines show low levels of miR-142-3p and low MGMT expressing cell lines show high levels of miR-142-3p [110]. In miR-142-3p overexpression experiments, no change in MGMT mRNA expression was observed, but a reduction in MGMT protein expression. This leads to the assumption that miR-142-3p directly interacts with the 30UTR of MGMT, which was further proven by luciferase reporter assay experiments [110]. Additionally, an increased sensitivity towards alkylating agents was determined using TMZ and BCNU in miR-142-3p overexpressing cell lines with a stronger effect when BCNU was added [110].

Previously, the same workgroup reported that miR-142-3p is suppressed by the oncogenic cytokine IL6 promoting GBM propagation, suggesting that miR-142-3p is a tumor suppressive miRNA [111].

Taken together, miR-142-3p regulates MGMT protein expression and sensitizes cells in vitro to alkylating agents, which might indicate a potential biomarker for individual GBM treatment [111].

#### 3.4.2. miR-181d

Zhang et al. were the first to identify a miRNA regulating MGMT. miR-181d post-transcriptionally regulates MGMT by direct interaction with the long 30UTR MGMT transcript [116,135]. In vitro experiments could show that transfection with miR-181d significantly downregulated MGMT mRNA as well as MGMT protein expression and sensitized cells to TMZ [116]. Further analysis of glioblastoma patient samples indicated that miR-181d is usually down-regulated and that transfection with a mimic in vitro inhibits cell proliferation by targeting K-ras, promotes G1 cell cycle arrest and induces apoptosis by targeting Bcl-2 [115]. Evaluation of clinical data also revealed that a higher miR-181d expression was associated with improved overall survival [116] and that miR-181d expression levels increased after either TMZ or irradiation alone and significantly increased after irradiation and TMZ treatment combined [136]. This suggests that miR-181d could act as a predictive biomarker for chemo- and radiotherapy outcomes.

Several studies have investigated the effect of miR-181d and MGMT expression and discovered similar results to Zhang et al. Interaction between miR-181d and other miRNAs, such as miR-409-3p [133], miR-648 and miR-661 [135], have been found to enhance the effect of MGMT down-regulation, suggesting that miR-181d is the key miRNA regulating MGMT expression.

Taken these factors together miR-181d as a tumor suppressive miRNA could be of great use in treating glioblastoma patients to increase sensitivity to TMZ by directly targeting MGMT mRNA [135]. To the best of our knowledge, miR-181d is the only miRNA that regulates MGMT and is associated with overall survival. Up to date, there are no clinical trials ongoing investigating miR-181d as an innovative treatment option.

#### 3.4.3. miR-221/222

miR-221/222 have been extensively studied in various cancers and were shown to be overexpressed in glioblastoma, prostate carcinoma, papillary thyroid carcinoma, hepatocellular cancer and pancreatic cancer [122]. Gillies and Lorimer demonstrated that miR-221/222 are upregulated in human glioblastoma and target p27, a cell cycle regulator [121]. Further targets include the Akt signaling pathway, PTEN, TIMP-3, as well as MMP-2 and MMP-9 [122,137]. In vitro overexpression of miR-221/222 resulted in the induction of p-Akt, MMP-2, and MMP-9 protein expression and hence increased cell proliferation and invasion. These results were confirmed in in vivo overexpression experiments, which also led to increased tumor growth as well as morphological changes towards a malignant phenotype [122].

A binding site of miR-221/222 was found at the 30UTR of MGMT and further confirmed in in vitro experiments [123]. Overexpression of miR-221/222 reduced MGMT levels in transfected human glioblastoma cell lines and increased the cells' sensitivity to TMZ [123]. It can be concluded that miR-221/222 are oncogenic miRNAs negatively influencing patients' survival, however, increasing sensitivity to TMZ in vitro by directly targeting MGMT.

#### 3.4.4. miR-370-3p

Peng et al. were the first to discover a suppressive potential of miR-370-3p in human glioblastoma [114]. miR-370-3p is significantly down-regulated in low- and high-grade gliomas (Grade II and IV) and also in glioblastoma cell lines. Upon transfection with a miR-370-3p mimic cell viability decreased, long-term proliferation was suppressed as well as the percentage of cells arrested in S and G2/M phase of the cell cycle decreased [114]. A direct post-transcriptional target was found in the 30UTR of ß-catenin, which is involved in the Wnt signaling pathway promoting cell proliferation and migration [114].

Gao et al. found similar results: in recurrent GBM miR-370-3p expression was significantly decreased compared to normal brain tissue, GBM cell lines showed low levels of miR-370-3p, as well as miR-370-3p-transfected cells showed decreased proliferation [132]. Cell lines expressing the lowest miR-370-3p were more resistant to TMZ compared to cell lines expressing higher levels of miR-370-3p [132]. Additionally, they could demonstrate a negative correlation between MGMT mRNA and miR-370-3p expression.

In 2018, Nadaradjane et al. postulated that miR-370-3p is a biomarker for the prediction of GBM treatment planning and therapy outcome. However, they found out that the expression level of miR-370-3p in the blood of GBM patients varies during standard treatment and is not associated with overall survival [127]. Still, they observed a longer patient survival when miR-370-3p overexpression lasted longer before relapse occurred. In vitro they could show that miR-370-3p overexpression leads to decreased MGMT mRNA and decreased MGMT protein levels. Further, miR-370-3p increased the cells' sensitivity to TMZ indicated by increased cell death after treatment. Subcutaneous tumors grown in mice and treated with a combination of TMZ and miR-370-3p significantly decreased in volume. In the resected tumors, a significant reduction of MGMT expression was observed [127].

Another target of miR-370-3p is FOXM1, which is involved in cell cycle progression. Upon miR-370-3p overexpression, FOXM1 expression reduced as well [127]. Hence, cell cycle progression was inhibited and cell death induced.

It can be concluded that miR-370-3p is a tumor suppressive miRNA in GBM by downregulating the mRNA and protein expression of MGMT as well as FOXM1 expression. miR-370-3p is not deterministic for patients' survival but can be used to sensitize to TMZ especially in MGMT unmethylated patients. However, no clinical trials are currently ongoing investigating miR-370-3p as an innovative treatment option.

#### 3.4.5. miR-409-3p

miR-409-3p was found 5-fold upregulated in human GBM samples compared to healthy brain tissue with an inverse correlation between MGMT and miR-409-3p expression [133]. Patient samples with low MGMT show high miR-409-3p levels, while high MGMT expressing samples show low miR-409-3p levels. In vitro transfection with a miR-409-3p mimic of the high MGMT expressing cell line T98G demonstrated a significant down-regulation of MGMT mRNA as well as MGMT protein. This suggests that miR-409-3p is a strong inhibitor of MGMT by the degradation of MGMT mRNA as well as by translational repression [133]. An even more enhanced effect of MGMT suppression was observed when miR-409-3p mimics were cotransfected with miR-181d mimics [133].

As miR-409-3p was found significantly downregulated in human GBM samples repressing MGMT expression, it can be concluded that miR-409-3p might be a potential therapeutic approach to sensitize MGMT unmethylated patients to alkylating chemotherapeutics. However, other targets of miR-409-3p are still unknown, but Khalil et al. suggested a possible protective role in pro-angiogenic and pro-metastatic processes [133].

#### 3.4.6. miR-603

miR-603 is found upregulated in glioblastoma samples and promotes cell proliferation as well as cell cycle progression [130]. Targets of miR-603 include WIF1 and CTNNBIP1 activating the Wnt/ß-catenin signaling pathway and promoting cell proliferation and migration [130]. Therefore, miR-603 can be considered an oncogenic miRNA.

MGMT is directly suppressed by the interaction of miR-603 with the 30UTR of MGMT [134]. Transfection with a miR-603 mimic significantly reduced MGMT mRNA levels and protein expression and further enhanced sensitivity to TMZ in vitro as well as in vivo [134]. Kushwaha et al. also showed that the combination of miR-181d and miR-603 most effectively regulated MGMT expression compared to either alone [134].

For innovative treatment options, miR-603 might be a promising candidate to inhibit MGMT and Wnt/ß-catenin signaling pathway activation. No clinical trials have been proposed yet.

#### 3.4.7. miR-648, miR-661 and miR-767-3p

When Kreth et al. discovered the presence of two MGMT isoforms either containing a long or a short UTR, they used target prediction software to determine miRNA with a binding site within the UTRs of MGMT [52]. They assumed that these miRNAs are expressed in human GBM and negatively correlate with MGMT expression. In human GBM samples, six miRNAs (miR-184, miR-183, miR-661, miR-370, miR-767-3p, and miR-648) were found binding exclusively in the long UTR, two (miR-1197 and miR-655) within the short UTR and one (miR-181-d) in both UTRs [135]. Upon cloning both UTRs into a reporter vector containing two luciferases and co-transfection with the miRNAs, they observed that the short UTR-binding miRNAs (miR-181d, miR-665, and miR-1197) did not show regulatory activity [135]. This indicates that the short UTR of MGMT is not regulated by these miRNAs. Five (miR-661, miR-370, miR-181d, miR-767-39, and miR-648) of the seven miRNAs possibly regulating the long UTR showed significant luciferase repression but only three miRNAs (miR-181d, miR-767-3p and miR-648) showed decreased MGMT protein expression. Here, miR-648 exerted the strongest MGMT protein reduction. In qPCR experiments, only two miRNAs (miR-181d and miR-767-3p) significantly reduced MGMT mRNA expression, indicating that those two regulate MGMT expression via direct degradation of the mRNA transcript and miR-648 might act via translational repression [135].

Further, they could show that miR-767-3p and miR-648 are significantly upregulated in human GBM samples and that cotransfection with all three miRNAs (miR-181d, miR-767-3p and miR-648) significantly increased the sensitivity to TMZ treatment [135]. These data correlate with data from Jesionek-Kupnicka et al., who also found an association between MGMT and miR-181d and miR-648 expression [138].

#### *3.5. miRNAs as Innovative Treatment Option for GBM*

To the best of our knowledge, no clinical trials are currently ongoing investigating the above-mentioned miRNAs as innovative treatment options for GBM patients nor have any miRNA-based therapies been approved by the FDA. Target specificity and tissue toxicities are major problems in the delivery of miRNA or miRNA inhibitors to their mRNA target.

Several invasive strategies have been postulated to enhance drug delivery across the BBB including intracerebral implants, disruption of the BBB, intra-arterial and intrathecal drug delivery, direct injections into the brain, catheters, pumps or microdialysis [139]. As all of these strategies require invasion into the brain tissue or tumor tissue, there is an increased risk for brain damage and other side effects, including toxicities, indicating the urgent need for non-invasive strategies. Therefore, biological strategies have been developed as innovative tools for drug delivery. These strategies include RNA interference, viral vectors, exosomes, antisense therapy, gene therapy, antibody conjugates, peptide carriers and other carriers [139,140]. Also, chemical systems have been developed, such as lipophilic analogues, prodrugs, efflux transporter inhibition, liposomes, nanoparticles, polymeric micelles and

dendrimers [139,140]. Both biological and chemical strategies allow for target specific delivery, as it is most important and challenging at the same time to deliver and internalize the drug or miRNA specifically to the tumor. The challenges of designing nanoparticles are reviewed elsewhere [141].

However, the most limiting factor in delivering these compounds into the brain and promoting restricted bioavailability is the BBB. Major issues are the enzymatic degradation of the miRNA or miRNA inhibitors themselves before the target can be reached as well as the inability of packaging molecules due to high molecular weight and polar functional groups [139]. The BBB is a natural barrier against toxins, harmful substances, and fluctuations in chemical concentrations [139]. It consists of endothelial cells, forming the walls of the capillaries and epithelial cells, creating the blood-cerebrospinal fluid barrier (BCSFB) [142]. The cerebrospinal fluid is secreted into the brain, while the interstitial fluid is secreted by the capillary endothelium [142]. These two fluids can communicate in order to regulate fluctuations and maintain a stable environment [142]. The avascular arachnoid epithelium is the enclosing layer sealing the extracellular fluids from the rest of the body [142]. Physical barriers such as tight junctions, transport barriers such as transporters, and metabolic barriers including enzymes, are found at all interfaces representing the protecting characteristics of the BBB. The most important factor thereby are tight junctions, significantly reducing the trespassing of polar solutes by blocking their penetration [142]. The only routes molecules and solutes can penetrate the BBB are via passive diffusion and ABC transporter efflux (lipid, soluble, non-polar molecules), via solute carriers (e.g., glucose, amino acids, small peptides), via transcytosis or receptor- and adsorptive-mediated (e.g., lipoproteins, insulin, glycosylated proteins, histones) or leukocytes via diapedesis [142].

However, all these strategies need further characterization, experimentation, and clinical trials to safely deliver molecules, miRNAs and other compounds to specific target sites. Up to date, only some miRNAs, including miR-122, -21, -155, -92 and -29 are currently tested in clinical trials as targeted therapy for Hepatitis C (HCV), nephritis, CLL, wounds and fibrosis [143,144]. Only two miRNAs are currently tested for the use in cancer therapy: a miR-16 mimic is involved in a Phase I trial for non-small-cell lung cancer [144] and another clinical trial testing a miR-34a mimic for hepatocellular carcinoma (HCC) has recently been terminated [145]. MRX34, a synthetic, 23 nt long double-stranded RNA encapsulated in a liposomal nanoparticle was administered to patients mainly suffering from HCC. Although pre-clinical studies in non-human primates showed promising results, severe adverse effects and also death of four patients due to the drug forced the phase I trial to be terminated [145]. Severe adverse effects were unlikely due to the liposomal carrier, but rather due to severe immune-related toxicities, which have yet to be resolved [145].

#### **4. Conclusions**

In the last decade, miRNAs have become promising tools as prognostic and diagnostic biomarkers as well as therapeutic targets for innovative and personalized cancer treatment [11,89,90]. Several miRNAs have been found differentially expressed and predictive for overall survival, progression-free survival or treatment outcome in several cancer entities. Some miRNAs such as miR-21, the miR-17 cluster and miR-221/222 are dysregulated in several cancer types, but most importantly, cancer type-specific miRNA signatures were also discovered [89,146–149].

With a survival rate of less than 3% [3], Glioblastoma multiforme presents an urgent need for new innovative and personalized treatment options. Patients with a wildtype IDH and an unmethylated MGMT promoter region have the poorest prognosis and the shortest survival [150], identifying these patients with the most urgent need for new treatment options. In this review, we focused on MGMT unmethylated patients and tried to identify possible miRNAs regulating MGMT expression, which could be used for personalized treatment in the future.

We identified eight promising miRNAs—miR-142-3p, -181d, -221/222, -370-3p, -409-3p, -603, -648, and -767-3p—negatively regulating MGMT expression either via mRNA degradation or translational repression. Five of these miRNAs (miR-142-3p, -181d, -221/222, -370-3p and -603) were positively tested

to increase sensitivity to alkylating agents such as BCNU and TMZ in vitro as well as in vivo [110]. miR-181d was the only miRNA found predictive for overall survival [116,136].

We present here miRNAs that could help reduce and repress MGMT expression by targeted treatment to sensitize the tumors against alkylating agents. However, target-specific delivery, especially into the brain, represents a challenging task, which has yet to be overcome.

**Funding:** This research received no external funding.

**Acknowledgments:** This project was funded in part by the German Consortium for Translational Cancer Research, Munich/TUM site, as well as by the Medical Faculty of TUM.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Review* **Key MicroRNA's and Their Targetome in Adrenocortical Cancer**

**Marthe Chehade 1,2, Martyn Bullock 1,2 , Anthony Glover 1,2,3 , Gyorgy Hutvagner 4,\* and Stan Sidhu 1,2,3,\***


Received: 14 June 2020; Accepted: 28 July 2020; Published: 6 August 2020

**Abstract:** Adrenocortical Carcinoma (ACC) is a rare but aggressive malignancy with poor prognosis and limited response to available systemic therapies. Although complete surgical resection gives the best chance for long-term survival, ACC has a two-year recurrence rate of 50%, which poses a therapeutic challenge. High throughput analyses focused on characterizing the molecular signature of ACC have revealed specific micro-RNAs (miRNAs) that are associated with aggressive tumor phenotypes. MiRNAs are small non-coding RNA molecules that regulate gene expression by inhibiting mRNA translation or degrading mRNA transcripts and have been generally implicated in carcinogenesis. This review summarizes the current insights into dysregulated miRNAs in ACC tumorigenesis, their known functions, and specific targetomes. In addition, we explore the possibility of particular miRNAs to be exploited as clinical biomarkers in ACC and as potential therapeutics.

**Keywords:** adrenocortical carcinoma; micro RNA; non-coding RNA

#### **1. Introduction**

Adrenal tumors are very common, affecting up to 10% of the general population, of which the large majority are benign non-functional adenomas [1]. Adrenocortical cancer (ACC), in contrast, is a rare endocrine malignancy with an incidence of 0.76 per million in the general population [2]. Approximately 60% of patients with ACC present with signs and symptoms of hormone excess [3], and approximately 20% present with mass associated symptoms, such as abdominal pain, early satiety, or abdominal fullness [4]. The remaining patients are incidentally diagnosed on abdominal imaging for other medical indications. As the clinical manifestations of hormone excess may be subtle and mass effect symptoms are vague, ACC is often diagnosed late. The median size of the primary tumor is 12 cm at diagnosis [5], and the rate of unresectable metastatic disease at diagnosis ranges between 30% [6] and 70% [7]. Complete surgical resection with oncologically clear margins affords the best chance of cure in ACC but even despite this, the rate of disease recurrence is high, and the prognosis is generally poor with five-year survival of less than 40% [8].

Management options for metastatic ACC are limited as cytotoxic chemotherapy affords only a marginal survival benefit, and mitotane, an adrenolytic agent, which is the only other approved systemic therapy for metastatic ACC, is poorly tolerated. The first and only randomized controlled

chemotherapy-based phase III clinical trial for advanced ACC (First International Randomized Trial in Locally Advanced and Metastatic Adrenocortical Carcinoma Treatment (FIRM-ACT)) was completed in 2010. This study compared mitotane administered in combination with either streptozocin or etoposide, doxorubicin, and cisplatin (EDPM) and demonstrated a modest improvement in progression-free survival in the EDPM arm, but no benefit in the overall survival [9]. ACC is, therefore, an orphan disease that presents challenges on both diagnostic and management fronts. ACC research is currently focused on developing methods for early detection and effective management of a metastatic disease. In particular, the discovery of novel approaches to the management of metastatic ACC is crucial to improving patient outlook.

MicroRNA (miRNA) are small non-protein-coding RNA molecules whose deregulation has been implicated in the pathogenesis of many human diseases, particularly cancer. Over the past two decades, miRNA research in cancer has focused on determining the miRNA expression signatures of different tumors in order to identify potential biomarkers for early diagnosis, as well as functional studies of specific miRNAs to determine their targets and function. The set of mRNAs targeted by a defined miRNA is known as its targetome. While multiple studies have profiled the miRNA signature of childhood and adult ACCs using various techniques and shown consistent deregulation in a set of candidate miRNAs, relatively fewer have demonstrated miRNA-target interactions. MiRNAs have also been identified to have both diagnostic and therapeutic potential in the cancer literature, broadening our understanding of their roles in tumor biology. In this review, we present a current summary of the mounting body of work describing miRNA dysfunction in ACC with the aims of highlighting their potential function and roles in modulating key oncogenic pathways.

#### **2. ACC Genetic Landscape and Associated Genetic Disorders**

ACC has a bimodal distribution with a worldwide childhood incidence of 0.2 per million [10], and an adult peak in the fifth decade of life. Childhood ACC differs from adult ACC, as 50–80% of childhood cases are associated with germline *TP53* mutations [11,12]. In contrast, most cases of adult ACC are sporadic, with germline *TP53* mutations being present in around three percent of patients [13].

In rare cases, ACC can be associated with specific germline mutations that cause hereditary cancer syndromes. ACC is a core malignancy in Li Fraumeni Syndrome (LFS) caused by the germline *TP53* mutation and affects ten percent of cases [14]. Notably, in a study by Soon and colleagues, sporadic ACC was associated with loss of heterozygosity (LOH) at the *TP53* gene locus 17p13.1 in 74% of cases compared with only 14% of adrenal adenomas [15]. ACC affects approximately seven percent of children with Beckwith–Weidemann Syndrome (BWS) [16], which is caused by mutations or epigenetic modifications at the genetic locus 11p15 containing the *Insulin-Like Growth Factor 2* (*IGF2*) gene. 11p15 LOH or *IGF2* overexpression were demonstrated in 93.1% of sporadic ACCs compared with only 8.6% of benign adrenal tumors [17] in a study by Gicquel and colleagues, highlighting the importance of this imprinted locus in the pathogenesis of ACC. Approximately three percent of patients with Familial Adenomatous Polyposis (FAP), caused by mutations in the Adenomatous Polyposis Coli (APC) gene, develop adrenocortical cancer as adults [18]. The APC protein is a negative regulator of β-Catenin, whose accumulation in the nuclei of primary ACCs has been associated with advanced tumor stage and poor prognosis [19]. ACC is also rarely associated with Lynch Syndrome, Neurofibromatosis Type 1, and Carney Complex, as well as Multiple Endocrine Neoplasia Type 1 (MEN1) in adults [20] (Table 1). Although mutations in *TP53*, *IGF-2*, and β*-catenin* genes have been established as drivers of sporadic ACC, the low penetrance of ACC in these genetic cancer syndromes indicates that mutations or the epigenetic regulation of the expression of nearby genes may play an important role in its etiology.


\* Epigenetic modifications to methylation of imprinting control regions or paternal uniparental disomy are more common than gene mutations. IGF2, Insulin-like growth factor 2; CDKN1C, Cyclin-dependent kinase inhibitor 1C; KCNQ1OT, Potassium voltage-gated channel subfamily Q member 1 antisense gene; APC, Adenomatous polyposis coli; MEN1, Multiple endocrine neoplasia Type 1; MLH1, MutL homolog 1; MutS homolog 2; MSH6, MutS homolog 6; PMS2, PMS1 homolog 2; NF1, Neurofibromatosis 1; CNC1, Carney complex type 1; PRKAR1A, Protein kinase A regulatory subunit 1-alpha; PI3K, Phosphatidylinositol-3-kinase; TGF-β, Transforming growth factor-beta; Wnt, Wingless-related integration site; MAPK, Mitogen-activated protein kinase; ERK, Extracellular signal-related kinase; cAMP, cyclic adenosine monophosphate.

#### **3. Key Genetic Drivers of ACC and Their Cellular Pathways**

It is now known that mutations in gene drivers alone do not completely explain the pathogenesis of ACC, and therefore non-coding gene mutations that lead to aberrant regulation of driver genes through their pathways can also contribute to tumor biogenesis. The following summary of key genetic drivers in ACC, therefore, serves to explore the extent to which ACC pathogenesis could be explained by them, and contextualize the importance of known miRNA targets within these pathways.

#### *3.1. Tumor Suppressor Protein 53 (TP53)*

The *TP53* gene encodes a homo-tetrameric transcription factor that mediates the cellular response to genotoxic stress and the activation of oncogenes by transcriptionally targeting many genes to ultimately activate cellular pathways involved in cell-cycle arrest and DNA damage repair. Where the cell fails to repair this damage, p53 induces cellular apoptosis via a p53-upregulated modulator of apoptosis (PUMA) to avoid propagating genetic mistakes. P53 is regulated by Human Double Minute 2 homolog (HDM2), an E3 ubiquitin-protein ligase, which targets p53 for cytosolic translocation, or proteosomal degradation when it is polyubiquitinated. HDM2 is, in turn, regulated by p53, forming a negative feedback loop [26].

*TP53* mutations in cancer are common and are present in more than half of human tumors [27]. ACC, despite its rarity, accounts for 11.9% of all human tumors harboring germline *TP53* mutations, after breast, soft-tissue, and brain tumors [28]. ACC's harboring somatic TP53 mutations are on average larger, more advanced in stage, and associated with shorter disease-free survival [29]. The majority of *TP53* mutations associated with ACC are loss-of-function mutations; however, many of these are predicted to result only in partial loss of p53 function [30]. Curiously, transgenic *TP53* knockout and mutant mouse models do not develop ACC despite developing multiple tumors [31,32]. Else and colleagues showed that transgenic mice carrying an inactivation mutation in the *tripeptidyl peptidase*

*1*/*ACD sheltering complex subunit and telomerase recruitment factor* (*Tpp*/*Acd*) in addition to a single wild type *TP53* allele do develop ACCs at low frequencies [33].

#### *3.2. Insulin-Like Growth Factor 2 (IGF2)*

IGF2 is a paternally imprinted critical growth factor in the development of many organ systems, including the adrenal cortex, where it is highly expressed in early fetal development [34]. Multiple studies have confirmed *IGF2* overexpression in between 83.3% and 90.9% of ACC's when compared with ACA and NAC [35–38]. The maternally imprinted long non-coding RNA *H19* gene located on the antisense strand of the *IGF2* gene is shown to be underexpressed in ACC compared with ACA and NAC in multiple studies [39–42]. Both LOH [17] and paternal uniparental disomy at the 11p15 locus result in *IGF2* overexpression and reduced expression of *H19* and *Cyclin-dependent kinase inhibitor 1C* (*CDKN1C*) in ACC [36], which are associated with poor prognosis and increased rates of recurrence [43].

IGF2 binds to the membrane tyrosine kinase receptor IGF Receptor Type 1 (IGF-1R), leading to receptor autophosphorylation and binding of the insulin receptor substrate 1 (IRS-1). Tyrosine phosphorylation of IRS-1 activates the phosphatidylinositol-3-kinase(PI3K)/serine/threonine protein kinase B (Akt) and mammalian target of rapamycin(mTOR) pathway as well as the Ras/Raf/mitogen-activated protein kinase (MEK)/extracellular signal-related kinase (ERK) pathways, potentiating cellular proliferation and viability in ACC cell models [44].

Transgenic mouse models of *IGF2* overexpression [45] and adrenal cortex-specific loss of imprinting at the *IGF2*/*H19* region [46] have demonstrated that these factors alone are not sufficient to initiate tumorigenesis. From a therapeutic approach, clinical trials of the IGF-1R small molecule inhibitors, Linsitinib and Figitumumab to treat advanced ACC failed to show benefit in progression-free survival or overall survival [47,48]. Another trial involving 26 ACC patients treated with the IGF-1R antibody Cixutumumab in combination with the mTOR inhibitor Temsirolimus achieved stable disease for at least six months in 42% of patients but did not lead to any partial or complete responses to therapy [49].

#### *3.3. Wnt*/β*-Catenin Signalling Pathway*

The Wnt signaling pathways are activated by Wnt-protein ligand binding extracellularly to a membrane Frizzled receptor. Canonical Wnt pathway activation leads to the accumulation of β-catenin in the cytoplasm, which ultimately translocates to the nucleus where it activates transcription. In the absence of Wnt, β-catenin is degraded by a protein complex formed by Axin, APC, protein phosphatase 2A (PP2A), glycogen synthase kinase 3 (GSK3) and casein kinase 1α (CK1α), by targeting it for ubiquitination and ultimate proteosomal degradation [50].

Assie and colleagues performed exome sequencing and single nucleotide polymorphism (SNP) analysis of 77 ACC tissues and showed alterations in the β-catenin pathway associated genes *zinc and ring finger 3* (*ZNRF3*) (21%), *cadherin-associated protein* β*1* (*CTNNB1*) (16%) and *APC* (2%) [51]. More strikingly, in a series of 50 ACC tissues, Maharjan and colleagues reported that the Wnt/β-catenin pathway was aberrantly activated in 62% [52].

Transgenic mouse models of constitutive β*-catenin* activation in the adrenal cortex produced aggressive adrenal tumors only in a subset of 17-month-old mice [53], while *APC* knockout mice displayed hyperplasia progressing to adrenal adenoma but not carcinoma [46].

Together these findings indicate that multiple genetic aberrations are required for the development of ACC, and combinatorial therapeutic strategies targeting multiple pathways may be effective.

#### **4. Overview of microRNA Structure, Biogenesis, and Function**

MiRNAs are short, single-stranded non-coding RNA molecules spanning between 17–25 nucleotides. Approximately 2,300 miRNAs have been identified in the human genome [54]. Their expression is tissue-specific, and they broadly act to negatively regulate the gene expression of at least 60% of human RNA transcripts through either translational inhibition or transcript decay [55]. While the majority of miRNA targets are mRNAs, other classes of RNA, including rRNA, tRNA, lncRNA, and other miRNAs make up 30% of all miRNA targets [56].

The RNAse enzymes DICER and DROSHA involved in miRNA maturation have been implicated in ACC. In a study which compared the expression of key miRNA processing factors between 29 ACC and 43 adrenocortical adenoma (ACA) tissues, Caramuta and colleagues showed that DICER, DROSHA, and TAR RNA-binding protein 2 (TARBP2) (a DICER cofactor required for miRNA processing [57]) were overexpressed at the mRNA and protein levels in ACC compared with ACA [58]. In addition, in vitro inhibition of *TARBP2* expression in the ACC cell line H295R resulted in decreased cellular proliferation and increased apoptosis [58]. This evidence suggests that dysregulation of the miRNA biogenesis pathway may potentiate tumorigenesis in ACC.

In the cytoplasm, the mature miRNA duplex unwinds from the thermodynamically less stable end, and the RNA strand that orients its 50 end in this direction known as the guide strand is loaded onto the RNA-induced silencing complex (RISC) [59]. RISC is a multiprotein complex containing one member of the Argonaute protein family [60]. miRNAs contain a seed region spanning 2–8 nucleotides at their 5' ends, which allows them to guide RISC mainly to the 3´ untranslated region (UTR) of their target mRNA through complementary base pairing. The degree of complementarity between miRNA and mRNA and the enzymatic properties of the Ago-2 protein determine whether mRNA silencing will be achieved through target cleavage or translational inhibition [61].

In cancer, the dysregulation of miRNA expression results from various mechanisms, including amplification or deletion of miRNA genes, dysregulation of transcriptional machinery, changes in methylation and histone modifications, as well as mutations and changes in expression of miRNA biogenesis-related proteins [62,63]. As miRNAs play an important role in the regulation of gene expression, their aberrant expression can lead to significant alterations in cellular phenotype. In their dysregulated state miRNAs can act as either oncogenes or tumor suppressors, affecting the cellular processes required for tumor initiation and progression.

#### *IsomiRs and Their Emerging Significance in Cancer*

Inaccurate cleavage by either DICER or DROSHA, nucleotide additions at the 3´ end, and nucleotide modifications could result in the production of miRNA isoforms (isomiRs) [64]. Increasing data shows that isomiRs have significant impacts on miRNA-mediated gene regulation [65]. Variations in the 5´ seed sequence could impact the specificity of miRNAs to their targets [66], and variations at the 3´ ends determine the stability of miRNA-mRNA binding [67]. The impact of a particular isomiR on miRNA function depends on the relative abundance and stability of the isomiR relative to its canonical miRNA, as well as its binding efficiency to its target. Chan and colleagues demonstrated this by showing that isomiRs of *miR-31* regulated the expression of known targets to varying degrees *in vitro*, and their binding capacity to the RISC complex determined using Ago-2 immunoprecipitation (IP) was also varied [68].

Recent advances in high-throughput RNA sequencing technologies have allowed tissue transcriptome profiling at the isomiR level. Telonis and colleagues [69] used The Cancer Gene Atlas (TCGA) miRNA sequencing data across 32 cancer types, including ACC, to determine whether the presence or absence of isomiRs could discriminate between the cancer types. By binarizing the isomiR expression data, they were able to successfully classify tumor datasets with an average sensitivity of 90% and false discovery rate (FDR) of 3%, which was superior to wild type miRNA expression (average sensitivity 83%, FDR 5%) [69]. More recently, Wang and colleagues used the same data to demonstrate that isomiRs that share their seed region (50 isoforms) could similarly discriminate between tumors [70]. In 2018, Lan and colleagues published the first study to use breast cancer TCGA small RNA sequencing expression data to show that the isomiR expression-based classification was superior to gene expression profiling at distinguishing between breast cancer subtypes [71]. Together these findings suggest that miRNAs may carry out significant functional roles in tumorigenesis at the isomiR level.

A study that investigated isomiR expression in adrenal tissues using RNA sequencing technology published in 2017 found that 411 miRNAs existed as 1763 various isoforms in a cohort of 14 ACC, 18 ACA, and 18 NAC samples [72]. These isomiRs contained 520 various seed sequences, of which 38% were non-canonical. Over a quarter of all expressed miRNAs in the ACC, ACA, and NAC groups produced isomiRs with two or more seed regions that were predicted to target a different set of mRNAs, but these were not investigated further. Therefore, the diagnostic and clinical significance of differentially expressed isomiRs in ACC remains unknown, and further research is needed to elucidate this.

While technical challenges related to the identification and quantification of isomiRs exist, recent developments in data sharing and technology are helping to overcome these. Often isomiR sequencing data is drawn from small sample numbers making it difficult to make general conclusions, but this is being overcome by the release of small RNA sequencing data from large sample databases such as TCGA. This has allowed researchers to study isomiR expression in more detail than previously possible. Also, recently developed specialized techniques such as photoactivatable ribonucleoside-enhanced crosslinking and Ago immunoprecipitation (Ago PAR-CLIP) ensures that identified isomiRs are biologically active and not degradation products, improving the reliability of the experimental data in this field. The utilization of such advances will facilitate research that will unlock a deeper understanding of the role of isomiRs in ACC.

#### **5. The Unique microRNA Expression Signature of ACC and Its Clinical Significance**

#### *5.1. The microRNA Expression Signature of ACC Tissues*

Sporadic ACC is a genetically heterogeneous malignancy that can be classified into distinct groups based on transcriptomics and clinical behavior. Several published studies have profiled differential miRNA expression in ACC tissue samples compared with either ACA and/or normal adrenal cortex (NAC) tissue using microarray data [73–77], TaqMan Low-Density Arrays (TLDA) [78,79], RT-q-PCR [80], or RNA sequencing [51,72,81] (Table 2). Across these studies, *miR-483-5p*, *miR-503-5p*, *miR-210*, and *miR-483-3p* were overexpressed in ACC compared with ACA or NAC in multiple datasets, and *miR-195*, *miR-497*, and *miR-335* were underexpressed. Of note, *miR-483-5p* was overexpressed in eight of eleven studies, and has been associated with poor prognosis in ACC [73]. Earlier studies using microarray and RT-q-PCR techniques could only investigate known miRNAs, whereas later studies which utilized RNA sequencing could identify differentially expressed miRNAs which had not previously been characterized. Hence, more numerous microRNA candidates have been identified with RNA sequencing, and novel candidates like *miR-508-3p* were only identified and validated in these later studies [51,81].

*Cancers* **2020**, *12*, 2198


**Table 2.** Studies investigating differential expression of microRNAs in ACC Tissues compared with ACA and NAC.

#### *5.2. Circulating microRNAs as Diagnostic Biomarkers in ACC*

Diagnosing patients with ACC continues to present challenges as no preoperative blood-borne tumor marker for the disease exists, and suspicion is often raised on imaging. MiRNAs are stable in bodily fluids within extracellular vesicles shed from tumor cells or in protein complexes and are relatively protected from enzymatic degradation in the circulation [82], making them attractive as potential noninvasive diagnostic markers for cancer. Despite a significant body of research confirming differentially expressed circulating miRNAs in various diseases, none have translated into clinical use.

A number of studies have reported on the expression and diagnostic utility of circulating miRNAs in ACC, whose findings have been summarized in a recent review by Decmann and colleagues [83]. Several of these studies have attempted to define candidate diagnostic circulating miRNA biomarkers for ACC in serum [76,84,85], but both relatively low sensitivity and specificity values have limited their clinical application. Although circulating *miR-483-5p* relative expression is a reliable differentiator between aggressive and non-aggressive ACC in serum, as demonstrated by Chabre and colleagues (AUC 0.929) [76], it is unable to differentiate between ACC and adrenocortical adenoma (AUC 0.74) [85]. Circulating *miR-483-5p* is also overexpressed in hepatocellular [86] as well as head and neck cancers [87], and has been proposed as a diagnostic marker in these diseases as in ACC. The lack of specificity of this miRNA as a biomarker further limits its clinical utility in ACC, and is characteristic of oncological miRNAs across different tumor types [88].

At present, various methods are in use for the quantification of circulating miRNAs in preclinical research. The lack of standardization in sample collection, storage, and processing introduces significant variation in the data, which limits the generalizability of differential expression results [89]. In addition, common reference RNA genes used as calibrators in comparing miRNA expressions are differentially expressed in the serum of patients with various diseases, leading to challenges in data normalization for analysis [90]. To overcome this, the addition of synthetic RNA during RNA extraction as a spike-in control is a method widely used for technical normalization. Such an exogenous control is helpful as it undergoes the same processing as endogenous RNA in the sample, but this does not correct for variables such as the serum miRNA fraction [91]. Standardization of sample collection techniques, developments in vesicle-associated miRNA quantification, and the use of absolute quantification methods that do not rely on housekeeper genes could help overcome the obstacles to the clinical adoption of circulating miRNAs as diagnostic biomarkers in the future.

#### *5.3. Tissue microRNA Expression as a Prognostic Tool in ACC*

In patients with ACC, the prescribed adjuvant clinical management and follow-up regimen is informed by the estimated risk of disease recurrence. Even in patients with unresectable disease, systemic therapy depends on tumor biology. While clinical and pathological prognostic indicators such as tumor stage, pathological grade, Ki67 proliferation index, and resection status are helpful in estimating survival, more recently, genomic and transcriptome based studies have identified molecular markers that can help predict recurrence-free survival as well [92]. Various RT-q-PCR studies have shown that tissue relative expression levels of *miR-210*, *miR-483-5p*, *miR-195*, *miR-503*, *miR-1202*, and *miR-1275* are associated with overall survival in ACC [73,74,80]. In addition, ACC tissue *miR-9* relative expression has been shown to correlate with recurrence-free survival as well as overall survival [93], as has the serum relative expression of *miR-483-5p* and *miR-195* [76].

In an RNA sequencing study of 45 ACC and three NAC tissues, Assie and colleagues used consensus clustering to classify tumors into three groups based on their microRNA profiles. The cluster most distinct from NAC, Mi1, was also characterized by consistent 14q32 LOH and *maternally expressed 3* (*MEG3*) long non-coding RNA promotor methylation. The 14q32 cytogenetic band contains 54 miRNAs, one of the largest miRNA clusters in the human genome, and 38 of these were underexpressed in the good prognosis of Mi1 tumors. The Mi2 group was characterized by weak overexpression of the *miR-506-514* cluster, known to have an oncogenic role in melanoma [94], while the Mi3 group was strongly correlated with the poor prognosis transcriptome cluster C1A. Interestingly, while

miRNA expression was maximally deregulated in Mi1 and Mi2 cluster tumors, ACC driver pathway alterations were more consistently associated with Mi3 cluster tumors [51]. This study suggests that the integrated analysis of miRNA expression is likely to be a superior approach to single miRNA prognostic biomarkers for ACC.

#### **6. Computational and Experimental Methods of miRNA Target Identification**

Based on the existing understanding of the interactions between miRNAs and target mRNAs, various software tools have been developed to predict endogenous miRNA targets for experimental validation [95]. Recently, Ab Mutalib and colleagues reviewed the thirty-nine computational tools currently available for miRNA target prediction [96], of which only one, DeAnnIso, allows for target prediction of isomiRs [97]. These bioinformatic prediction tools are limited as they cannot predict miRNA binding to non-coding RNAs, nor do they account for non-canonical mRNA binding sites [98]. In addition, bioinformatics methods may give false-positive results, and miRNA target predictions do not always account for the tissue specificity of miRNA expression [99]; therefore, computational target predictions should always be validated experimentally.

The various experimental methods available for miRNA target validation in biological systems have been comprehensively reviewed elsewhere [100–102]. They include indirect methods such as expression profiling or stable isotope labeling by amino acids in cell culture (SILAC) following miRNA overexpression or inhibition, as well as direct methods such as reporter assays, biotinylated miRNA pulldown assays, and RISC component pulldown assays. In reporter assays, direct evidence of miRNA regulation is established when a mutated mRNA target site results in loss of miRNA regulation. In these experimental approaches, miRNAs are often overexpressed to supraphysiological levels, resulting in the saturation of RISC complexes at the expense of other endogenous miRNAs, and false-positive results that allow low-affinity targets to appear functionally relevant [103]. Nevertheless, these low-affinity targets, while irrelevant to the endogenous functioning of the miRNA in question, continue to be important when considering the cellular effects of miRNAs as potential therapeutics. Caution must also be exercised when extrapolating results from miRNA target validation experiments in particular cellular environments across tissue types, as the failure to detect cell-specific natural targets may ensue [100].

#### **7. Functional miRNA Target Relationships in ACC**

In ACC, the evidence for miRNA functional targets comes largely from reporter assays in combination with the cellular effects of modulation of miRNA expression in cell culture.

#### *7.1. Overexpressed miRNAs and Their ACCs*

Several miRNAs that are overexpressed in ACC relative to NAC have proven oncogenic roles *in vitro*, as well as defined molecular targets, that they regulate (Table 3). *miR-9* [104], *miR-21* [105], *miR-483-3p* [106], and *miR-483-5p* [106] have been well described in the literature as 'oncomiRs' across multiple mammalian cell types, which is consistent with their role in ACC. In contrast, *miR-139-5p* is an established tumor suppressor in head and neck/oral, breast, and gastric cancers [107], but is overexpressed in aggressive ACCs compared with non-aggressive ACCs [108].



DFS, Disease-Free Survival; PDCD4, Programmed cell death protein 4; NDRG, N-myc downstream-regulated gene; PUMA, p53 upregulated modulator of apoptosis.

#### 7.1.1. miR-9 Regulates LIN28

*miR-9* has diverse actions in cancer, and whether it acts as a tumor suppressor or oncomiR is tissue dependent [118]. In ACC, aggressive phenotypes overexpress *miR-9* in comparison with non-aggressive phenotypes [51], and *miR-9* overexpression is associated with poor prognosis in clinical datasets [93]. Luciferase-based assays have demonstrated direct binding between *miR-9* and LIN28, an RNA binding protein that regulates miRNA biogenesis via the miRNA *let-7* in HeLa and A2780 ovarian carcinoma cell lines [111]. Aggressive ACCs have been demonstrated to have weak LIN28 protein expression on immunostaining [93], lending evidence to the hypothesis that LIN28 expression is regulated by *miR-9* in this cellular environment.

#### 7.1.2. MiR-21 Regulates PDCD4

*MiR-21* is the most commonly overexpressed miRNA in cancer and is generally associated with an aggressive phenotype and poor prognosis. *miR-21* expression is negatively correlated with *Programmed Cell Death Protein 4* (*PCDC4*) expression across many solid tumors [105]. *PCDC4* is upregulated during apoptosis and inhibits translation of particular genes, including *p53*, by competitively binding translation initiation factors [119]. Luciferase reporter assays in HeLa cells as well as colorectal and thyroid cell lines, have established *miR-21* as a direct regulator of *PCDC4* expression [114,115]. In ACC, in vitro gene-specific silencing of *miR-21* resulted in increased *PCDC4* expression and reduced cellular proliferation [113], which suggests that this regulatory relationship between *miR-21* and *PCDC4* is also present in ACC.

#### 7.1.3. miR-483-3p Regulates PUMA

*miR-483-3p* is an oncomiR in ACC, promoting cellular proliferation and inhibiting apoptosis in in vitro cell models [74]. Reporter assays in three different cell lines, including human embryonic kidney (HEK293), liver cancer (HepG2), and colon cancer (HCT116), demonstrated that *miR-483-3p* directly inhibits *PUMA* expression [117]. *PUMA* is a downstream target of p53, which antagonizes the anti-apoptotic B-cell lymphoma 2 (Bcl-2) family proteins and consequently induces apoptosis [120]. *PUMA* expression was found to be inversely correlated with *miR-483-3p* expression in ACC, but not in ACA or NAC tissue [74]. Given that *miR-483-3p* is a proven regulator of *PUMA* expression in various cell models, this relationship can be extrapolated to ACC.

#### *7.2. Underexpressed miRNAs and Their ACC Targets*

The molecular targets of a number of underexpressed tumor suppressor miRNAs in ACC have also been characterized (Table 4). These miRNAs include *miR-7* and *miR-205* which have demonstrated tumor suppressor activity in in vivo xenograft ACC models [121,122]. They also include *miR-195* and *miR-497*, which are members of the tumor suppressor *miR-15* family and share the same seed sequence [123], as well as *miR-99* family members *miR-99a* and *miR-100*, which are known to target the mTOR signaling pathway [124].


**Table 4.** Underexpressed microRNAs and their regulated targets in ACC.

\* Regulatory interaction demonstrated by reporter assays for Raf-1 and EGFR only. Raf-1, Rapidly Accelerated Fibrosarcoma-1; EGFR, Epidermal Growth Factor Receptor; PAK1, p21 activated kinase 1; CKS2, CDC28 Protein Kinase Regulatory Subunit 2; CDK1, Cyclin-Dependent Kinase 1; IGFR1, Insulin-like Growth Factor 1 Receptor; mTOR, Mechanistic Target of Rapamycin Kinase; TARBP2, TAR RNA-binding protein 2; DICER1, Dicer 1 Ribonuclease III; Ago-2 IP, Argonaute-2 immunoprecipitation; ZNF367, Zinc Finger Protein 367; Bcl-2, B-cell lymphoma 2; MTDH, metadherin; ZEB1, Zinc finger E-box binding homeobox 1; MALAT1, metastasis-associated lung adenocarcinoma transcript 1; eIF4E, eukaryotic translation initiation factor 4E; SFPQ, splicing factor proline and glutamine-rich.

#### 7.2.1. *miR-7* Regulates *Raf-1, EGFR, CDK1, PAK1, CKS2*

In the human genome, *miR-7* is encoded on three separate loci whose different DNA sequences can all be processed into the same mature *miR-7* sequence [130]. *miR-7* is an almost ubiquitous tumor suppressor, being underexpressed in malignancies that range from those derived from brain tissue to a myriad of solid tumors as well as leukemias [131]. In ACC, in vitro overexpression of *miR-7* decreases proliferation and induces G1 cell cycle arrest and decreases the expression of *p21 activated kinase 1* (*PAK1*), *CDC28 protein kinase regulatory subunit 2* (*CKS2*), and *cyclin-dependent kinase 1* (*CDK1*) mRNA [121]. PAK1 activation induces apoptosis, while CKS2 is an essential co-factor for CDK proteins that regulate the cell cycle. Luciferase reporter assays in H295R cells demonstrated the regulatory relationship between *miR-7* and *rapidly accelerated fibrosarcoma-1* (*Raf-1*) as well as the *epidermal growth factor receptor* (*EGFR*) [121]. *miR-7* targeting of *EGFR* has also been demonstrated in breast, lung, gastric, and ovarian cancers, as well as glioma and schwannoma tumors. In schwannomas and breast cancer, *miR-7* has also been shown to target *PAK1* [131].

#### 7.2.2. miR-99a/100 Regulates IGFR1, mTOR

Both *miR-99a* and *miR-100* were discovered as underexpressed relative to NAC in childhood ACC tissue samples, where their expression was inversely correlated with both *mTOR* and *insulin-like growth factor 1 receptor* (*IGFR1*) mRNA. *miR-100* specific knockdown in in vitro ACC cell models was associated with increased mTOR and IGFR1 protein expression, and furthermore, luciferase reporter assays in HEK293 cells showed that both miRNAs could regulate mTOR and IGFR1 [125]. The regulatory relationship between the *miR-99* family and mTOR has been well studied in cardiovascular disease [132] as well as in wound healing [133]. In cancer, *miR-99* regulation of mTOR has been demonstrated to enhance radiation sensitivity in urothelial carcinoma [134] as well as non-small cell lung cancer [135].

#### 7.2.3. miR-205 Regulates Bcl-2

In ACC, *miR-205* was shown to be underexpressed in a clinical cohort with RT-q-PCR. The subsequent gain of function studies carried out using SW13 cells (a cell line derived from adrenocortical metastasis of unknown origin) showed that *miR-205* promoted apoptosis and impaired cellular proliferation *in vitro*, and in vivo mouse SW13 xenograft studies showed that it could inhibit tumor growth [122]. Luciferase reporter assays also carried out in SW13 demonstrated a direct regulatory relationship between *miR-205* and B-cell lymphoma 2 (Bcl-2) protein, which is known to regulate the intrinsic apoptotic pathway in cancer [136].

#### 7.2.4. *miR-375* Regulates MTDH

*miR-375*, a known tumor suppressor in multiple cancers, is underexpressed in ACC [78], and in aldosterone-producing adrenal adenomas, its expression is correlated with the tumor size [127]. In vitro overexpression of *miR-375* reduces cellular proliferation and suppresses metadherin (MTDH), which functions to promote tumor invasion, metastasis, and chemoresistance. It also acts via the PI3K/Akt and Wnt/β-catenin pathways to promote cellular proliferation, invasion, and survival [137]. Luciferase reporter assays in H295R cells show that *miR-375* directly binds *MTDH* mRNA and regulates its expression in vitro in ACC.

#### 7.2.5. miR-431 Regulates ZEB1

In ACC clinical samples, *miR-431* is differentially expressed in chemosensitive tumors compared with chemoresistant tumors. The gain of function studies in H295R and primary ACC cells showed that *miR-431* overexpression decreased the IC50 of both doxorubicin and mitotane to inhibit cellular proliferation. In cells treated with doxorubicin, *miR-431* reversed epithelial-to-mesenchymal transition (EMT) [128]. Zinc finger E-box binding homeobox 1 (ZEB1), a protein that induces EMT in cancer cells, had already been established as a direct target of *miR-431* in hepatocellular carcinoma [138]. Both *ZEB1* mRNA and protein expression decreased in doxorubicin treated H295R cells overexpressing *miR-431* [128], indicating that this regulatory relationship between *miR-431* and ZEB1 is active in ACC.

### 7.2.6. miR-497 Regulates TARBP2, DICER1, MALAT1, eIF4E, SFPQ

*miR-497* expression is dysregulated in many solid organ tumors, which suggests that it may play an important tumor suppressor role. Multiple studies have confirmed *miR-497* underexpression in ACC and its genomic location in a region of frequent LOH (17p13.1-13.3), in close proximity to the *p53* locus, indicates that it may play a role in ACC tumorigenesis [15]. In vitro H295R gain of function studies have shown that *miR-497* decreases cellular proliferation, increases apoptosis, and also induces G1 cell cycle arrest [74,129]. *miR-497* has been shown to directly target the miRNA biogenesis related proteins DICER1 and TARBP2 in Ago-2 IP assays, along with *miR-195*. This was confirmed with gain of function studies that showed an inverse correlation between *miR-497* expression and *DICER1*, as well as *TARBP2* mRNA and protein expression in H295R cells [58]. In a separate study, which was the first to demonstrate miRNA targeting of long non-coding RNAs in ACC, luciferase reporter assays demonstrated that *miR-497* regulates the expression of *metastasis-associated lung adenocarcinoma transcript 1* (*MALAT1*) [129]. *MALAT1* is overexpressed in numerous types of tumors, including ACC, and is known to promote cellular proliferation, apoptosis, migration, and invasion [138]. In H295R, *miR-497* overexpression and *MALAT1* knockdown inhibit the expression of *eukaryotic translation initiation factor 4E* (*eIF4E*), which directs ribosomes to the cap structure of mRNAs and is, therefore, essential for protein synthesis [129]. *miR-497* gain of function and *MALAT1* knockdown studies further demonstrated the reciprocal inhibitory relationship between them in in vitro ACC models.

Within the limits of the caveats previously outlined, we can infer from the above studies that miRNAs modulate many protein targets that are involved in key driver pathways in ACC.

#### **8. miRNA Modulation of ACC Driver Pathways**

A significant proportion of the identified ACC miRNA molecular targets play various roles in the established ACC driver pathways. This supports the notion that miRNA modulation of protein expression, which in healthy cells helps to finetune and maintain the homeostatic balance, can potentiate oncogenesis when dysregulated.

#### *8.1. miRNA Modulators of the p53 Pathway in ACC*

The overexpressed oncomiR *miR-483-3p* and the underexpressed tumor suppressors *miR-7* and *miR-205* all regulate downstream targets of p53 (Figure 1). *miR-483-3p* suppression of PUMA expression and the alleviation of *miR-205* modulated Bcl-2 inhibition of Bax, act synergistically to inhibit p53-mediated apoptosis. P53 is known to transcriptionally downregulate CDK1, and thus, initiates G2 cell cycle arrest. Constitutive activation of CDK1 resulting from the loss of *miR-7* targeted suppression overrides p53 mediated G2 arrest, leading to uncontrolled proliferation.

protein synthesis [129]. *miR-497* gain of function and *MALAT1* knockdown studies further

Within the limits of the caveats previously outlined, we can infer from the above studies that

A significant proportion of the identified ACC miRNA molecular targets play various roles in the established ACC driver pathways. This supports the notion that miRNA modulation of protein expression, which in healthy cells helps to finetune and maintain the homeostatic balance, can

The overexpressed oncomiR *miR-483-3p* and the underexpressed tumor suppressors *miR-7* and *miR-205* all regulate downstream targets of p53 (Figure 1). *miR-483-3p* suppression of PUMA expression and the alleviation of *miR-205* modulated Bcl-2 inhibition of Bax, act synergistically to inhibit p53 mediated apoptosis. P53 is known to transcriptionally downregulate CDK1, and thus, initiates G2 cell

demonstrated the reciprocal inhibitory relationship between them in in vitro ACC models.

miRNAs modulate many protein targets that are involved in key driver pathways in ACC.

**8. miRNA Modulation of ACC Driver Pathways** 

potentiate oncogenesis when dysregulated.

*8.1. miRNA Modulators of the p53 Pathway in ACC* 

overrides p53 mediated G2 arrest, leading to uncontrolled proliferation.

**Figure 1.** miRNA modulators of the p53 signaling pathway in Adrenocortical Carcinoma. **Figure 1.** miRNA modulators of the p53 signaling pathway in Adrenocortical Carcinoma.

#### *8.2. miRNA Modulators of the mTOR Pathway in ACC*

The loss of *miR-7* regulation of Raf-1 and EGFR expression leads to downstream mTOR activation in ACC. The underexpression of *miR-99a*/*100* also leads to mTOR activation, ultimately potentiating protein synthesis, which is further enhanced by the loss of *miR-497* mediated eIF4E regulation. The loss of *miR-99a*/*100* mediated IGFR1 expression also promotes cell survival via the PI3K/AKT signaling pathway (Figure 2).

*8.2. miRNA Modulators of the mTOR Pathway in ACC* 

*8.2. miRNA Modulators of the mTOR Pathway in ACC* 

PI3K/AKT signaling pathway (Figure 2).

The loss of *miR-7* regulation of Raf-1 and EGFR expression leads to downstream mTOR activation in ACC. The underexpression of *miR-99a/100* also leads to mTOR activation, ultimately potentiating protein synthesis, which is further enhanced by the loss of *miR-497* mediated eIF4E

The loss of *miR-7* regulation of Raf-1 and EGFR expression leads to downstream mTOR activation in ACC. The underexpression of *miR-99a/100* also leads to mTOR activation, ultimately potentiating protein synthesis, which is further enhanced by the loss of *miR-497* mediated eIF4E

*Cancers* **2020**, *12*, x 14 of 22

**Figure 2.** miRNA modulators of the mTOR signaling pathway in Adrenocortical Carcinoma. **Figure 2.** miRNA modulators of the mTOR signaling pathway in Adrenocortical Carcinoma. **Figure 2.** miRNA modulators of the mTOR signaling pathway in Adrenocortical Carcinoma.

*8.3. miRNA Modulators of the Wnt/Β-Catenin Pathway in ACC 8.3. miRNA Modulators of the Wnt*/*B-Catenin Pathway in ACC 8.3. miRNA Modulators of the Wnt/Β-Catenin Pathway in ACC* 

In ACC, the loss of *miR-431* regulation allows ZEB1 to activate Wnt, consequently activating βcatenin, which potentiates cell cycling. The loss of *miR-375* mediated MTDH suppression upstream of MAPK modulates the Wnt/β-catenin pathway to promote cell cycling (Figure 3). In ACC, the loss of *miR-431* regulation allows ZEB1 to activate Wnt, consequently activating β-catenin, which potentiates cell cycling. The loss of *miR-375* mediated MTDH suppression upstream of MAPK modulates the Wnt/β-catenin pathway to promote cell cycling (Figure 3). In ACC, the loss of *miR-431* regulation allows ZEB1 to activate Wnt, consequently activating βcatenin, which potentiates cell cycling. The loss of *miR-375* mediated MTDH suppression upstream of MAPK modulates the Wnt/β-catenin pathway to promote cell cycling (Figure 3).

**Figure 3.** miRNA modulators of the Wnt/β-catenin signaling pathway in Adrenocortical Carcinoma. **Figure 3. Figure 3.**  miRNA modulators of the Wnt miRNA modulators of the Wnt/ /β β-catenin -catenin signaling pathway in Adrenocortical Carcinoma. signaling pathway in Adrenocortical Carcinoma.

#### **9. Future Directions and Conclusions**

miRNAs play an important role in the modulation of ACC related target protein expression. The dysregulation of miRNA expression disturbs the homeostatic balance of proteins that participate in the pathways controlling cell cycle progression, cellular proliferation, apoptosis, and chemoresistance. The overexpression of oncogenic miRNAs and underexpression of tumor suppressor miRNAs thus potentiate tumorigenesis. The role of miRNA regulation in ACC remains an area of active research with the potential to further our understanding of its tumor biology and the molecular pathways involved. Small RNA sequencing of isomiRs and further refining our understanding of the miRNA signature of

ACC provides the opportunity to improve diagnostic accuracy with techniques such as miRNA liquid biopsy. With continuing advances in functional techniques that allow molecular interactions to be clearly established, it will be possible to explore novel miRNA-based therapeutic approaches with the aim of improving the current poor prognosis of these patients.

**Author Contributions:** Conceptualization, M.C., S.S. and G.H.; writing—original draft preparation, M.C.; writing—review and editing, M.B., A.G., S.S., G.H.; visualization, M.C.; supervision, G.H. and S.S. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** Charbel Azzi for graphics support.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
