A review of Dynamin 2 involvement in cancers highlights a promising therapeutic target

Dynamin 2 (DNM2) belongs to a superfamily of large GTPases including three classical dynamins and several dynamin-like proteins sharing propensity to self-assemble into polymeric structures [1]. DNM2 is ubiquitously expressed and is encoded by the DNM2 gene producing four major splice isoforms using a combination of two alternative splice sites [2]. In addition, a muscle-specific isoform was identified [3]. DNM2 is a central actor of many membrane remodeling processes including clathrin-mediated endocytosis [4], clathrin-independent endocytosis [5, 6], coat-independent endocytosis [7, 8], intracellular vesicle trafficking [9‐13], and exocytosis [14]. For its function of vesicle formation, DNM2 oligomerizes in a helical structure around the neck of nascent vesicles [4], and GTP hydrolysis leads to modifications in the helical structure associated with the release of the vesicles [15]. A function of regulator of actin [16‐18] and microtubule [19, 20] cytoskeletons is now well recognized for DNM2 which sustains, in particular, its participation to mitosis and cell cycle progression [8, 21‐23]. Heterozygous mutations in the DNM2 gene are responsible for autosomal dominant forms of three human diseases, i.e. centronuclear myopathy [24], Charcot-Marie-Tooth disease [25], and hereditary spastic paraplegia [26] and one homozygous mutation was shown to cause a lethal congenital syndrome [27]. DNM2 has also been described as a susceptibility gene for late-onset Alzheimer disease [28]. In addition, numerous studies are now available involving DNM2 in several cancers. The purpose here is to review involvement of DNM2 in pathophysiology of cancers and highlight the opened perspectives in medicine using DNM2 as a therapeutic target in these conditions.

Acute lymphoblastic leukaemia (ALL) is the most common malignancy in B cells, immature T lymphocytes or lymphoid progenitors. Somatic heterozygous mutations in the DNM2 gene were identified in children affected by early T-cell precursor acute lymphoblastic leukaemia (ETP-ALL), a form of ALL associated with a high risk of treatment failure [29]. In this study, DNM2 mutations have been identified in 17 patients (13 out of 64 ETP patients and 4 out of 42 non ETP ALL patients) including 2 cases with biallelic mutations. The 19 mutations include 3 frameshifts, 3 splice site mutations, 6 missense mutations, 2 nonsense mutations, and 5 in frame deletions or insertions not found in centronuclear myopathy, Charcot-Marie-Tooth disease and hereditary spastic paraplegia. Thereafter, 4 other DNM2 mutations (1 nonsense and 3 missense) have been described in adult T cell ALL patients extending the DNM2 involvement in ALL [30]. Clinical analysis of this small cohort of patients suggested that DNM2 mutations are associated with a poor prognosis.

In addition to the DNM2 mutations, an increased DNM2 expression was also identified in several cancer types (summarized in Table 1). DNM2 expression is increased in ALL, especially in B-ALL in which overexpression is associated with leukaemia cell proliferation and a poor prognosis [31]. DNM2 overexpression was also demonstrated in stem cells and progenitor cells of chronic myeloid leukemia (CML), i.e. a clonal myeloproliferative disorder originating from hematopoietic stem cells [32], in multiple ovarian cancer data sets [33], in bladder tumors in which the overexpression is correlated with the grade progression [34], in papillary thyroid cancer in which high DNM2 expression is associated with unfavourable prognosis (tumor recurrence, overall survival rate, distant metastases) [35], and in prostate cancer in which high DNM2 expression is associated with higher aggressiveness (Gleason score) and mortality [36, 37]. DNM2 overexpression was also reported in cervical cancer. In this cancer, the initial transformation of normal epithelium to preneoplastic cervical intraepithelial neoplasia (CIN) may be followed by a transformation to invasive cancer. DNM2 is not expressed in normal cervical epithelium but is overexpressed in preinvasive low grade lesions (CIN grade 1). With cancer progression, DNM2 expression decreases through an unidentified mechanism leading to low or no expression in high-grade lesions (CIN grades 2 and 3) especially, in case of deep tumor invasion and lymph node metastasis [38, 39]. These results establish DNM2 expression as a biomarker in grading of CIN, with a negative correlation between DNM2 expression and the severity of lesions, and suggest a crucial impact of DNM2 overexpression in the earliest steps of neoplasia of cervix. Both expression and localization of DNM2 were studied in tumor and adjacent normal tissue from 113 patients affected by several types of breast cancers [40]. All these breast tumor tissue samples exhibit a higher DNM2 expression compared to normal tissues. Among these cases, cytoplasmic overexpression was correlated with a specific cancer sub-type (invasive ductal carcinoma) whereas membranous DNM2 staining was associated with vascular invasion, an indicator of aggressiveness of breast cancer. Intriguingly, an increased expression of DNM2 was also noticed in the nucleus which is positively correlated with the histological grade and the tumor stage. This non-conventional nuclear localization also suggests a gain-of-function in the nucleus which needs to be deepen explored. In a particular sub-group of breast cancer, i.e. the triple negative breast cancers, defined by absence of estrogen receptor, progesterone receptor and without overexpression or amplification of EGF receptor 2, the level of DNM2 overexpression was negatively correlated with positive chemotherapy outcome [41]. The greater relapse rate in the patients exhibiting the higher DNM2 expression in tumor was attributed to the role of DNM2 in homology-directed repair of DNA, involving a DNM2 mediated trafficking of the RAD51 recombinase, which may mitigate over time the benefit of DNA-damaging chemotherapy in case of DNM2 overexpression. Finally, DNM2 expression was also shown to be markedly upregulated in tumors and metastases of patients affected by pancreatic ductal adenocarcinoma and this overexpression contributes to lamellipodia extension, cell migration and invasion [42]. In addition, patients with the highest levels of DNM2, associated with high level of α-actinin 4, had lower mean survival times compared to patients with lower expression [43]. Altogether, these data highlight the frequent up-regulation of DNM2 in cancers which can be used as a marker of poor prognosis.

The mechanisms underlying DNM2 overexpression in cancers have not been extensively studied and remains an important open question. Of note, Ikaros was shown to bind to the DNM2 promoter in B cells [31] (Fig. 1). Ikaros is a DNA-binding zinc finger protein essential for normal haematopoiesis and immune development and acting as a tumor suppressor gene in acute B- and T-cell ALL. The correlation between low level of Ikaros mRNA and high level of DNM2 mRNA in ALL cells reinforces the hypothesis that Ikaros depletion may explain the high DNM2 expression in this pathological condition [31]. In addition, it was shown in ovarian cancer cell lines that DNM2 levels are transcriptionally down-regulated by the hypoxia-induced factor 1 (HIF-1) which directly binds to the DNM2 promoter [44]. Consequently, HIF-1 inhibition may participate to DNM2 overexpression. The finding that DNM2 overexpression conversely reduced HIF-1α expression [44] may create a vicious circle maintaining a high level of DNM2 transcription. Finally, a regulation of DNM2 expression by CD9, a member of the tetraspanin family playing important functions in signal transductions from the plasma membrane, was demonstrated in pancreatic cancer cell lines [45]. In these cells, overexpression of CD9 led to the upregulation of the DNM2 expression by an unsolved molecular pathway.

Other potential regulators of DNM2 transcription in cancers can be inferred from studies relative to another disease in which deleterious DNM2 overexpression occurs. The myotubular myopathy is the severe X-linked form of centronuclear myopathy due to mutations in the MTM1 gene encoding Myotubularin 1. Patients and mouse model of myotubular myopathy exhibit an increased DNM2 expression and a forced DNM2 reduction restores phenotypes and lifespan in a mouse model of the disease [46, 47]. A recent study identified overexpression and activation by phosphorylation of the transcription factor STAT3 (signal transducer and activator of transcription 3) as responsible for the DNM2 upregulation in this pathological condition [48]. STAT3 is a key regulator of cell proliferation, survival and apoptosis and is constitutively activated in most human cancers including prostate, pancreas, breast cancers and leukaemia [49‐51]. It is tempting to speculate that STAT3 activation is involved in DNM2 upregulation in cancers and that DNM2 overexpression may be an effector of features associated with abnormal STAT3 activation in cancers such as tumor growth and metastasis leading to poor prognosis [49‐51]. In addition, the demonstration that tamoxifen, known to bind estrogen receptors, normalizes DNM2 expression and improves the phenotype of the Mtm1-deficient mouse model suggests that estrogen receptors may be involved in the regulation of the DNM2 expression [52].

Besides the above-mentioned transcriptional regulations of the DNM2 gene, post-transcriptional mechanisms may be also involved in DNM2 upregulation in cancers. DNM2 is a target of miR-133a-1 and miR-133a-2 (Fig. 1) and silencing of miR-133a in mouse induces an increase in DNM2 in skeletal muscle and a centronuclear myopathy phenotype [53]. It is tempting to speculate that downregulation of miR-133a, as reported in prostate, pancreas and bladder cancers [54], may participate to the DNM2 deregulation in cancers.

Several DNM2-dependent pathways have been identified in cancer cell lines and may be involved in cancer pathomechanisms. These mechanisms, summarized in Fig. 2, may be classified in 2 groups: mechanisms promoting cell migration and metastasis, and deregulation of intracellular signaling pathways linked to cell proliferation and survival.

Benefit of DNM2 inhibition was largely documented for almost all the DNM2-related phenotypes in tumor cells in vitro. The function of the DNM2 in cytokinesis was targeted in order to influence the cell cycle progression and reduce proliferation of tumor cells. With this objective, pharmacological inhibitors of DNM2 [74, 75] were used to induce cytokinesis failure at the membrane abscission step leading to growth arrest and death of cervical, lung, and leukaemia cancer cell lines [76‐78]. This induced cancer cell death following cytokinesis failure results from the intrinsic caspases 3 and 9-mediated apoptotic pathway and is more efficient in cancer cell lines harbouring low level of the anti-apoptotic Bcl-2 and Mcl-1 proteins [79]. Interestingly, nontumorigenic fibroblasts treated with DNM2 inhibitors appear less sensitive to cell death than cancer cells [76, 77]. A benefit a DNM2 inhibition to reduce cell proliferation and/or induce apoptosis was confirmed in a wide range of cancer cells including cervical epithelial cancer cells [80], prostate cancer cells [36, 42], non-small-cell lung cancer cells [81], glioblastoma cells [82], chronic myeloid leukemia cells [32], B- and T-ALL cells [31] and hepatocellular carcinoma cells [83]. The similar benefit using pharmacological DNM2 inhibitors [74, 75, 82, 84‐88] or DNM2 gene silencing demonstrates the requirement of the DNM2 GTPase activity in the phenotypes occurring in these cancer cells. In hepatocellular carcinoma cells, dynamin inhibition accelerates degradation of c-Met, a tyrosine kinase receptor involved in hepatocellular carcinoma development and progression, and decreases the c-Met downstream signaling [83]. DNM2 inhibition can also reduce the invasive phenotype of tumor cells in vitro. The benefit of pharmacological inhibitors [75, 84, 85, 89] or DNM2 mRNA silencing on cell invasion in prostate, lung, bladder, and pancreatic cancer cells is provided through an impact on actin dynamics and formation of the actin-based protrusions (lamellipodia, filopodia, and invadopodia) and/or stabilizing focal adhesion which impede invasive behaviour [36, 42, 43, 57, 58, 61, 62, 89]. In melanoma and breast cancer cell lines, DNM2 inhibition impacts cell invasion by reducing extracellular matrix remodelling and the number of degrading cells [63, 65]. DNM2 inhibition was also achieved by targeting the DNM2 gene transcription. Indeed, due to the role of Ikaros in DNM2 expression in ALL cells, effect of 4,5,6,7-Tetrabromobenzotriazole (TBB), an enhancer of Ikaros tumor suppressor activity, was investigated. Interestingly, TBB increased Ikaros binding to the DNM2 promoter and inhibited expression of DNM2 mRNA in ALL cells [31]. A better understanding of the mechanisms underlying the DNM2 overexpression in the different types of cancers will be helpful for developing similar therapeutic approaches.

Importantly, the proof of concept of therapy by modulating DNM2 expression was also achieved in vivo in several animal models (Table 2). In order to assess whether lowering DNM2 levels would increase sensitivity to chemotherapy, breast cancer cells expressing an inducible shRNA to silence DNM2 by RNA interference were implanted into mammary fat pads of mice. Reduced expression of DNM2 significantly improved tumor response to cyclophosphamide, a drug widely used in chemotherapy of breast cancers, leading to reduced tumor volume [41]. A proof of concept was also done using pancreatic cancer cells in vivo [42]. Orthotopic injection of cells overexpressing DNM2 promotes tumor cells dissemination distal from the injection area compared to mice injected with pancreatic cancer cells with basal DNM2 expression. Inhibition of DNM2 by overexpressing phospho-deficient mutant in the injected pancreatic cancer cells drastically reduces the number of large tumors outside the injection site. A third study in which prostate cancer cells were injected in prostate of mice showed that tumor weight and lymph node metastases are reduced when DNM2 is inhibited by RNA interference in the injected cells [36]. A similar result was reported for the PDGFRα-stimulated glioma cell growth and invasion in brain of mice [60]. In this case, injection of DNM2-depleted cells is associated with decrease in tumor cell proliferation and increase in apoptosis. Proofs of concept were also achieved using pharmacological compounds. In a prostate cancer mouse model in which formation of subcutaneous tumors are induced after injection of human prostatic adenocarcinoma cells, intratumoral injection of a DNM2 inhibitor (N’-[4-(dipropylamino)benzylidene]-2-hydroxybenzohydrazide, DBHA) rapidly reduces tumor size without apparent adverse effect [89]. In the same line, continuous delivery of the CyDyn4-36 DNM2 inhibitor by subcutaneous osmotic pumps reduces the size of established tumors formed by prior intracranial injection of glioma stem cells [82] and intraperitoneal injection of the Dynole 34.2 DNM2 inhibitor leads to progressive exhaustion of pre-leukemia stem cells in a mouse model of acute leukemia [90]. Dynole 34.2 also reduces the number of leukemic cells in the bone marrow and spleen of mice after T-ALL and acute myeloid leukemia cells xenografts [90]. DBHA, Dynole 34.2 and CyDyn4-36 represent a new generation of DNM2 inhibitors opening the way for future clinical use. DBHA, is an analogue of the non-competitive inhibitor dynasore, identified through a screening for inhibitors of dynamin GTPase activity, which suppress actin dynamics and cancer cells migration more efficiently with less cytotoxicity compared to dynasore [85, 89]. Dynoles are also non-competitive inhibitors of dynamin GTPase activity [75] and the Dynole 34–2 was shown as the most effective at causing cancer cell death [77]. CyDyn4-36 is a next-generation Dynole designed to be more brain penetrant through a reduction in polar surface area and the number of hydrogen bond donors and acceptors [82]. Altogether, these data highlight Dynamin 2 as a promising molecular target for the development of anti-invasive agents and the proof of concept of therapy already achieved by reducing DNM2 expression in animal models is an important step of preclinical development.

Numerous studies are now available demonstrating the interest of DNM2 as a biomarker for prognosis of tumor invasion and metastasis, as exemplified for cervical cancers, or to predict therapy efficacy as shown in a sub-group of breast cancer. The available data strongly suggest that DNM2 participates to the maintenance of the necessarily favourable environment for tumor to grow, invade and migrate. Interestingly, modulation of DNM2 expression may prove a valuable therapeutic target with proof of concept achieved in vitro and in vivo. Molecules targeting DNM2 expression and/or activity may expand the panoply of anti-mitotic and anti-metastatic agents in cancer treatments. Therapeutic potential of DNM2-based approaches may be of particular interest to counteract metastasis which mark the transition from a benign tumor to a lethal, malignant cancer with dissemination of tumor cells. However, future therapeutic intervention should avoid to excessively reduce DNM2 expression for maintaining the beneficial aspects of DNM2 function in cancer cells as DNA repair already mentioned [41] or the metastasis suppressor function of NME proteins [91]. Indeed, when re-expressed or overexpressed in metastatic tumor cells, NME proteins suppress cell motility and migration in vitro and metastatic colonization in vivo through a mechanism requiring DNM2-dependent endocytosis [91]. The importance to preserve DNM2 expression is also highlighted by the reported analysis of 131 hepatocellular carcinoma showing that patients with low DNM2 expression displayed a significantly worse overall survival [92]. This study concludes that drastic DNM2 downregulation by siRNA or drug in hepatocellular carcinoma cell lines provides colony formation, migration and invasion of tumor cells by reducing the role of DNM2 as a negative regulator of epidermal growth factor (EGF) signaling through the endocytosis of the EGF receptor [92]. Overall, the challenge of future DNM2-based therapies will be to simultaneously maintain DNM2 positive roles and counteract its deleterious functions in tumor cells.