cancers Article Cx43 and Associated Cell Signaling Pathways Regulate Tunneling Nanotubes in Breast Cancer Cells Alexander Tishchenko 1,†,‡ , Daniel D. Azorín 1,†,§ , Laia Vidal-Brime 1 , María José Muñoz 1 , Pol Jiménez Arenas 1 , Christopher Pearce 1 , Henrique Girao 2,3,4 , Santiago Ramón y Cajal 1,5,6 and Trond Aasen 1,6,7, * 1 Patologia Molecular Translacional, Vall d’Hebron Institut de Recerca (VHIR), Vall d’Hebron Hospital Universitari, Vall d’Hebron Barcelona Hospital Campus, Passeig Vall d’Hebron 119-129, 08035 Barcelona, Spain; alexander.tishchenko@ugent.be (A.T.); d.dominguezazorin@dkfz-heidelberg.de (D.D.A.); laia.vidal.b@gmail.com (L.V.-B.); maria.munoz@vhir.org (M.J.M.); pol.jimenezarenas@gmail.com (P.J.A.); chrispearce99@yahoo.co.uk (C.P.); sramon@vhebron.net (S.R.yC.) 2 Biomedical Research (iCBR), Faculty of Medicine, University of Coimbra, Azinhaga de Santa Comba, Celas, 3000-548 Coimbra, Portugal; hmgirao@fmed.uc.pt 3 Center for Innovative Biomedicine and Biotechnology, University of Coimbra, 3000-548 Coimbra, Portugal 4 Clinical Academic Centre of Coimbra, CACC, 3000-548 Coimbra, Portugal 5 Anatomía Patológica, Vall d’Hebron Hospital Universitari, Vall d’Hebron Barcelona Hospital Campus, Passeig Vall d’Hebron 119-129, 08035 Barcelona, Spain 6 CIBER de Cáncer (CIBERONC), Instituto de Salud Carlos III, Avenida de Monforte de Lemos 3-5, 28029 Madrid, Spain 7 Universitat Autònoma de Barcelona, 08193 Bellaterra, Spain * Correspondence: trond.aasen@vhir.org; Tel.: +34-93489-4168 † These authors contributed equally to this work. ‡ Current address: Department of Virology, Parasitology and Immunology, Faculty of Veterinary Medicine, Ghent University, 9820 Merelbeke, Belgium. § Current address: Clinical Cooperation Unit Neuro-Oncology, German Cancer Consortium (DKTK), German Cancer Research Center (DKFZ), 69120 Heidelberg, Germany.  Received: 7 September 2020; Accepted: 25 September 2020; Published: 29 September 2020  Simple Summary: The gap junction protein connexin 43 (Cx43) facilitates direct intercellular communication and displays a complex dichotomous role in cancer. Here, we used breast cancer cell lines and CRISPR/Cas9 technology to dissect some of the poorly understood multimodal functions of Cx43 in this setting. Our data show that Cx43 directly modulates the formation of pro-tumorigenic tunneling nanotubes (TNTs), which are cellular protrusions that connect non-adjacent cells to facilitate cellular communication, including the direct cell–cell transfer of organelles. We identify several important cancer signaling pathways that affect TNT formation and show that this effect is strongly influenced by the presence or absence of Cx43. Finally, we demonstrate that Cx43 can affect TNT formation by modulating the cellular secretome. This work provides important insight into the pro-tumorigenic role of Cx43 and its interconnections with TNTs. Abstract: Connexin 43 (Cx43) forms gap junctions that mediate the direct intercellular diffusion of ions and small molecules between adjacent cells. Cx43 displays both pro- and anti-tumorigenic properties, but the mechanisms underlying these characteristics are not fully understood. Tunneling nanotubes (TNTs) are long and thin membrane projections that connect cells, facilitating the exchange of not only small molecules, but also larger proteins, organelles, bacteria, and viruses. Typically, TNTs exhibit increased formation under conditions of cellular stress and are more prominent in cancer cells, where they are generally thought to be pro-metastatic and to provide growth and survival advantages. Cx43 has been described in TNTs, where it is thought to regulate small molecule diffusion through gap junctions. Here, we developed a high-fidelity CRISPR/Cas9 system to knockout Cancers 2020, 12, 2798; doi:10.3390/cancers12102798 www.mdpi.com/journal/cancers Cancers 2020, 12, 2798 2 of 25 (KO) Cx43. We found that the loss of Cx43 expression was associated with significantly reduced TNT length and number in breast cancer cell lines. Notably, secreted factors present in conditioned medium stimulated TNTs more potently when derived from Cx43-expressing cells than from KO cells. Moreover, TNT formation was significantly induced by the inhibition of several key cancer signaling pathways that both regulate Cx43 and are regulated by Cx43, including RhoA kinase (ROCK), protein kinase A (PKA), focal adhesion kinase (FAK), and p38. Intriguingly, the drug-induced stimulation of TNTs was more potent in Cx43 KO cells than in wild-type (WT) cells. In conclusion, this work describes a novel non-canonical role for Cx43 in regulating TNTs, identifies key cancer signaling pathways that regulate TNTs in this setting, and provides mechanistic insight into a pro-tumorigenic role of Cx43 in cancer. Keywords: connexin 43; gap junctions; cancer; intercellular communication; breast cancer; cell signaling; tunneling nanotubes; cell–cell communication; tumor microtubes 1. Introduction Twenty-one human connexin proteins form the main structural units of gap junctions, which are intercellular channels that allow the direct diffusion of ions and small molecules between neighboring cells [1]. The critical importance of connexins and gap junctional intercellular communication (GJIC) in normal tissue physiology is evident from the multitude of human diseases associated with connexin mutations and the dysregulation of gap junctions observed in many pathological conditions [2]. More than 50 years of research show that GJIC and connexins are implicated in cancer progression and often act as tumor suppressors [3]. This has led to a significant drive toward the use of connexins as prognostic markers and therapeutic targets in cancer [4]. Although considered tumor suppressors, emerging evidence suggests that connexins can promote tumor growth and survival, particularly in advanced metastatic disease states [3–5]. In addition to facilitating GJIC, connexins are thought to possess channel-independent functions. Numerous such functions have been described for connexin 43 (Cx43), the most widely expressed and studied connexin. Notably, many protein–protein interactions in the Cx43 C terminus are thought to regulate important cancer signaling pathways such as c-Src [6]. Interestingly, Cx43 can take up residence in unexpected intracellular compartments, including the inner membrane of mitochondria and the nucleus [7], and Cx43 has been ascribed a functional role in exosomes [8,9]. Furthermore, we and others have shown that Cx43 produces a single mRNA transcript that is capable of synthesizing various N-terminally truncated isoforms through a unique mechanism of direct internal translation initiation [10–12]. The major truncated form, named GJA1-20k, corresponds to the C-terminal tail of Cx43 and can interact with numerous proteins and regulate malignant features in cancer cells, including proliferation, migration, and epithelial-to-mesenchymal transition (EMT) [6]. Notably, GJA1-20k modulates the motility and biogenesis of mitochondria [13,14] and regulates actin dynamics [15]. GJA1-20k has also been described in exosomes [16] and, surprisingly, in the nucleus, where it may act as a transcriptional regulator of N-cadherin during EMT [17]. These emerging non-canonical functions of Cx43 are providing new biological insight into cancer, as well as therapeutic opportunities [4,18]. However, dissection of the channel-dependent and -independent functions driving both the anti- and pro-tumorigenic pathways remains a major challenge. Tunneling nanotubes (TNTs), first described in 2004 [19], are thin, actin-based, membrane bridges enclosed in a lipid bilayer that connect cells over distances of up to several cell diameters [20,21]. A wide variety of TNT-mediated cargo transport has been described, including that of ions [22], miRNAs [23], cytosolic proteins [24], and prions [25–27], pathogens, such as viruses [28,29] and bacteria [30], and organelles, such as the endoplasmic reticulum [31], mitochondria [32], Golgi vesicles [31], endosomes [33], and lysosomes [33]. Due to the heterogeneous nature of these transient structures, Cancers 2020, 12, 2798 3 of 25 the terms “thin” and “thick” have been used to differentiate between TNTs with smaller (<0.7 µm) and larger (>0.7 µm) diameters, respectively [33,34]. “Thin” TNTs contain mainly F-actin, whereas “thick” TNTs tend to contain both actin and microtubules. TNTs may originate when two joined cells move apart from each other and remain connected by a thin membranous thread [35] or when filopodia-like protrusions from one cell dock with a nearby cell [19]. These models are not mutually exclusive. The underlying molecular mechanisms are not fully understood and likely depend on the specific cell type and its environment. Indeed, it appears that several types of TNT-like structures exist, including what are known as tumor microtubes, which are longer and seemingly more stable [36]. A better understanding of the molecular mechanisms underlying the formation and functions of TNTs is necessary to fully understand their physiological and pathological roles. Cellular stress, such as hypoxia or a low-serum hyperglycemic medium, seems to stimulate TNT formation. In particular, TNTs have been described in several cancer cell lines, as well as between cancer cells and stromal cells (reviewed in [37,38]). TNTs have also been found in primary human tumors [39–41]. It is becoming clear that TNTs facilitate a pathway of intercellular conduits in tumors [42] and may promote numerous malignant features such as invasion, growth, chemo- and radiotherapy resistance, and surgery-induced tumor growth [43–45]. Connexins localize to TNTs and regulate intercellular communication [44,46,47]. Wang et al. [46] identified TNT-mediated voltage-sensitive electrical coupling that was inhibited by the gap junction inhibitor meclofenamate. Cx43 immunoreactivity was observed at one end of the TNTs. Lock et al. [48] showed that TNTs between HeLa cells created cytosolic continuity allowing inositol triphosphate-mediated calcium signaling, but only if a pair of cells were both expressing Cx43. More recently, human immunodeficiency virus (HIV) infection was shown to stimulate TNT formation and connect human macrophages. These TNTs contained Cx43 at the tip, and treatment with the GJIC inhibitors 18-alpha-glycyrrhetinic acid or octanol prevented TNT-mediated cell–cell communication and viral cell–cell spread [49]. Although these inhibitors are nonspecific and affect several other targets and channel proteins, the findings suggest that Cx43 channels may be required for the functional establishment of TNTs that then allow the transfer of larger structures (including HIV) that, due to size constraints, do not pass through gap junctions. Notably, recent work suggests that Cx43 regulates TNT formation in normal human trabecular meshwork cells [50]. Yao et al. showed that Cx43 mediates TNT formation and mitochondrial transfer from human-induced pluripotent stem cell-derived mesenchymal stem cells and mouse epithelial cells [51]. Mechanistically, Fykerud et al. [52] showed that Cx43 remodeling was driving the formation of actin-based plasma membrane bridges (termed “mitotic nanotubes”) with adjacent cells during mitosis, facilitating vesicle transfer between cells. In vivo, glioma cells have been described to form a network of TNT-like structures called microtubes that contribute to tumor invasion in the brain [40]. Cx43-containing gap junctions within this network are thought to turn the tumor into a syncytium of interconnected cells that is highly resistant to radiation therapy, presumably by distributing calcium between cells to prevent apoptosis upon the radiation-induced release of intracellular calcium [40]. Brain tumors appear to employ a similar system of microtubes or TNTs to establish connections with neurons to “feed off the brain” [53]. Similarly, breast cancer cells that metastasize to the brain benefit from the formation of pseudo-tripartite synapses between cancer cells and glutamatergic neurons to promote metastatic growth [54,55]. Indeed, several studies have described the presence of TNTs in breast cancer cells and between breast cancer cells and macrophages, facilitating drug resistance, invasion, and metastasis [56–59]. A large body of data suggests that Cx43 is intricately linked to breast cancer development and may act as a tumor suppressor, particularly at early stages (see reviews [60–64]). However, increasing evidence suggests that some breast cancers overexpress Cx43, which may facilitate certain malignant and metastatic capabilities. Most notably, the expression of Cx43 was shown to create a syncytium between breast cancer cells and astrocytes, forming a cyclic guanosine monophosphate–adenosine monophosphate (cGAMP)-driven paracrine feedback loop that supports tumor growth and chemoresistance [65]. Cancers 2020, 12, 2798 4 of 25 Consequently, the GJIC inhibitor meclofenamate is under clinical trial for breast cancer patients with brain metastasis (https://clinicaltrials.gov/ct2/show/NCT02429570). Considering the emerging evidence linking Cx43 and TNTs, we sought to further explore this interplay in breast cancer cells by knocking out Cx43 expression using CRISPR-Cas9 technology. We show that Cx43 significantly regulates the formation and length of TNTs in breast cancer cells. Notably, TNT stimulation was more potent with conditioned medium derived from Cx43-expressing cells than with that from Cx43 knockout (KO) cells. Furthermore, several pathways that are regulated by Cx43 were identified to induce TNTs. Some of these pathways are novel TNT modulators. These pathways also regulate Cx43 function, and thus the direct association needs to be explored further. Specifically, Cx43 KO cells were more sensitive to drug-induced TNT formation than wild type (WT) cells. Further elucidation of the role of Cx43 in this setting is an exciting cutting-edge challenge that can provide mechanistic insight into the pro-metastatic features of Cx43 and the regulation of TNTs. 2. Results 2.1. GJA1 Knockout Reduces TNT Number and Length in BT549 Breast Cancer Cells To achieve the full and long-term elimination of Cx43 expression from cells, we developed an all-in-one transient CRISPR/Cas9 expression plasmid that targets GJA1 (encoding Cx43) and truncated forms of the protein such as GJA1-20k. The well-established PX458 plasmid allows the expression of CRISPR targeting sequences together with an EGFP-linked form of Cas9 to control transfection and fluorescence-activated cell sorting (FACS) selection [66]. To minimize putative off-target cleavage effects, we replaced the standard Cas9 with the VP12 high-fidelity Cas9 version [67]. Specific sequences targeting the coding exon of GJA1 were introduced. We transfected various cell lines and used FACS to isolate single cells transiently expressing Cas9-EGFP. Clones were screened by western blot to detect loss of Cx43 expression. This tool has allowed us to generate a wide variety of cancer cell lines lacking Cx43 expression to evaluate various phenotypic changes related to malignancy. We noticed that the triple-negative breast cancer cell line BT549 tends to display an extensive TNT network (Figure 1a) that stains for F-actin (Figure 1b). Time-lapse microscopy clearly highlighted this network and showed that the formation of these TNTs may occur through the protrusion of filopodia-like structures that connect to a distant cell and/or through two adjacent cells migrating apart from each other (Supplementary Video S1). We previously showed that this cell line expresses very high levels of full-length Cx43, as well as the 20-kDa truncated form GJA1-20k [10]. We efficiently eliminated GJA1 expression from this cell line using our CRISPR-Cas9 plasmid (Figure 1c). This led to a marked but not complete loss of GJIC, suggesting that other connexin channels may operate in this cell line (Figure S1). Morphologically, we observed a noticeable reduction in the number of TNTs in knockout (KO) cells compared with WT cells. Detailed quantification demonstrated that the loss of Cx43 expression significantly reduced the average numbers of TNTs, both 24 (Figure 1d) and 48 (Figure 1e) h after seeding. Next, we decided to deliberately stress cells by culturing them in low-serum hyperglycemic media, which was a condition previously shown to stimulate TNT formation [39]. Using such TNT-inductive media (im+), the number of TNTs noticeably but nonsignificantly increased (Figure 1f). The difference between WT and KO cells was significant under im+ conditions (Figure 1f), altogether suggesting that the response to this specific condition of cellular stress is similar in WT and KO cells. In addition to having noticeably thicker TNTs, BT549 cells also display some very thin TNTs at 20× magnification or higher (see white arrow, Figure 1a). After the inclusion of these “thin” TNTs via the analysis of 20× images, the difference between WT and KO cells remained similar and significant (Figure 1g), together with a significant (and expected) increase in the number of TNTs per cell (Figure 1h). Similar to our 10× analysis, our 20× analysis suggested that low-serum hyperglycemic conditions stimulated TNT formation, albeit nonsignificantly (Figure 1i). The results of the detailed 20× analysis seemed to Cancers 2020, 12, 2798 5 of 25 strongly correspond with our 10× analysis in terms of differences between cells and were not repeated for subsequent analyses. Since there was a significant decrease in the number of TNTs in KO cells, we also compared lengths. WeCancers found2020, a significant 12, x reduction in TNT length in the absence of Cx43 (Figure 1j). This was noticeable 5 of 24 but not significant at 48 h, possibly because cells are at a higher density at this stage (Figure 1k). noticeable but TNT-inductive not significant medium at 48appeared (im+) also h, possiblytobecause cells increase theare at a higher length density of TNTs at this in both WTstage (Figure and KO cells 1k). TNT-inductive (Figure 1l). medium (im+) also appeared to increase the length of TNTs in both WT and KO cells (Figure 1l). Figure 1. Tunneling nanotubes (TNTs) are regulated by connexin 43 (Cx43) in BT-549 cells. (a) Figure 1. Tunneling nanotubes (TNTs) are regulated by connexin 43 (Cx43) in BT-549 cells. Morphological examples of TNTs in BT549 cells (20×). The black arrow denotes a typical “thick” TNT, (a) Morphological examples of TNTs in BT549 cells (20×). The black arrow denotes a typical whereas the white arrow denotes a typical “thin” TNT that is often only visible at higher “thick” TNT, whereas the white arrow denotes a typical “thin” TNT that is often only visible at magnifications. (b) Confocal image of BT-549 cells stained with F-actin and the nuclear stain 4′,6- higher magnifications. (b) Confocal image of BT-549 cells stained with F-actin and the nuclear stain diamidino-2-phenylindole (DAPI) demonstrating a long, thin, F-actin–positive TNT. (c) Western blot 40 ,6-diamidino-2-phenylindole (DAPI) demonstrating a long, thin, F-actin–positive TNT. (c) Western with independently isolated samples of three wild type (WT) and three CRISPR-Cas9 Cx43 knockout (KO) BT549 cells showing the presence or absence (respectively) of both Cx43 (43 kDa) and the truncated GJA1-20k form (20 kDa), as well as some other weakly expressed truncated forms of Cx43. The blot was co-incubated with a vinculin antibody (124 kDa) as an internal loading control. (d and e) Average numbers of TNTs per cell in WT and Cx43 KO BT549 cells at 10× (counting thick TNTs) Cancers 2020, 12, 2798 6 of 25 blot with independently isolated samples of three wild type (WT) and three CRISPR-Cas9 Cx43 knockout (KO) BT549 cells showing the presence or absence (respectively) of both Cx43 (43 kDa) and the truncated GJA1-20k form (20 kDa), as well as some other weakly expressed truncated forms of Cx43. The blot was co-incubated with a vinculin antibody (124 kDa) as an internal loading control. (d,e) Average numbers of TNTs per cell in WT and Cx43 KO BT549 cells at 10× (counting thick TNTs) (d) 24 h after seeding, (e) 48 h after seeding with cells, and (f) 24 h after seeding followed by 24 h refreshed Cancers 2020,with 12, x the same medium or in a low-glucose, low-serum inductive medium (im+) to stimulate 6 of 24 TNT formation. (g) Average numbers of TNTs per cell in WT and Cx43-KO BT549 cells at 20× (counting thick(d) and24thin h after TNTs)seeding, 24 h (e) 48 seeding. after h after seeding with cells, and (h) Comparison (f) 24 of the h after seeding numbers of TNTsfollowed countedbyat24 h and 10× refreshed with the same medium or in a low-glucose, low-serum inductive medium 20×. (i) Average numbers of TNTs at 48 h with or without TNT-inductive medium (im+) for the last (im+) to stimulate 24 h. TNT (j–l) formation. Average lengths(g) Average of TNTsnumbers of TNTs at (j) 24 and (k) per48cell in WT h and and (l) 48 Cx43-KO h with BT549 cells or without at 20× with treatment (counting thick and thin TNTs) 24 h after seeding. (h) Comparison of the numbers of TNTs counted TNT-inductive medium (im+) for the last 24 h. The means of individual experiments (calculating the at 10× and 20×. (i) Average numbers of TNTs at 48 h with or without TNT-inductive medium (im+) average number of TNTs per cell [TNTs/cell] in n = 5 images) are highlighted and linked by color for the last 24 h. (j–l) Average lengths of TNTs at (j) 24 and (k) 48 h and (l) 48 h with or without and shape to facilitate the tracking of trends between individual experiments. Error bars denote the treatment with TNT-inductive medium (im+) for the last 24 h. The means of individual experiments mean(calculating and SEM of thethree independent average biological number of TNTs experiments. per cell [TNTs/cell] in Significance n = 5 images)was aretested with two-way highlighted and linkedt-test Student’s by colorforand single shapecomparisons and to facilitate the one-way tracking ANOVA of trends betweenwithindividual Tukey’s experiments. post-hoc analysis Error for multiple bars comparisons denote the mean < 0.05, (* pand SEM ** < 0.01, of pthree ns = not biological independent significant).experiments. Significance was tested with two-way Student’s t-test for single comparisons and one-way ANOVA with Tukey’s post-hoc 2.2. GJA1analysis Knockout Reducescomparisons for multiple the Number(*of p