Transmembrane voltage potential of somatic cells controls oncogene-mediated tumorigenesis at long-range

The microenvironment is increasingly recognized as a crucial aspect of cancer. In contrast and complement to the field's focus on biochemical factors and extracellular matrix, we characterize a novel aspect of host:tumor interaction – endogenous bioelectric signals among non-excitable somatic cells. Extending prior work focused on the bioelectric state of cancer cells themselves, we show for the first time that the resting potentials of distant cells are critical for oncogene-dependent tumorigenesis. In the Xenopus laevis tadpole model, we used human oncogenes such as mutant KRAS to drive formation of tumor-like structures that exhibited overproliferation, increased nuclear size, hypoxia, acidity, and leukocyte attraction. Remarkably, misexpression of hyperpolarizing ion channels at distant sites within the tadpole significantly reduced the incidence of these tumors. The suppression of tumorigenesis could also be achieved by hyperpolarization using native CLIC1 chloride channels, suggesting a treatment modality not requiring gene therapy. Using a dominant negative approach, we implicate HDAC1 as the mechanism by which resting potential changes affect downstream cell behaviors. Based on published data on the voltage-mediated changes of butyrate flux through the SLC5A8 transporter, we present a model linking resting potentials of host cells to the ability of oncogenes to initiate tumorigenesis. Antibiotic data suggest that the relevant butyrate is generated by a native bacterial species, identifying a novel link between the microbiome and cancer that is mediated by alterations in bioelectric signaling.

Keywords: cancer, tumors, ion channels, bioelectricity, resting potential, Vmem, transmembrane potential, HDAC, butyrate, microenvironment, bacteria

Normal embryonic development, as well as repair and dynamic maintenance of complex structures throughout the lifespan, both depends upon a set of signals that keeps individual cell activities orchestrated toward the large-scale anatomical goals of the host. Morphogenesis and remodeling can be challenged by cancer, which can be viewed as a process in which cells escape or become isolated from the normally tight morphogenetic control of the organism [1-6]. To achieve robust development, signaling pathways need to affect their target cells with sufficient spatiotemporal resolution to integrate organ sculpting, anatomical polarity, tissue identity, and growth rates appropriate to the large-scale order maintenance within the body. In the context of cancer, it has been well documented that healthy neighboring cells help to stabilize aberrant cell behavior [7-9] to control tumorigenesis by tissue-level organization that adheres to the proper patterning needs of the host [10-14]. Indeed, fascinating classical and recent data show that actively patterning environments, such as embryos and regenerating amphibian limbs, can normalize and reprogram tumors [15-31]. Thus, in addition to any cell-autonomous properties that may have gone awry in cancer cells, it is crucial to understand the non-cell-autonomous patterning signals that may be exploited to prevent and treat cancer.

Endogenous bioelectric signaling among all (not just excitable) cell types is one component of the microenvironment, and is now known to mediate instructive information for large-scale pattern formation [32, 33]. Bioelectrical processes, such as extracellular electric fields and transmembrane resting potentials produced by ion channel and pump proteins, regulate cell activity [34-36]. In particular, proliferation, differentiation, migration, and apoptosis are all important cell behaviors relevant to cancer that are guided partly by bioelectric signals, such as resting voltage potential (Vmem) in non-excitable cells [37-39]. It was recognized long ago that the electrical properties of mature tumors differ from that of healthy tissue [40-42], and we recently showed that monitoring of a distinctly depolarized resting potential allows early detection of tumors in vivo. Moreover, forced hyperpolarization via a number of different channels prevents those cells from forming tumors [43], revealing resting potential as an instructive cue and not merely a marker [39]. Indeed, a number of ion channels are now known to be bona fide oncogenes [44-54], and ion channel blockers are an important area for cancer drug development [55-58].

However, it is important to note that ion channels are not simply molecular markers or targets within the cancer cells themselves. Gradients of resting potential are a long-range, global system for exerting patterning control, and thus a tractable candidate for manipulation of the crosstalk by which microenvironment suppresses aberrant cell behavior [59]. For example, selective depolarization of glycine-gated chloride channel-expressing cells in the Xenopus model results in a metastatic conversion of melanocytes, which over-proliferate, acquire an arborized shape, and invade blood vessels and soft tissues in an MMP-dependent manner [46, 60, 61]. This effect occurs at considerable distance, via a serotonergic pathway, and demonstrates how disruption of long-range endogenous physiological signaling can activate a cancer phenotype in the absence of carcinogen exposure or DNA damage. The data suggest that bioelectric gradients are a fascinating new aspect of cancer:host interaction, and that the notion of microenvironment may need to be expanded to account for long-range interactions mediated by Vmem changes throughout tissues.

The existing literature on roles of ionic signaling in cancer leaves open a number of fundamental questions. What role might remote Vmem modulation play in oncogene-induced carcinogenesis? While distant changes of resting potential can trigger a metastatic conversion [60, 61], it is not known whether or how such signaling plays a role when canonical oncogenes initiate tumorigenesis. A better understanding of the microenvironment and long-range aspects of aberrant growth control by endogenous developmental patterning mechanisms would have significant implications for design of novel therapeutic approaches to prevent and reprogram cancer.

Here, we use mRNA encoding human tumor inducers (Gli1, XRel3, KRAS) in Xenopus laevis embryos to initiate growth of tumor like structures (ITLS) that highly resemble classic tumors. Remarkably, forced hyperpolarization (by misexpression of specific ion channels) is able to suppress ITLS formation, despite high levels of oncogene protein, even when the hyperpolarized cells are at a considerable distance from the oncogene-expressing tumor site. The suppression effect can also be exerted by native chloride intracellular channel 1 (CLIC1)-mediated hyperpolarization, revealing an endogenous target for bioelectric control of abnormal growth that does not require transgene expression. We show that the suppression effects of distant hyperpolarization are mediated by voltage control of a butyrate and histone deacetylase mechanism. Together, these data reveal the first mechanistic details of Vmem as a powerful, tractable regulator of long-range signaling between cancer cells and the host and suggest several new entry points for biomedical strategies.

To study the role of bioelectric events to oncogene-mediated tumorigenesis, we took advantage of Xenopus laevis embryos – a model system that is ideal for molecular biophysics approaches and also has been increasingly used for cancer research[43, 61-66]. We injected mRNAs encoding Xrel3 and human KRASG12D into Xenopus embryos; these oncogenes are known to cause morphologically apparent ITLS in up to 50% of injected embryos without any other developmental defects. ITLS have previously been characterized as having some of the hallmarks of human tumors – increased proliferation, vasculature attraction, lack of differentiation, invasiveness and transplantability to healthy recipients [43, 67-69]. We began by further investigating the pathology of the affected tissue to confirm and expand the relevance of Xenopus as a medically-relevant cancer model.

Clinically-relevant tumors exhibit a surplus of proliferative capacity [70]. To characterize the cell proliferation dynamics in ITLS loci, we monitored dynamic patterns of cell cycle progression in ITLS and unaffected regions using FUCCI (fluorescent ubiquitination-based cell cycle indicator) pair: mKO2-Cdt1 and mAG-Geminin [71, 72]. The accumulation of these two cell cycle regulators – in nuclei of transfected cells – in a mutually exclusive manner between G1(Cdt1) and G2/S/M (Geminin) phases, allowed us to monitor spatial dynamics of cell cycle progression: while the number of ITLS cells in G1 and control cells in G1 and G2/S/M phases are statistically the same, twice as many cells in G2/S/M phases were present in ITLS by comparison (N=8 per category, ANOVA, P<0.001) (Fig. 1A).

Next, we examined the extent of hypoxia within ITLS loci, since hypoxia is known to be present in tumors [73-75]. Treatment of ITLS bearing embryos (N=7) with the hypoxia marker pimonidazole revealed a 12-fold increase (t-test; p<0.0001) for immunoperoxidase staining for hypoxia (Fig. 1B). This is consistent with data obtained from pimonidazole-based qualitative and quantitative assessment of tumor hypoxia, where an increase in the frequency of detecting pimonidazole adducts is reported [76-78].

We next examined histological features of the ITLS using hematoxylin and eosin (H&E) [79]. Cells within the ITLS exhibited disorganized growth patterns, including misplaced mesodermal and endodermal cells, and mesodermal cells that were larger than those present in unperturbed regions, revealing features consistent with neoplasia (Fig. 1C).

We next examined a change in nuclear size, which is another well-known aspect of tumor tissue [80, 81]. Computer-aided image analysis of the area of 2371 nuclei each from 16 ITLS and 16 control regions of oncogene-injected tadpoles showed on average a 10% increase in nuclear size of ITLS, mirroring similar phenomenon observed between different stages of human breast cancer cells [82] (Fig. 1D). Another marker of tumorigenesis is the abnormal acidity of tumor tissue [83, 84]. Using BCECF, a dual excitation pH indicator, fluorescence intensity ratios at two different wavelengths of ITLS and control regions revealed that ITLS have a more acidic (pH 6.6) intracellular environment compared to that of control regions, which have a pH average of 7.41 (N=8 for both treatments, t-test, P<0.05) (Fig. 1E).

Finally, to investigate host immune response to tumor formation, immunohistochemistry with anti-XL2 (anti-Xenopus leukocytes) was performed on cross sections of tissue bearing ITLS. We observed recruitment of leukocytes to the site of transformed cells (Fig 1F), consistent with the known role of leukocyte migration as an anti-tumor immune response [85].

Taken together, these characteristics of ITLS are consistent with previous findings of increased proliferation within ITLS, ability to form in internal tissues (not just the epidermis), and propensity to attract vasculature[43], and demonstrate the relevance of human oncogene-induced tumors in Xenopus for understanding the basic mechanisms of carcinogenesis in vivo.

To test the hypothesis that hyperpolarization affects oncogenic transformation of cells located at a distance, we performed functional experiments using sets of oncogenes and hyperpolarizing reagents. The use of multiple oncogenes and hyperpolarizing agents allowed us to confirm that there is a generalized suppression effect due to changes in Vmem, one that is not tied to a particular channel protein or ion. We used Xrel3 tagged with tdTomato to track the presence of oncogenic protein (Fig. 2A, I-III). To modulate Vmem, we used Kv1.5 (Potassium voltage-gated channel, shaker-related subfamily, member 5) [86], a well-characterized hyperpolarizing channel, whose overexpression in our model system has been used to alter developmental patterning [87].

To probe the effects of long-range hyperpolarization on tumor formation, we injected mRNA's encoding Xrel3-tdTomato and Kv1.5-GFP3 in two separate blastomeres of 16-cell stage embryos, and scored for the presence of ITLS and hyperpolarizing channel/oncogene protein by stage 34. Compared to embryos receiving only oncogene injections, embryos with a remote hyperpolarized region (Fig. 2B, IV&V) often displayed lack of morphologically apparent ITLS (Fig. 2B, I) despite strong presence Xrel3-2A-tdTomato protein (Fig. 2B, II&III). Kv1.5-induced long-distance hyperpolarization was also able to significantly suppress Xrel3 and Gli1 ITLS formation by 34.9% (N=225; t-test, P<0.05) and 29.4% (N=292; t-test, P<0.05), respectively (Fig. 2C), suggesting the effect is not specific for just one type of oncogene. Treatment with high chloride (whose uptake by native Cl− channels hyperpolarizes cells) resulted in a 37.5% decrease (N=349; t-test P<0.05) in KRASG12D ITLS, demonstrating that the suppressive effect is not specific for Kv1.5 nor for potassium, as mediators of the crucial Vmem change (Fig. 2C).

Collectively, these data reveal a long-range signal that is initiated by hyperpolarization and has influence over tumorigenesis at remote sites in the tadpole. Moreover, these results demonstrate that ITLS suppression through Vmem modulation is possible to achieve by taking advantage of endogenous channels that are not necessarily located within oncogene expressing cells.

Having shown a functional role of long-range hyperpolarization in the context of tumorigenesis, we proceeded to characterize the native channel responsible for chloride-based inhibition of ITLS formation. We performed an inverse drug screen [88] with chloride channel blockers to see which ones would abrogate the chloride-dependent ITLS suppression: Anthracene-9-Carboxylic Acid (ACA), inadyloxyacetic acid 94 (IAA), and 5-Nitro-2-(3-phenylpropylamino) benzoic acid (NPPB) [89-92] (Fig. 3A). After the injection of KRASG12D and treatment with high chloride (N=164), ITLS incidence dropped significantly (χ2, P<0.05) compared to their counterparts in a low chloride concentration (N=162). This anti-tumor effect of high chloride was completely abolished with IAA treatment (N=186; χ2, P<0.05) (Fig 3B), while the other two inhibitors did not significantly affect the rate of ITLS incidence. We then proceeded to validate this result using molecular reagents targeting IAA-sensitive channels.

A known target of IAA is the intracellular chloride channel 1 (CLIC1) [93]. Given its pharmacological profile and ubiquitous presence during every stage of the Xenopus laevis development [94], we focused on CLIC1 in this study. To test the hypothesis that CLIC1-mediated hyperpolarization affects oncogenic transformation of cells, we performed functional experiments involving CLIC1 (wildtype and mutant) and KRASG12D injections. Wildtype CLIC1 was tagged with GFP, while a dominant negative form of CLIC1 was generated by mutating cysteine 24 to serine, in order to form an inactive multimer channel [95], and tagged it with tdTomato to track localization of the protein product. mRNA's encoding CLIC1-2A-GFP3/KRASG12D-tdTomato (Fig. 4B, I&II) and CLIC1C24S-2A-tdTomato/KRASG12D-GFP3 (Fig. 4B, III&IV) were injected in two different blastomeres of 16-cell stage embryos, treated with high Cl−, and scored for the presence of ITLS and CLIC1 channel/oncogene proteins. As a control to which CLIC injected embryos can be compared, KRASG12D-only injected embryos were also treated with high Cl− (N=372), resulting in 31.5% fold decrease in ITLS incidence when compared to KRASG12D-only injected embryos (N=341). The rate of ITLS suppression was further amplified (44.3% suppression) when the wildtype CLIC1 was overexpressed non-locally in oncogene-injected embryos that were then raised in 70mM Cl− (N=364). In contrast, overexpression of the mutant CLIC1 (N=336) effectively blocked ITLS suppression despite the presence of high Cl− (Fig 4C). Together, these data suggest a role for native CLIC1 in mediating hyperpolarization-induced long-range ITLS suppression.

Bioelectric signals are transduced into transcriptional and epigenetics responses via a number of second-messenger pathways. In order to determine the transduction mechanism by which hyperpolarization inhibits tumorigenesis, we conducted a pharmacological suppression screen [63] and molecular-genetic loss-of-function of several well-characterized candidate pathways that have been shown to mediate the actions of Vmem change in other patterning contexts [96]. Our strategy was to block, one at a time, the candidate transduction mechanisms and see which ones abrogated the ability of hyperpolarization to reduce tumor incidence. Targeted pathways include: movement of serotonin through the Vmem-dependent serotonin transporter SERT [97] (inhibited via 10 µM fluoxetine); Vmem-guided transport of signaling molecules through gap junctions [98, 99] (disrupted by 0.5 ng injection of a dominant negative connexin H7); voltage gated calcium channels [100, 101] (blocked via 0.1 mM cadmium chloride), and Vmem-modulated transport of small molecule inhibitors of histone deacetylase 1 [102-104] (targeted by 0.5ng injection of mRNA encoding a dominant negative HDAC1) (Fig 5A). HDAC1 is an especially attractive target because it is known to control levels of histone acetylation, which determines cell cycle progression, rates of proliferation, apoptosis and differentiation of cancer cells [105-107]. Administering oncogene, hyperpolarizing agent, and Vmem-sensing pathway disruptors (Fig 5B) showed that blocking HDAC1 activity had the most significant effect (χ2, P<0.05) on CLIC1-mediated suppression of tumors: the rate of ITLS formation in DN-HDAC1 injected embryos in high Cl− was similar to that of oncogene-only injected embryos - introducing DN-HDAC1 non-locally to the site of KRASG12D increased ITLS formation incidence by 35.6% (χ2, P<0.05) (Fig.

5C), suggesting a role for histone deacetylase 1 activity controlling cell behavior at a distance.

Having previously identified butyrate as an endogenous HDAC inhibitor involved in local ITLS suppression [43], we sought to identify the source of butyrate in this system. Reasoning that it would be produced by native bacteria, we conducted a screen for antibiotics that target butyrate producing bacteria. We first tested novobiocin, which has been shown to be effective against gram-positive bacteria [108]. At 78 μM, novobiocin sodium salt treatment significantly increased ITLS incidence by 32.1% (χ2, P<0.01) (Fig. 6). It has been determined in other studies that the mechanism of action for novobiocin involves the inhibition of the bacterial DNA gyrase [109, 110]. Given also nalidixic acid's well-known inhibition of DNA gyrase [111], we tested its effect (at 43 μM) on ITLS incidence. While an 8.2% increase in ITLS was observed, it was not statistically significantly different from ITLS incidence in oncogene-only injected embryos (Fig. 6). Finally, a cocktail of antibiotics (gentamicin, metrodinazole, vancomycin hydrochloride, and clindamycin hydrochloride, all at 20 μM concentration) targeting gram-positive bacteria resulted in a significant increase in ITLS incidence by 26.2% (χ2, P<0.05) (Fig. 6). Bulk analysis of embryos exposed to this cocktail of antibiotics does indeed show a significant drop in total butyrate concentration (data not shown).

Together, these data suggest that the control of tumorigenesis by Vmem is mediated by HDAC1, via butyrate derived from gram-positive bacteria.

It has been well-documented that bioelectric properties control individual cell behaviors that are highly relevant to the processes of cancer [35, 36, 96, 112, 113]. For example, electric fields generated by voltage-gated sodium channels (VGSC) provide motility cues to highly metastatic rat prostate cells [114], and many cancer cells are galvanotactic [115-119]. Distinct from extracellular electric fields, resting membrane potentials established by ion channels and pumps control cell proliferation, transformation, and growth in number of tumor forming cells [120-125]. Moreover, the actions of these translocators have allowed for the development of predictive physiological markers [43, 61, 125] and control of proliferative and metastatic behaviors [39, 46, 60, 126]. Here, we show for the first time that resting potentials in remote cells have a profound influence on whether or not a cell expressing human oncogenes will go on to form a tumor in vivo.

The induced tumor like structures (ITLS) were shown to exhibit striking similarities with classic tumors, including increased proliferation as monitored by visualizing the cell cycle dynamics, disorganization of the normal developmental architecture revealed by H&E stains, increased hypoxia, increased nuclear size, acidic intracellular microenvironment, and ability to illicit innate immune response (Fig. 1A-F). Given the increased recognition of bioelectric signals as endogenous patterning cues that keep cellular activities orchestrated towards anatomical goals [32, 33, 127], our data confirm Xenopus embryos as a tractable model system in which to investigate the role of bioelectric signals in host:cancer interaction, and in which to ask fundamental questions about cancer as a disorder of developmental cues [5, 6, 10, 12, 128, 129].

The spatial distance over which carcinogenesis can be predicted and controlled have been addressed in a few studies. One example is illustrated by the ability to assess cancer risk of readily-inaccessible organs using accessible surrogate sites – functionally or anatomically similar, but not necessarily contiguous with the target organ – that make up the cancer field [130-132]. Another is the activation of melanoma-like transformation in the melanocyte population [60] by Vmem modulation of even a few instructor cells at considerable distance from the melanocytes themselves [61]. However, to our knowledge this is the first time that modulation of distant cells' resting potentials has been shown to impact the formation of discrete tumor foci. It is tempting to speculate that the long-range connections are bi-directional: not only does remote electrical state of tissue matter for tumor growth, but perhaps tumors also emit bioelectrical information that could be detectable at a distance. This is compatible with classical data [131, 133-135] and will be tested using molecular reagents in the future.

In our previous work [43], we showed that hyperpolarized cells were themselves resistant to transformation [136-138]. To analyze long-range effects of hyperpolarization, we established an assay in which Xenopus 16-cell embryos were injected with oncogene in one blastomere and subjected to a hyperpolarizing treatment, non-locally, via ectopic-expression of hyperpolarization channel or manipulation of ionic contents of the media. Remarkably, the data show that hyperpolarization significantly suppresses the formation of ITLS, regardless of which oncogene was used (Fig. 2B, C), even when the oncogene-bearing cells and the cells that have been hyperpolarized are on opposite ends of the body. Complete suppression of ITLS was not achieved, because our reagents had to be tittered down to avoid perturbing the normal developmental processes of the organism. Future work is needed to refine the strategy for optimal management of the bioelectric crosstalk that goes awry in cancer; it is likely that a more nuanced strategy, which manages spatial relationships at higher resolution, will be required in clinical practice. Another important area for future investigation is slow time-dependent variability of Vmem, as we only explored continuously-acting reagents and it is possible that important patterning information is encoded in the time profile of slowly-changing bioelectrical states [37, 139-142]. Optogenetics – the use of light to regulate ion channels with high spatio-temporal specificity - [143, 144] is a promising technology for both of these directions.

Resting potential changes can result from genomic, transcriptional, or post-translation control of ion channels and pumps. Importantly, we obtained tumor suppression using either Cl− or K+-based hyperpolarization, just as we showed previously for cell-autonomous tumorigenesis and metastatic activation by depolarization [43, 46, 60, 61]. This rules out signaling pathways triggered by specific ion channels or even limited to specific ion types. Thus, in complement to the gene-focused idea of specific ion channels being intrinsic oncogenes or tumor suppressors [44, 48, 145], we suggest that a systems-level physiological property – resting potential – can be a powerful causal factor in regulating these processes.

A large body of literature reveals differential ion channel and transporter expression between tumor cells and their untransformed counterparts [146-150]. These native ion translocators are important not only as markers [52, 145, 151-153] but also as targets for interventions designed to manage the bioelectric state of cancer and surrounding tissue without needing gene therapy with heterologous channels. Exploiting native channels may allow manipulation of bioelectric signals merely by changing ionic composition of the cellular environment; indeed, we observed that hyperpolarization by high Cl− reduces ITLS incidences (Fig. 3B) without additional channel misexpression. The alteration of tumorigenic transformation by ionic conditions in the medium is consistent with previous data showing an association between the (depolarizing) level of sodium and cancer [154-156], the control of cell differentiation by sodium levels [157-159] and pH [160], and more recent work on stem cell reprogramming by acid baths and streptolysin-mediated membrane permeability change [161]. Reprogramming of cell state by physiological cues is certainly of relevance to cancer [162-165]. However, two important novel aspects of our data are the focus on Vmem per se, not specific ions as obligate signals, and the non-local nature of the suppression mechanism.

A pharmacological loss-of-function experiment using chloride channel blockers (Fig. 3A) and a gene-specific dominant negative (Fig. 4) implicated chloride intracellular chloride channel 1 (CLIC1) as a likely candidate for the channel required for hyperpolarization in the long-range ITLS suppression. Roles for CLIC1 in carcinogenesis – albeit cell autonomous – have been shown previously in several studies. A search for ion channels in the cancer profiling database, Oncomine, turns up CLIC1 as one of the most upregulated genes [166], and chloride current associated with increased CLIC1 expression is present in progenitor cells isolated from human glioblastomas, and is responsible for promoting proliferation, clonogenicity, and tumorigenic capacity [167]. Channels like this are of high interest as a therapeutic modality [56], but it should be noted that it is not enough to look for blockers (a standard loss-of-function genetic strategy) – the key is to modulate Vmem, which may mean opening or closing specific channels depending on the cells' surrounding milieu and its ion gradients. An understanding of Vmem as a regulatory agent may also explain the anti-cancer activity of a number of agents such as ivermectin [57], bafilomycin [168, 169], and salinomycin [170].