When left does not seem right: epigenetic and bioelectric differences between left- and right-sided breast cancer

In the light of this, DNA of the three P1 L/R paired xenograft tumors was selected for Reduced Restricted Bisulfite Sequencing (RRBS) assays. After establishing an arbitrary cut-off of 10% absolute DM, (see section ‘Methods’ for more details), we decided to discard inflammation-response genes, since they can increase their expression after surgical manipulation (Russell et al. 2015). Besides, after selecting one CpG site per gene (with the highest DM), 1288 genes remained (Additional file 1: Table S1), on which gene enrichment analyses were performed by Metascape. The analyses revealed cellular (GO:0032989) and tissue morphogenesis (GO:0048729), several developmental pathways (e.g., GO:0001655; GO:0021700; GO:0021675), morphogenesis of a branching structures (GO:0001763) and chemical synaptic transmission (GO:0007268) as the main GO biological processes (Fig. 1B). Furthermore, Genomic Region Enrichment Analyses (GREAT) (McLean et al. 2010) showed that the overall differences were mainly involved in the GO biological processes embryonic camera-type eye development (GO:0031076), epidermis development (GO:0008544), mammary gland development (GO:0030879), and regulation of neuronal synaptic plasticity (GO:0048168) (data not shown).

So very interestingly, the generated animal model revealed biological processes following what we previously had found in human in-silico data of breast tumors, indicating that L/R differences were consistently associated with embryogenic and neuronal features.

The role of electrochemical gradients in neurons is well known. However, an increasing amount of literature is revealing the role of electrochemical gradients in the regulation of diverse functions of non-neuronal cells, including morphogenesis of numerous embryonic and adult structures (Bates 2015; Harris 2021; Levin et al. 2017). Consistently, the GO term ion transport had appeared among the in-silico main enriched pathways. We therefore decided to search whether ion-channel genes (ICH) were included in our DM lists.

We found 33 ICH genes matches (using as reference the Human Gene Nomenclature Committee (HGNC) ion channel list (https://www.genenames.org/data/genegroup/#!/group/177) among the in-vivo DM and 77 ICH genes in the in-silico data (Table 1). In Table 1, we have indicated for each gene an increased methylation in L-sided tumors as “more methylated left” -MML-; and decreased methylation in L as “more methylated right” -MMR-. Even though slighty, we see a tendency for decreased percentage of hyperpolarizing channels among the MMR genes, as compared to MML ones (53.8% vs 66.6%, OR 0.5, 95%CI 0.11–2.13 in the in-vivo experiment; 49.01% vs 50%, OR 0.96, 95%CI 0.35–2.59 in the in-silico experiment) (Fig. 2). Even though the differences did not reach statistical significance, it suggested a possible non-stochastic pattern.

We reasoned that it could be possible that the L/R breast tumor differences occurred at bioelectric levels, preserving a consistent voltage change direction probably in a non-specific gene manner.Table 1L–R differential methylation of ion channels in in-vivo and in-silico studiesMore methylated rightMore methylated leftGene symbol*Channel functionGene symbol*Channel functionIon channels from in-vivo assays CHRNEDepolarizationASIC2Depolarization HTR1ADepolarizationCACNA1ADepolarization PIEZO2DepolarizationCACNA2D2Depolarization RYR3DepolarizationCATSPERDDepolarization TRPC7DepolarizationPKD1L1Depolarization TRPM8DepolarizationTRPM4Depolarization ANO2HyperpolarizationANO3Hyperpolarization ATP1A3HyperpolarizationANO5Hyperpolarization ATP6V1C2HyperpolarizationCLCN1Hyperpolarization ATP6V1HHyperpolarizationCLIC5Hyperpolarization GABBR1HyperpolarizationGABBR2Hyperpolarization KCNA7HyperpolarizationGABRA5Hyperpolarization KCNB1HyperpolarizationGABRDHyperpolarizationGABRG1HyperpolarizationKCNH2HyperpolarizationKCNIP3HyperpolarizationKCNJ18HyperpolarizationKCNK9HyperpolarizationKCNN1HyperpolarizationVDAC2HyperpolarizationIon channels from in-silico assays CACNA1DDepolarizationCACNA2D2Depolarization CACNA1HDepolarizationCACNA2D4Depolarization CACNA1IDepolarizationCHRNA1Depolarization CACNB2DepolarizationCNGA1Depolarization CACNG4DepolarizationHCN1Depolarization CACNG6DepolarizationP2RX4Depolarization CHRNA6DepolarizationPKD2Depolarization CHRNB1DepolarizationRYR1Depolarization CHRNB2DepolarizationSCN11ADepolarization CNGA3DepolarizationSCNN1ADepolarization GRIA1DepolarizationTRPM3Depolarization HVCN1DepolarizationCFTRHyperpolarization LRRC8DDepolarizationCLCN1Hyperpolarization LRRC8EDepolarizationCLCNKBHyperpolarization MCOLN2DepolarizationKCNA1Hyperpolarization MCOLN3DepolarizationKCNA5Hyperpolarization PKD2L2DepolarizationKCNC3Hyperpolarization SCN3BDepolarizationKCNH7Hyperpo

larization SCN9ADepolarizationKCNH8Hyperpolarization SCNN1GDepolarizationKCNJ9Hyperpolarization TRPA1DepolarizationKCNMA1Hyperpolarization TRPC2DepolarizationKCNS2Hyperpolarization TRPC3Depolarization TRPM2Depolarization TRPM6Depolarization TRPV3Depolarization TRPV4Depolarization TRPV6Depolarization CLIC3Hyperpolarization GABRA2Hyperpolarization GABRG1Hyperpolarization GABRPHyperpolarization GABRR1Hyperpolarization GLRA3Hyperpolarization GLRBHyperpolarization KCNA2Hyperpolarization KCNA6Hyperpolarization KCNA7Hyperpolarization KCNG1Hyperpolarization KCNH1Hyperpolarization KCNH4Hyperpolarization KCNH6Hyperpolarization KCNJ15Hyperpolarization KCNJ2Hyperpolarization KCNJ5Hyperpolarization KCNJ6Hyperpolarization KCNK3Hyperpolarization KCNK4Hyperpolarization KCNK5Hyperpolarization KCNN4Hyperpolarization KCNQ3Hyperpolarization KCNQ4Hyperpolarization KCNS1Hyperpolarization KCNS3Hyperpolarization KCNT2Hyperpolarization*HUGO Gene Nomenclature Committee (HGNC)Fig. 2L–R comparison of methylated hyper vs depolarizing ion channels. Data from in-vivo and in-silico analyses. Tendency shows in both approaches R-sided tumors with subtle decreased proportion of methylation in hyperpolarizing channels, as compared to L ones (53.8% vs 66.6% in-vivo; 49% vs 50% in-silico)

Taken together so far, L/R epithelial carcinomas presented methylation differences in genes involved in embryogenic and neuronal processes, suggesting a pattern in ion channel genes, by decreased DNA methylation of hyperpolarizing genes on R-sided tumors. Inferring from our xenograft experiments, we could discard that these differences were original of the tissue where the tumor started (since mice-genomic interferences had been filtered), which allowed us to adventure that they were acquired during the tumor progression. To establish the environmental role in acquiring bioelectric and DNA methylation differences, we further continued with in-vitro studies.

To establish how (or if) the mammary gland microenvironments contributed to the L/R voltage differences, we set up an in-vitro model where cellular extracts of healthy L/R human mammary tissue were used to induce changes in cultured cells. From surgical reductions of healthy L/R mammary glands, we included samples from four women (median age 34) in this study. Paired L/R cellular extracts from one female donor (W1) were first used to treat MDA-MB-231 breast cancer cells for 5 days, and afterward measure Ca2+ concentration with a calcium fluorophore by confocal microscopy. We counted between 50 and 100 cells for each condition and confirmed that the L/R W1 extracts had a different effect on cells, showing an increased Ca2+ concentration in the L-treated cells (Fig. 3) (unpaired T-test with Welch’s correction, p < 0.003). With this, we confirmed that the developed in-vitro model was sensitive and reproducible to test the effect of L/R extracts on cellular electricity. Furthermore, the generated difference in Ca2+ concentration suggested that L-treated cells had relatively depolarized the membrane.Fig. 3L/R Ca2+ concentration comparison. MDA-MB-231 cells treated with L/R normal breast extracts from donor W1 and measurement of intracellular calcium concentration with Fluo3 AM. A Representative calcium fluorescence in two-dimensional (a and b) and three-dimensional images-surface plot- (a′ and b′). Pseudo-color from black to red represents low to high intracellular calcium concentration, respectively. B Fluorescence quantification was performed on still images (×20 magnification), by calculating the fluorescence mean of 50–100 cells fields. The analysis reveals a significant increment of intracellular calcium concentration in the L-treated cells.

Values represent the mean ± SEM from 3 independent experiments, 50–100 cells were analyzed per experiment (unpaired T-test with Welch’s correction, p < 0.003). A.U artificial units

To understand the underlying Ca2+ dynamics, we designed a stepwise approach, by which independent wells were each day incubated with Fluo3 and the fluorescence values were calculated as means of still images of 15–70 cells per field. By this, we observed that R-treated cells decreased the Ca2+ concentration at early times (at day 1), as compared to L-treated and to untreated control cells (Kruskal–Wallis test for: L/R differences: p < 0.0001;Control/R differences: p < 0.001). On the contrary, L-treated cells maintained a similar initial Ca2+ concentration and did not differ from the control cells. Even though some fluctuations occured during day 2–3–4, the L/R significant difference was maintained along the 5 days (Kruskal–Wallis test: Control/L differences: day 1, 2, 3 and 5: ns, day 4 p < 0.01) (Fig. 4A, B).Fig. 4Calcium dynamics in MDA-MB231 cells treated with L- and R-extracts. A Mean of intracellular Fluo-3 fluorescence measured day-by-day in a Nikon TE300 Inverted Fluorescence Phase Contrast Microscope, in L, R and Control conditions. Still Images of cohorts of 15–75 cells per condition were analyzed with Microsoft Excel and Image J (National Institutes of Health, USA, RRID:SCR_003070) during 5 days. Fluorescence data was calculated as mean/area. As can be observed, in the R-treated cells intracellular Ca2+ concentration decreases at day 1, (as compared to controls and to L-treated cells) and this difference is maintained along the 5 days (Kruskal–Wallis test for: L/R differences: p < 0.0001;Control/R differences: p < 0.001, indicated by asterisks). While L-treated cells, despite some fluctuations during days 2–3–4, do not significantly change from the untreated cells (Kruskal–Wallis test: Control/L differences: day 1, 2, 3 and 5: ns, day 4 p < 0.01).

B The Δ fluorescence of A (initial fluorescence − final fluorescence) was measured at time 0 and 1440 min (24 h). This shows how R-treated cells decreased the intracellular Ca2+ concentration, which is not observed in the other conditions. C Representative Ca2+ fluorescence plot in real time, of single cells in L- and R-conditions, during 3 days. D Representative Ca2+ fluorescence plot in real time and representative single cell spatiotemporal fluorescence changes in the first 1440 min (24 h). Representative pictures of different times are shown (0, 60 and 550 min). The pseudocolor from black to red represents low to high concentration, respectively. E Scatter plot of real-time fluorescence in L- and R-conditions shown in C. F Plots of slopes of Ca2+ increments during the first 360 min (6 h) shown in C. The L-slope is markedly greater than R-slope

In addition, we incubated the cells with Fluo3 together with the L/R extracts and collected images with a real-time approach every 15 min during 3 days (we were not able to extend the experiment to 5 days, since cells did not survive longer in a reduced chamber and glass-surface) using a Zeiss-Axio-Observer microscope. This experiment confirmed how L- and R-treated cells show differential Ca2+ uptake dynamics during the first hours of treatment and maintain a differential concentration along the measured days (Fig. 4C, E). Focusing on the first 24 h (Fig. 4D), it can be observed how the Ca2+ dynamism slopes differ, since L-treatment provoked a rapid and strike increment in the Ca2+ fluorescence while R-treated cells respond slower showing a lower fluorescence increment (Fig. 4F).

The voltage-sensitive dye specific for plasma membrane potential (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Delta {\psi }_{p}$$\end{document}Δψp) Bis-(1,3-Dibutylbarbituric Acid) Trimethine Oxonol (DiBAC4(3)) was used to measure the effect of L/R mammary tissue extracts on MDA-MB-231. The dye (negatively charged) accumulates into depolarized cells. It has been previously established by others that this method is reliable for bioelectric studies in non-neuronal cancer cells which are known to be less polarized than normal cells (Bonzanni et al. 2020).

The L/R extracts of three female donors (W2, W3, W4) were used to perform treatment replications. After 5 days of treatment, the DiBAC4(3) signal was measured by flow-cytometry (Raw data available in Additional file 3: Table S3). Interestingly, the L-treated cells displayed an increment in the fluorescence signal, indicating a less polarized state, in line with our previous assumption (L/R-fluorescence ratio, One-sample T-test with hypothetical R value = 1, p = 0.04, Fig. 5A). This observation was consistent for each extract, although not all reached the statistical significance (W2: p = 0.01, mean of difference 1.65; W3: p = 0.25 and W4: p = 0.41, 2–3 technical replicates, Fig. 5B). When we mixed the L and R extracts in a pool and compared the effect, the generated difference was statistically significant (L/R-fluorescence ratio, One-sample T-test with hypothetical R value = 1, p = 0.03, mean of difference: 2.53, three technical replications, Fig. 5B). With this, we confirmed that: (i) the extracts had a differential bioelectric effect on the treated cells, (ii) the effect was independent of the donor, and (iii) the L-extracts generated a depolarized state as compared to R-extracts. Having confirmed this, we chose the pooled extracts for further studies to avoid possible donor-specific bias.Fig. 5L/R membrane potential comparison in conditioned MDA-MB-231 cells. A The results are expressed as the L-median of DiBAC4(3) fluorescence relativized to R. The histogram shows the L/R and R/R ratio of the signal. The value of R/R ratio is 1. The mean value of L/R ratio is 2.65, N = 3 experimental replications (treatments with extracts W2, W3, W4; each with 2–3 technical replications). B Individual analyses of the effect of L/R normal breast tissue extracts on cells. The results are expressed as the L DiBAC4(3) fluorescence relativized to R, as explained for A.

The three extracts present increased DiBAC4(3) fluorescence in L-treated cells, although not all significant. W2: p = 0.01, mean of difference 1.65; W3 and W4: L-treated NS increased tendency; pool: p = 0.03, mean of difference: 2.53. One-sample T-test with hypothetical R value = 1, for all experiments 2–3 technical replicates were performed. C L and R DiBAC4(3) fluorescence medians are expressed as percentages of the median fluorescence of high-KCl treated cells. L-treated cells present higher fluorescence as compared to the R ones (unpaired T test, p = 0.02). D Median fluorescence of high-KCl, L and R treated cells are compared. Only the R-treated ones differ significantly from the completely depolarized (One-way ANOVA + Dunnett post-test, p = 0.0037). Raw data are available in Additional file 3: Table S3

Our following aim was to establish the magnitude of the L/R \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Delta {\psi }_{p}$$\end{document}Δψp differences. For this, to normalize the potentials to the maximum possible depolarized state (which we established as a 100% depolarization reference value), we treated cells with a depolarizing agent (65 mM KCl, as suggested by Bonzanni et al. to depolarize MDA-MB-231 (Bonzanni et al. 2020). When expressing the DiBAC4(3) results as a percentage of the values of completely depolarized cells, we found that L-treated cells showed 69% (95% CI 56.67–81.35) vs R-treated 48% (95% CI 25.72–70.42), difference which was statistically significant (Unpaired T test, p = 0.02, Fig. 5C, Raw data available in Additional file 3: Table S3). Interestingly, however, no statistical difference was observed between L-treated cells and the KCl-treated ones, while R-treated cells did differ significantly (One-way ANOVA + Dunnett post-test, p = 0.0037, Fig. 5D). So we could conclude by this that the L-treated cells reached a similar depolarization as the maximum depolarized cells.

It has been well documented that epigenetics has a role in the adaptive regulation of gene expression. Specifically in neurons, in the dynamic expression of ion channels it has been reported that enzymes involved in DNA cytosine methylation have a crucial participation (Meadows et al. 2016). The process is catalyzed by DNA methyltransferases (DNMTs) and most commonly occurs at cytosines followed by a guanine, called CpG sites. DNMT3A is a de novo DNMT that methylates cytosines on unmethylated CpG sites, while DNMT1 is a maintenance DNMT that methylates cytosines on an unmethylated CpG with a methylated opposite strand. The inverse de-methylation process is regulated by the ten-eleven translocation (TET1, 2, and 3) family enzymes which oxidate the 5-methylcytosine to 5-hydroxymethylcytosine. We wondered whether these enzymes were differentially expressed in L/R mammary tumors. Previous work by others have reported that the expression of DNMTs is associated with the total DNMT activity (Maugeri et al. 2019; Tolg et al. 2011). From the TCGA-dataset of the Xena Functional Genomics Explorer (https://xena.ucsc.edu/), 1168 primary breast tumors (584 L and 584 R) were analyzed for DNMTs gene expression. L breast tumors presented increased expression of the 3 DNMT types (non-normally distributed data, Welch unpaired T-test, DNMT1: p = 0.01; DMT3a: p = 0.04; DMT3b: p = 0.001) (Fig. 6A). Instead, the demethylating enzymes TET1, 2 and 3 did not present any difference associated with laterality in 1095 primary breast tumors (571 L and 524 R) (Fig. 6B, non-normally distributed data, Welch unpaired T-test, p > 0.05). As control, normal tissue was analyzed where none of the studied enzymes presented L/R differences. When comparing normal vs tumoral, all the enzymes had altered expression in tumors.

The 3 DNMTs presented significantly increased expression (Unpaired T test, p < 0.0001), TET 1 and 2 decreased expression (Unpaired T test, p < 0.001), and TET3 increased expression (Unpaired T test, p < 0.001), as compared to their side-respective normal tissue.Fig. 6L–R epigenetic modulators expression comparison. A Gene expression data from 1168 TCGA L–R breast tumors for DNMTs and 1095 for TETs. DNMTs present significant increased expression on L-sided tumors, as compared to the R ones. This is not accompanied by the TET genes (B), which do not differ in their expression regarding the side

The observations suggest that, independently of the TET enzymes, the DNMTs are increased on the L-sided tumors, when compared to the R-ones. When deepening on the TET/DNMT relationship, it has been recently shown that TETs do compete with DNMTs in promoters of genes associated primarily to development and morphogenesis (Zhou et al. 2018). TETs act maintaining a hypomethylated state in these promoters, only in the absence of DNMTs. Based on these recent findings, one could ask whether increased methylation of hyperpolarizing ICH in L-tumors is due to enhanced DNMT activities, or/and if in R-tumors, with less DNMT activity, the TET enzymes are more actively demethylating the hyperpolarizing ICH.

A cell that needs to divide enters the cell cycle, and the regulation of the progression from one phase to the next one has been proposed to be coupled to environmental conditions so that this occurs only when it is necessary (Inzé and Veylder 2006). It is also known that the activity and expression of ion channels change during the cell cycle, and that Ca2+ concentration increases at the 3 cell cycle checkpoints and the membrane depolarizes between G2 and Mitosis, as reviewed by Rosendo-Pineda et al. (2020). We decided therefore to analyze whether the L/R tumors presented proliferation differences. The protein KI67 is widely used as a proliferation marker in different types of tumors. In-silico databases contain RNAseq values of KI67, obtained by Illumina HiSeq RNA Sequencing. We searched in breast in-silico datasets the expression of KI67 and matched it with the tumoral laterality information. Of 1060 primary breast tumors of the TCGA breast cancer dataset, we found a significant increment of KI67 expression in L-sided tumors (L: 10.76 ± 0.05259; R: 10.52 ± 0.05836; Unpaired T-test, p = 0.002, Fig. 7A). Normal L/R breast tissue did not present differences in KI67 expression (Unpaired T-test, p > 0.5). Both observations suggest a subtle increment of proliferation in L-sided tumors, as compared to the R-sided. This is consistent with our bioelectric findings since a depolarized state is necessary for cells to enter in mitosis.Fig. 7L–R proliferation rate comparison by KI67 and mitotic index. A Comparison of L–R KI67 expression in 1060 primary breast tumors from TCGA. L expression is subtle but significantly increased (10.76 ± 0.05259 N = 571) as compared to R (10.52 ± 0.05836 N = 524) (Unpaired T-test, p = 0.002). B Proportion of 95 L–R IDC breast tumors from local female patients with high and low mitotic count (Fisher’s exact Test, p = 0.002)

From our previous publication (Campoy et al. 2016) we counted with a database of clinic-pathological information of 95 local female patients with breast cancer (mean age 54, range 31–86). Revising the data, we analyzed if side correlated with the tumoral mitotic count. To establish the mitotic index, 10 different areas had been counted and cells in metaphase, anaphase or telophase were considered in mitosis. We classified the tumors as low mitotic index with a mean of up to 19 mitotic cells/area and high mitotic index with a mean of 20 or more mitotic cells/area. In concordance with in-silico observations, we found significantly more tumors with high mitotic index on the L side (Fisher’s exact Test, p = 0.002, Fig. 7B).

Bioelectric fields are produced naturally in all living tissue. Not only excitable nerve and muscle cells, but all cell-collectives that are organized in a functional network generate bioelectric signals to communicate among each other. Long before neurons existed, evolution exploited bioelectric networks to regulate morphogenesis and behavior (Fields et al. 2020; Martinez-Corral et al. 2019). When multicellular organisms appeared, the same efficient mechanism has been conserved for long-distance communication at different levels of the whole body. Bioelectric gradients are involved in embryogenic processes, such as eye development (Pai et al. 2012b), brain shape (Pai et al. 2015), antero-posterior and L/R axes (Levin et al. 2002), and the control of appendage size and shape (Belus et al. 2018; Perathoner et al. 2014; Lanni et al. 2019; Dahal et al. 2012). Endogenous bioelectric properties are seen to be critical due to numerous channelopathies in human and model systems, and there is an increasing realization that ion channels can also be oncogenes (reviewed in Lanni et al. 2019; Chernet and Levin 2013; Srivastava et al. 2021; Prevarskaya et al. 2018; Becchetti 2011; Rao et al. 2015; Fraser et al. 2014). In 1938, Burr et al. showed that tumorigenic processes in the mammary glands of mice were associated with disrupted bioelectric patterns in the chest (Burr et al. 1938). Since the last decades, when the tools to study bioelectricity increased significantly, many others have associated cancer with bioelectric alterations (reviewed in Moore et al. 2017) and discovered that the tumor microenvironment impacts on the bioelectric tumor pattern (Chernet and Levin 2014).

Epigenetics is also a key player in the interaction between cells and microenvironment. A rapid gene-expression shift is many times required to respond on time to the variable environment. We propose here a connection between epigenetics, environment, and bioelectric changes that the tumor cell senses, uses, and copes-with to shape a survival strategy.

In this work we have identified differences in methylation profiles and epigenetic regulators associated with distinct microenvironments (L/R), in addition to different bioelectric states and proliferation markers. We have found that L tumors present an increased expression of DNA methylation enzymes, an increased proportion of methylated hyperpolarizing ICH genes, a more depolarized membrane potential, and an increment in proliferation markers or mitotic index. These results can complement related observations of other biological and medical fields. For example, in the clinic it is well known that breast cancer has a slightly lower incidence on right sides (Roychoudhuri et al. 2006). And others have explored that hyperpolarization decreases tumor incidence (Sundelacruz et al. 2009; Levin 2012; Chernet and Levin 2014). Our results can connect both descriptions, by proposing that the more polarized state of R-sided tumors could explain the lower tumor incidence.

How this interplay between methylation, ion channels, voltage changes and proliferation occurs, in which order they are related or whether one is causative of the other are open questions for next studies. Are the methylation profiles responsible for the bioelectric differences? We did not find a strong inverse correlation between methylation and expression of the involved ICH genes in TCGA. The expression profiles of the ion channels which were found methylated did not reveal laterality differences in in-silico data. However, we think that this is explainable by the fact that the bioelectric differences are not gene-specific. So, probably it is not possible to establish a fixed panel of ion channel genes to study L/R differences. It is also possible to think on an inverse relation between bioelectricity and DNA methylation, where the epigenetic profiles are not causative but instead are a consequence of the bioelectric alterations, as has been proposed previously by others in neurons (Cortés-Mendoza et al. 2013; Penas and Navarro 2018) and development (Tseng and Levin 2012). In any case, it is worth to state that the transcriptional profiling undertaken here is meant to characterize one important and tractable input into bioelectrical differences, however not claimed to be the only source of asymmetry. Future work will explore other possible inputs as well, such as physiological gating dynamics.

Our in-vitro model has shown to be a reliable experimental tool to electrochemically transdifferentiate cells with L/R extracts (even though not sensitive enough to reach significant gene expression differences). Although it is generally accepted that experiments in culture do not recapitulate the complexity of the cellular surroundings, our model produced repeatable and consistent bioelectric results in concordance with what was observed in-silico and in animals. This encourages to postulated it as an efficient study tool for this purpose. Again, many questions remain. What components of the L/R extracts are producing different polarization in cultured cells? Morphogens? Small molecules? Neurotransmitters? Ions?

If further studies establish that general tumors on bilateral organs differ in their membrane potential, it could open new candidate therapeutic options by, for example, designing cocktails of channel openers/blockers, which are widely used in the clinic (Levin et al. 2019). The promising perspective is that, as proposed in Levin (2019), the interference with (or restoration of) bioelectric communication among tumor cells should be able to suppress carcinogenesis. Our work has opened new focuses based on L/R epigenetic and bioelectric differences in breast cancer, which could serve as prove of principle for other bilateral cancers like kidney, lung, testis, ovary and brain.

Additional file 1: Table S1. Raw of 1288 differentialy methylated genes in Left and Right xenograft tumors.Additional file 2: Table S2: Raw data of 2997 differentialy methylated genes from Left and Right in-silico breast tumors.Additional file 3: Table S3. Raw data of DiBAC fluorescence measured by flow cytometry in Left and Right extract-treated cells.