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

During embryogenesis lateral symmetry is broken, giving rise to Left/Right (L/R) breast tissues with distinct identity. L/R-sided breast tumors exhibit consistently-biased incidence, gene expression, and DNA methylation. We postulate that a differential L/R tumor-microenvironment crosstalk generates different tumorigenesis mechanisms.

We performed in-silico analyses on breast tumors of public datasets, developed xenografted tumors, and conditioned MDA-MB-231 cells with L/R mammary extracts.

We found L/R differential DNA methylation involved in embryogenic and neuron-like functions. Focusing on ion-channels, we discovered significant L/R epigenetic and bioelectric differences. Specifically, L-sided cells presented increased methylation of hyperpolarizing ion channel genes and increased Ca2+ concentration and depolarized membrane potential, compared to R-ones. Functional consequences were associated with increased proliferation in left tumors, assessed by KI67 expression and mitotic count.

Our findings reveal considerable L/R asymmetry in cancer processes, and suggest specific L/R epigenetic and bioelectric differences as future targets for cancer therapeutic approaches in the breast and many other paired organs.

The online version contains supplementary material available at 10.1186/s10020-022-00440-5.

Some organs, such as the heart or viscera are asymmetric: their structures to the left and right of the body mid-plane are consistently different in all normal individuals (Monsoro-Burq and Levin 2018). Most other tissues are often believed to be symmetrical. However, major knowledge gaps exist about the degree to which paired structures could exhibit not only the fluctuating asymmetry of developmental noise but consistently biased asymmetry that might impact structure and function. Despite the general assumption that mammary glands are mere copies of one another, each gland has its own identity and presents left–right (L/R) asymmetries. During embryogenesis of bilateral organisms, lateral symmetry is broken at very early stages in a programmed and consistent way (Vandenberg and Levin 2010; Levin et al. 2016; Ma 2013). The establishment of the L/R axis is the start of a regulated patterning, through which asymmetric sides arise at morphological, functional and molecular levels (Monsoro-Burq and Levin 2018; Sutherland and Ware 2009; McDowell et al. 2016). Alterations in laterality decisions during development give rise in humans not only to a reversed laterality (situs inversus) but also to an increased susceptibility to other diseases (Soofi et al. 2021). In particular, tumors in bilateral organs such as breast, colon, kidney or lung, show subtle but significant differences at morphological, genetic, molecular and incidence levels (Robichaux et al. 2014; Yang et al. 2018). Research data from our group and others (of diverse disciplines such as embryogenesis, development, molecular oncology, or cellular biology), have proposed that the asymmetric tumor microenvironment of bilateral organs could be part of the explanation for the L/R differences in cancer (Levin et al. 2016; Ma 2013; Robichaux et al. 2014; Campoy et al. 2016; Atiya et al. 2019; Yi et al. 2021).

Tumor cells sense the environment and fire in consequence internal signals. By this, the tumor transcriptome differs from the surrounding normal tissue and acquires specific features. The gene expression shift that the tumor applies to face the environmental challenges depends on the surrounding tissue signals, especially during the initial tumorigenesis stages. Thus, it can be said that the microenvironment contributes to the decision-making strategy of a tumor to reach the cancer hallmarks (Hanahan and Weinberg 2011). In this context, epigenetics and bioelectricity have a crucial role since both constitute vehicles by which external signals reach and modulate the transcriptome in an experience-dependent and dynamic way.

Epigenetic modifications highly influence the biology of cancer. A key feature of cancer cells is to respond rapidly to environmental challenges, and this is mainly attributed to the dynamic plasticity of the epigenetic mechanisms. Epigenetic regulators have both writing and erasing capacities, so are therefore able to maintain a flexible transcriptome which is crucial for tumor development and survival [reviewed in Dawson (2017)]. In addition, epigenomes are also defined as the bridges between the environment and the phenotype (or transcriptome) (Tammen et al. 2013). Being more dynamic and reversible than the genome, epigenomic variations can rapidly provoke a transcriptomic shift without changing the genomic sequence. DNA methylation, the most studied epigenetic modification, presents a specific signature associated with some cancer types, suggesting a distinct interplay between the tumor epigenome and the surrounding tissue. Based on this, specific aberrant DNA methylation patterns have been proposed as predictive and prognostic markers for several cancer types (Okugawa et al. 2015; D'Errico et al. 2020; Singh et al. 2019; Almeida et al. 2019). Specifically, in breast cancer, previous work of our group has identified that the DNA methylation profiles of tumor suppressor genes correlate with prognosis index (Marzese et al. 2010, 2012), with tumor subtypes (Branham et al. 2012), migration and metastasis capacity (Urrutia et al. 2015; Marzese et al. 2014), benign mammary lesions (Marzese et al. 2011), and, more relevant for the topic of this study, with the laterality of the tumors (Campoy et al. 2016). In summary, tumor epigenome is influenceable by the microenvironment and can be associated with differential tumor behaviors.

Bioelectric gradients are considered epigenetic mediators in a broad sense of the word, since they can modify the transcriptome following environmental signals (Cortés-Mendoza et al. 2013; Tseng and Levin 2012; Penas and Navarro 2018). The flow of ions (inside the tumor and between the tumor and the microenvironment) enables the transmission of membrane potential patterns, which are maintained as information for survival decisions in response to external challenges (Levin et al. 2019; Levin 2021). Like epigenetics, bioelectric control is reprogrammable, rapid and dynamic, and is driven by physiological states that are not 1:1 mapped to specific genes (Levin 2021). Bioelectric states are acquired by ion flux through channels and pumps in the membrane and are transmitted to neighboring cells via gap-junctions. The current flux produces changes in membrane potentials, which in turn generate downstream signaling to regulate different cellular processes, e.g. proliferation (Blackiston et al. 2009), migration, differentiation, or gene expression. Therefore, it is accepted that cells of the same tissue share similar bioelectric states, which is maintained as non-genetic information. This is also applicable to the L/R sides of bilaterian bodies, where metabolic (Onjiko et al. 2016), epigenetic, bioelectric, and gene-expression differences (Mittwoch 2008) have been reported. Just to highlight an example, L/R bioelectric differences have been observed in Xenopus and chicken embryos, finding consistent voltage and ion transporter asymmetries as early as the 2nd cell division (Levin et al. 2002). These bioelectric differences subsequently regulate asymmetric gene expression to control the sidedness of asymmetric organs and paired structures such as eyes and neural crest derivatives (Pai et al. 2012a).

Differences in L/R bilateral tumors have been reported for several cancer types, like breast (Kenney et al. 2004; Garfinkel et al. 1959), colon (Yang et al. 2018; Baran et al. 2018; Sponholz et al. 2021), kidney (Ni et al. 2021; Guo et al. 2019), brain (Połczyńska et al. 2021), ovary (He et al. 2021), and eye (Hussain 2021). Particularly in breast tumors, in addition to the largely known increased L-side incidence (Busk and Clemmensen 1947; Roychoudhuri et al. 2006), interesting differences in gene expression (Robichaux et al. 2014; Mittwoch 2008), gland microbiota (Klann et al. 2020), mitochondria distribution (Mishra and Chan 2014) and methylation profiles (Campoy et al. 2016) have been reported. It is reasonable to postulate that L/R adult glands conserve memories of their asymmetric embryogenic development, maintaining different L/R bioelectric patterns. These patterns are, in part, constituted by “attractors” (such as morphogens, neurotransmitters, small molecules) that trigger downstream different signaling pathways and change transcription regulation (Levin 2021).

In this work, we hypothesize that tumorigenic breast processes face environmental challenges that differ between L/R sides, establishing a side-dependent tumor-microenvironment crosstalk reflected in bioelectric and epigenetic differences. A serendipitous finding during our previous research gave rise to the present work, when we discovered that DNA methylation patterns of female patient’s breast tumors clustered in two groups based on whether they were located on the L or R gland (Campoy et al. 2016). This striking observation opened the questions of whether these L/R differences were reproducible in an in-silico, in-vivo and/or in-vitro model, and if functional differences were associated with these epigenetic profiles. In this work, we developed in-silico, in-vivo and in-vitro approaches to address these questions.

For gene methylation analyses, Illumina Infinium Human Methylation-450 information was obtained from breast cancer TCGA dataset, available in the public platform cBioportal for Cancer Genomics (https://www.cbioportal.org/, repository Firehose Legacy of the Broad Institute). The DNA methylation data is found in the repositories as beta values, which are continuous ratios between 0 and 1, indicating the intensities between methylated and unmethylated alleles (0 being unmethylated and 1 fully methylated). For laterality data, clinical datasets were also retrieved from the same platform. We used 782 primary breast tumors for which DNA methylation data of ~ 16,000 genes plus anatomical location (L/R gland) was available. After curating the information, we calculated the L/R DNA methylation mean for each genomic region and ranked their absolute differences (called from now on differential methylation, DM) (note: for all experiments, differences are calculated as left minus right; since data are beta values, the L–R differences are between 0 and 1).

To find the cellular and molecular functions in which the most differentially methylated genes were involved, we performed gene enrichment analyses with the public tools Metascape (https://metascape.org, RRID:SCR_016620) and EnrichR (https://maayanlab.cloud/Enrichr/enrich, RRID:SCR_001575). For Metascape tool, the enrichment analyses were set as: minimum overlap = 3 and p-value cutoff = 0.001, with Gene priorization by Evidence Countins (GPEC). To establish the potential functional consequences of the proximal (up to 2.5 kb up and downstream) as well as the distal (up to 1 Mb up and downstream) genomic context of the differential methylated CpGs, GREAT analyses were performed (Genomic Regions Enrichment of Annotations Tool v4.0.0) (McLean et al. 2010). We used the basal plus extension configuration as a background setup “whole genome”, as recently shown (Emran et al. 2017).

For gene expression analyses for selected genes of interest, Illumina HiSeq 2000 RNA Sequencing platform of the University of North Carolina was obtained from 1168 primary breast cancer TCGA dataset for DNMTs, 1095 for TETs and 1060 for KI67, available in the UCSC (University of California Santa Cruz) Xena Functional Genomics explorer (http://xena.ucsc.edu/, RRID:SCR_018938). The RNA-Seq data are shown in the dataset as normalized log2 (x + 1) values and indicate an estimated gene expression level.

The highly immunosuppressed Nod Scid Gamma mice (NOD.Cg-PrkdcscidIl2rgtm1Wjl/SzJ, NSG) (RRID:IMSR_JAX:005557) were obtained from Jackson Laboratory and were housed in a pathogen-free condition throughout the experimental duration. All procedures were performed following the consideration of animal welfare and were approved by the Institutional Committee for Care and Procedures of Laboratory Animals (CICUAL in Spanish) of the National University of Cuyo, Mendoza, Argentina. To perform the xenograft experiment, 6-week-old female NSG (20 g) mice were anesthetized with isofluorane 4% in O2, and injected with 1 × 106 MDA-MB-231 cells (suspended in physiologic solution) in the 4th L/R breast glands. Mice were closely monitored, and tumor size was measured weekly. Five weeks after cell inoculation, the mice were sacrificed in a CO2 camera, and tumors were excised. Part of the tumors was set apart and frozen at − 80 °C for further DNA and RNA extractions (labeled as passage 0). The remaining parts were reimplanted in small pieces in 3 NSG mice, maintaining laterality (labeled as passage 1). The complete procedures were repeated in 3 more NSG mice to generate tumors passage 2.

DNA was extracted from xenograft tumor tissues and from MDA-MB-231 cells, using PureLink® Genomic DNA Kits, Mammalian Tissue and Mouse/Rat Tail Lysate (Catalog Numbers K1820-02, Invitrogen), following manufacturer’s protocol. RNA was extracted from MDA-MB-231 cells using a Trizol based protocol (TRIzol® Reagent (Life technologies, Catalog Numbers 15596-026).

To assess the methylation status of 50 CpG sites located on 40 genes, the MS-MLPA kits ME001 and ME002 (Catalog Numbers ME001-025R, ME002-025R) were used. The MS-MLPA assays were performed basically according to manufacturer’s recommendations (MRC-Holland, Amsterdam, The Netherlands, www.mrc-holland.com) (Nygren et al. 2005), introducing subtle modifications (i.e., extended restriction enzyme incubation time, separated ligation and digestion steps), to avoid background signals (Marzese et al. 2010). The fluorescent-labeled PCR products were separated by capillary electrophoresis (3500 Genetic Analyzer for Fragment Analysis, Applied Biosystems) and analyzed by GeneMarker v1.75 software (RRID:SCR_015661). A cutoff of 8% fluorescence signal was established to consider the site significantly methylated.

To assess an extended methylation analysis involving most of the genome CpG sites, a reduced restricted bisulfite sequencing (RRBS) assay was performed with the technical and bioinformatic assessment of the Genomic Unit–Consortium CATG-National Institute of Agricultural Technology (INTA) in Buenos Aires, Argentina. For this, 3 matched left and right passage-1 xenografted tumors were selected. The experimental steps consisted on: preparation of the libraries with Diagenode’s Premium RRBS kit (Diagenode, Cat. No. C02030032), sodium bisulfite conversion of the DNA samples and PCR amplification and sequencing of the generated fragments on an Illumina NextSeq 550 equipment. Quality control of sequencing reads was performed using FastQC® (Babraham Bioinformatics®, RRID:SCR_014583). Adapter removal was done using Trim Galore® version 0.4.1 (Babraham Bioinformatics®, RRID:SCR_011847). Reads were then aligned to the reference genome GRCh38 using Bismark v0.22.1.® (Babraham Bioinformatics®, RRID:SCR_005604), followed by methylation calling using the corresponding bismark functionality. The comparison between the RRBS data sets was carried out using methylKit® (Bioconductor®, RRID:SCR_005177), with the GRCh38 refGene and CpG island annotation from UCSC (University of California Santa Cruz, RRID:SCR_006553) genome browser. Bioinformatic filters were applied on the raw results, to select only human sequences aligned with the human reference genome GRCh38, discarding possible mice genomic interference. After methylation calling, and difference calling with Bioconductor 3.9, L/R DM with more than 10% difference were found in 2219 sites. For enrichment analysis (by Metascape) we eliminated duplicated genes (with more than one methylated site per gene), leaving 1288 genes for further analyses (Additional file 1: Table S1).

Human breast cancer cell line MDA-MB-231 (ATCC, RRID:CVCL_0062) was kindly provided by Dr. Matias Sanchez (IMBECU Institute, Mendoza, Argentina) and passages 20–30 were used for this work. The cells were routinely tested for mycoplasma contamination. In general, cells were cultured in DMEM medium (Gibco by Life Technologies, Grand Island, NY, USA, # 112800-058) supplemented with 10% fetal bovine serum (Internegocios S.A, Mercedes, BA, Argentina), 100 U/mL of penicillin and 100 μg/mL streptomycin (Gibco by Life Technologies, Grand Island, NY, USA, #1796440), at 37 °C in a humidified atmosphere containing 5% CO2. For the extract-conditioned cultures, fetal bovine serum was reduced to 1%.

Healthy L/R breast glands were obtained from plastic surgeries, provided by Dr. Cataneo from the Clinic of Plastic Surgery of Mendoza, after patients signed an informed consent previously approved by the Ethics Committee of the Medical School of the National University of Cuyo. Tissues were first disaggregated with a scalpel and the pieces were suspended in 25 mL of DMEM medium with Penicillin/Streptomycin 1% and incubated in a shaker for 24 h at 37 °C. Next, samples were centrifuged to remove the solid fat and the remaining suspension was filtered with cell strainers of first, 100 μm and afterwards 40 μm, to eliminate residual tissue parts. The obtained liquid-phase extracts were L/R labeled and stored for further experiments at − 20 °C.

MDA-MB-231 were conditioned with a cocktail consistent of 49% DMEM with Penicillin/Streptomycin, 1% Serum Fetal Bovine and 50% left or right liquid-phase extract.

Cells were cultured on 30 mm glass coverslips for performing Ca2+ imaging. Coverslips with cells attached were mounted in a chamber and incubated at 37 °C and protected from the light for 30 min in a culture medium containing 3 μM Fluo3-AM (Invitrogen, Cat# F1242). After incubation, cells were washed two times with PBS 1× and bathed in DMEM medium supplemented with 10% fetal bovine serum, 100 U/mL of penicillin, and 100 μg/mL streptomycin for 5 min before Ca2+ measurements were made on an inverted Olympus FV 1000 confocal microscope (Olympus Corporation, Tokyo, Japan). Images were collected using the Fluoview FV-1000 software and an Olympus 20X lens (UPlanSApo 20X/0.75). Fluo 3 fluorescence was detected using the filter cube U-MWB2 (excitation BP 460–490 nm and emission LP 520 nm).

Still Images of cohorts of 50–100 cells were analyzed with Microsoft Excel and Image J (National Institutes of Health, USA, RRID:SCR_003070). Fluorescence data was calculated as mean/area.

To determine the time-window at which L/R treated cells acquired differences in their Ca2+ fluorescence curves, two experimental approaches were used. A first one, with a day-by-day stepwise assay, collecting still images of Fluo 3 every day, for 5 days with a Nikon TE300 Inverted Fluorescence Phase Contrast Microscope using ×20 magnification (Objective: PLAN FLUOR 20x/0.45), filter cube B-2A, 450–490 nm for excitation, and 500 nm dichroic and 515 nm barrier. The exposition time was 150 ms. The images were taken with a LUCA ANDOR EMCCD camera. Still Images of cohorts of 15–75 cells were analyzed with Microsoft Excel and Image J (National Institutes of Health, USA, RRID:SCR_003070). Fluorescence data of L-treated, R-treated and non-treated cells (control) was calculated as mean/area. With a second approach based on real-time experiments on a Zeiss-Axio-Observer microscope at 37 °C with 5% CO2, images were collected every 15 min with ×20 magnification during 3 days (since the cells did not survive longer on this system). In the stepwise approach, independent wells were each day incubated with Fluo 3, and the fluorescence values were calculated as means of still images of a 50–100 cells per field. In the real-time experiments, the fluorescent probe was first loaded and afterwards the L/R extracts were added to cells. Fluorescence data was recorded for 15–75 single cells per condition, over the 3-day-experimental time.

For Δψp measurements, 104 MDA-MB-231 cells were plated and conditioned with L/R extracts for 5 days as described above, and then incubated for 30 min with 1 μM DiBAC4(3) (Bis-1,3-Dibutylbarbituric AcidTrimethine Oxonol, a fluorescent probe for membrane potential determination) (Invitrogen by Thermofisher Scientific, Cat. No. B438) at 37 °C and 5% CO2. Afterwards, cells were trypsinized and fluorescence was measured by flow cytometry (FACSARIA-III, BD-Biosciences®) with a BP 530/30 emission filter. Results were analyzed using FlowJo v X.0.7® software (RRID:SCR_008520).

To provoke maximum depolarization (considered as 100% depolarization), cells were first treated with 65 mM KCl for 5 min at 37 °C and afterwards incubated for 30 min with the fluorescent probe DiBAC4(3) as described above.

From a previous work of our group (Campoy et al. 2016), we counted with a database of 95 breast cancer female patients (mean age 54, range 31–86) who had previously signed an informed consent approved by the Ethics Committee of the Medical School of the National University of Cuyo, Mendoza, Argentina. The database included information of the tumor mitotic index provided by the same anatomo-pathologist. In brief, at least 10 different areas had been counted and cells in metaphase, anaphase or telophase were considered in mitosis as indicated in Ogston et al. (2003). We dichotomized the data 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.

Differences between 2 proportions of hyper/depolarizing ICH were calculated as Odds Ratios (OR), with the corresponding 95% CI. To compare means and medians of fluorescent-probe concentrations, unpaired T-test was applied with Welch’s corrections when variance was not equal among L/R data. L/R ratio differences were analyzed by One sample T-test with hypothetical Right value = 1 (assigning the values of Right as reference). When more than two groups were compared, one or two-way ANOVA test were applied (with Dunnett post-test). Finally, Fisher’s exact test was used to compare categorical data. P values below 0.05 were considered as statistically significant.

The methylation profile of ~ 16,000 genomic regions were analyzed in 782 primary breast carcinomas (394 L and 388 R). We calculated the L/R DNA methylation mean for each genomic region and ranked their absolute differences (called from now on differential methylation, DM), which ranged between 10–7 and 5%, with a median value = 0.03%. We decided to focus on the top 2997 genes with > 1.24% difference (Additional file 2: Table S2). Gene Enrichment analyses performed on the selected genes by Metascape (https://metascape.org) revealed that the main pathways in which they were involved were related to regulation of ion transport (GO:0043269), trans-synaptic signaling (GO:0099537), and embryonic morphogenesis (GO:0043598) (Fig. 1A). By the tool EnrichR (https://maayanlab.cloud/Enrichr/enrich), also embryonic digestive tract development (GO:0048566), chemical synaptic transmission (GO: 0007268), calcium ion transport (GO: 0006816), and positive regulation of ion transport (GO:0043270) appeared as significantly involved pathways (adjusted p values < 0.05) (data not shown).Fig. 1Gene enriched pathways of L/R differential methylated regions. A DNA regions with more than 1% DM between L–R primary breast tumors from TCGA public dataset were analyzed by Metascape for enriched GO Biological processes (p value cut-off 0.001). B DNA regions with more than 10% DM between 6 L/R xenograft tumors were analyzed by Metascape for enriched GO Biological processes (p value cut-off 0.001). Ion/neural and embryonic pathways are highlighted in red

To study whether the L/R DM was reproducible in an animal model, we generated synchronic L/R breast tumors in Nod-Scid-Gamma (NSG) immune depressed mice by inoculating the human breast cancer cell line MDA-MB-231 simultaneously in both 4th mammary glands. After generating three sequential passages (labeled as P0, P1, and P2), the tumors were first analyzed in a reduced number of CpG sites by Methyl-Specific-MLPA (50 CpG sites located on 40 tumor suppressor genes). The three passages showed subtle L/R differences in several genes, and we chose P1 as the passage with the major DM. In 15 CpG sites of 11 genes (RASSF1A, ESR, IGSF4, CDH13, MGMT, TP73, WT1, MSH6, PAX6, GATA5, and RARB) the L/R DM per site were from − 6.8 to 11.17%. These observations were considered only useful for choosing the best tumor passage to scale up experimentally and perform a whole genome methylation analysis.