Concurrent micro- to macro-cardiac electrophysiology in myocyte cultures and human heart slices

The contact cardiac electrogram is derived from the extracellular manifestation of cellular action potentials and cell-to-cell communication. It is used to guide catheter based clinical procedures. Theoretically, the contact electrogram and the cellular action potential are directly related, and should change in conjunction with each other during arrhythmogenesis, however there is currently no methodology by which to concurrently record both electrograms and action potentials in the same preparation for direct validation of their relationships and their direct mechanistic links. We report a novel dual modality apparatus for concurrent electrogram and cellular action potential recording at a single cell level within multicellular preparations. We further demonstrate the capabilities of this system to validate the direct link between these two modalities of voltage recordings.

In terms of fundamental physics, the origin of the contact EGM is the superposition or summation of the electric field of charged ions in the vicinity of the electrode. Both the ion flux across cell membranes, producing the action potential, and the propagation of the action potential within the tissue lead to variations in electric field strength at the electrode and the subsequent EGM morphology 12 . Therefore, specific features of the EGM morphology could be used to infer the presence of, and potentially quantify, particular cellular electrophysiological factors responsible for specific EGM features, thereby providing a method to identify potential therapeutic targets more specifically. However, to date the direct relationship between characteristics of the EGM, myocardial structure and cellular and tissue level electrophysiology remains unclear, due to the lack of ability to extract meaningful causal relationships with cellular factors by simultaneously recording cellular activity alongside contact EGMs, particularly in intact preparations. The need for micro-to macro-electrophysiology extends past cardiac electrophysiology and is relevant to any electrophysiologically active organ, in particular the brain 13 .
Several methodological approaches exist for measuring electrophysiological activity in an in vitro or ex vivo setting. Commercially available research microelectrode array (MEA) systems allow recording of extracellular unipolar EGMs in in vitro models with a high signal to noise ratio 14 . The use of MEAs for the study of cardiac myocyte and embryonic stem cell-derived cardiomyocyte electrical activity is well established [14][15][16] with recent applications on human-induced pluripotent stem cells (hiPSCs) 17 .
In contrast, optical mapping of transmembrane voltage is an electrophysiological technique used to study cellular action potentials (AP) with high spatiotemporal resolution after staining cardiac myocytes or the whole heart 18 with voltage-sensitive dyes 19 . It gives the ability to measure both activation and repolarisation times from individual and multiple sites 19 . Optical mapping was developed as a technology to facilitate the investigation of APs in systems where the use of microelectrodes was not convenient or possible, due to scale, topology or complexity reasons 18 and does not require complex technical skills associated with traditional electrophysiological techniques. Development of new molecular probes and advancements in optical imaging technology have increased the efficiency of optical mapping 18 . Among the technical difficulties that have been overcome over the recent years is the improved spatial and temporal resolution. For this reason, new charge-coupled device (CCD) cameras, photodiode arrays (PDA) and complementary metal-oxide semiconductor (CMOS) cameras have been developed [20][21][22][23] . CMOS technology preserves the quantum efficiency of CCD cameras, while allowing high-speed image acquisition 18 facilitating single cell resolution and the subsequent analysis of AP heterogeneities.
Here we describe a novel non-invasive methodology which combines optical mapping at a single cell resolution within multicellular preparations with MEA recordings, for obtaining simultaneous APs and EGMs to sub-millimetre accuracy, compared to previous work 3, 6,7 , using in vitro neonatal rat ventricular myocytes (NRVMs) and ex vivo human cardiac slices. We present this unique dual modality apparatus and demonstrate applications of the technique by characterisation of EGM morphology alterations due to the blockade of specific ion channels or gap junction uncoupling and how these are related to corresponding modifications of AP morphology. This provides the first direct and concurrent validation of how AP changes at a single cell level can manifest on EGM morphology on multicellular preparations.

Synchronisation of dual modality concurrent action potential and electrogram recordings.
During dual modality recordings, the MEA plate is connected to the amplifier of the MEA system and the amplifier is loaded on the stage of the custom made optical mapping system (Fig. 1). The MEA stimulator and the optical mapping camera are connected through a BNC sync cable, to provide a blank frame when the stimulator fires to give a false stimulation artefact on the optical recording for time correlation between the two recording modalities. Representative correlations between electrical and optical mapping signals and stimulus artefacts are shown in Fig. 1d. MEA recordings were performed for 10s and during this time period 5s ratiometric voltage optical mapping recordings were simultaneously obtained. Further details are described in the Methods section. Signals were recorded at the baseline sample state and after the administration of ion channel blockers or the gap junctional blocker, carbenoxolone (CBX), at IC 50 at increasing pacing rates until loss of capture. Dual modality recordings were carried out on cardiac myocyte cultures and human ventricular myocardial slices. In addition to the entire NRVM culture, individual cells overlaying an electrode were also studied.

Single cell level correlations of action potential timings and electrogram fractionation.
To determine whether gap junction uncoupling affects both AP upstroke timing heterogeneity and EGM fractionation, dual modality recordings before and after the addition of 15.8 μmol/L CBX (IC 50 ) were analysed for individual myocytes coming from the same group of cells on top of an electrode. The AP traces, obtained from single cells before and after the addition of CBX, showed that the time point of upstroke peak is similar at baseline, but showed a larger distribution range across an electrode after gap junction uncoupling (Fig. 6b). CBX administration also led to increased EGM fractionation (Fig. 6d). Linear regression analysis shows that the time difference (ΔT) between the upstroke peak time points of two cells located at equal points across an electrode is directly correlated with EGM fractionation (r 2 = 0.74, slope = 0.57, p < 0.0001, n = 14) (Fig. 6c).

Discussion
We present the first application of a unique technique, adequate for identifying single cells within multicellular preparations, for the simultaneous recording of contact EGMs and optical action potentials in 2D preparations providing direct mechanistic insight into how changes in AP morphology directly manifest in unipolar EGM morphology. Our novel method enabled us to obtain high resolution data using electrodes which have an order of magnitude smaller diameter and intra-electrode distance less than half of that presented on previous systems 6,7 . We further demonstrate application of this novel technique using both cell monolayer and intact tissue slice models. Although EGM correlates of AP changes have previously been characterised [3][4][5][6][7][8] , this methodology allows direct and comprehensive analyses of all factors affecting AP morphology and its propagation, and allows concurrently recorded consequent EGMs to be determined. The simplicity of the MEA and optical mapping techniques allows these electrophysiological analyses to be carried out without the need for specific expertise in complex electrophysiological techniques such as patch clamp 3,4,24 . Moreover, the technique that we present here allows the simple colocalisation of optical mapping and electrogram recordings, and as a result it can be easily applied in a biological lab with relatively low cost, compared to previously published devices 8,25 , making it more accessible. This system allows analyses of summated EGMs alongside the cumulative single cell APs contributing to that EGM, allowing for investigation of how cellular level heterogeneities impact EGM morphology.
Because of the clinical importance of contact EGMs 26 in both the diagnosis and treatment of arrhythmias, there is a strong interest in associating abnormalities in the EGM morphology with modifications occurring at the tissue and cellular levels and numerous studies have attempted to clarify this [27][28][29] , however these findings are each in isolation and there is a lack of comprehensive mechanistic understanding. The use of this novel dual modality system to comprehensively characterise the direct and concurrent relationships between EGMs and APs will provide added value to future studies where only non-invasive EGMs can be recorded. As proof of concept we modulated a number of AP characteristics to determine their impact on the EGM.
In all preparations a delay between the stimulus artefact and response was observed. The stimulation voltage was kept at 120% of the threshold voltage, which is low enough so as not to damage the preparations with excess energy, however, as a result, cell response is delayed both in EGMs and action potentials. Delayed initial depolarisation when using low voltage stimulation was first described by Hodgkin and Huxley 30 . In addition to this, when recording from an electrode which is located at a distance from the stimulus electrode, the time taken for propagation to occur through the preparation needs to be taken into account. The above phenomena together can explain the delay between stimulation artefact and recorded action potential/EGM initiation.
We have demonstrated that lidocaine, a class 1B antiarrythmic agent and Na + -channel blocker, led to conduction delay in NRVMs, which is in general agreement with previous studies 31 . Our dual modality data show both prolonged upstroke and EGM duration concurrently in the same cells, providing evidence of a mechanistic link. Longer upstroke duration due to I Na blockade leads to slower conduction velocity and since EGM duration is the time needed for the activation wavefront to pass across an electrode, slower conduction velocity leads to longer EGM duration. This phenomenon was also observed in the human slices, though to a lesser extent, because the rapid upstroke showed little variation within the temporal resolution of the mapping system. HERG K + current and ventricular I to are blocked by 4-AP 32,33 . This compound was used to block I to in NRVM cultures and subsequently modulated APD 90 . Although APD 80 and APD 70 were also increased due to 4-AP activity, their correlation with FPD was weaker than for APD 90 . FPD has previously been described as an EGM feature on MEAs, defined as the time between the depolarisation peak (minimum voltage) and the repolarisation peak (maximum voltage), and is thought to be an index of APD 90 34 . Not only this I to current modulation, but all functional modulations investigated in this study validated that APD 90 is best correlate of FPD and this relationship is retained with modulation of APD 90 .
Nifedipine is well studied for its effects on L-type Ca 2+ -channel blockade 35,36 , and was used here to modulate APD 90 . The simultaneous MEA and optical mapping recording showed that FPD can be used as a predictor of APD 90 , since they are both reduced simultaneously on the administration of nifedipine. A direct relationship between FPD and APD 90 is also demonstrated with CBX administration. CBX, a gap junction blocker, has been studied for its effects on conduction slowing and EGM fractionation 37-39 , but it still remains unclear how it affects AP morphology. It has been suggested that gap junction uncoupling by CBX can either lead to prolonged APD in perfused adult rat hearts 40 or it has no effect on the APD of rabbit ventricular myocytes 38,39 . However, the AP data we collected concurrently with EGM recordings confirmed a direct correlation between FPD and APD 90 . Therefore we have validated the use of our technique in correlations of action potential depolarisation, repolarisation and propagation, demonstrating the capabilites to investigate all aspects of cellular electrophysiology.
Recently, intact myocardial slices are increasingly being used, due to the retention of structural and functional properties of the native myocardium 34 compared to in vitro preparations. There have been attempts in the past to measure the EGMs on myocardial slices derived from human, guinea pig or rabbit heart 34,41 . However, this study demonstrated the capabilites of the dual modality system to record concurrent APs and MEAs from the same area of human heart failure slices, in order to correlate EGM and AP morphology modifications. We obtained data from endocardial and epicardial left ventricular slices showing that there is a direct correlation between FPD and APD 90 . The dual modality results corroborated that this relationship was present even after modulation of gap junction coupling using CBX on epicardial slices. This further validated the results obtained from the in vitro model in a more complex ex vivo intact tissue model and demonstrated the translational benefit of this method. The activation maps acquired from the slice preparation show non-uniform propagation, even during baseline conditions. This may be due to the line of stimulus not being parallel to the fibre orientation, or due to the presence of scar, as the source of human slices was patients with heart failure. However, conduction velocities were calculated as average inter-electrode velocities in the direction of propagation, and therefore, are still an accurate marker of conduction velocity in these preparations. Since MEA recordings are currently applied on brain slice preparations 42 and the necessity of micro-to macro-electrophysiology techniques for measuring the neuronal activity have previously been justified 13 , the application of our set up for simultaneous optical imaging and MEA recordings on heart slices indicates that it could be also used for studying the neuronal activity on brain slices.
We have demonstrated the potential capability of our technique to elicit correlations between EGM and AP morphological charateristics, not only in groups of cells subjected to pharmacological manipulation, but also at the level of action potentials of individual cells located on top of an electrode, for the first time. This was possible due to the fact that the Zyla sCMOS camera, used for optical mapping of our NRVM preparations, incorporates complementary metal oxide semiconductor technology that is proved to be superior to other detector technologies in terms of spatiotemporal resolution because of the high pixel density and fast data-streaming rates 43 . show that APD 90 and FPD are strongly correlated in both sites (r 2 = 0.64, p < 0.0001 for both cases). (c) Raw trace obtained during optical mapping from human epicardial ventricular slice and the concurrent EGM, as both derived from the same electrode presented on the left. It is highlighted that APD 90 and FPD are similar (500 ms). Effect of 15.8 μmol/L CBX on APD 90 (d) and FPD (e) (Control: n = 2; Carbenoxolone: n = 2). APD 90 and FPD differed significantly both among pacing rates (p < 0.01 and p < 0.001 respectively) and APD 90 only changed significantly due to gap junction uncoupling (p < 0.001, ordinary two-way ANOVA). (f) Isochronal maps presenting the AP propagation before (14.2 cm/s) and after (9.2 cm/s) gap junction uncoupling in adjacent tissue slices. The markedly slow conduction in the centre of the control slice is due to the presence of a fibrotic area within the slice preparation. (g) Linear regression correlating APD 90 , APD 80 and APD 70 with FPD using data before and after the CBX activity on neighbouring heart failure tissue slices (APD 90 : r 2 = 0.64, slope = 0.39; APD 80 : r 2 = 0.63, slope = 0.4; APD 70 : r 2 = 0.59, slope = 0.39; p < 0.0001). All bar charts represent mean ± SEM; **p < 0.01; ***p < 0.001. (h) Linear regression analysis correlating conduction velocity measured before and after gap junction uncoupling with the EGM duration of the data obtained under the same conditions (r 2 = 0.14, slope = −0.14, p = 0.1).
Scientific RePORTS | (2018) 8:6947 | DOI:10.1038/s41598-018-25170-9 Concurrently recorded AP and EGM morphology were studied before and after the administration of CBX on individual myocytes, in order to investigate the morphology alterations due to gap junction uncoupling. We showed an increase in the spread of the AP upstroke timings in cells located across an electrode due to CBX administration. These results agree with previous in vitro studies which show that gap junction uncoupling leads to intracellular activation delays and multiphasic optically recorded AP upstrokes 44,45 . The delay shown in our study correlated with the generation of fractionated EGMs, while no fractionated EGMs were recorded before the administration of CBX. This observation may be explained by the discontinuity in AP propagation 38 associated with gap-junctional uncoupling. Our correlation of ΔT of action potential upstroke and EGM fractionation index suggests that in these paced preparations with linear propagating wavefronts, discontinuous propagation due to cellular uncoupling is sufficient to lead to EGM fractionation, and they can be directly correlated.
We have demonstrated this novel methodology and its possible outcomes using in vitro cell and slice models. However, with the availablity of flexible electrode arrays (both research and clinical) and whole heart optical mapping, it should be possible to apply our described methods on ex vivo whole heart models to gain insights into the cellular mechanisms leading to abnormal electrograms including the complexities of 3D structure and far-field signals.
In conclusion, we have demonstrated the utililty of a novel method combining optical mapping with MEA recordings for simultaneous recording of action potentials and EGMs with single cell resolution from the same group of cells. This technique has a number of applications in investigating electrophysiological abnormalities, as demonstrated using in vitro and ex vivo models, and can give a unique insight into cellular mechanisms leading to EGM changes. A variety of functional modulations led to different EGM changes and showed strong correlations with the corresponding AP morphology characteristic, indicating that EGM features could be used as predictors for specific changes on AP morphology from the single cell to the tissue level.

Methods
Cell Cultures. All reagents were sourced from Sigma Aldrich (St Louis, USA) unless otherwise stated.
NRVMs were obtained from Sprague-Dawley rats (0 to 3 days old). All procedures were conducted according to the standards set by the EU Directive 2010/63/EU and were approved by the Imperial College London Ethical Committee. Anaesthesia by isoflurane was followed by euthanasia carried out by cervical dislocation, and assessed by cessation of circulation after which the heart was removed and dissected immediately. The vessels and atria were removed to isolate the ventricles which were further dissected into 1-2 mm 3 pieces. Enzymatic digestion of ventricles was carried out using the gentleMACS neonatal heart dissociation kit (Miltenyi Biotec GmbH). The tissue and enzyme mix was incubated three times at 37 °C for 15 minutes each time before being attached to the gentleMACS dissociator for gentle agitation. The digested sample was resuspended in 10.5 mL M199 10% cell culture medium (100 mL M199, 10 mL neonate serum, 10 μM/mL penicillin-streptomycin, 0.68 mM L-glutamine, 2 μg/mL vitamin B 12 ), then passed through a 70 μm pre-separation filter and centrifuged at 1000 rpm for 5 min. The cell pellet was re-suspended in 20 mL M199 10% culture medium and fibroblasts were removed by preplating cells for 1 hour at 37 °C/1% CO 2 . The remaining suspended cell population consisting of NRVMs only was extracted after a final filtration through a 70 μm pre-separation filter. 200, 000 cells were plated on microelectrode array (MEA) dishes (MultiChannel Systems MCS GmbH, Germany) coated with 20 μL 0.2 mg/mL collagen over the electrode matrix. Cultures were incubated at 37 °C/1% CO 2 in M199 10% for the first 24 hours and then in M199 5% and measurements were recorded between the third and fourth day in culture.
Collection and preparation of tissue samples. Human left ventricular transmural tissue samples were prepared from end-stage heart failure patients. The mean age of the patients was 57 ± 10.5 years. All samples were immersed in ice-cold cardioplegia solution (Martindale Pharmaceuticals, UK) immediately after explantation and transported to the laboratory within one hour. Samples with an approximately 6 × 6 mm surface area were obtained and mounted with epicardium down onto the specimen holder of a high precision vibrating microtome (7000 smz, Campden Instruments Ltd., UK). Samples were in cold (4 °C) oxygenated (100% O 2 ) cutting Tyrode solution (140 mM NaCl, 6 mM KCl, 10 mM glucose, 10 mM HEPES, 1 mM MgCl 2 , 1.8 mM CaCl 2 , 3 g/L 2,3-butanedione monoxime (BDM), pH 7.4) during the preparation of 300 μm slices derived from the endocardial and epicardial sample areas. The advancement speed of the steel blade was 0.03 mm/s, the amplitude 2 mm and the vibration frequency was 80 Hz. Slices were incubated in oxygenated ice-cold cutting Tyrode solution for at least 30 minutes prior to being placed in the centre of a MEA plate and obtaining the electrophysiological recordings in Tyrodes solution (140 mM NaCl, 4.5 mM KCl, 10 mM glucose, 10 mM HEPES, 1 mM MgCl 2 , 1.8 mM CaCl 2 , 1 g/L BDM). In order to study the effects of gap-junction uncoupling on tissue slices, 15.83 μmol/L carbenoxolone was added in the same recording Tyrodes solution. This study was supported by the supply of human tissue samples from the Cardiovascular Research Centre Biobank at the Royal Brompton and Harefield NHS Foundation Trust (NRES ethics number for biobank samples: 09/H0504/104 + 5; Biobank approval number: NP001-06-2015). Informed consent was obtained from each patient involved in this study. All procedures described in this manuscript were carried out in accordance to the Human Tissue Act 2004 (c30).
Microelectrode array recordings. The electrophysiological properties of cardiac myocytes were assessed using the USB-MEA60-Inv MEA system (MultiChannel Systems, Reutlingen, Germany). An MEA plate consists of 60 gold electrodes arranged on an 8 × 8 matrix (inter-electrode space: 700 μm, electrode diameter: 100 μm) with missing electrodes in the corners of the matrix. Stimulation was carried out using a STG stimulus generator programmed by MC Stimulus II software (version 3.4.4, MultiChannel Systems). A biphasic stimulus (2 ms duration; 120% of the threshold; voltage: 500-1000 mV) was applied for several minutes from the 6 electrodes located on one of the four external rows of the matrix to reach steady state before obtaining 10 sec recordings in incremental rates until loss of 1:1 capture. The temperature was kept at 37 °C during stimulation. The signals were recorded at a sampling frequency of 25 kHz. The responsiveness of cells to lidocaine (1-100 μmol/L), 4-aminopyridine (0.5-1.5 mmol/L), nifedipine (0.03-10 μmol/L) and carbenoxolone (5-100 μmol/L) dissolved in Hank's buffered saline solution (HBSS) supplemented with 1 mM MgCl 2 and 1.5 mM CaCl 2 was tested 3-4 days after plating cells on MEA dishes. All chemical compounds were sourced by Sigma-Aldrich. Control recordings with NRVMs in 1 mL of HBSS were obtained before changing the control solution for 1 mL of incremental concentrations of each pharmacological agent. Signals were displayed and data were analysed offline using MC Rack software (v4.6.2, MultiChannel Systems). No filtering occurred during the post-processing analysis.
EGMs were manually annotated to extracted QRS complex duration and extracellular field potential duration (FPD). FPD was defined as the time elapsed between the initial deflection of the field potential and the maximal point of the T wave.
Optical Mapping. Optical mapping was carried out in combination with MEA recordings. Figure 1 is a schematic representation of the custom-made optical system used for simultaneous membrane voltage and MEA recordings (supplied by Cairn Research, UK). The system was built around an upright microscope (Eclipse FN1, Nikon Instruments Europe B.V.) with a modified stage height to hold the amplifier of the MEA system. Excitation light (470 nm) was supplied by an OptoLED system (Cairn Research, UK) which provided controlled illumination and modulation. The light was passed through an emission filter of 470/40 nm. Light was collimated on the MEA dish by a water dipping objective lens with magnification 20× (NA: 1.0 20×, XLUMPLFLN20XW PL FLUORITE OBJ, Olympus). Samples were stained with 40 μM di-8-ANNEPS (Molecular Probes ® , Invitrogen) diluted in 1 mL HBSS with 2.5 μL Pluronic ® F-127 (Life Technologies, USA). The emitted fluorescence was collected by the same objective lens and passed through dichroic mirrors. Using a 560 nm edge BrightLine ® single-edge dichroic beamsplitter, located in an Optosplit II 'LS' emission image splitter (x1.0 magnification), the fluorescent light was divided into two beams that were passed through emission filters (525/36 nm and 628/32 nm). Subsequently, the light was focused onto a complementary metal-oxide semiconductor camera (Zyla 10-tap sCMOS, Andor Technologies Ltd., Belfast, UK) for the detection of the dual wavelength optical signal with a spatial resolution of 400 × 885 pixels at 525.39 frames/sec. The light intensity measurements were recorded using the Andor Solis software platform (version 4.23.30008, Andor Technologies Ltd.). Optical mapping data were spatially smoothed on GraphPad Prism (version 6.0 f for Mac OS X, GraphPad Software, La Jolla California, USA) using 3 averaging neighbours and 4 th order of the smoothing polynomial. Action potential data were undertaken with emission ratiometry 46 . The ratiometric signal analysis was achieved by using the fluorescence signals of opposing ΔF (green and red fluorescence). The ratio signal, collected at both wavelengths, was extracted for the attenuation of any motion artefacts, since motion artefacts can appear as a common change on both wavelengths and consequently they can be cancelled out by the ratiometric calculation 46 . As a result, the introduction of excitation-contraction uncouplers, such as blebbistatin, is not necessary with our system. Blebbistatin may have unwanted effects, such as formation of blebbistatin precipitate, and it may cause changes in metabolic state 47 . Analyses were carried out using a custom-written macro in GraphPad Prism for measurement of action potential duration at 90% of repolarisation (APD 90 ) and upstroke duration. Post processing of the raw data included drift removal and normalisation before calculation of AP morphology.

Synchronisation.
During dual modality recordings, the MEA plate was connected to the amplifier of the MEA system and the amplifier was loaded on the stage of the custom made optical mapping system (Fig. 1). The MEA stimulator and the optical mapping camera were connected through a BNC sync cable to generate a blank frame when the stimulator fired. This produced a false stimulation artefact on the optical recording which enabled accurate time correlation between the two recording modalities. MEA recordings were made for 10s and during this time period 5s ratiometric voltage optical mapping recordings were simultaneously obtained.
Statistical Analysis. Data are expressed as mean ± SEM. Statistical analysis was carried out using GraphPad Prism software (version 6.0 f for Mac OS X, GraphPad Software). Statistical significance was evaluated using student's t-test (two-tailed), one-way analysis of variance (ANOVA) for unpaired data or ordinary two-way ANOVA followed by Bonferroni multiple-comparison post hoc analysis where appropriate. Baseline and post-modification data are unpaired, unless otherwise stated. The comparison of MEA recordings with optical mapping was validated with linear regression analysis (confidence interval: 95%) for correlation of corresponding characteristics measured with each technique. The APD 90 /FPD, APD 80 /FPD and APD 70 /FPD relationships were compared between the control and modulated data groups by correlation coefficient. All numerical data are presented as mean ± SEM and p-value of <0.05 was considered as significant.
Data availability. All data supporting the findings of this study are available within the article and from the corresponding author on reasonable request.