Non-invasive phenotyping and drug testing in single cardiomyocytes or beta-cells by calcium imaging and optogenetics

Identification of drug induced electrical instability of the heart curtails development, and introduction, of potentially proarrhythmic drugs. This problem usually requires complimentary contact based approaches such as patch-clamp electrophysiology combined with field stimulation electrodes to observe and control the cell. This produces data with high signal to noise but requires direct physical contact generally preventing high-throughput, or prolonged, phenotyping of single cells or tissues. Combining genetically encoded optogenetic control and spectrally compatible calcium indicator tools into a single adenoviral vector allows the analogous capability for cell control with simultaneous cellular phenotyping without the need for contact. This combination can be applied to single rodent primary adult cardiomyocytes, and human stem cell derived cardiomyocytes, enabling contactless small molecule evaluation for inhibitors of sodium, potassium and calcium channels suggesting it may be useful for early toxicity work. In pancreatic beta-cells it reveals the effects of glucose and the KATP inhibitor gliclazide.


Introduction
Disrupting the ion channels responsible for depolarisation and repolarisation of cardiomyocytes by gene mutation or off-target drug effect increases the risk of sudden arrhythmic death [1]. This pillar of safety pharmacology is principally studied using animal explant material and manual approaches requiring direct contact with the cell. Improvements to increase human relevance and experimental throughput are currently being evaluated in the Comprehensive PLOS ONE | https://doi.org/10.1371/journal.pone.0174181 April 5, 2017 1 / 17 a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 followed by a 1mg/ml Collagenase type II solution (250 units/mg, Worthington Biochemical Corporation) for 9 minutes. The digested tissue was subsequently mechanically agitated in enzymatic solution for an additional 5 minutes and the cells were collected by centrifugation at low speed (500 rpm).

ChETA TC -myc and R-GECO imaging in fixed cells
Adult ventricular cardiomyocytes were infected at an MOI of 5 as for live imaging and kept in storage solution in a humidified incubator at 37˚C, 5% CO2. At 48 hours an aliquot was taken and paraformaldehyde to a final concentration of 4% was added, at 10 minutes, cells were pelleted (200rpm, 2min) in a benchtop centrifuge, washed with 1xPBS and resuspended in PBS. Cells were spun (Cytospin, Thermo Scientific) onto glass slides, and ringed with a PAP pen (Sigma). Permeabilisation with 0.1% Triton-X-100 in Tris Buffered saline for 10 minutes at room temperature was followed by blocking (0.2% albumin in permeabilisation buffer) for 20 minutes. Primary antibodies (mouse monoclonal 9E10 anti-myc (Santa-Cruz), and rabbit polyclonal anti-DsRed (Clontech) were diluted 1:200 in blocking buffer. Three hours after primary incubation cells were washed in permeabilisation buffer, and counter stained with Alexa-488 anti-mouse, and Alexa-568 anti-rabbit fab fragment secondaries (Invitrogen), nuclear counterstaining was with Topro3 (Invitrogen), for an hour before washing and mounting (Vectashield, Vector labs). Images were acquired on a Leica SP5 confocal microscope with a 63x oil immersion lens. hSC-CM's were plated onto 0 thickness coverglass, infected at an MOI of 5. Cells were fixed and stained 48 hours after infection as above.

Live cell imaging microscope
An inverted IX81 frame (Olympus, Japan), with a custom 7 LED array (Cairn Research, Faversham, UK), automated stage (Prior Scientific, Cambridge, UK), lens turret, and filter wheel is housed in a custom heated, humidified chamber (Solent Scientific, Segensworth, UK) with image collection on two C-1900 EMCCD cameras (Hamamatsu, Japan) mounted on a beam splitter (Photometrics, UK Stem cell derived cardiomyocyte and chemical calcium dye imaging iPS derived cardiomyocytes were plated out as above, and then loaded with 5μM Fluo-4-AM (Thermo-Fisher, UK) at room temperature for 20 min, free dye was washed off by media replacement with pre-heated culture media, followed by imaging with C-1900 EMCCD (Hamamatsu, Japan) camera using a 5msec 488nm LED illumination pulse. The GFP filter set (DS/FF02-485/20-25, T495lpxr dichroic mirror, and ET525/50 emission filter) was used for Fluo-4 observation.

Image processing & statistics
Raw image data was extracted using CellR, and processed in Excel (Microsoft) MATLAB (Mathworks) and OriginLab7.5 (Origin), unfiltered traces are shown. Numerical data is presented as mean +/-standard deviation. Raw movies were exported to Fiji for processing for publication, movies were compressed and time-stamped using VideoMach (Gromada.com). Samples were compared by Students T-test, significance values P<0.05 is shown by Ã , or P<0.005 by ÃÃ where applicable.

Chemicals
Were purchased from Sigma-Aldrich (Dorset, UK) diluted in DMSO, and diluted in culture medium to final concentrations as stated in the text. Cells were observed 30 minutes after drug addition.

Results and discussion
Optical control and calcium imaging in primary adult cardiomyocytes Although not currently possible, changes to hSC-CM differentiation and culture aspire to produce increasingly adult ventricular phenotypes, hence initial method development was undertaken in primary ventricular cardiomyocytes. We felt the brief in vitro survival of adult cells should bias tool selection to the brightest of the indicator/control tool combinations previously tested in neurons. An adenovirus was made containing the optical control tool, ChETA TC [24] engineered for large photocurrents and rapid inactivation, together with the brightest red shifted calcium indicator R-GECO [21,25]. To prevent unbalanced expression of two transgenes in a single cell [26] a rapid self-cleaving 2A peptide [27] was used (S1 Fig). Guinea pig ventricular cardiomyocytes survive 72 hours once isolated allowing 48 hours for reporter production ( Fig 1A) after infection. Electrical stimulation confirmed the ability of the calcium  (Fig 1B and 1C, S1 Movie). Laser illumination enables control of the size, intensity, and duration of the stimulating light spot. Discrete observation and control windows in single cells can therefore be made ( Fig 1D) overcoming imaging artefacts previously tackled by syncytial approaches where optical control, and reporter tools are expressed in different cell populations [14,28].

Optical control and calcium imaging in stem-cell derived cardiomyocytes
The growth of human genomes represented in iPS repositories, and the ease of hSC-CM differentiation, provides a bridge between computer models and patients [29]. Existing differentiation strategies produce cells with APD and cycle length spanning an order of magnitude [30] particularly at low cell density [31]. This variability may hinder identification of small effect sizes relevant to human health, for example a QT interval of 450msec is normal, whereas 500msec is pathological [32]. Although adult cardiomyocytes do not spontaneously depolarise, the hSC-CM's do (those used here beat at 0.52Hz +/-0.15Hz). Some cells have lower (<0.2Hz) rates of spontaneous activity. We reasoned this group might represent an electrically consistent population for further testing as spontaneous activity would not break through an applied optical pacing regime. Programmed optical stimulation at lower frequencies ( Since lower triggering frequencies minimise potential phototoxicity, and photoactivation artefact, while maximising signal to noise, low frequency stimulation at 0.3Hz and 0.6Hz was used in further work. Drug testing can both be done by paired measurements before and after compound addition using the same cell as its own internal control [14,16]; or as an unpaired assessment compared to a reference population (S6 Fig, and approximately half of the patient disease models summarised in [3]). The first approach might improve sensitivity to small drug effects, whereas the second approach is simpler but vulnerable to intrinsic variability in a sample which may preclude this option. We find that the cell selection strategy combined with optical stimulation limits baseline variability and makes the second approach feasible, although either is possible (S6 Fig).
Application of flecainide (sodium (INa) channel inhibitor), dofetilide (potassium (hERG) channel inhibitor), cisapride (serotonin receptor inhibitor with off-target hERG block), or the voltage dependent L-type calcium channel inhibitors verapamil (has off target hERG inhibition), nifedipine, and diltiazem were tested at clinically relevant concentrations ( . All calcium channel blockers suppress the triggered intensity change of the calcium indicator suggesting this alone may identify such compounds. However loss of indicator brightness represents a weakness in this strategy as although the anticipated CTD shortening with nifedipine was seen, the opposite trend due to off-target hERG block by verapamil was missed even though this was detected in parallel patching studies. Cisapride and dofetilide (Fig 4A, 4C and 4D) both cause dose-dependent prolongation of CTD, with early ( Fig 4B1) and after (Fig 4B2) depolarisation transients at higher doses. Flecainide demonstrated a complex response, reducing the activation slope of the calcium transient, and also prolonging it. hERG inhibition by flecainide [33] at this clinical dose is reported, at higher doses this effect is less marked (Fig 4E and 4F) reflecting the integrated sum of other off-target effects.

Optical control and calcium imaging in pancreatic beta-cells
The need for simultaneous calcium transient phenotyping and control is not a unique cardiomyocyte problem suggesting the method may be useful in other excitable cell types. The betacell of the endocrine pancreas uses a glucose triggered voltage change to produce a calcium transient enabling insulin release by exocytosis. Glucose causes ATP levels to rise, increasing the inhibition of an ATP dependent potassium channel K ATP reducing repolarisation, and causing calcium release. Augmentation of insulin release by chemical inhibition of K ATP using  the sulphonylurea class of drugs has been a mainstay of Type II Diabetes management for 60 years [34]. Optical control and calcium imaging has been tried in this model [35,36] but although insulin accumulation could be measured biochemically the underlying calcium transients were either not detected, or drift upward as the excitation spectrum of the calcium indicators (Fluo-4-AM, and Fura2-AM respectively) overlaps with the blue-green activation spectrum of the optical control tool causing unintended ChR2 activation during imaging.
As proof of concept that simultaneous optical control and calcium imaging could be improved the immortalised rat insulinoma beta-cell model (INS-1) was infected and imaged as above in the presence or absence of the sulphonylurea gliclazide at low (3mM) and activating (9mM) glucose concentrations. In the absence of optical control, a chaotic pattern of random calcium activation is apparent (S3 Movie, S7  (Fig 5D) producing coordinated datasets from which parameters equivalent to the cardiomyocyte can be derived (Fig 5E, Table 3). When exposed to 10μM gliclazide prolongation of CTD50 and CTD90 was observed at 9mM glucose, at 3mM glucose only a change in CTD50 was seen, peak calcium intensity increases in both 3mM and 9mM glucose, as summarised in Table 3 and Fig 5F and 5G.   Fig 3. Small molecule ion channel inhibitors applied to hSC-CMs. hSC-CMs were plated out at low density and exposed to clinically relevant doses of known small molecule inhibitors of the sodium channel NaV1.5 (flecainide 0.5μM), the potassium channel KCNH2/hERG (dofetilide, 5nM, and cisapride, 30nM), and the voltage gated calcium channel CACNA1.2 (nifedipine, 100nM; verapamil 1μM, diltiazem 1μM). Traces obtained at 0.3Hz, and 0.6Hz are shown for the annotated small molecule as before with the average response shown by the thick line. Analysis of the data is presented in Tables 1 and 2. https://doi.org/10.1371/journal.pone.0174181.g003

Discussion
The principle limitations of the method in its current form is that although it identifies that cellular behaviour is altered, unlike patching, it is unable to distinguish effects of individual ion currents through which effects might be occurring. As such it is not a direct replacement for patching. Similarly at the moment transgenes are delivered by virus. This means that primary cardiomyocytes infected in vitro are often used towards the end of their natural life as time for gene expression is required with inevitable loss of quality. This can be offset by in vivo infection, and delayed cell isolation, but at the expense of additional procedures such as direct myocardial injection which limits the general utility of the approach. Furthermore, viral infection produces a host cell response that may alter the overall performance of any cell. This should be possible to overcome either via transgenic approaches of gene targeted knock-in or perhaps use of alternative viral strategies engineered to elicit less host cell reaction.
Improving the maturity of stem-cell derived cardiomyocytes will lead to pacing dependence. As these tools are genetically encoded, and equivalents have already been utilised in transgenic models [9,37], genome knock-in lines with constitutive and homogeneous expression may become useful adjuncts to this approach. Indeed, further refinement identifying particular cardiomyocyte subtypes by targeting promoters active in the atrium or ventricle, as demonstrated for lentiviral promoter:transgene combinations in SC-CMs [38] may be possible.
Methodological improvements may arise by combining hyperpolarising and depolarising optical control tools to completely suppress intrinsic activity independently of innovations in differentiation or cell maintenance. Although this would increase the number of cardiomyocytes suitable for study, compatible indicators are currently limiting for this approach [39]. Alternatively, developments allowing multi-parameter (voltage, calcium and contraction) measurements under optical stimulation would enable creation of tools to explore the interaction between these aspects of cardiac physiology in health and disease states and how those changes can be influenced by small molecules.

Conclusion
This method provides a contactless all genetic single cell assay with temporal stimulation control over the physiological range of cardiomyocytes and pancreatic beta-cells. Anticipated small molecule effects were detected during brief experimental periods on low cell numbers in two model systems. The approach is quick, simple, and can be applied to microscopes with conventional blue/green/red imaging capabilities, using a single virus and isolated cells. However it increases data storage, data processing requirements, slows down throughput and may be more vulnerable to phototoxicity. An alternative strategy compares drug exposed cells to a reference population. This is more useful when cells show a consistent behaviour. We find that the cell selection strategy combined with optical stimulation enables either approach, even at the lowest (0.3Hz) stimulation frequency where variability is greatest. In the paired experiment CTD90 rises from 0.82 +/-0.13s to 1.72 +/-0.37s (p<0.005), in the unpaired experiment it rises from 0.535+/-0.18s to 1.32 +/-0.34s (p<0.005).