Optogenetic confirmation of transverse-tubular membrane excitability in intact cardiac myocytes

T-tubules (TT) form a complex network of sarcolemmal membrane invaginations, essential for well-coordinated excitation-contraction coupling (ECC) and, thus, homogeneous mechanical activation of cardiomyocytes. ECC is initiated by rapid depolarization of the sarcolemmal membrane. Whether TT membrane depolarisation is active (local generation of action potentials; AP) or passive (following depolarisation of the outer cell surface sarcolemma; SS) has not been experimentally assessed in cardiomyocytes. Based on the assessment of ion flux pathways needed for AP generation, we hypothesise that TT are excitable. We therefore explored TT excitability experimentally, using an all-optical approach to stimulate and record trans-membrane potential changes in TT that were electrically insulated from the SS membrane by transient osmotic shock. Our results establish that cardiomyocyte TT can generate AP. These AP show electrical features that differ substantially from those observed in SS, consistent with differences in the density of ion channels and transporters in the two different membrane domains. We propose that TT-generated AP represent a safety mechanism for TT AP propagation and ECC, which may be particularly relevant in pathophysiological settings where morpho-functional changes reduce the electrical connectivity between SS and TT membranes. KEY POINTS Cardiomyocytes are characterized by a complex network of membrane invaginations (the T-tubular system) that propagate action potentials to the core of the cell, ensuring synchronous and uniform cell contraction. In this study, we investigate whether the T-tubular system is able to generate action potentials autonomously, rather than following depolarization of the outer cell surface sarcolemma. For this purpose, we developed a fully optical platform to probe and manipulate the electrical dynamics of sub-cellular membrane domains. Our findings demonstrate that T-tubules are intrinsically excitable, revealing distinct characteristics of self-generated T-tubular action potentials. This active electrical capability may serve as a protective mechanism against voltage drops occurring within the T-tubular network.

Despite indications of non-homogeneous distribution of ion flux pathways and chemical gradients in SS and TT compartments, there are no significant differences between action potential (AP) profiles recorded in SS and TT membranes of healthy cardiomyocytes, presumably because the two membrane compartments are electrically well-coupled (Sacconi et al., 2012).The characteristic length constant of electrotonic potential decay in TT is one order of magnitude larger than a typical cardiomyocyte radius (Scardigli et al., 2017), indicating a remarkable safety factor for co-ordinated voltage changes across the entire sarcolemma in healthy cardiomyocytes.This may be impaired, however, in cardiac diseases where substantial TT remodelling has been observed (Crocini et al., 2014;Sacconi et al., 2012).
This poses an interesting and as yet experimentally unresolved question: is TT membrane depolarization, and hence well-co-ordinated initiation of ECC across the entire cell volume, dependent on active AP generation in TT, or is it driven by passive electrotonic effects from AP generation in the SS membrane?
We therefore implemented an experimental investigation which involved, first, electrical insulation of TT from SS membranes and, second, an all-optical approach to locally depolarize TT or SS membranes at the same time as tracking changes in their trans-membrane potentials.Electrical insulation of TT from SS membranes was achieved by transient exposure of cells to an acute osmotic shock (Ferrantini et al., 2014;Kawai et al., 1999), which ties off TT near their junctions with the SS membrane.Using random-access multiphoton (RAMP) microscopy and cardiomyocytes containing a combination of either CheRiff or ChR2 (optogenetic actuators) (Hochbaum et al., 2014) and di-4-AN(F)EPPTEA (voltage-sensitive dye; VSD) (Yan et al., 2012), we found that TT membranes are electrically excitable (i.e.able to generate local AP even after electrical insulation from the SS).

Animals
We used 12-week-old female transgenic mice (ChR2-mhc6-cre +) with cardiomyocyte-specific expression of ChR2 (H134R variant) and age-matched mice from the same strain (C57BL/6J).Animal handling and procedures were performed in accordance with guidelines in Directive 2010/63/EU of the European Parliament (protection of animals used for scientific purposes).The experimental protocol was approved by the Italian Ministry of Health (authorization number 944/2018-PR).

Adeno-associated virus-based gene transfer of CheRiff
Mouse hearts were driven to express the optogenetic actuator CheRiff (Hochbaum et al., 2014) coupled with enhanced green fluorescent protein (eGFP) by adeno-associated virus (AAV) based transduction using AAV2/9.45.GgTnnt2.CheRiff∼eGFP.WHVPRE.SV40pA particles.This AAV (Pulicherla et al., 2011) contains a transgene consisting of the chicken cardiac troponin T promoter, the coding sequence of the blue light-activatable cationic channelrhodopsin CheRiff tagged at its C terminus with eGFP, the woodchuck hepatitis virus post-transcriptional regulatory element and the simian virus 40 polyadenylation signal.Production, purification and quantification of these vector particles was carried out as described in Nyns et al. (2017).Systemic delivery of AAV particles was performed as described in Yan et al. (2012).Briefly, AAV particles were injected into the left jugular vein of mice anaesthetized with isoflurane (3% for induction and 1.5% for maintenance, in oxygen).Body temperature was maintained at 37°C using a homeothermic blanket system (#50300; Stoelting Co., Wood Dale, IL, USA) and mice were monitored using respiratory rate and toe pinch response tests throughout the procedure.Mice were placed on a stereotaxic apparatus and shaved over the chest to allow identification of the jugular vein underneath the mouse skin, after which the shaved area was sterilized with betadine and 70% ethanol.Next, lidocaine solution was applied to the skin surface and a 5 mm long cervical incision was made to expose the jugular vein.An insulin syringe (30G × 1/6 ; Biotekne, Casalecchio di Reno, Italy) was used to slowly inject 200 μL of the injection mix composed of ∼2 × 10 11 genome copies of AAV diluted in phosphate-buffered saline (P4417; Merck, Rahway, NJ, USA) containing 2.5 mm MgCl 2 and 1 mm KCl.After stopping of the bleeding with surgical swabs, the skin was sutured with Vicryl Plus Ethicon (VCP394H; Johnson & Johnson Medical NV, Machelen, Belgium) and mice were placed back into their cages.Tail vein injection was also explored.To this end, tails were cleaned and sterilized with 70% ethanol and a bent needle of an insulin syringe was inserted into one of the lateral veins, followed by the slow administration of the same injection mix.At 4 weeks post-transduction, hearts were excised and used for the isolation of ventricular cardiomyocytes as described below.

Quantification of T-tubular density
Isolated cardiomyocytes, both intact (CTRL) and formamide-treated (DETUB) were incubated for 20 min with 5 μm of Di-3-ANEPPDHQ (Thermo Fisher Scientific, Waltham, MA, USA) membrane dye.This approach allowed us to investigate the fraction of detached T-tubules because only those still connected with the surface were labelled.CRTL and DETUB cells were resuspended in calcium-free HEPES-buffered Tyrode's solution, placed into a glass-bottomed chamber (P50G-0-14-F; MatTek, Ashland, MA, USA) and observed using a 63× oil immersion objective on a TCS5 confocal microscope (Leica, Wetzlar, Germany).The excitation wavelength was 488 nm and the detection window was set to 504-649 nm.TT quantitative analysis was performed using AutoTT (Guo & Song, 2014) on confocal images, as previously described (Crocini et al., 2016).In brief, AutoTT skeletonizes the global architecture of the TT system to extract morphological patterns, discriminating transverse and axial elements of the system.To determine the densities of transverse and axial elements, the total number of pixels in each of these TT subcategories was divided by the total number of pixels of the intracellular region imaged.

Patch clamp experiments
To perform patch clamp recordings, viable ventricular cardiomyocytes were resuspended with HEPES-buffered Tyrode's solution containing 1.8 mm CaCl 2 at RT. Patch clamp data recording and analysis were performed using a Multiclamp700B amplifier in conjunction with pClamp10.0and a DigiData 1440A AD/DA interface (Molecular Devices, San Jose, CA, USA).AP recordings and K + current measurements were conducted in whole-cell ruptured patch mode.The pipette solution contained (in mm): 130 K-aspartate, 0.1 Na-GTP, 5 Na 2 -ATP, 11 EGTA, 5 CaCl 2 , 2 MgCl 2 and 10 HEPES, pH 7.2 adjusted with KOH; free [Ca 2+ ] was calculated using maxchelator (https://somapp.ucdmc.ucdavis.edu/pharmacology/bers/maxchelator/) and was equal to 243 nm.For K + current measurements, 0.3 mM CdCl 2 was added to the bath solution.AP trains were electrically elicited with short depolarizing stimuli (3 ms square current pulses, 500-1000 pA) at 1 Hz pacing rate.Recordings of global potassium current (I K ) were carried out using depolarization steps of 900 ms, from −50 mV to 0 mV, immediately after eliciting the Na current.I K-peak was calculated as the maximum current, whereas I K-end is the residual outward current at 900 ms.For L-type calcium (I CaL ) and sodium (I Na ) current recordings, the pipette solution contained (in mm) 120 CsCl, 10 HEPES, 5 Na 2 ATP, 10 TEACl, 5 MgCl 2 , 4.3 CaCl 2 and 10 EGTA (calculated free [Ca 2+ ] = 226 nm), pH 7.2, adjusted with CsOH, whereas the external solution contained (in mm) 135 NaCl, 5.4 CsCl, 10 glucose, 10 HEPES, 1 MgCl 2 , 1.8 CaCl 2 , 0.33 NaH 2 PO 4 and 0.3 BaCl 2 , pH 7.4, adjusted with CsOH.Peak sodium current (I Na ) was elicited with a 15 ms depolarizing step from −80 mV (holding potential) to −45 mV, whereas I CaL was elicited with a longer (200 ms) depolarization step from −45 mV to 0 mV, immediately eliciting the Na current.I Na and I CaL were induced at a repeat rate of 1 Hz.

Modelling
The model used in the present study is based on a previously published model of rat ventricular cardiomyocytes incorporating TT (Pasek et al., 2006) and was modified to reproduce the key properties of the mouse ventricular cardiomyocytes (Bondarenko et al., 2004).Modifications included: (1) adjustment of geometric parameters of the model cell; (2) reformulation of ion flux pathways in SS and TT membranes; (3) adaptation of sarcoplasmic reticulum Ca 2+ pump activity; and (4) adjustment of ion exchange between the TT lumen and the extracellular space.All modifications are described in detail in the appendix.The model was implemented in MATLAB, version 7.2 (MathWorks, Inc., Natick, MA, USA).Computation of the system of 47 non-linear differential equations was performed using the solver for stiff system ode15s (MATLAB).Model equations were solved simultaneously, using a variable time-step (0.4 × 10 −6 -0.1 s) adjusted to keep the estimated relative error of inner variables below a threshold value of 10 −6 .Steady-state behaviour of the model was achieved by running it for 600 s of equivalent cell lifetime at the specified stimulation frequencies.The MATLAB code of the model is available at https://www.it.cas.cz/en/d3/l033.

RAMP microscope and optogenetics stimulation
The basic design of our RAMP microscope has been described elsewhere (Crocini et al., 2014;Sacconi et al., 2012;Scardigli et al., 2017).In the present study, the system was implemented with a fibre-based optical stimulation system for optogenetic pacing.A LED (M470F3; Thorlabs, Bergkirchen, Germany) operating at 470 ± 25 nm was employed as a fibre-coupled light source and a 3-D micromanipulator was used to place the distal end of the multimode optical fibre (FG200UEA; Thorlabs) in proximity of the cell.The numerical aperture (N.A.) of the optical fibre and the angle of the fibre with respect to the main optical axis were chosen to avoid direct collection of blue light by the RAMP excitation/detection objective (40×, water immersion, N.A. 0.8; Olympus, Tokyo, Japan).This is crucial, given that blue light can generate unwanted phosphorescent signals inside the objective glass that may affect optical recordings (not shown).The RAMP fluorescence signal was detected using a photon-counting module, based on the GaAsP photomultiplier tube (PMT) with an ultra-fast gate circuit (H11706P-40; Hamamatsu, Shizuoka, Japan).An emission filter of 655 ± 20 nm was used for fluorescence detection.A trigger distributor switched off the high-voltage power supply circuit of the PMT during optical stimulation.This avoids the possibility that endogenous or exogenous fluorescence signals overload the PMT during blue light illumination.Cells were optically paced using pulses of blue light with a duration of 3 ms and a light intensity of 10 mW mm −2 at the output of the optical fibre.Alternatively, cells could be electrically field-stimulated using two parallel platinum wires (diameter 250 μm), placed in the bath at a distance of 6.3 mm.Square pulses of 10-20 V (swapping anode and cathode), lasting 3 ms, were used to trigger AP.In a typical RAMP measurement, we performed one line-scan along the SS, and two to four line-scans on different TT membrane segments.The length of scanned lines ranged from 5 to 10 μm, with an integration time per membrane pass of ∼200 μs, leading to a temporal resolution of 0.4-2 ms.Optical recordings in detubulated cells were performed within 30 min after the formamide treatment to limit a potential luminal ion concentrations variation of sealed-off TT.Opto-electronic components of the setup were computer-controlled with custom-made software developed in LabVIEW 7.1 (National Instruments, Austin, TX, USA).Optical recordings were analyzed with a custom-made software written in LabVIEW 2013 (National Instruments).

Statistical analysis
For each experimental condition, we assess whether measured data came from a normal, log-normal or other distribution, performing a Shapiro-Wilk test of all data before and after their logarithmic transformation.The median was used as the best representation of central data values.In patch clamp experiments, data from one cell were averaged, and this average was used for comparison and statistical analysis.One-way analysis of variance (ANOVA) tests were used to compare electrophysiological features between CTRL and DETUB.In optical recordings, two-way ANOVA tests were used to compare electrophysiological features between SS and TT at different stimulation modalities, the Bonferroni post hoc analysis was used.AP optical stimulation efficiency in CTRL and DETUB was compared using Fisher's exact test.Statistical analysis was performed using OriginPro 2018 (OriginLab, Northampton, MA, USA).

Ion flux pathways across sarcolemma in mouse cardiomyocyte and in silico assessment of TT excitability
Structural and patch clamp data from ventricular cardiomyocytes which underwent acute detubulation through J Physiol 602.5 osmotic shock (DETUB) were compared to intact cells (CTRL) to assess AP-relevant ion current distributions between TT and SS (Fig. 1).Our data indicate that L-type Ca 2+ channels are expressed predominantly in the TT membrane (note: I CaL density was reduced by ∼0.5 in DETUB, compared to CTRL cells), whereas Na + and K + channels densities appear to be uniform across the whole cell membrane.Consistent with the high I CaL density in TT, a reduction of AP duration at 50% and 90% repolarization was found in DETUB cells, compared to CTRL, whereas no significant differences in AP amplitude or resting membrane potential were detected between the groups.
Measured ion flux pathways were incorporated into an adapted computer model of mouse ventricular cardiomyocytes with two ECC domains: SS and TT (for a detailed description, see the Appendix).The model predicts that, based on available electrophysiological data, TT are excitable and AP duration of isolated TT is longer than in SS alone (Fig. 2A).Nevertheless, the model also showed that co-ordinated ECC could be ensured both with active AP generation, or with passive electrotonic conduction, in TT (Fig. 2B) and that it is not possible to distinguish between these two possible mechanisms, based on available spatially resolved data on ECC and contraction dynamics in cardiac myocytes.

Optical induction and recording of AP in SS and TT of CTRL cells
RAMP microscopy was combined with optogenetic stimulation, employing an ultra-fast gated detection system (Fig. 3A).The experimental approach was first validated in intact ventricular myocytes, constitutively expressing a light-gated ion channel (ChR2) and labelled with VSD.RAMP rapidly scanned linear segments on SS and TT membranes during electrical field stimulation or whole-cell optical stimulation.Both stimulation modes allowed AP induction in SS and TT membranes of CTRL cells (Fig. 3B and C).As expected, electrically and optically stimulated AP amplitudes and kinetics in TT are not significantly different from those in SS (Fig. 3D-F).
Comparing AP features between the two stimulation modalities, we found a substantially longer time-to-peak AP (∼5-fold, 15 ms instead of 3 ms; TTP in Fig. 3E) during optical compared to electrical stimulation, as previously reported (Bruegmann et al., 2010).Effects of ChR2 activation on AP repolarization (e.g.APD 50 in Fig. 3F) were less pronounced or absent.These results indicate that a full-optical approach can be used to quantitatively assess AP dynamics in different sarcolemmal domains.
Of note, AP amplitudes are lower than what was found in our previous investigation (∼20%) (Crocini et al., 2016).This is not unexpected considering that, in the present mouse line (Zaglia et al., 2015), the 1064 nm light excites both the VSD and ChR2-conjugated td-Tomato fluorescent protein (Drobizhev et al., 2011).In a subset of experiments where cardiomyocytes from CTRL mice were transfected with light-activated channels fused with eGFP (not excited at 1064 nm), we found (Fig. 4) that the relative  fluorescence change at the AP peak is restored to previously reported values, both upon electrical and optical stimulation.

Optical induction and recording of AP in SS and TT of DETUB cells
Next, we employed the same approach on ChR2-expressing cardiomyocytes that underwent acute detubulation.As indicated in Fig. 5A, cells were first stained with the VSD and then acutely detubulated by osmotic shock.Thereafter, AP were optically recorded from SS and TT membranes upon electrical or optical stimulation (ten pulses each) (Fig. 5B).As expected, upon electrical stimulation, TT that were detached from the SS did not exhibit AP (grey traces), showing no membrane voltage variation above the shot-noise (∼5 mV).This means that structurally detached and electrically uncoupled TT are not excited by either the local extracellular field stimulation potential (which is Faraday-screened by the SS) or passively depolarized by AP elicited in SS membrane.Crucially, however, some of electrically disconnected (from the SS) TT can be triggered to generate AP using optical stimulation (AP+) (Fig. 5B, left).The amplitude distribution of optically induced AP in SS and TT is depicted in Fig. 5C.Although AP amplitude in SS can be described by an unimodal distribution, TT membrane potential exhibits a pronounced bimodal profile, where most TT are silent (AP-) (Fig. 5B, right) but a fraction of TT (AP+) are intrinsically excitable.This bimodal distribution of AP amplitudes in TT suggests that AP require activation of voltage-dependent channels with threshold for activation (e.g.NaV1.5), rather than simply reflecting ChR2-induced inward currents.This is further confirmed in patch clamp experiments on single cardiomyocytes, where the effect of ChR2-induced photocurrents on the membrane potential was evaluated in the presence or absence of Na + and Ca 2+ channel blockers (Fig. 5D).
In keeping with in silico predictions, isolated TT that are able to generate AP (AP+) show slower repolarization compared to SS (increased AP duration at 50% repolarization), unravelling intrinsic features of TT-generated AP (Fig. 5E).
The fraction of TT that were optically excitable after osmotic detubulation (AP+) was only 10-15% of the structurally disconnected TT (Fig. 6A).We exclude the possibility that a low photo-current density would account for the lack of TT excitability in the majority of TT, based on the homogenous sarcolemmal distribution of ChR2 measured in this transgenic mouse model (Fig. 6B).However, a reduction of membrane excitability in sealed-off TT is not unexpected: luminal ion concentrations are no longer clamped to bulk extracellular levels, and this can affect the local electrochemical gradient and thus channel/transporter functionality.

Discussion
In the present study, we exploit the utility of light to probe crucial electrophysiological properties of SS and TT membranes in living cardiomyocytes.Our approach combines the advantages of RAMP microscopy, in terms of spatio-temporal resolution for local AP recording, with optogenetics for precise spatio-temporally controlled stimulation of subcellular membrane domains.Applying this methodology to cardiomyocytes where TT were structurally disconnected from the SS membrane by DETUB, we confirm that TT membranes are electrically excitable, such that they are able to generate AP, locally and independently from the SS.We further identify dissimilarities in AP shape between the two membrane domains that are in keeping with differences in their Ca 2+ current density: AP duration at 50% repolarization is longer and AP-shape is more triangulated in TT than in SS.
We found that only a fraction (10-15%) of sealed-off TT is optically excitable.Although the full set of transporters and channels is present in the TT membrane of ventricular mouse cardiomyocytes (Berry et al., 2007;Bossuyt et al., 2009;Despa & Bers, 2007;Gadeberg et al., 2017a;Swift et al., 2007;Thomas et al., 2003), in the limited luminal volume of vacuolated TT, ion concentrations can fluctuate not only during cell activity, but also during periods of quiescence such as required to implement the protocol for the present study (15-30 min).There are many reasons for which a sealed membrane compartment may lose excitability.Indeed, accumulation of K + and depletion of Ca 2+ have been predicted and experimentally measured in sealed-off TT (Moench & Lopatin, 2014;Pasek et al., 2006).To illustrate this, ion fluctuations were phenomenologically introduced by increasing TT Na + background conductance in our in silico model.Figure 7 shows how a background Na + conductance may lead to time-dependent variations of TT ion concentrations and resting membrane potential.For example, if TT background Na + conductance is enhanced, TT resting potential depolarizes, so that the fraction of inactivated voltage-gated Na + channel increases, and TT excitability is disabled.In vacuolated TT, local luminal Na + concentration is continuously reset by the balance between passive Na + efflux (through voltageor stretch-activated Na + channels and forward-mode of NCX) vs. active Na + influx (mainly through Na + -K + ATPase pump).The behaviour of these channels and transporters is difficult to predict and potentially complicated by the presence of neuronal isoforms of voltage-gated Na + channels (i.e.Nav1.1, Nav1.3 and Nav1.6) (Brette & Orchard, 2006;Maier et al., 2002;Maier et al., 2004;Vermij et al., 2020).Similar variations of vacuolated TT membrane resting potential can be predicted if K + fluxes are dysbalanced as a consequence of K + accumulation in the TT lumen (with consequent depolarization of the K + equilibrium potential).
TT membrane excitability had previously been reported for skinned fibres of vertebrate skeletal muscle (Macdonald & Stephenson, 2004;Ortenblad & Stephenson, 2003;Pedersen et al., 2005;Posterino et al., 2000).In those experiments, the SS of muscle fibres was mechanically removed by rolling back the surface sarcolemma with forceps, forming a sort of 'cuff' , whereas part of the TT system sealed off.In our report on isolated living cardiomyocytes, electrical uncoupling between SS and TT was achieved using transient osmotic shock that physically detaches TT membranes from the cell surface and, at the same time in principle, maintaining overall cell integrity, with intact SS and TT system components that are still operational and a near-native intracellular environment.
By contrast to skeletal muscle, cardiomyocyte activation in vivo occurs via electrotonic interaction with neighbouring cells, AP duration is longer (by one to two orders of magnitude, depending on species) and ECC involves trans-sarcolemmal calcium fluxes, making cells highly responsive to changes in local ion concentrations and regulatory influences on ion flux pathways.For these reasons, we consider that the osmotic shock-based approach is more appropriate for cardiomyocytes compared to membrane stripping.
Our study shows that TT do not merely represent a 'current sink' compartment that is driven by the SS 'source' , but rather that TT are part of an overall excitable membrane system in which SS and TT affect each other.We are unable to assess whether this is a general feature that applies to the entire TT membrane system.However, even if TT membrane excitability was non-uniform, any presence of excitable TT segments would affect AP conduction within the TT because such regions would act as repeater stations, reconditioning the AP signal amplitude.Several considerations regarding the physiological and patho-physiological consequences of TT excitability can be made.
The capacity of TT to generate AP may help to sustain spatio-temporal homogeneities of membrane depolarization within the TT network, which may be challenged when the TT system is structurally remodelled, such as in disease.In healthy cardiomyocytes, the characteristic length constant of electrotonic conduction along TT membranes is one order of magnitude larger than cell radius (Kong et al., 2018;Scardigli et al., 2017), although the utility of this length constant as a measure to predict spatial voltage uniformity within the TT network is questionable in supra-threshold voltage ranges (i.e. because membrane resistance markedly varies during the AP).The scenario becomes even more complex in atrial cardiomyocytes, where the transverse tubular system is less developed, whereas axial elements are more dominant (Brandenburg et al., 2016;Brandenburg et al., 2018;Richards et al., 2011).Furthermore, in healthy ventricular cardiomyocytes, no significant voltage gradient is expected, whereas, in a pathologically remodelled tubular system (Ibrahim et al., 2011), the resistance of the TT system may significantly increase and some tubular elements can be poorly coupled or even electrically uncoupled from the SS (Sacconi et al., 2012).
The pathophysiological relevance of TT excitability remains to be elucidated.Interestingly, behaviour similar to that observed in the present study has been reported previously in pathological conditions (Crocini et al., 2014;Crocini et al., 2016) where, in electrically isolated TT, local supra-threshold voltage transients were observed.These spontaneous TT depolarizations may trigger large local SR Ca 2+ releases (voltage-associated Ca 2+ sparks, V-sparks) and potentially initiate arrhythmias.Future investigations will focus on exploring the relevance of TT excitability in cardiomyocytes for which ECC homogeneity has been challenged as a consequence of a pathologically remodelled TT system, employing the experimental approach described here to dissect TT and SS excitability and AP features.At present, however, we can unequivocally confirm that TT membranes in intact cardiomyocytes can be electrically excitable.

Appendix
A biophysically detailed computer model of a mouse ventricular cardiomyocyte, incorporating both surface sarcolemma (SS) and t-tubular membrane (TT) was used (Fig. A1).

Geometric parameters of cell and of cellular
compartments.The total volume of the model cell (V c ) was set to 10 pL.The fractional volumes of cytosolic space and dyadic space, as well as of network and junctional (release) compartments of the sarcoplasmic reticulum (NSR, JSR) are as in the original rat ventricular cardiomyocyte model (Pasek et al., 2006): 0.585, 75 × 10 −6 , 0.0315 and 0.0035, respectively.The fractional volume of TT was set to 3.2% (Clark et al., 2001).The conclusion that about half of the total sarcolemmal surface is in TT in mouse ventricular cardiomyocytes is consistent with a previous optical study by Forbes et al. (1984).Applying such data to a model with a total membrane area of 10,000 μm 2 and a specific membrane capacitance of 1 μF cm -2 , the modelled TT and SS have an identical capacitance of 50 pF.
Membrane currents.To simulate more closely the behaviour of mouse ventricular septal cardiomyocyte (Anumonwo et al., 2001), the following modifications of membrane transport systems were implemented.
(i) The conductance of I Na -channels (g Na ) was increased from 10 to 13.3 mS cm -2 to adjust the maximum upstroke velocity to ∼150 V s -1 as observed experimentally (Anumonwo et al., 2001).Additionally, the membrane Na + transport was supplemented by a formulation of late Na + current (I NaL ), adopted from the mouse ventricular myocyte model of Morotti et al. (2014).The corresponding conductance (g NaL ) was set to 6.5 μS cm -2 .(ii) The permeability of I Ca -channels (P Ca ) was decreased from 2.1 × 10 −4 to 0.7 × 10 −4 cm s -1 and the time constant of voltage-dependent activation of I Ca (τ Ca,VDA ) was adjusted to reduce the rate of I Ca deactivation in later phases of AP repolarization (negative to −20 mV), allowing the model AP duration at 75% repolarization to approach the value of 20.9 ± 1.3 ms, measured in mouse ventricular cardiomyocytes at 1 Hz stimulation (Guo et al., 1999).(iii) The conductance of I Kto -channels (g Kto ) was decreased from 0.35 to 0.146 mS cm -2 and gating characteristics of the channels were reformulated to mimic the parameters of early AP repolarization (APD 25 = 4 ± 0.2 ms and APD 50 = 8.5 ± 0.3), measured at 1 Hz (Guo et al., 1999).(iv) The formulations of I Ks -channels (non-inactivating steady-state K + current) and of the newly added I Kur -channels (ultra-rapid component of the delayed rectifier K + current) were adopted from the model of mouse ventricular myocytes developed by (Pandit, 2003) that is available in the CellML model repository (https://models.cellml.org/cellml).3 mS cm -2 0.5 g Nab 1.2 μS cm -2 0.5 g NaL 6.5 μS cm -2 0.5 g Cab 0.367 μS cm -2 0.5 P Ca 0.7 × 10 −4 cm s -1 0.87 g Kb 1.380 μS cm -2 0.5 g Kto 0.146 mS cm -2 0.5 k NaCa 0.18 nA cm -2 mM -4 0.5 g Kss 0.075 mS cm -2 0.5 I NaK,max 1.4 μA cm -2 0.5 g Kur 0.085 mS cm -2 0.5 I pCa,max 0.21 μA cm -2 0.5 g K1 0.159 mS cm -2 0.5 f x,t represents the TT fraction of the ion transporter underlying current I X .
The related conductances g Kss and g Kur were set to 0.075 and 0.085 mS cm -2 , respectively.(v) Other changes to the original rat model include: decrease of g K1 from 0.24 to 0.159 mS cm -2 , omission of hyperpolarization-activated current (I f ), increase of g Nab (background Na + conductance) from 0.8 to 1.2 μS cm -2 , decrease of g Cab (background Ca 2+ conductance) from 0.648 to 0.367 μS cm -2 (Anumonwo et al., 2001), increase of I NaK,max (maximal density of Na + -K + ATPase current) from 1 to 1.4 μA cm -2 and of Na + -half-saturation constant for I NaK from 10 to 21 mm (Anumonwo et al., 2001), and decrease of I pCa,max (maximal density of sarcolemmal Ca 2+ -pump current) from 0.85 to 0.21 μA cm -2 (summarized in Table A1).
Consistent with the absence of changes in I Na , I K , and membrane resting potential after detubulation as well as with excess of I Ca at the TT membrane (Fig. 1), we simplified the distribution of ion transporters by setting all fractions to 0.5 (uniform distribution between the TT and SS membranes), with the exception of L-type Ca 2+ current (f Ca ).
f Ca was estimated on the basis of measurements of membrane capacitance and corresponding current densities in intact and detubulated cardiomyocytes (C CTRL , C DETUB , I x,CTRL , I x,DETUB ) and relations derived in Pasek et al., 2017;Pasek et al., 2021): Using the mean values of I Ca , CTRL and I Ca,DETUB (6.5 and 3.26 pA/pF, respectively) and inserting these data into eqns 4 and 5 gives I Ca,ratio of 6.9 and f Ca of 0.87, which corresponds well with the previous estimates done by Gadeberg et al. (2017b) suggesting predominant localization of I Ca -channels in TT of mouse ventricular cardiomyocytes.The electrical properties of all ion transport systems used in the model are summarized in Table A1.Intracellular Ca 2+ -handling.To simulate the increase of the systolic and diastolic maxima of intracellular Ca 2+ transients with stimulation frequency observed in mouse ventricular cardiomyocytes by Ito et al. (2000), the forward rate parameter of SR Ca 2+ ATPase was decreased from 0.4 to 0.2 mm s -1 and the slope factor controlling the [Na + ] i dependence of Na + /K + pump was increased from 1.5 to 2.15.
Rate of ion exchange between TT and external space.
The time constant of K + exchange between the TT and external space (τ K,te ) was set to 50 ms to comply with the average rate of tail current decay observed at the end of depolarizing steps from −80 mV to the voltages between −30 and 40 mV (Clark et al., 2001).τ Na,te was set to the same value as τ K,te and τ Ca,te was set to 167 ms to respect the ratio τ Ca,te / K,te in the original rat model (Pasek et al., 2006).
Basic behaviour of the model.The basic behaviour of the mouse ventricular cell model is illustrated in Fig. A2.It shows superimposed action potential (AP) shapes, principal membrane currents, and ion concentrations in intracellular and extracellular compartments during 1, 3 and 5 Hz steady-state stimulation (after 600 s of stabilization).Similarly to the model by Bondarenko et al. (2004), the characteristics of AP reproduced by our model (APD 25 = 4 ms, APD 50 = 8 ms and APD 75 = 16 ms at 1 Hz) agree well with experimentally measured values (Guo et al., 1999) and change only slightly with stimulation frequency (Guo et al., 2000).The model reconstruction of experimental data of Ito et al. (2000), showing a frequency-dependent increase in the systolic and diastolic maxima of intracellular Ca 2+ transients, is illustrated in Fig. A3.

Figure 1 .
Figure 1.Patch clamp recordings on CTRL and DETUB cells A, two representative confocal images of intact (CTRL) and detubulated (DETUB) cardiomyocytes.Sarcolemma stained with Di-3-ANEPPDHQ.Bottom: correspondent images of the same cells showing a skeletonized TT system, obtained with AUTO-TT.Right: TT density expressed as a percentage (%) in CTRL (black box; N = 3 animal, n = 28 cells) and DETUB (grey box; N = 3, n = 32) cardiomyocytes.B, cell capacitance (pF), measured by dividing membrane time constant by membrane resistance, recorded in CTRL (N = 6, n = 30) and DETUB (N = 6, n = 31) cardiomyocytes.C, representative traces and average values of L-Type Ca 2+ current density (I CaL ) at 0 mV from CTRL (N = 5, n = 18) and DETUB (N = 5, n = 13) cardiomyocytes.D, representative traces and average values of peak Na + current density (I Na-peak ) at 0 mV from CTRL (N = 5, n = 14) and DETUB (N = 5, n = 13).E, representative traces of outward K + current density (I K ) and average values of peak outward K + current density (I K-peak ) and residual outward K + current density at 900 ms (I Kend ) at 0 mV from CTRL (N = 5, n = 9) and DETUB (N = 5, n = 12).F, representative AP recorded in CTRL and in DETUB cardiomyocytes, paced at 1 Hz.G, duration at 20%, 50% and 90% of repolarization from AP peak (APD 20 , APD 50 and APD 90 , respectively) recorded in CTRL (N = 6, n = 30) and DETUB (N = 6, n = 31) cardiomyocytes, stimulated at 1 Hz.H, AP amplitude expressed as the difference between the resting membrane potential (RMP) and the maximum potential recorded, stimulated at 1 Hz.I, RMP.Single data are reported as grey points and box plots with a range of 25th to 75th percentile are superimposed.Median values are represented by the line in box plots.P values were calculated with one-way ANOVA and the Bonferroni test.N indicates the number of animals and n the number of cells in each data series.Raw data are available at: https://doi.org/10.6084/m9.figshare.23742231.[Colour figure can be viewed at wileyonlinelibrary.com]

Figure 2 .
Figure 2. In silico prediction of TT membrane excitability A, Left: simulation of 1 Hz steady-state AP shapes in CTRL (black) and DETUB cells (grey) assuming a residual TT fraction equal to zero.Intracellular ion concentrations were fixed at experimentally clamped values: [Na + ] i = 10 mM, [K + ] i = 130 mM and [Ca 2+ ] i = 200 nM.Extracellular ion concentrations were: [Na + ] e = 140 mM, [K + ] e = 5.4 mM and [Ca 2+ ] e = 1.8 mM.Right: simulation of TT (green trace) and SS (red trace) AP after separation of TT.TT lumen ion concentrations were close to the extracellular ion concentration: [Na + ] TT = 140 mM, [K + ] TT = 5.4 mM and [Ca 2+ ] TT = 1.8 mM.B, simulated AP shapes in the presence (Excitable TT) and absence (Non-excitable TT) of the fast sodium current in TT (dashed green).Simulated cytosolic Ca 2+ -transient in the presence of excitable TT (dark orange) and non-excitable TT (dashed light orange).Raw data are available at: https://doi.org/10.6084/m9.figshare.23742231.[Colour figure can be viewed at wileyonlinelibrary.com]

Figure 3 .
Figure 3. All-optical AP recording in TT and SS membrane sites of CTRL cardiomyocytesA, scheme of the random access multiphoton (RAMP) microscope: 1064 nm fibre laser, an acousto-optic modulator (AOM), two orthogonally mounted acousto-optic deflectors (AOD-x and AOD-y), scanning lens (SL), tube lens (TL), dichroic mirror (DM), water immersion objective (Obj), band-pass filter (BPF), condenser lens (CL) and photo-multiplier tube (PMT).Cells were optically stimulated using a fibre-coupled LED, placing the distal end of the multimode optical fibre in the proximity of the cell (highlighted in the inset).A trigger distributor was used to switch-off the PMT (red traces) during optical stimulation (blue trace).B, two-photon fluorescence image of a ventricular myocyte constitutively expressing ChR2 and labelled with VSD.The yellow lines mark the probed membrane sites (SS, sarcolemma; TT 1 and TT 2, T-tubules).C, normalized fluorescence traces ( F/F 0 ) recorded from the three scanned membrane domains (average of 10 sequential runs).An AP (elicited at 200 ms, black arrowheads) is clearly visible in all three sites, regardless of whether the cell was stimulated electrically or optically.AP elicited using electrical field are represented in grey, whereas APs evoked using optical stimulation are in blue.D, AP amplitude ( F/F 0 at peak), (E) time-to-peak (TTP) and (F) APD 50 , measured upon electrical (black)

Figure 4 .Figure 5 .
Figure 4. eGFP-Cheriff characterization Columns showing (A) AP amplitudes ( F/F 0 at peak), (B>) time-to-peak (TTP) and (C) AP duration at 50% repolarization (APD 50 ) mean values measured in eGFP-Cheriff cells using electrical field stimulation (black) compared to optical stimulation (blue).Single cell data are reported as grey points and box plots with a range of 25th to 75th percentile are superimposed.Median values are represented by the line in box plots.Data from 7 SS and from 18 TT recordings (N = 3 animals, n = 7 cells).P values were calculated with two-way ANOVA and the Bonferroni test.Raw data are available at: https://doi.org/10.6084/m9.figshare.23742231.[Colour figure can be viewed at wileyonlinelibrary.com]

Figure 6 .
Figure 6.Loss of excitability of detached t-tubules A, percentage of TT in which optical stimulation can trigger AP in CTRL and DETUB cells.Fisher's exact test was applied.B, average value and representative traces of IChR2 density provoked using 0.4 mW of light intensity from CTRL (N = 3, n = 12) and DETUB (N = 3, n = 12).Raw data are available at: https://doi.org/10.6084/m9.figshare.23742231.[Colour figure can be viewed at wileyonlinelibrary.com]

Figure 7 .
Figure 7. Changes of TT ion concentrations Simulation of changes of TT concentrations and resting membrane voltage over time in detached TT in control conditions (cyan line) and when background Na + current (g Nab ) was increased by 1.5× (light blue line), 2× (blue line) and 2.5× (dark blue line) at [Na + ] e = 132 mM, [K + ] e = 4 mM and [Ca 2+ ] e = 1 mM.The model takes into account the experimentally used different levels of external ion concentrations during detubulation ([Na + ] e = 133 mM, [K + ] e = 4.8 mM and [Ca 2+ ] e = 0.5 mM) and during the recordings ([Na + ] e = 132 mM, [K + ] e = 4 mM and [Ca 2+ ] e = 1 mM).The transmembrane voltage at the SS is denoted by red line.Raw data are available at: https://doi.org/10.6084/m9.figshare.23742231.[Colour figure can be viewed at wileyonlinelibrary.com] Membrane capacitance.According to our experimental data, the mean values of membrane capacitance (C) in intact (C CTRL ) and detubulated (C DETUB ) cells were 193 and 126 pF, respectively, and the mean fraction of TT resisting detubulation (TT res ) was 0.3.Assuming that the specific membrane capacitance is comparable in both membrane domains, the following equations allow

Figure A1 .
Figure A1.Schematic diagram of the mouse ventricular myocyte modelThe description of electrical activity of SS and TT comprises the formulation of the following ion currents: fast Na + current (I Na ), late Na + current (I NaL ), L-type Ca 2+ current (I Ca ), time-dependent transient outward K + current (I Kto ), ultrarapidly activating delayed rectifier K + current (I Kur ), non-inactivating steady-state K + current (I Kss ), time independent inward rectifier K + current (I K1 ), background Na + current (I Nab ), background Ca 2+ current (I Cab ), background K + current (I Kb ), Na + -Ca 2+ exchange current (I NaCa ), Na + -K + ATPase pump current (I NaK ) and sarcolemmal Ca 2+ pump current (I pCa ).The intracellular space contains the subspace, the network and junctional compartments of the sarcoplasmic reticulum (NSR, JSR) and the Ca 2+ buffers calmodulin (B cm ), troponin (B htrpn , B ltrpn ) and calsequestrin (B cs ).The small filled rectangles in JSR membrane represent ryanodine receptors.The small bi-directional arrows denote Ca 2+ diffusion and the thick dashed arrow represents ionic diffusion between the TT and the external bulk spaces.[Colour figure can be viewed at wileyonlinelibrary.com] DETUB C DETUB I x,CTRL C CTRL TT res − I x,DETUB C DETUB • C CTRL TT res − C DETUB C CTRL − C DETUB (4) f x = I x,ratio C SS /C TT + I x,ratio(5)where the I x , ratio represents the ratio between current densities in the TT and SS membranes (I x,TT /I x,SS ), TT res stands for the fraction of TT resisting detubulation, andCTT and C SS represent the membrane capacitance of the respective membrane domains.Assuming that C TT ≈ C SS , and taking into account the mean values of C CTRL and C DETUB in a subpopulation of cells used for our I Ca measurement (C CTRL = 162 pF, = 97 pF), we conclude that the TT res was ∼0.2 in the corresponding cells (see eqns 1 and 2).

Figure A3 .
Figure A3.Model reconstruction of the frequency dependence of intracellular Ca 2+ transient amplitudes in mouse ventricular myocytes Black and red empty circles represent minimal (diastolic) and maximal (systolic) values of Ca 2+ transients in experiments (Ito et al. 2000), respectively.Black and red full circles represent corresponding data from the model.Extracellular ion concentrations in experiment and model: [Na + ]e = 137 mM, [K + ]e = 3.7 mM and [Ca 2+ ] e = 1.5 mM.[Colour figure can be viewed at wileyonlinelibrary.com]