1

Beta-adrenergic stimulation maintains cardiac function inSerca2 knockout miceSander Land1, William E. Louch3,4, Steven A. Niederer2, Jan Magnus Aronsen3,4, Geir Christensen3,4,Ivar Sjaastad3,4, Ole M. Sejersted3,4, Nicolas P. Smith2,∗

1 Department of Computer science, University of Oxford, UK2 Biomedical Engineering Department, King’s College London, UK3 Institute for Experimental Medical Research, Oslo University Hospital Ullev̊al, Norway4 Centre for Heart Failure Research, University of Oslo, Norway.∗ E-mail: nicolas.smith@kcl.ac.uk

Abstract

Previous studies on Serca2 knockout mice showed that cardiac function is sustained in vivo for severalweeks after knockout, while SERCA protein levels decrease and calcium dynamics are significantly im-paired. In this study we reconcile observed cellular and organ level contractile function using a cardiacmultiscale model. We identified and quantified the changes in cellular function that are both consistentwith observations and able to compensate for the decrease in SERCA. Calcium transients were usedas input for multiscale computational simulations to predict whole-organ response. While this responsematched experimental PV measurements in healthy mice, the reduced magnitude calcium transients ob-served in knockout cells were insufficient to trigger ventricular ejection. To replicate the effects of elevatedcatecholamine levels observed in vivo, cells were treated with isoproterenol. Incorporation of the resultingmeasured β-adrenergically stimulated calcium transients into the model resulted in a close match withexperimental PV loops. Changes in myofilament properties, when considered in isolation, were not ableto increase tension development to levels consistent with measurements, further confirming the necessityof a high β-adrenergic state. Modelling additionally indicated that increased venous return observed inthe KO mice helps maintain a high ejection fraction via the Frank-Starling effect. Our study shows thatincreased β-adrenergic stimulation is an essential compensatory mechanism by which cardiac function ismaintained in Serca2 knockout mice, producing the increases in both systolic and diastolic calcium, con-sistent with the observed contractile function observed in experimental pressure-volume measurements.

Author Summary

The SERCA2 pump is a vital component of heart cells, and is responsible for changes in cellular calciumwhich drive the contraction of heart cells. Removal of its associated Serca2 gene is known to be lethalin mouse embryos. However, previous research shows that a knockout of this gene in adult mice is notimmediately lethal. In fact, although measurements on isolated heart cells show serious failure of thecalcium signal, measurements in living mice only shows a moderate decrease in heart function. Usinga computational model we have connected data from experiments on isolated cells and tested differenthypothesis to explain this discrepancy. We found that the calcium signal is too small for the heart functionto be explained by changes in cellular function that would make cells more efficient at generating force.Instead, we found that the calcium signal is boosted by an adrenaline-based response, as activated infight-or-flight situations. When this response was induced in isolated heart cells, our model predicted thatthe resulting measurements were compatible with the measured heart function. These results show theimportance of computational models in making exact connections between different types of experimentaldata, and has implications for understanding specific types of heart failure.

2

Introduction

Within the computational biology community, multi-scale modelling has long been proposed as an effectivestrategy for integrating experimental data focused on characterising function at different scales. Arguably,this approach has most effectively and significantly been demonstrated in the development of models of theheart and in particular, the simulation of electrical activation, mechanical contraction and the couplingbetween these two processes [1,2]. However, what is typically assumed in this process is that the isolatedcellular function, or more specifically the environment within which the cell interacts, is largely unchangedbetween scales. With consistent data sets (species, temperature and even genetic background) becomingavailable for both cellular and whole organ cardiac systems, such assumptions can now be tested. Incases where measurements at one scale are not able to be fully explained at another, differences betweenthe in-vivo and in-vitro function have the potential to be separated, providing more insight into theunderlying physiology of the system as a whole. Specifically, in this study we demonstrate the value ofthis approach by applying it to understand a murine model of genetically induced heart failure and itseffects on cardiac excitation contraction coupling at both the cellular and whole organ scales.

In our genetically modified mice, the Serca2 gene is knocked out, leading to a major disruption inthe function of the sarcoplasmic reticulum (SR). The SR is an intracellular compartment which functionsas a calcium store, and SR function accounts for up to 90% of calcium cycling in mice [3]. Uptake ofcalcium into this store is regulated by the sarco(endo)plasmic reticulum Ca2+ATPase 2 (SERCA2) ionpump, and decreased SERCA2 function has been associated with both systolic and diastolic heart failureacross species [4]. Since a full knockout of the Serca2 gene is lethal in utero [5], we use a cardiomyocytespecific gene knockout in adult mice [6], which results in mice which are viable for 7-10 weeks after geneknockout [4]. However, even though protein content is reduced to less than 5% by 4 weeks after knockout,observed cardiac function is maintained at near normal levels. Only after 7 weeks do the mice developmore severe cardiac dysfunction [7].

This sustained function of the heart, despite a large reduction in SERCA2 function, suggests theexistence of significant compensatory mechanisms which may be important in understanding SERCArelated heart failure in humans. Previous experiments on isolated myocytes, combined with mathematicalmodelling, has shown that calcium transients are compensated for by upregulation of transmembranecalcium dynamics, most importantly the L-type calcium channels and NCX [8]. Furthermore, there areadaptive changes in the cell geometry, including T-tubule remodelling and a decrease in SR volume [9].However, despite these compensatory mechanisms, calcium transients in cells from knockout mice remainvery small, an observation that is fundamentally inconsistent with the whole organ response.

We hypothesise that this discrepancy appears because previous observations in isolated myocytes wereperformed mainly in the absence of other factors present in vivo, and with many experiments performed atlow frequencies. The importance of this issue is highlighted by the significant differences between observedcellular function compared to in vivo function due to complex feedback mechanisms, in particular theinfluence that length- and velocity dependent feedback has on tension generation throughout the heartcycle [10]. Investigating these feedback mechanisms is difficult in isolated cells. However, as outlinedabove, computational modelling can provide a biophysically consistent framework with the means toincorporate the effects of regulatory pathways on calcium and contractile dynamics.

Applying this approach, we show that despite the several compensatory mechanisms previously iden-tified in regulating calcium dynamics, the small calcium transients previously measured in single cellsare still not compatible with the observed whole-organ function. Furthermore, we demonstrate that theaddition of isoproterenol to induce a high β-adrenergic state in knockout cells causes significant changesin measured calcium transients. These changes can fully explain the observed alterations in cardiac func-tion after knockout, in turn providing new insight into the progression of heart failure with decreasedSERCA2 function. The changes in calcium are consistent with the known effects of β-adrenergic stim-ulation, which include increased SERCA and L-type calcium influx, changes in calcium sensitivity, andmyofilament properties [11, 12]. This result is also consistent with previous observations of increased

3

norepinephrine levels in KO mice [7], which have previously not been taken into account in experimentson isolated myocytes at physiological pacing rates, and are difficult to quantify in vivo. Through theapplication of data-driven computational modelling we quantify the whole-organ response to decreasedcalcium transients in knockout mice and show that the disparities observed require the significant changesin calcium transients induced by β-adrenergic stimulation to reconcile cellular and whole-organ function.

Methods

Ethics statement

Experiments were performed in accordance with the Norwegian Animal Welfare Act, which conforms toNIH guidelines (NIH publication No 85-23, revised 1996). Mice were housed at 55% humidity on a 12 hlight/dark cycle, with food pellets (RM1, 801151, Scanbur BK) and water freely available.

Experimental methods

We use genetically modified Serca2flox/flox Tg(α-MHC MerCreMer) mice, in which injection of tamoxifeninduces disruption of the Serca2 gene in cardiomyocytes [7]. Measurements were performed 4 and 7 weeksafter injection, in separate groups of mice, referred to as KO4 and KO7 mice. Control flox/flox mice atthe same age as KO7 mice are denoted by FF. Pressure-volume loops were measured in FF, KO4 andKO7 mice as previously described [10]. Calcium transients were measured during 6 Hz stimulation usingwhole-cell fluorescence with and without 200 nmol/L isoproterenol. For more details on the experimentalmethods, see the electronic supplement.

Computational modelling methods

Metric FF KO4 KO7EDV adj. (µl) 58.9 54.6 55.0EDP (kPa) 1.02 1.66 3.02SEP (kPa) 10.5 9.8 11.1Min. P (kPa) 0.08 1.28 2.35

Control [Ca2+]refT50(µmol/L) 0.578 0.469 0.423

+ISO [Ca2+]refT50(µmol/L) 0.427 0.588 0.661

Table 1. Simulated parameters taken from experimental pressure-volume loops. EDV adj.is end-diastolic volume adjusted by 15µl to account for muscles included in simulation cavity volumebut not in PV volume (as previously determined by comparison with MRI data [10]). EDP isend-diastolic pressure. SEP is the pressure at start of ejection, used to determine end of isovolumetriccontraction. Min. P is the minimal pressure, used to determine end of isovolumetric relaxation. Theparameter [Ca2+]refT50 is the half-activation of calcium binding to TnC at resting sarcomere length, andwas fitted such that EDV matches within 0.1µl when prescribing EDP. Calcium sensitivity[Ca2+]refT50was chosen such that EDV matches experimental data when EDP is prescribed, seperately incontrol and +ISO cases.

Simulations on a left-ventricular geometry were performed using the mechanics framework previouslydescribed [10, 13], with calcium transients prescribed from experimental measurements. We performedsimulations for calcium transients measured in FF, KO4 and KO7 mice, both with and without iso-proterenol (referred to as ‘+ISO’ and ‘control’). In all cases the calcium sensitivity [Ca2+]refT50, which

4

determines the half-maximal activation of calcium binding to TnC, was set to match end-diastolic vol-ume at end-diastolic pressure, as this affinity is known to change in vivo depending on regulatory factors.Other contractile parameters were unchanged from previous work, where they were determined to matchhealthy mouse tension generation at physiological temperature and frequency. Hemodynamic parametersfor transitions from isovolumetric contraction to ejection, and from isovolumetric relaxation to diastolewere taken from representative traces of experimental PV data, summarized in Table 1. Due to rapidmurine electrical activation, tension development was assumed to initiate homogeneously. As heart fail-ure was observed in these animals without hypertrophy [7], the LV geometry and constitutive propertiesused were identical to previous work [10].

Since comparing pressure-volume loops can obscure temporal differences in results, we also inves-tigated whole-organ tension development. Performing such comparisons requires an estimate of activetension from experimental results. This mean active tension was estimated by prescribing pressure andnumerically determining the level of active tension required in in the mechanics solution to match exper-imentally measured volume. using the PV measurements described in the previous section.

To support the simulations above, where only the calcium sensitivity [Ca2+]refT50 was changed, we alsoinvestigated alternate compensatory mechanisms in the contraction model parameters. In each of thesesimulations, the calcium sensitivity [Ca2+]refT50 was fitted individually as in previous simulations, aftervariation of fundamental contractile parameters: the Hill coefficient for cooperative crossbridge actionnxb, the reference maximal isometric tension Tref, and the magnitude of length-dependent activationeffects β1. We also investigated changing the rate constants, the unbinding rate of calcium from TnC,kTRPN and the scaling factor for rate of crossbridge binding kxb. These simulations were only performedfor the KO4 and KO7 mouse models, as the FF model already shows realistic contraction under controlconditions.

For more details on the mathematical models and active tension estimation, see the electronic sup-plement.

Results

Representative calcium transients are shown in Fig. 1, along with details of diastolic and peak calciummeasurements. In untreated cells, we have observed diastolic calcium is unaltered after knockout acrossa range of frequencies (approximately 200 nmol/L at 6 Hz [4]), and these values were used to calibratemeasurements in the present study. In FF mice, treatment with isoproterenol significantly reduceddiastolic calcium levels and increased peak calcium. In KO mice, peak levels of intracellular calcium werealso augmented by isoproterenol. However, due to compromised SERCA function, diastolic calcium levelsincreased (all significant, unpaired t-tests p < 0.05). Rather than maintaining diastolic calcium as seenin control mice before and after knockout, the calcium transients for KO4+ISO and KO7+ISO have peakcalcium levels not significantly different from each other and that of the FF control (ANOVA p = 0.945).

These six representative calcium transients were used as input for simulations of LV function. Theresulting PV loops from simulations were compared with experimental data in Fig. 2. These resultsdemonstrate that control calcium transients in KO mice are incompatible with observed whole heartfunction, as the predicted maximum pressures of 7-8 kPa are not sufficient to initiate ejection. How-ever, calcium transients recorded in KO cells stimulated with isoproterenol reproduced experimental PVloops. For FF mice, PV loops obtained with control calcium transients matched experimental data, whiletransients recorded during isoproterenol resulted in an ejection fraction significantly larger than seen inexperimental data.

Fig. 3 shows cellular tension estimated from experimental PV data compared to active tension fromthese six simulated cases. These estimated tension traces demonstrate more clearly the unphysiologicalresults produced by using the +ISO calcium transient in the FF mouse model. After the usual fastdecrease in tension development at the beginning of ejection due to the velocity dependent effects of fast

5

FF KO4 KO7Control Min Ca calibrated to 200 nmol/LControl Peak Ca (nmol/L) 442 ± 25 250 ± 5 236 ± 5+ISO Min Ca (nmol/L) 152 ± 11 251 ± 23 315 ± 17+ISO Peak Ca (nmol/L) 990 ± 135 459 ± 84 430 ± 29

Figure 1. Calcium transients without (‘control’) and with (‘+ISO’) isoproterenol. Panelsshow results as mean±SE in FF (n = 7), KO4 (n = 12) and KO7 (n = 10) mice from left to right.Accompanying table shows details of peak and minimum calcium concentrations. In control mice,diastolic calcium was maintained, while peak calcium decreases significantly after knockout. Withisoproterenol, diastolic calcium decreases in FF due to upregulated SERCA, but increases in KO due toa lack of SERCA and increased calcium influx. Peak calcium increases with isoproterenol to reach levelsapproximately equal to FF control levels.

Figure 2. Pressure-volume loops from simulations based on calcium transients without(‘control’) and with (‘+ISO’) isoproterenol, compared to experimental PV data. Panelsshow results in FF, KO4 and KO7 mice from left to right. For simulations which do not reach ejection,maximal pressure is indicated by an ‘x’.

6

Figure 3. Active tension from simulations based on calcium transients in the absence andpresence of ISO, compared to active tension estimated from PV data. Panels show results inFF, KO4 and KO7 mice from left to right, all aligned by end-diastole. Heart rates for simulations, andexperimental data for KO4 and KO7 are 6 Hz, and experimental data for FF is at 7.5 Hz.

shortening, there is a second phase of tension development, leading to a peak tension twice as high asestimated from experimental data. In comparison, the control calcium transients in KO mice were toosmall to result in a simulated model contraction consistent with the observed measurements.

Fig. 4 shows the predicted deformation for FF, KO4 and KO7 mice from the relevant simulations.These results show the striking similarities in ejection, especially between FF and KO4, regardless ofdifferences in IVC and IVR. In KO7, ejection is markedly decreased even following a more rapid IVCphase due to higher residual tension.

To investigate the necessity of beta-adrenergic stimulation in light of the assumptions inherent in acontraction model, we also performed an analysis of alternate means to generate more tension in themodel that was consistent with the observed experimental data, using changes in contraction modelparameters. Table 2 and Fig. 5 shows the results for alternate contraction models in the KO4 and KO7control cases. In the KO4 model, doubling the rate constants kTRPN, kxb, simulating increased reactionrates for troponin C and crossbridge binding respectively, has little effect. Increasing the cooperativityof crossbridge binding from nperm = 5 to nperm = 7 has the greatest effect, but simulation results arestill far from experimental traces. Furthermore, this change leads to an F-pCa Hill curve fit with aHill coefficient of 9.1, and a biphasic fit with coefficients 10.8 and 5.1 for the lower and upper halfrespectively, which exceeds experimentally measured Hill coefficients [14]. Increasing the reference tensionto 240 kPa results in only a small increase in the the maximum pressure, as it requires a lower calciumsensitivity to match end-diastolic state. Finally, maximum pressure increased by decreasing the magnitudeof length-dependent activation (LDA) factor β1. This possibly counterintuitive mechanism is related tothe fact that calcium sensitivity has to be decreased significantly at higher LDA to match residualtension. Lowering LDA thus allows for higher sensitivity to calcium, and slightly higher developedtension. This lower length dependence also results in a cellular response significantly lower than thelength dependence seen in experiments. In KO7 mice the balance between required changes in calciumsensitivity and maximum pressure changes in some cases. However the overall conclusions remain thesame: only changes in reference tension and cooperativity inconsistent with observed experimental valueshave the potential to significantly increase maximum pressure. In the KO7 case there is no ejection in anysimulation, which further demonstrates that the control calcium transients are incompatible with whole-organ function observed. This shows the necessity of β-adrenergic stimulation in maintaining cardiacfunction in knockout mice, as even large compensatory changes in myofilament properties did not result

7

Figure 4. Deformation from simulations based on FF control, KO4+ISO and KO7+ISO.Surfaces are colored according to active tension, and fiber direction arrows are colored according tofiber strain, as shown in the legend. Top row shows FF at end-diastole, mid IVC, beginning, mid, endejection, and beginning of diastole. Middle row shows KO4, and bottom row KO7.

in viable cardiac output.

Figure 5. Pressure-volume loops for parameter variations Pressure-volume loops from the‘original’ and nxb = 7 simulations shown in Table 2, compared to experimental PV loops. Forsimulations which do not reach ejection, maximal pressure is indicated by an ‘x’.

8

KO4 KO7

Parameter [Ca2+]refT50 Max. pressure [Ca2+]refT50 Max. pressureoriginal 0.469 7.5 kPa 0.423 8.0 kPakTRPN = 0.2/ms 0.470 7.5 kPa 0.429 7.7 kPakxb = 0.2/ms 0.469 7.5 kPa 0.426 7.9 kPanxb = 7 0.420 10.0 kPa 0.399 8.8 kPaTref = 240 kPa 0.511 8.6 kPa 0.469 8.6 kPaβ1 = −0.75 0.446 7.7 kPa 0.412 7.4 kPa

Table 2. Whole-organ response to changes in the contraction model, using control calciumfor KO4 and KO7 mice. Shown are the necessary change in [Ca2+]refT50 to match end-diastolic state,and resulting maximal pressure developed. The relevant original contraction model parameters are: Hillcoefficient for cooperative crossbridge action, nxb = 5. Unbinding rate of calcium from TnC,kTRPN = 0.1. Scaling factor for rate of crossbridge binding, kxb = 0.1/ms. Reference maximal isometrictension, Tref = 120 kPa. Magnitude of length-dependent activation effects, β1 = −1.5.

Discussion

Through a combination of novel experimental data and data-driven multi-scale mathematical modellingat both the single cell and whole-organ levels we have been able to explain the compensation for heartfailure in Serca2 knockout mice. Our results show that β-adrenergic stimulation is an essential factor inexplaining the observed changes from FF to KO mice and the relatively slow development of heart failureafter a major disruption in normal cellular function.

In FF mice, predicting cardiac function using control calcium transients shows realistic pressure andtension development and ejection consistent with experimental measurements and previous modelling[10]. For the FF animals, added isoproterenol produced unrealistic tension development due to velocitydependent effects during fast shortening, as well as an ejection fraction significantly greater than valuesmeasured experimentally. Furthermore, a higher calcium sensitivity is needed to explain the diastolictension observed experimentally when diastolic calcium is lowered by upregulated SERCA. In constrast,β-adrenergic stimulation generally causes lower calcium sensitivity due to the effects of PKA on TnI andTnC [12].

In KO4 and KO7 mice, control calcium transients are too small to result in the observed ejectionfraction, as they can not sustain the systolic tension observed even when diastolic tension is significantlyelevated. Furthermore, the observed reduction in the magnitude of the calcium transients of > 70% istoo large to be explained by minor changes in contractile function. With β-adrenergic stimulated calciumtransients, simulations matched observed PV data, showing realistic ejection and tension development.Matching these experimental results also required a decrease in calcium sensitivity between control and+ISO contraction in the model to be consistent with diastolic tension, which is due to higher diastoliccalcium in the absence of sufficient upregulation of SERCA. The small decrease in calcium sensitivityfrom FF control to KO+ISO is also consistent with the effects of β-adrenergic stimulation.

From these results we can now provide a novel interpretation of the progression of heart failure andcompensatory mechanisms which result from knockout of the Serca2 gene. Several electrophysiologicalmechanisms compensate for the decrease in SERCA. First, in the absence of β-adrenergic stimulation,transmembrane calcium dynamics are upregulated to increase the calcium transient [8]. Increased acti-vation of the L-type calcium channel through β-adrenergic stimulation increases the calcium transientfurther, preserving systolic calcium levels to be equal to those normally observed in healthy mice. How-ever, in the absence of SERCA this upregulation of the L-type calcium channel through multiple mech-anisms causes net transmembrane removal of calcium through NCX and PMCA to be insufficient tomaintain diastolic calcium at healthy levels. When systolic calcium is maintained, and as the size of

9

the calcium transient decreases with increasing diastolic calcium, either diastolic tension must increase,or systolic tension must decrease. Consistent with maintaining cardiac pump function and preventingsystolic heart failure, systolic tension is maintained while diastolic tension increases significantly. Finally,the resulting higher diastolic tension is compensated for through higher minimum and end-diastolic pres-sure, which indicates a higher venous return pressure. Venous return is also known to be increased byβ-adrenergic stimulation, through venodilation and lower splanchnic venous resistance [15]. This furtherensures systolic tension can be maintained through effective length dependence of tension development,by maintaining end-diastolic volume even in the presence of this significantly higher diastolic active ten-sion. Increasingly severe heart failure develops as the β-adrenergic state and venous return pressure canno longer be increased, while the calcium transient keeps decreasing, possibly further exacerbated byacidosis, Na+ accumulation, and increased ATP consumption [4, 8].

Interestingly, although the Serca2 KO mice develop heart failure, it appears that beta-adrenergicstimulation is advantageous in this particular situation, and likely necessary for survival at late stages.Contraction in KO cells is facilitated by a marked increase in systolic calcium during β-adrenergic stim-ulation. Importantly, through actions in the kidneys, β-adrenergic stimulation promotes fluid retentionand increased venous return, which are traditionally considered deleterious in heart failure. However, inthe KO mice such an increase in pre-load appears to be essential in permitting filling of the ventricledespite increased diastolic active tension. In heart failure, β-adrenergic stimulation is generally associ-ated with arrhythmogenesis due to overloading of the SR, and thus spontaneous calcium release [16].However, in the KO mice, such spontaneous release events are inhibited due to the low SR content, andthus the pro-arrhythmic danger of β-adrenergic stimulation appears to be avoided. Finally, a detrimentalincrease in heart rate expected with greater β-adrenergic tone does not occur in KO mice; in fact wehave observed a mild decrease in heart rate in these mice [7]. This may be due to the role of SERCA2 insinoatrial node cells [17], which are partially affected by the knockout. Thus, there appears to be severaladvantages to increased β-adrenergic stimulation in the Serca2 KO mice, without the largely negativeeffects associated with sympathetic stimulation in other types of chronic heart failure. In humans, theeffects of sympathetic stimulation are often blocked with β-blockers, making them an effective treatmentfor a range of heart diseases [18]. By contrast, KO animals appear dependent on continuous β-adrenergicstimulation for survival.

A possible limitation of our experimental data is that we have only considered very high and noβ-adrenergic stimulation. Although ejection predicted from KO4 +ISO calcium data is slightly highercompared to PV data, within the limits of uncertainty of the experimental data, it is difficult to saywhether β-adrenergic stimulation is already near maximal in KO4 mice. Previously measured bloodcatecholamine levels are slightly elevated in KO7 compared to KO4 [7], but significantly elevated in bothcompared to FF. As blood catecholamines are spillover from nerve terminals, and have great variability,β-adrenergic stimulation may be maximal at both time points.

The close match of model predictions to data demonstrates the effectiveness of careful species-specificmodelling at physiological temperatures and pacing rates. However, our novel tension estimation methodalso reveals some small but striking differences between model predictions and observed tension. Mostnotably relaxation of tension during IVR is slower, even considering the small disparity in heart rate for theFF data. This is likely to be related to complex length- and velocity-dependent effects, such as shortening-induced cooperative deactivation of the filaments [19], which are not included in the contraction modelas their mechanisms remain difficult to quantify. Effects of β-adrenergic stimulation on the myofilamentscould also play a role [12], although doubling the rate constant for crossbridges kxb did not improverelaxation, suggesting that this is dominated by the calcium transient decay rate in simulations.

In multi-scale modelling, cell models based on isolated cells are often assumed to perfectly describewhole-organ behaviour. As we have shown, even when using a carefully species-specific model at aphysiological temperature and pacing frequency, such an assumption may be incorrect. In this context,our experimental data also reveals some of the disadvantages of using isolated cell preparations. While the

10

cell environment is carefully controlled, isolated cell behaviour can be significantly different in comparisonto in vivo function, as important regulatory factors of calcium dynamics, such as β-adrenergic stimulation,changing sarcomere length and circulatory factors, are missing. Our results also show an important rolefor computational modelling in quantifying the connection between single cell and whole organ function.Whereas the decreased calcium transients seen in the knockout mice may initially seem compatible withthe observed decrease in heart function, modelling results demonstrate that more significant changes areneeded, which can only be induced by high β-adrenergic stimulation.

In conclusion, through the integration of novel experimental data collected from both vitro cellularpreparations and live animals a we have been able to identify a key component of cardiac function which isessential for in-vivo function. Specifically these results explained the progression of heart failure in Serca2knockout mice, and highlight the role of increased β-adrenergic stimulation as the major compensatorymechanism.

11

References

1. Smith NP, Crampin EJ, Niederer SA, Bassingthwaighte JB, Beard DA (2007) Computationalbiology of cardiac myocytes: proposed standards for the physiome. Journal of Experimental Biology210: 1576–1583.

2. Nordsletten D, Niederer S, Nash M, Hunter P, Smith N (2011) Coupling multi-physics models tocardiac mechanics. Progress in Biophysics and Molecular Biology 104: 77–88.

3. Bers DM (2002) Cardiac excitation-contraction coupling. Nature 415: 198–205.

4. Louch WE, Hougen K, Mørk HK, Swift F, Aronsen JM, et al. (2010) Sodium accumulation promotesdiastolic dysfunction in end-stage heart failure following serca2 knockout. The Journal of Physiology588: 465–478.

5. Periasamy M, Reed TD, Liu LH, Ji Y, Loukianov E, et al. (1999) Impaired cardiac performance inheterozygous mice with a null mutation in the sarco(endo)plasmic reticulum ca2+-ATPase isoform2 (SERCA2) gene. Journal of Biological Chemistry 274: 2556–2562.

6. Andersson KB, Finsen AV, Sj̊aland C, Winer LH, Sjaastad I, et al. (2009) Mice carrying a condi-tional serca2(flox) allele for the generation of ca(2+) handling-deficient mouse models. Cell calcium46: 219–225.

7. Andersson K, Birkeland J, Finsen A, Louch W, Sjaastad I, et al. (2009) Moderate heart dysfunctionin mice with inducible cardiomyocyte-specific excision of the serca2 gene. Journal of Molecular andCellular Cardiology 47: 180–187.

8. Li L, Louch WE, Niederer SA, Aronsen JM, Christensen G, et al. (2012) Sodium accumulation inSERCA knockout-induced heart failure. Biophysical journal 102: 2039–2048.

9. Swift F, Franzini-Armstrong C, Øyehaug L, Enger UH, Andersson KB, et al. (2012) Extremesarcoplasmic reticulum volume loss and compensatory t-tubule remodeling after serca2 knockout.Proceedings of the National Academy of Sciences .

10. Land S, Niederer SA, Aronsen JM, Espe EKS, Zhang L, et al. (2012) An analysis of deformation-dependent electromechanical coupling in the mouse heart. The Journal of Physiology 590: 4553–4569.

11. Saucerman JJ, McCulloch AD (2006) Cardiac β-adrenergic signaling. Annals of the New YorkAcademy of Sciences 1080: 348–361.

12. Metzger JM, Westfall MV (2004) Covalent and noncovalent modification of thin filament actionthe essential role of troponin in cardiac muscle regulation. Circulation Research 94: 146–158.

13. Land S, Niederer SA, Smith NP (2012) Efficient computational methods for strongly coupledcardiac electromechanics. Biomedical Engineering, IEEE Transactions on 59: 1219 –1228.

14. Dobesh DP, Konhilas JP, de Tombe PP (2002) Cooperative activation in cardiac muscle: impact ofsarcomere length. American Journal of Physiology - Heart and Circulatory Physiology 282: H1055–H1062.

15. Green JF (1977) Mechanism of action of isoproterenol on venous return. The American journal ofphysiology 232: H152–156.

12

16. Venetucci LA, Trafford AW, Eisner DA (2007) Increasing ryanodine receptor open probability alonedoes not produce arrhythmogenic calcium waves threshold sarcoplasmic reticulum calcium contentis required. Circulation Research 100: 105–111.

17. Maltsev VA, Lakatta EG (2010) Yin and yang of the cardiac pacemaker clock system in healthand disease. Heart rhythm: the official journal of the Heart Rhythm Society 7: 96–98.

18. Frishman WH, Saunders E (2011) β-adrenergic blockers. The Journal of Clinical Hypertension 13:649–653.

19. Hanft LM, Korte FS, McDonald KS (2008) Cardiac function and modulation of sarcomeric functionby length. Cardiovascular Research 77: 627 –636.