Altered Sympathetic Control of Myocardial Infarct Border Zones: High-Resolution Mapping with Automated Structure Detection

Intramyocardial sympathetic nerve remodeling after myocardial infarction (MI) has been implicated in adverse outcomes such as sudden arrhythmic death, yet the underlying mechanisms are poorly understood. We sought to examine microstructural remodeling of ventricular myocardium and cardiac sympathetic nerve �bers after chronic infarction and to correlate this remodeling with perturbations in electrical conduction. We developed a high-resolution pipeline for anatomically precise alignment of optical action potential maps with structural maps of myo�ber and nerve �ber features detected by customized computer vision algorithms. Using this integrative approach in a mouse model of chronic MI, we identi�ed distinct structure-function correlates of discontinuous electrical propagation to objectively de�ne the infarct border zone. During sympathetic activation induced by tyramine administration, we also discovered regional patterns of altered impulse propagation directly associated with altered neuroeffector junction distribution, pointing to potential substrates for neurally mediated arrhythmogenesis. This study establishes a synergistic framework for examining structure-function relationships after MI with unprecedented spatial precision, which has implications for advancing our understanding of arrhythmogenic substrates and mechanisms.


Introduction
Myocardial infarction (MI) and its consequent cardiac arrhythmias are leading causes of mortality in the world 1,2 .Following MI, injured myocardium creates a substrate for discontinuous electrical propagation 3- 5 , and concomitant neural remodeling leads to dysregulation at multiple levels of the cardiac autonomic nervous system [6][7][8] .Together, these pathophysiological changes can lead to lethal arrhythmias.
0][11][12][13] These ndings have been vital to the advancement of pharmacologic therapies, which have had signi cant, though incomplete, success in reducing cardiac morbidity and mortality. 14As anti-arrhythmic therapies also focus on structural substrate modi cation 15 , elucidating upstream organ-level neurocardiac control is crucial to bridging the bench-to-bedside gap.
Previous studies using immunostaining of heart Sect. 16and, more recently, tissue clearing of whole hearts 17 have shown structural changes in sympathetic nerve bers innervating the myocardium after MI, with both regions of denervation due to ischemic injury and hyperinnervation due to nerve sprouting.
While post-MI structural changes have been well described in several animal models [17][18][19][20] and in humans 16,21 , their mechanistic signi cance in arrhythmogenesis remains unclear.Functional studies using multielectrode arrays 18,19 and optical mapping [22][23][24][25][26] have demonstrated perturbations in impulse propagation with stimulation of the sympathetic nervous system following MI.However, the inability to directly correlate high-resolution structural and functional data from the same heart has impeded our understanding of how structural remodeling of nerves impacts functional regulation of the heart post-MI.
Our previous work established an important technical basis for high-resolution imaging and semiautomated analysis of global innervation patterns in healthy hearts 27 .In this study, we apply our prior techniques to structural heart disease and add functional mapping to establish a novel, multi-modal pipeline which allows direct structure-to-function correlation in a mouse model of chronic MI.By merging functional electrical maps obtained by optical mapping with global structural maps of the heart obtained by tissue clearing, we examine post-MI neurocardiac dynamics at unprecedented resolution.

High-resolution mapping and alignment of cardiac structure with electrical function
To evaluate structure-function relationships and changes in these relationships after MI, we developed a pipeline of aligning optical maps of myocardial impulse propagation directly with high-resolution images of myo ber and nerve ber structure in the same hearts after tissue clearing and semi-automated ber tracing (Fig. 1a).With a two-camera system, we rst optically mapped action potentials (APs) in normal (sham) hearts and chronic MI hearts and obtained a simultaneous bright eld image of surface vascular features (Fig. 1b-c).We then xed these hearts, immunolabeled them with the sympathetic nerve marker tyrosine hydroxylase (TH), and performed tissue clearing using a modi ed, immunolabeling-enabled 3dimensional imaging of solvent-cleared organs (iDISCO) method 28 to allow for high-resolution confocal imaging and semi-automated structural analysis 27 (Fig. 1d-f).To align the structural images back to the optical electrical maps, we used vascular ducial points (venous bifurcations) clearly visible on both the bright eld heart images and the confocal images of muscle auto uorescence (Fig. 1g-h).We found this transformation was su cient to account for tissue deformation after clearing, bringing ducials into good alignment over most of the surface visible in the bright eld image.Thus we were able to overlay structural data such as global nerve ber features (Fig. 1i) directly and precisely onto functional electrical data from the same hearts.

Structure-function alignment precisely correlates global ventricular electrical propagation with myo ber orientation
After developing our structure-function alignment pipeline, we validated it in a proof-of-concept analysis to demonstrate the relationship between ventricular myo ber orientation and directionality of electrical propagation in the same heart.Ventricular conduction vectors were calculated from activation maps with basolateral left ventricular (LV) pacing (Fig. 2a), exported as angular data matrices, and visualized as vector orientation colormaps (Fig. 2c).High-resolution muscle auto uorescence images were used for automated tracing of myo ber orientations (Fig. 2d-e), which were then also visualized as colormaps (Fig. 2f).We validated this automated myo ber tracing by comparison of muscle auto uorescence to viral labeling of cardiomyocytes (Supplemental Fig. 1).
Once the pairs of functional and structural maps from four representative sham hearts were aligned, we rst qualitatively assessed the association between myo ber orientation and conduction vector orientation by creating cosine similarity maps (Fig. 2g).Of note, our method of alignment detected the area of lowest cosine similarity around the right ventricular (RV) insertion point (where RV wall attaches to the interventricular septum anteriorly), where there is almost a 90-degree shift in myo ber orientation.Next, we quantitatively analyzed the degree of structure-function concordance in each heart by calculating angular correlation coe cients between myo ber and conduction data matrices, and then testing for matrix similarity (Fig. 2h).We found signi cantly close correlation between structural myo ber orientation and functional conduction vector orientation, thus validating our alignment method and experimentally demonstrating the cable theory of myocardial impulse propagation 5,29,30 at the global ventricular level.
A composite metric of myo ber anisotropy and tissue activation time uniquely de nes infarct border zones To begin studying how structural remodeling alters electrical function after MI, we segmented both sham and MI hearts into anatomical regions of interest (ROIs) using speci c structural criteria to maintain consistency across hearts.Dense scar and infarct border zone (BZ) were de ned using intensity of muscle auto uorescence, and LV basal, LV apical, and RV regions were determined using anatomical landmarks (Supplemental Figs.2-4).Dense scar data were excluded from these analyses due to lack of surviving myocytes (Supplemental Fig. 5, Supplemental Movie 1).Quantitative data from these ROIs were extracted from aligned myo ber structure maps (Fig. 3a) and activation maps (Fig. 3b) and were used to calculate tissue activation times and anisotropy indices of myo ber disorder.Conduction vectors at structurally de ned BZ regions displayed discontinuous electrical propagation (Fig. 3c).Compared to the isotropic activation curve of sham LV apex, BZ regions displayed activation curves consistent with anisotropic conduction and conduction block 31 (Fig. 3d).In plots of myo ber anisotropy index versus tissue activation time, we found that ROIs from sham hearts were all tightly clustered in the lowanisotropy, fast-activation-time region of the plots (Fig. 3e, Spearman r = 0.0667, p = 0.8801, n = 9 regions from 3 mice).In contrast, the BZ in MI hearts were signi cantly distinct from other ROIs and localized to the high-anisotropy, slow-activation-time region of the plot (Fig. 3f, Spearman r = 0.833, p = 0.0083, n = 9 regions from 3 mice).Thus, by integrating regionally speci c structural and functional data, we establish a novel quantitative metric which precisely de nes the BZ.

Chronic MI induces altered patterning of ventricular neuroeffector junctions
Having established perturbed myocardial structure-function relationships in our chronic MI model, we next turned our attention to assessing sympathetic nerve remodeling.Automatically detected nerve ber tracings, from confocal microscopic images of whole-heart TH staining, were binned into small, medium, and large bers according to previously reported diameters [32][33][34] .The same anatomically segmented ROIs used in the aforementioned myo ber analyses were applied to extract and quantify regional nerve ber lengths (Fig. 4a-f).Qualitatively, there was obvious denervation with absent TH staining at the LV apex (dense scar) of MI hearts compared to sham (Fig. 4a,d).Dense scar data were excluded from these analyses due to the extremely low amount of surviving nerve bers (Supplemental Fig. 5).In both sham and MI hearts, the LV base tended to have signi cantly more large bers than medium and small (Kruskal-Wallis, p = 0.0048 for sham, p = 0.0005 for MI, n = 4 mice per group), while the RV had fewer large bers than medium and small (Kruskal-Wallis, p = 0.0132 for sham, p = 0.0031 for MI, n = 4 mice per group) (Fig. 4h,i).
The post-MI changes in small-size bers were of special interest, as these are both closest to, and include, the neuroeffector varicosities which interface with myocytes to control cardiac function.In MI hearts, the infarct BZ showed signi cant increase in small ber prevalence compared with sham LV apex (Mann-Whitney, p = 0.0286), as well as decrease in medium ber prevalence (Mann-Whitney, p = 0.0286) (Fig. 4g).This was visually apparent on high-magni cation images of BZ versus sham LV apex (Fig. 4b,e) and was detectable by our automated ber tracing algorithm (Fig. 4c,f).Interestingly, the LV base in MI hearts also displayed a decrease in small ber prevalence (Mann-Whitney, p = 0.0286) compared with sham LV base (Fig. 4h).Taken together, these data establish a regional pattern of nerve sprouting at the infarct BZ along with small-ber denervation at the remote LV base, speci cally indicative of perturbed neuroeffector junction topography.
Post-MI changes in neuroeffector junction patterning underlies regional heterogeneity in sympathetic control of impulse propagation Given the altered neuroeffector junction distribution we discovered after chronic MI, we next examined whether these regional neural changes had functional effects on myocardial impulse propagation.Using our alignment technique, we overlaid neural structural data with optical mapping data from the same hearts and assessed regional changes in repolarization after sympathetic stimulation with tyramine, which stimulates norepinephrine release from neuroeffector terminals.For previously discussed reasons, dense scar was excluded from these analyses.
We found that in sham hearts, tyramine infusion caused an expected initial prolongation of eighty percent of action potential duration (APD 80 ) 35 , in an evenly distributed fashion across the whole heart (Kruskal-Wallis p = 0.7463) (Fig. 5a,b,h).In contrast, MI hearts exhibited signi cant regional variation in APD 80 prolongation after tyramine infusion (Fig. 5c,d,h, Kruskal-Wallis p = 0.0132).Speci cally, there was more APD 80 prolongation of the RV in MI hearts compared to the LV base and infarct BZ (Fig. 5e-g,h, Mann-Whitney p = 0.0286, n = 4 mice per group).When we correlated small-ber distribution after MI to these functional repolarization changes (Fig. 5i), we found that while there was a positive correlation between small-ber prevalence and tyramine-induced APD 80 prolongation at the LV base and RV regions (Spearman r = 0.7381, p = 0.0458), the BZ notably lacked this functional correlation despite having the highest prevalence of small bers (Spearman r = 0.021, p = 0.956).These data demonstrate a direct, anatomically precise relationship between regional small ber content -a surrogate index of neuroeffector junction quantity -and sympathetic control of myocardial repolarization, with the interesting exception of the functionally distinct infarct BZ.

Discussion
We developed a high-resolution platform for precisely aligning functional maps of electrical propagation to structural maps of post-MI neurocardiac remodeling, using optical mapping, state-of-the-art intact-heart imaging, and computer vision algorithms for semi-automated feature detection.Using this platform, we report several novel ndings: 1) direct spatial correlation of ventricular myo ber structure to directionality of AP propagation at the global ventricular level; 2) a mathematically precise de nition of the infarct BZ that integrates both its distinctive microstructural and functional features; 3) perturbed neuroeffector-junction topography of the whole post-MI ventricle; and 4) a direct relationship between post-MI neuroeffector-junction distribution and altered sympathetic control of impulse propagation.
Our structure-function alignment method demonstrates close, global concordance between myo ber architecture and conduction vector elds for the rst time in intact ventricles, a relationship which had previously only been studied at the single-myocyte level 29,36 or through computational modeling of myocardial function [37][38][39] .The high spatial resolution of our alignment pipeline also allows structurally precise regional analyses, which we utilized to de ne the infarct BZ with a novel composite metric that encompasses both myo ber anisotropy as well as discontinuous impulse propagation.The relationship between myo ber disorder and conduction block has been demonstrated previously in computational models 38 and low-resolution electrode recordings or optical maps from grossly approximated BZ regions 19,40−43 , but these studies utilized methods of localizing the BZ that are highly variable and subjective.Our ndings represent the rst experimental correlation of perturbed myo ber architecture to disordered electrical propagation at this degree of microstructural resolution and mathematical precision.This integrative approach to de ning the BZ region by both structure and function offers unparalleled anatomical consistency for studies of its pathophysiology.
Previous studies examining the post-MI distribution of cardiac sympathetic innervation relied on manual quanti cation of total nerve immuno uorescence 17,44−46 and thus lacked speci city for nerve endings versus larger pass-through bers.In contrast, our microstructural feature detection algorithms identi ed regional patterns of size-speci c nerve ber remodeling, allowing focus on the functionally important neuroeffector junction.Because we were able to automatically detect and de ne small ber dimensions speci cally by the size of sympathetic neuroeffector terminals [32][33][34] , we revealed the novel and important nding of small-ber predominance at the infarct BZ and small-ber decrease at the remote LV base.This post-MI perturbation of small-ber topography suggests an altered neural-myocardial interface, with regional loss of neuroeffector terminals at the LV base and nerve sprouting at the BZ.
Moreover, we discovered that the altered small nerve ber pattern after MI has a direct relationship to altered sympathetic control of ventricular repolarization.Speci cally, we found that chronic MI hearts display a correlation between regional variation in tyramine-mediated APD prolongation (higher in RV compared with LV base) and the spatial distribution of small ber prevalence (also higher in RV compared with LV base).This nding is especially important, as it establishes a potential neuralstructural substrate for the sympathetically driven increase in regional heterogeneity of repolarization, which may lead to arrhythmogenic gradients 5,47 .
Interestingly, the infarct BZ did not display higher APD prolongation compared with other regions, despite the higher small-ber prevalence suggestive of nerve sprouting.While nerve sprouting has previously been shown to be localized at the infarct BZ and to correlate with sudden death 20,48 , the precise pathophysiological processes remain unclear.Several possible mechanisms may underlie our nding.These sprouts may be dysfunctional in tyramine uptake via the norepinephrine transporter 49 , which has been previously shown to be downregulated after MI 44,49 .Alternatively, the sprouts may have altered neurotransmitter release functions 45 , or the cardiomyocytes in this region may have altered adrenergic receptor pro les 39,50 .That nerve sprouting at the BZ does not align with tyramine-mediated in uence on APD points to the functional distinctiveness of this boundary between surviving myocardium and dense scar, and promotes the generation of highly speci c hypotheses regarding sympathetically driven arrhythmias after MI.
Taken together, these data generated from our novel platform for structure-function alignment establish an important framework for understanding how structural cardiac diseases such as MI perturb speci c myocardial electrical functions, as well as the neural substrates and mechanisms which control these functions.While most of the existing neurocardiac literature focuses on molecular and cellular alterations in arrhythmogenic heart disease 51,9,50,39,10,11 , current clinical therapies for arrhythmia actually depend heavily on anatomical substrate modi cation 15 and neuromodulation at multiple structures of the autonomic nervous system 52 .These therapies have bene ted greatly from advancements in clinical imaging 53 to localize potential arrhythmogenic substrates, yet spatially correlating these substrates to their functional roles remains an important challenge.Thus, our study addresses the crucial need to understand post-MI neurocardiac dynamics at the whole-organ level, while still offering the high spatial resolution necessary to target the microstructural features underlying arrhythmogenic processes.
Overall, the synergistic neural-myocardial framework we present in this study is vitally important to elucidating the pathophysiology leading to sudden cardiac death.Our approach to structure-function alignment could feasibly incorporate emerging technologies, such as spatial detection of interstitial neurotransmitter levels with fast-scanning cyclic voltammetry 54 and optical norepinephrine tracers 55 , to generate additional mechanistic insights.Ultimately, a combination of such techniques will be needed to enable the development of more powerful and targeted neuromodulatory therapies for heart disease.

Animals
Animal experiments complied with all relevant ethical regulations and institutional regulations of the UCLA Animal Research Committee (Protocol #16-033).All mice used were male, C57BL/6J strain, obtained from the Jackson Laboratory.Survival surgeries to create chronic MI were performed when mice were 12 weeks (±5 days) of age (weighing 22-28g), and terminal optical mapping experiments occurred approximately 4 weeks after MI, when mice were 16 weeks (±4 days) of age.

Creation of chronic MI mouse model
Mice were anesthetized with iso urane (2%), endotracheally intubated, and mechanically ventilated.A small thoracotomy incision was made in the left 7 th or 8 th intercostal space to access the heart, the pericardium was opened with ne forceps, and the left coronary artery (analogous to the human left anterior descending artery) 56 was ligated with 8-0 silk suture at the mid-level of the LV.Acute transmural ischemia was con rmed by visualization of myocardial blanching and ST elevation on electrocardiogram (ECG).The incision was then closed in two layers (muscle and skin), and the animal was extubated and allowed to recover on a temperature-controlled surface.Carprofen (5mg/kg, intraperitoneal injection every 24 hours) and buprenorphine (0.02mg/kg, intraperitoneal injection every 8 hours) were given for pain control on the day of and for 48 hours after surgery.Sham surgeries included all steps except coronary artery ligation.

Optical mapping of action potentials
In Langendorff-perfused hearts, optical mapping of V m was performed as previously described 57,58 .Brie y, mice were sacri ced per protocol by anesthesia with 5% iso urane followed by cervical dislocation.Hearts were removed immediately and perfused via the aortic root with Tyrode's solution (130mM NaCl, 1.25mM CaCl 2 , 5mM KCl, 1.2mM NaH 2 PO 4 , 1.1mM MgCl 2 , 22mM NaHCO 3 , and 50mM dextrose).Hearts were immobilized and immersed in a Tyrode's solution bath within a 3-D printed chamber to reduce motion artifact.Perfusate and bath temperature were maintained at 36.6-37°C.
Hearts were stained with bolus injections of voltage-sensitive dye RH237 (8-10 µl of 2 mg/ml in DMSO, Thermo Fisher Scienti c, S1109) into the coronary perfusate.Blebbistatin (Cayman Chemical, 13186) was added to the perfusate at a concentration of 1.7ug/mL for excitation-contraction uncoupling.
Light from two collimated ultra-high-power LED (Prizmatix, UHP-T-520-EP) guides was focused on the ventral epicardial surface of the heart for excitation.Emitted uorescence was collected using a tandemlens arrangement of Nikon NIKKOR 50mm f/1.2 camera lenses and split with a 635 nm dichroic mirror (Edmund Optics, 87064). 59The V m signal was ltered at 690 ± 50 nm (Chroma ET690/50m), and a simultaneous bright eld image for vascular visualization and alignment was taken using the shorterwavelength ltered light at 590 ± 33nm (Chroma ET590/33m).The emitted Vm signals and bright eld images (for vascular alignment) were recorded using 2 CMOS cameras (SciMedia, MiCAM N256) with a sampling rate of 1.03 kHz and 256 x 256 pixels with a 14 x 14 mm eld of view.Pixel resolution of the images was approximately 55 x 55 µm.Data were acquired in 2-second intervals before and after addition of tyramine to the perfusate at a concentration of 5 µM.Data acquisition was done using BV Workbench software version 1.7.10 (SciMedia).
For conduction velocity analyses, epicardial pacing was performed from the basolateral LV wall at a cycle length (CL) of 167ms (with current of 1.1-1.3mA and pulse width of 0.8 ms), using a Transonic Scisense 1.1F mouse EP catheter (FTS-1113A-0518).For analyses of tyramine effect on repolarization, pre-and post-tyramine time points were taken in sinus rhythm, just before heart rate increase, to allow comparison at the same CL (Supplemental Figure 6).

Optical mapping data analysis
Optical mapping data was analyzed using the open-source software ElectroMap 31 .V m activation maps were displayed as isochronal maps generated from points of maximum upstroke (dF/dt) max as well as depolarization midpoint of optical APs.Repolarization maps were generated from points of APD 80 .A minimum of 4 beats were averaged at baseline and after tyramine infusion.A 3x3 Gaussian spatial lter, Top-hat, and Savitzky-Golay lters were applied to correct for baseline drift and noise.Maps were exported as 256 x 256 data matrices for alignment with structural data and quantitative analyses.

Immunohistochemistry and tissue clearing
After optical mapping, whole mouse hearts were xed by immersion in 4% paraformaldehyde/phosphatebuffered saline (PBS) overnight at 4°C, then washed three times for 1h in 0.01M PBS at room temperature (RT).Hearts were stained and cleared using a modi ed iDISCO protocol 28 .Fixed hearts were dehydrated by graded methanol treatments (20%, 40%, 60%, and 80% methanol in H 2 O (vol/vol), each for 1 h at RT), washed twice with 100% methanol for 1 h at RT, and chilled at 4°C.Hearts were then immersed in 66% dichloromethane/33% methanol overnight at RT with agitation, washed twice in 100% methanol for 1 h at RT, and chilled to 4°C.Next, hearts were bleached with 5% H 2 O 2 in methanol (vol/vol) overnight at 4°C.After bleaching, hearts were rehydrated with graded methanol treatments, followed by one wash with 0.01 M PBS and 2 washes with 0.01 M PBS with 0.2% Triton X-100, each for 1 h at RT. Hearts were permeabilized with 0.01 M PBS with 0.2% Triton X-100, 20% DMSO, and 0.3 M glycine and blocked with 0.01 M PBS with 0.2% Triton X-100, 10% DMSO, and 5% normal donkey serum, each for 2 d at 37°C with agitation.Hearts were incubated in sheep anti-TH (EMD Millipore, AB1542, 1:200) and/or rabbit antiperiostin (Abcam, ab14041, 1:200) diluted in 0.01 M PBS with 0.2% Tween-20 and 10 mg/ml heparin (PTwH) for 5-7 days at 37°C with agitation.Hearts were then washed 4-5 times in PTwH overnight at RT before incubating in secondary antibodies donkey anti-rabbit Cy3 (Jackson ImmunoResearch, 711-165-152, 1:300) and/or donkey anti-sheep Alexa Fluor 647 (Jackson ImmunoResearch, 713-605-147, 1:300) diluted in PTwH for 5-7 days at 37°C with agitation.Primary and secondary Ab were replenished approximately halfway through incubation period.Hearts were then washed several times in PTwH overnight at room temperature.For clearing, stained hearts were dehydrated with a graded methanol series and incubated in 66% dichloromethane/33% methanol for 3 h at room temperature with agitation.Hearts were then washed twice in 100% dichloromethane for 15 min at room temperature.Hearts were stored in benzyl ether (Millipore Sigma, 108014 ALDRICH; refractive index: 1.55) for up to 7 days prior to imaging.

Confocal imaging
Hearts were mounted in benzyl ether with adhesive plastic spacers (Sunjin Labs, IS012 and IS012).Images were acquired on a confocal laser scanning microscope (Zeiss, LSM 880) tted with the following objectives: Fluar 5x/0.25 M27 Plan-Apochromat (working distance 12.5mm) and 10x/0.45M27 (working distance 2.0 mm).Images were taken at both 5x and 10x magni cations for speci c ROIs, such as RV and LV base, prior to the whole hearts being imaged at 5x in tiles with XY-resolution of 1.661mm and Zresolution of 8.29mm.

Image processing and automated structural mapping
All image processing was performed using Zeiss Zen 2.1 v11, NIH ImageJ, Fiji 60 , and custom Matlab scripts (available upon request from corresponding author).Computational tracing of nerve bers was performed using a customized version of the open-source software neuTube 61 .neuTube software was originally developed for tracing morphology of single cells.To trace nerves in large image volumes required additional pre-and post-processing including: (1) partitioning large volumes into smaller tiles, tracing each tile, and reassembling the traced morphologies, and (2) ltering out spurious junctions between parallel bers and inaccurate ber diameter estimates arising due to background staining.To quantify myo ber orientation distributions, we utilized confocal images of muscle auto uorescence.This was validated using comparison of auto uorescence with virally labeled myocyte imaging (Supplemental Figure 1).We computed the image gradient orientation at each point and then smoothed the gradient orientation eld using a Gaussian weighted moving average window of size σ=100mm.
Structural images were aligned to functional images using vascular ducial points from bright eld images obtained during optical mapping.For each sample, 5-10 ducial points (branches in vasculature, sutures or scars) visible in both bright eld optical mapping and confocal images were used to t a perspective warping (homography) between the two images.Only structural data from the outer 100mmthick "shell" of each heart were used for alignment and correlation with optical mapping data.This depth was determined empirically by light penetration experiments (Supplemental Figure 7).

Quantitative data analysis
Conduction velocity and activation curves were calculated using ElectroMap.Regional myo ber anisotropy was a normalized index de ned as the coe cient of variation (angular standard deviation over the angular mean) of ber angles, divided by the total surface area of the segmented ROI: Per prior reports [32][33][34] , nerve ber size bins were de ned by the following diameters: small bers = 1.2-3 µm, medium bers = 3-5 µm, large bers = 5-100 µm.1.2µm was used as the lower limit of small bers to minimize detection of non-speci c background staining.Fiber prevalence was a normalized index de ned as the proportion of a particular size ber in a ROI, divided by the proportion of that ber size in the whole heart:

Figures
Figures

Figure 1 Optical
Figure 1 mapping and tissue clearing pipeline to align electrical and structural maps.(a) Schematic of optical mapping, clearing, imaging, and automated feature tracing steps in the alignment pipeline.(b, c) Bright eld image taken simultaneously with optical action potential map showing activation in sinus rhythm.(d) Maximum intensity projection (MIP) image of tyrosine hydroxylase (TH)-positive nerve bers on the ventral surface of the same heart after immunohistochemistry (IHC), tissue clearing, and confocal imaging.(e, f) Zoomed insets of (d) with TH staining alongside nerve ber tracing by computer vision, color-coded by ber diameter.(g, h) Venous bifurcations (magenta points) on MIP confocal shell image of a cleared heart alongside bright eld image of same heart were used as ducial anchors for alignment.(i) Automated global nerve ber tracing aligned with bright eld image allows spatial correlation with optical action potential data.Scale bars are 1mm (b-d, g-i) and 100µm (e, f).

Figure 2 Structure
Figure 2

Figure 3 A
Figure 3