Engineered Myovascular Tissues for Studies of Endothelial/Satellite Cell Interactions

Abstract

In the native skeletal muscle, capillaries reside in close proximity to muscle stem cells (satellite cells, SCs) and regulate SC numbers and quiescence through partially understood mechanisms difficult to study in vivo. This challenge could be addressed by the development of a 3-dimensional (3D) in vitro model of vascularized skeletal muscle harboring both a pool of quiescent SCs and a robust network of capillaries. Still, studying interactions between SCs and endothelial cells (ECs) within a tissue-engineered muscle environment has been hampered by the incompatibility of commercially available EC media with skeletal muscle differentiation. In this study, we first optimized co-culture media and cellular ratios to generate highly functional vascularized human skeletal muscle tissues (“myovascular bundles”) with contractile properties (~10 mN/mm²) equaling those of avascular, muscle-only tissues (“myobundles”). Within one week of muscle differentiation, ECs in these tissues formed a dense network of capillaries that co-aligned with muscle fibers and underwent initial lumenization. Incorporating vasculature within myobundles increased the total SC number by 82%, with SC density and quiescent signature being increased proximal (≤20μm) to EC networks. In vivo, at two weeks post-implantation into dorsal window chambers in nude mice, vascularized myobundles exhibited improved calcium handling compared to avascular implants. In summary, we engineered highly functional myovascular tissues that enable studies of the roles of EC-SC crosstalk in human muscle development, physiology, and disease.

Graphical Abstract

1. Introduction:

Skeletal muscle comprises an average of 30-50% of the human body by mass and generates contractile forces responsible for locomotion and respiration. The high regenerative capacity of skeletal muscle is enabled by the existence of muscle stem cells, termed satellite cells (SCs), which in healthy muscle express the transcription factor Pax7, are negative for MyoD, and reside in a quiescent, non-cycling (Ki67⁻) state in specialized niches abutting myofibers [1-3]. Upon injury, SCs activate (expressing MyoD), proliferate, and either (1) return to quiescence or (2) commit to differentiation to repair damaged or form new myofibers [4, 5]. SC fate in skeletal muscle is regulated by cell-matrix interactions and both juxtacrine and paracrine signals from myofibers and resident non-muscle cells [1, 4-6]. In particular, SCs and endothelial cells (ECs) are found in close proximity [7-9] where reciprocal juxtacrine [10] and paracrine [7, 9, 10] interactions have been shown to maintain SC quiescence [7, 9, 10] and capillary patterning [10] in homeostasis, or promote synergistic angio-myogenesis during regeneration [11-13]. Additionally, SC density is directly correlated with vascular density in vivo [9, 14, 15] and dysfunction of one cell type impairs function of the other via paracrine signals [16, 17]. However, direct mechanistic studies of SC-EC interactions in vivo are hindered by the inability to study muscle biology in the absence of vasculature and by local and systemic confounding variables present in the in vivo environment.

The modular nature of in vitro tissue-engineered platforms permits tightly controlled cellular crosstalk studies of ECs with multiple cell types in a physiologically relevant, three-dimensional (3D) culture environment. In our original studies, we have described engineering of highly functional 3D neonatal [18] and adult-derived rat skeletal muscle tissues harboring a quiescent SC pool with regenerative capacity that could be augmented by the inclusion of macrophages [19]. More recently, we have developed engineered human skeletal muscle tissues (“myobundles”) using primary or iPSC-derived muscle progenitor cells which allowed us to study human muscle response to exercise [20, 21], disease [20, 22], injury [23], and drug treatment [22, 24]. Recently, we have shown that human primary SCs, which are typically activated during in vitro culture, undergo deactivation in 3D myobundles to attain a quiescent (Pax7⁺/MyoD⁻/Ki67⁻) but heterogeneous phenotype akin to that of the native SCs in vivo [25]. Yet, interactions between SCs and microvasculature have not been studied within homeostatic tissue-engineered skeletal muscles, though in 2D culture, dysfunctional ECs have been shown to impair SC proliferation and differentiation [17].

To date, in vitro vascularization of skeletal muscle tissues has been mainly utilized to accelerate their perfusion, viability, and engraftment efficiency after implantation in vivo [26-29]. However, functionality of such vascularized muscle models has been significantly hindered by inability to define proper coculture media. Simple mixing of EC- and muscle-specific growth media does not resolve this issue, yielding suboptimal EC viability and vascular tube formation and decreased muscle differentiation and function [30, 31]. Recently, human plasma-like media (HPLM), comprised of human physiologically relevant macro- and micronutrient levels, has been developed to more accurately model human circulating factors [32]. Compared to traditional media, HPLM better replicates clinical drug responses [32-35] and T cell function in vitro [34], but its potential to support coculture of incompatible cell types has not been established. Alternatively, EC-muscle cocultures have been engineered to mimic the compartmentalization of fluid within vasculature [30, 36]. In these models, a hollow EC channel surrounded by engineered muscle tissue allows for use of cell-specific media inside and outside the channel, which improves muscle functionality compared to use of mixed media [30, 36]. Still, compared to native muscle tissue, ECs in these models only line the muscle surface, limiting their utility for studies of complex EC-SC interactions found in the native muscle. As such, the roles of ECs in SC niche formation and function are yet to be studied in a homeostatic 3D muscle tissue environment in vitro.

To study the effect of vasculature on the human SC niche, we first optimized media for coculture of primary myogenic cells and ECs and generated vascularized myobundles with contractile function comparable to that of avascular myobundles. In these “myovascular tissues”, we then evaluated the effect of vasculature on abundance of Pax7⁺ cells, their quiescence signature, and proximity to formed vessels. Finally, avascular and myovascular tissues were implanted into mouse dorsal skin-fold window chambers to assess the longevity of implanted vascular networks and the effects of in vitro pre-vascularization on myobundle perfusion throughout the duration of implantation and Ca²⁺ handling 2 weeks post-implantation.

2. Materials and Methods

2.1. Human Myogenic Cell Isolation and Culture

Human skeletal muscle tissue was acquired from surgical waste of deidentified donors with no known pathology per Duke University Institutional Review Board approved protocols (Pro00052540, Pro00054162) and myogenic cells were isolated as previously described [24, 25]. Briefly, freshly harvested human muscle tissue was washed in PBS with 1X Anti-Anti and finely minced with a scalpel. Tissue fragments were partially enzymatically digested in 0.05% Trypsin for 30 minutes on a rocker at 37°C. Following digestion, trypsin was neutralized using Fetal Bovine Serum (FBS) (Hyclone), the remaining fragments were spun down and resuspended in DMEM Low Glucose (Sigma) supplemented with 1X Anti-Anti (Gibco), then pre-plated in 1 T-175 for 1 Hour at 37°C. The tissue fragment containing media was then collected, spun down, and the tissue fragments were resuspended in skeletal muscle 2D growth media (DMEM Low Glucose, 10% FBS, 0.4ng/mL Dexamethasone (Sigma), 10ng/mL epidermal growth factor (EGF, Prospecio), and 0.064mg/mL Penicillin, supplemented with 2X Anti-Anti) and plated on Matrigel (Corning) coated plates to allow for cell outgrowth for approximately 1 week (with media changes every 2 days). During and following outgrowth, SCs from muscle tissue fragments proliferate and spontaneously differentiate into myoblasts, which can be repeatedly passaged and cultured or frozen for downstream use. These myogenic cells were cultured on Matrigel coated flasks in skeletal muscle growth media, passaged with 0.025% Trypsin (Thermo Fisher) at ~85% confluency, frozen at earlier passages if needed, and used at passage 5 for all shown studies. Cells from three human donors were used, and each experiment was repeated using cells from at least two independent donors.

2.2. Human Endothelial Progenitor Cell Isolation and Culture

Human endothelial progenitor cells (EPCs) were generously provided by the Truskey Lab at Duke University. As previously described, EPCs were isolated from human cord blood (sourced from the Carolina Cord Blood Bank, Duke IRB protocol Pro00058324) [37]. The whole blood was diluted 1:1 with Hanks Balanced Salt Solution and Histopaque-1077 (Sigma). The suspension was spun at 740G for 30 minutes to create a density gradient and mononuclear cells were collected and plated in endothelial growth media (EBM2 (Lonza) supplemented with EGM-2 bullet kit and 10% FBS) on Collagen Type I (rat tail, BD Biosciences) coated flasks. Media was changed daily, and endothelial colonies were visible within 7-10 days of plating. EPC identity was confirmed via flow cytometry for CD31 and cells were passaged or frozen for later use. EPCs were cultured in endothelial growth media (EGM-2, Lonza) and were passaged with 0.05% Trypsin through passage 10.

2.3. Lentivirus Production and Cell Transduction

Lentiviral plasmids were designed, generated, and lentivirus was produced as previously described [38-40]. Briefly, to generate a CMV-mCherry lentivirus, a fluorescent mCherry reporter gene was inserted into a pRRL-CMV plasmid (a gift from Inder Verma, Salk Instute). To generate an MHCK7-GCaMP6 lentivirus, the muscle-specific promotor MHCK7 [41] and the fluorescent calcium reporter GCaMP6 [42] were inserted into the pRRL backbone plasmid. HEK293T cells were cultured in high glucose DMEM supplemented with 10% fetal bovine serum and 1% Penicillin/Streptomycin. At 70-80% confluency, HEKs were co-transfected with pRRL-CMV-mCherry or pRRL-MHCK7-GCaMP6, VSVG, and Pax2 plasmids using Jetprime. HEK media was changed 16 hours post-transfection and media containing virus was collected 2-3 days following the initial media change. Virus was purified by precipitation with addition of 33% LentiX- Concentrator at 4°C on an orbital shaker and pelleted by centrifugation at 1500xg for 45 minutes. Precipitated virus was then resuspended in PBS, aliquoted, and stored at −80°C. To label EPCs with mCherry flourescence, cells were transduced with the CMV-mCherry lentivirus (4.8 vp/cell) at the time of plating and media was changed 24 hours later. Transduced EPCs were cultured for 3-4 days to allow for visual confirmation of lentiviral expression, before being used for myovascular bundle fabrication. For GCaMP6 transduction of myoblasts, pRRL-MHCK7-GCaMP6 lentivirus was added to the cell suspension (0.17 vp/cell) at the time of myobundle fabrication.

2.4. Human Primary Myobundle Fabrication

Human myobundles were fabricated in polydimethylsiloxane PDMS molds with two 5-mm long, 1-mm wide semi-circular channels, as previously described [21, 22, 24, 25]. Briefly, molds were coated with 2% (w/v) pluronic F-127 (Invitrogen) for 30 minutes at room temperature to prevent myobundle adhesion. Cerex^® (nylon) frames were placed in the molds to serve as anchor points for the myobundles and support uniaxial tension. A cell solution (375K myogenic cells – with or without 18.75K EPCs – in 9.3μL media and 1μL of 50 U/mL thrombin (Sigma) in 1% BSA in PBS per myobundle) and a gelling solution (1.75μL media, 5μL Matrigel, and 5μL of 22mg/mL fibrinogen (Sigma) in PBS per myobundle) were prepared separately on ice in vials to make 4 myobundles at a time. Cell solution was added to the gelling solution, pipetted into the channels of the PDMS molds, and left to polymerize for 30 minutes at 37°C prior to the addition of media. Myobundles were then cultured dynamically on a rocker in in skeletal muscle 3D growth media (2D growth media supplemented with 1.5mg/mL 6-aminocaproic acid (ACA, Sigma) and 10ng/mL human VEGF-165 (Peprotech)) for 4 days, after which media was switched to custom-designed human plasma like media (HPLM) (Sup Table 1) supplemented with 10 ng/mL human VEGF-165 for an additional 7-10 days of differentiation. At culture day 2, frames with myobundles were removed from PDMS molds and cultured in free-floating conditions [43]. Media was changed every other day through both the growth and differentiation phases.

2.5. Force Testing

Contractile function of myobundles was tested between days 7 and 9 of differentiation using a custom force measurement set-up, as previously described [21, 24, 25, 44]. Briefly, individual myobundles were placed in the testing chamber in DMEM Low Glucose media and maintained at 37°C, pinned to a fixed PDMS block on one end and a PDMS float attached to a force transducer (mounted on a computer-controlled linear actuator (Thor Labs)) on the other. The myobundle frame was cut to allow myobundle stretching (via the linear actuator) and tissue was stimulated with 90V/cm 5ms electrical pulses using parallel platinum electrodes submerged in the chamber. Twitch contraction was measured at 0, 5, 10, and 15% stretch. At the stretched length yielding the maximum twitch contractile force, myobundles were stimulated at frequencies of 1, 5, 10, 20, and 40Hz to record force-frequency response. Recorded force traces were analyzed using a custom MATLAB script to determine the maximum active force, time to peak twitch, and half-relaxation time [21, 24, 25, 44].

2.6. Ca²⁺ Transient Imaging

At the time of myobundle formation, myoblasts were transduced with an MHCK7-GCaMP6 lentivirus, as previously described [19, 24, 38]. Electrically stimulated Ca²⁺ transients were quantified after 1 week of differentiation or after additional 2 weeks post-implantation into mouse dorsal skinfold window chambers [19]. Myobundles were transferred into a custom glass-bottom chamber filled with 37°C Tyrode’s solution and stimulated using parallel carbon electrodes. GCaMP6 fluorescence movies were acquired through a FITC filter set at 50 fps rate under 4X magnification using an Andor iXon camera affixed to a Nikon microscope. Upon background subtraction, relative change in fluorescence ΔF/F₀ within a region of interest encompassing the majority of the myobundle was calculated as (F-F₀)/F₀ with F₀ being the resting fluorescence level. Ca²⁺ transient amplitude was calculated as the peak ΔF/F₀ induced by electrical stimulation. All analysis was performed using the Andor Solis software and a custom Matlab script [19].

2.7. Immunohistology and Image Analysis

Myobundles were fixed with 2% paraformaldehyde in PBS overnight at 4°C. Following fixation, samples were washed with PBS and either immunostained as whole tissues or embedded in Optimum Cooling Temperature (OCT) Compound (Sakura) for cryosectioning and cross-sectional immunostaining. For cross-sectional images, myobundles were embedded in OCT perpendicular to the cutting surface, snap frozen in liquid nitrogen, cut in 10 μm sections, and mounted on slides. For antigen retrieval (M-cadherin staining), slides were submerged in citrate buffer containing slide holders and placed in a 2100 Antigen Retriever (Aptum Biologics) for the duration of the retrieval cycle. Samples were then blocked for 30 minutes at room temperature with a solution of 5% donkey serum, 0.1% Triton X-100 and 0.03% Sodium Azide in PBS. Primary antibodies were applied overnight at 4°C followed by secondary antibodies applied for 2 hours at room temperature. For whole-tissue stains, samples were blocked overnight at 4°C with a solution of 5% donkey serum (Abcam), 0.3% Triton X-100 (Thermo Fisher), and .03% Sodium Azide in PBS. Both primary and secondary antibodies (Sup Table 2) were applied overnight. Samples were then washed and mounted for imaging using ProLong Glass Antifade Mountant (Invitrogen) and were imaged using an Andor Dragonfly spinning disk confocal microscope.

Immunoflourescent images were analyzed using Image J or custom CellProfiler Pipelines. Generally, primary objects were identified using Huang global thresholding (Image J) or robust background adaptive thresholding (CellProfiler). For whole myobundle quantifications (CD31⁺ Area, %Pax7⁺ Nuclei, Pax7⁺ cell distance to vessels, %Ki67⁺ SC), images were captured through the depth of the myobundle in 1mm Z-slices. A minimum of 3 representative Z-slices were then quantified and averaged for each tissue. For 2D myotube diameter measurements, 20 manual measurements of diameter were averaged per monolayer. Each diameter measurement was made perpendicular to the longitudinal axis of the myotube and intersected a minimum of 1 myogenin⁺ nucleus. Vascular morphology was quantified using the ImageJ tool “Angiogenesis Analyzer” [45], which identified vascular structures from thresholded whole-mount CD31-stained images and calculated total vessel length. The average vessel length was calculated by dividing the total vessel length by the number of independent vascular structures, counted manually. The vessel alignment was quantified using the ImageJ “Directionality” tool to analyze alignment of vessel “trees” generated by Angiogenesis Analyzer. Three representative Z slices were quantified and averaged for each tissue. For concentric vessel density analysis myobundle cross-sectional area was outlined manually and automatically segmented into three segments with equal thickness (center, middle, and outer) using a custom Cell Profiler pipeline. Vessel density in each segment was calculated by dividing CD31+ area within the segment by the segment area.

2.8. Dorsal Skinfold Window Chamber Implantation

Animal experiments were approved by the Duke University Institutional Animal Care and Use committee and followed all clinical and ethical regulations as detailed by the NIH Guide for Care and Use of Laboratory Animals. MHCK7-GCaMP6 transduced avascular (control) and myovascular bundles were implanted into athymic nude mice dorsal skinfold window chambers as previously described [18, 19, 38]. Briefly, mice (~10 weeks of age; 25-35g; male) were anaesthetized via an intraperitoneal injection of ketamine/xylazine (100 mg kg⁻¹/ 10 mg kg⁻¹) and were attached to a temporary “C-frame” which was used to demarcate where the forward-facing skin needed to be resected. All layers of the forward-facing skin (cutis, subcutis, retractor and panniculus carnosis muscles, and associated fascia) were dissected away and the titanium dorsal skinfold window chamber was assembled around the dissected skin. The frame with myobundles was then laid over the panniculus carnosis muscle of the rear-facing skin and a sterile cover glass was placed over the implanted tissue while the chamber was superfused with sterile PBS. The mice were then removed from the “C-frame”, the cover glass was secured with a ring clamp, and the chamber was secured with sutures along the top line. Pre-operatively, mice were injected subcutaneously with buprenorphine ER and were left to recover on a heat pad under observation. The mice were checked daily for signs of discomfort or infection with antibiotics administered orally as needed.

2.9. Intravital Imaging of Implanted Myobundles

Intravital imaging of implanted avascular and vascularized myobundles was performed on days 4, 7, 11, and 14 post-implantation in dorsal window chambers using our previously published methods [18, 19, 46]. Briefly, mice were anesthetized by isoflurane and positioned onto a custom frame mounted above the microscope objective. The body of the mouse was covered with a 37°C heating pad. All images were captured by a 2.5X objective (Zeiss), a tunable filter (Cambridge Research & Instrumentation, Inc.), and a DVC Camera (DVC-1412M, ThorLabs). Brightfield images were captured at 510 nm and mCherry images were captured at 610 nm. To quantify blood vessel density (BVD), images were processed using local contrast enhancement in Image J (CLAHE, FIJI), myobundle borders were manually identified, and large blood vessels located behind the implanted myobundles were manually excluded. BVD was calculated by dividing total vessel area by the myobundle area.

2.10. Statistical Methods

Data was analyzed using Graphpad Prism Version 10.2.0 and shown as mean ± SEM. For immunohistochemistry quantification and contractile force measurements, individual data points represent separate myobundles or 2D cultures. Outlier tests (ROUT, Q=2%) were run on all data points. Paired or unpaired T-tests were used for comparisons between 2 groups with normal distributions depending on experimental design. Mann-Whitney tests were used for comparison between 2 groups with non-normal distributions. 1- or 2-way ANOVA’s were used for multiple comparisons (based on experimental design) followed by Sidak post-hoc testing as applicable. Ex vivo GCaMP quantifications were analyzed using a 2-way ANOVA main effects model with repeated measures.

3. Results

3.1. Optimized culture conditions enable the formation of vascular networks with lumen-like structures within engineered myobundles

Generation of high-fidelity in vitro vascularized engineered muscle tissues has long proved challenging due to differing EC and muscle media requirements [30, 31]. To enable the culture of vascularized human myobundles, we first optimized media conditions for both human primary endothelial progenitor cells (EPCs) and human primary myogenic cells in 2D monoculture. The effects of 3D skeletal muscle growth media (3DG) on EPC viability and proliferation were assessed by culturing EPCs in 2D for 4 days (Sup Fig 1a) to mimic the 4-day growth phase of myobundle culture (Fig 1a). While commercially available EGM-2 EC media resulted in continuous EPC proliferation, 3DG media resulted in progressive cell death (Sup Fig 1b, c). This cell death was prevented by supplementing 3DG with 10 ng/mL of VEGF-165, the most abundant and biologically active VEGF isoform [47, 48], which maintained EPC numbers through at least 4 days of culture (Sup Fig 1b,c).

Figure 1: Development of aligned and lumenized endothelial cell (EC) vessels within vascularized myobundles.

a, Experimental schematic of myovascular bundle formation, culture, and analysis (IHC, immunohistochemistry). b-e, Representative whole-mount images of avascular, muscle-only (Mu) and vascularized (Mu+EC) myobundles just prior to switch to HPLM differentiation media (D0, b) and on day 7 of differentiation (D7, c; note co-alignment with F-actin⁺ myofibers) and corresponding quantifications of CD31⁺ (EC) area (n=14 myobundles, ***p<0.001, d) and vessel length (n=14 myobundles, ****p<0.0001, e). f-i, Representative images of Mu and Mu+EC myobundle cross-sections (f) and corresponding quantifications of myobundle cross-sectional area (n=14-15 myobundles, ns: p>0.05, g), nuclei count per cross-section (n=9-10 myobundles, ns: p>0.05, h), F-Actin area per nucleus (n=9-10, ns: p>0.05, i), and regional CD31⁺ (EC) density in myovascular bundles (n=11 myobundles ns: p>0.05, *p<0.05, **p<0.01, j) on D7 of differentiation. k-l, Representative image of myovascular bundle cross-section showing formation of CD31⁺ lumens (white arrow, k) and corresponding quantification of lumen density at D0 and D7 (n=11-12, **p<0.01, I).

Typically, vascularized engineered muscle tissues have been generated by mixing skeletal muscle and EC media in varying ratios resulting in compromised engineered muscle functionality [36]. To overcome this limitation, we developed a serum-free human plasma-like media (HPLM) that better mimics nutrients and metabolites found in healthy plasma (Sup Table 1). Specifically, by utilizing the human blood metabolome database [49], we identified 34 metabolites expressed at greater than 1μM in healthy adult blood and included those with physiological levels of amino acids and established levels of inorganic salts, trace elements, and vitamins previously utilized in serum-containing HPLM [32-34]. We then identified and added 25 additional hormones, antioxidants, lipids, and growth factors that support differentiation of human myoblasts including IGF-1 [50] and circulating factors such as bFGF [51], EGF [52, 53], thyroid hormone [54-58], and heparin [59, 60], which modify growth factor activity and support both myogenesis [50, 57-59] and vasculogenesis [51-55, 60]. Addition of VEGF to HPLM supported the survival of EPCs for at least 6 days of 2D monoculture (Sup Fig 1b), with cell numbers that were comparable to cell numbers during culture in 3DG+VEGF media (Sup Fig 1d).

We next determined the effects of the tested culture media on primary human myogenic cell differentiation for 5 days in 2D culture by immunostaining for sarcomeric α-actinin (SAA) and the terminal muscle differentiation transcription factor myogenin. We found that EGM-2 alone limited fusion and differentiation capacity of human myogenic cells, as evident from reductions in percentage of myogenin⁺ nuclei incorporated into SAA⁺ myotubes and average myotube diameter (Sup Fig 2a-c). Furthermore, use of EGM2 media yielded the formation of non-contractile myobundles (data not shown). In contrast, use of HPLM enabled robust myoblast fusion as shown by the formation of greater than 60 μm thick SAA+ myotubes (Sup Fig 2c). Similar to previous studies with use of traditional muscle differentiation media [61], mixing of EGM-2 with HPLM resulted in a dose-dependent loss of myoblast fusion capacity (Sup Fig 2b) and decrease of myotube diameter (Sup Fig 2c).

For subsequent 3D culture, we characterized the composition of cells used for myobundle fabrication by staining for muscle specific transcription factors Pax7 and MyoD, the EC marker CD31, and the pericyte marker PDGFRβ. This cell population consisted of predominantly myogenic progenitors (90.4±1.65% MyoD⁺ or Pax7⁺, Sup Fig 3a,b), a small fraction of PDGFRβ⁺ cells (1.6±0.24%, Sup Fig 3c-e), no CD31⁺ ECs (Sup Fig 3f-h), and the reminder of vimentin⁺ cells (Sup Fig 3i,j). To generate 3D vascularized myobundles (“myovascular bundles”), EPCs were lentivirally transduced with mCherry to facilitate identification during culture and supplemented into the myogenic cell suspension used for myobundle fabrication (Fig 1a). We first optimized the EPC: ratio and found that the 1:20 ratio (5%) enabled formation of vascular networks (Sup Fig 4a,b) without detrimental effects on muscle contractile function (Sup Fig 4c). We then compared EC network formation and muscle force generation between HPLM and our traditional (N2) muscle differentiation media [25] both supplemented with VEGF. In 1-week differentiated myobundles, HPLM supported greater EC network density (Sup Fig 5a,b) and contractile force amplitude (Sup Fig 5c) compared to traditional muscle differentiation media. Based on these findings, we proceeded to characterize myovascular bundles comprised of 5% EPCs and cultured in 3DG growth media followed by HPLM differentiation media, both supplemented with VEGF-165. On day 0 of differentiation (D0), densely packed F-actin⁺ differentiating myotubes and developing mCherry⁺ EPC networks were present within the myovascular bundles but lacked the organization present in native muscle (Fig 1b). However, by day 7 of differentiation (D7), F-actin and SAA immunostaining revealed uniaxially aligned, cross-striated multinucleated myotubes present in in both avascular and vascularized myobundles (Fig 1c, Sup Fig 6a,b), indicating that the presence of vasculature did not impact muscle architecture or sarcomere organization. Additionally, mCherry⁺ EPC networks aligned within myobundles starting from D0 (Sup Fig 6c), thus recapitulating the structure of native muscle-resident vasculature [28]. We further found that CD31⁺ area in vascularized myobundles increased from D0 to D7 (Fig 1d), confirming the findings from our 2D studies that optimized media conditions prevent loss of EC viability. Additionally, with time of culture, organization of formed vascular networks was improved, as evident from the increased average vessel length at D7 of differentiation (Fig 1e).

To further assess morphological differences between avascular and vascularized myobundles, we performed cross-sectional stainings of D7 tissues (Fig 1f) and found no differences in gross tissue morphology, including tissue cross-sectional area (CSA) (Fig 1g), nuclear density (Fig 1h), F-Actin⁺ CSA (Sup Fig 6d), or F-Actin⁺ CSA per nucleus (indicating myotube size, Fig 1i). In myovascular bundles, we found CD31⁺ vessels to be the most abundant at the outer tissue region (Fig 1f,j), indicating potential roles of oxygen and mass transfer gradients in the vessel formation. Further analysis of CD31⁺ cells in cross-sections of D0 and D7 tissues revealed an EC network density of 185.4±60.34 CD31⁺ cords/mm² at D7 (Sup Fig 6e), comparable to capillary density of native skeletal muscle [62, 63]. Capillary development in vivo is characterized by the formation of vascular cord-like structures which subsequently mature to form lumens and ultimately support blood flow [64]. Notably, from D0 to D7 of differentiation, we observed the formation of lumen-like structures at a density of 24.23±21.61 lumens/mm² (Fig 1k, l), which despite being significantly lower than in native muscle [62, 63], suggested progressive maturation of vascular networks with time of culture.

3.2. Myovascular bundles maintain contractile function comparable to avascular myobundles

To assess the effects of vascularization on muscle function, we next compared isometric force and Ca²⁺ transient generation in D7 avascular and myovascular tissues. We found no difference in the maximum active force, maximum specific force, or passive force generation in avascular vs. vascularized myobundles (Fig 2a-d). Analysis of twitch kinetics revealed a small but significant decrease in time to peak contractile force in vascularized myobundles (Mu: 75.45 ± 1.872 ms, Mu+EC: 68.00 ± 1.214 ms) (Fig 2e), while no changes were observed in half-relaxation time (HRT, Fig 2f). We next measured electrically induced Ca²⁺ transients in D7 myobundles and myovascular bundles transduced with a muscle-specific MHCK7-driven GCaMP6 lentivirus, as previously described [18, 19, 38]. In agreement with similar muscle mass and contractile strength of myobundles and myovascular bundles, we found no differences in maximal Ca²⁺ transient amplitude or kinetics (not shown) during twitch and tetanic contraction (Fig 2h-j). Collectively, these findings indicate successful vascularization of human myobundles with no detrimental effects on high contractile strength achieved in muscle-only tissues.

Figure 2: Functional characterization of myovascular bundles.

a-f, Representative traces of twitch (1Hz stimulation) and tetanic (40Hz stimulation) contractions in avascular (Mu) and myovascular (Mu+EC) tissues at differentiation day D7 (a) and corresponding quantifications of maximum active (tetanic) force (n=28-31 myobundles, ns: not significant, b), specific tetanic force (n=10-12, ns: not significant, c), passive force generation at 5%, 10%, and 15% stretch (n=7-12 myobundles, ns: not significant, d), time to peak twitch (n=22-23, **p<0.01, e), and half-relaxation time (HRT, n=22-26, ns: not significant, f). g, Representative snapshots of baseline and peak GCaMP6 fluorescence during twitch and tetanic contraction in Mu and Mu+EC myobundles. h-j, Representative traces of twitch and tetanic Ca²⁺ transients (shown as relative GCaMP6 fluorescence, ΔF/F₀, h) and corresponding quantifications of twitch (n=16, ns: not significant, i) and tetanic (n=16, ns: not significant, j) Ca²⁺ transient amplitudes shown relative to Mu group.

3.3. Vascularization of myobundles results in increased number, juxtavascular positioning, and decreased proliferation of Pax7⁺ SCs

In native skeletal muscle, ECs and SCs correlate in numbers [9] and capillaries reside near SCs to form juxtavascular niches where they regulate SC quiescence and activation [10]. In our engineered tissues, the presence of Pax7⁺ cells expressing the niche marker M-cadherin (Sup Fig 7a,b) suggested the formation of SC niches [25]. Quantitative image analysis further revealed that compared to avascular bundles, myovascular bundles exhibited an 82% increase in Pax7⁺ SC number by D7, while no significant differences were observed at D0 (Fig 3a,b). We next assessed Pax7⁺ cell cycle activity (a hallmark of SC quiescence) by immunostaining for the proliferation marker Ki67 and found no difference in SC Ki67 expression between control and vascularized myobundles on either D0 or D7 (Fig 3c). As SCs are typically found within 20μm of ECs in healthy native muscle [9], we next examined if this preferential SC-EC proximity was replicated in myovascular bundles by quantifying SC density within and beyond 20μm of the nearest CD31⁺ vessel (Fig 3d). Consistent with average SC density results (Fig 3b), SC density in vascularized myobundles was increased both within and beyond 20μm of the nearest CD31⁺ vessel (Fig 3e). Additionally, it was significantly increased within vs. beyond 20μm of the nearest CD31⁺ vessel (Fig 3e). We then corroborated these findings by analyzing images of “mock” Mu+EC myobundles generated by overlaying representative CD31⁺ vessel masks derived from myovascular bundle images onto images of avascular myobundles stained for Pax7 (Sup Fig 7c). In this case, we found no preferential SC localization near overlaid vessels either through Pax7 density analysis (Sup Fig 7d) or when comparing distance histograms of mock vs. true myovascular myobundles (Sup Fig 7e,f). Furthermore, analysis of Pax7⁺ nuclei in myovascular bundles revealed significantly decreased cell cycle activity (assessed by the fraction of Ki67⁺/Pax7⁺ nuclei) within 20μm of the nearest vessel (Fig 3f), consistent with the enhanced quiescence of juxtavascular SCs observed in vivo [10]. Together, these findings show that the presence of EC networks within myovascular bundles increases SC numbers and decreases SC proliferation near formed vessels.

Figure 3: Effects of vascularization on abundance and location of Pax7⁺ cells in myobundles.

a-c, Representative whole-mount images of myobundles on D0 and D7 of differentiation and corresponding quantifications of percentage of Pax7⁺ nuclei (relative to Mu group, n=11-26 myobundles, ****: p<0.0001, ns: not significant, b) and Pax7⁺ cells that are Ki67⁺ (n=11-26, ns: not significant, c). d, Representative whole-mount images of Mu and Mu+EC myobundles. Yellow arrows in myovascular bundles denote Pax7⁺ nuclei in juxtavascular positions. e, Quantification of Pax7⁺ density in Mu myobundles and within 20μm and beyond 20μm from the nearest CD31⁺ vessels in vascularized myobundles, all shown relative to Mu group (n=16, *p<0.05, **p<0.01, ****p<0.0001), h, Quantification of percentage Pax7⁺ cells that are Ki67⁺ within 20μm and beyond 20μm from the nearest CD31⁺ vessels in vascularized myobundles (n=13, *p<0.05).

3.4. Factors secreted by vascularized myobundles do not increase SC abundance in avascular myobundles

EC regulation of SCs is known to be multifactorial involving both juxtacrine and paracrine signaling [7, 9, 10]. To assess if paracrine signaling was sufficient to increase SC numbers as observed in vascularized tissues, we treated avascular myobundles daily with media conditioned for 24hrs by either avascular myobundles (control) or myovascular bundles (Fig 4a). Whole-tissue immunostaining revealed no effect of myovascular bundle conditioned media on SC abundance (Fig 4b,c) or cell cycling (Fig 4d) in avascular myobundles. Furthermore, unlike previous reports suggesting that paracrine signaling from ECs can improve engineered muscle contractility [65], we found no effects of conditioned media from either control or vascularized myobundles on contractile force amplitude (Fig 4e) or kinetics (Sup Fig 8).

Figure 4: Effects of vascular paracrine signals on SC abundance in myobundles.

a, Schematic of experimental design for media conditioning experiments. b-e, Representative whole-mount images of Mu myobundles, Mu+EC myobundles, Mu myobundles conditioned with Mu myobundle media, and Mu myobundles conditioned with Mu+EC myobundle media (b) and corresponding quantifications of the percent of Pax7⁺ nuclei shown relative to Mu group (n=13-16 myobundles, **p<0.01, c), percentage of Pax7⁺ cells that are Ki67⁺ (n=13-16, ns: not significant, d) and maximum tetanic force (40Hz stimulation) shown relative to Mu group (n=14-16, ns: not significant, e).

3.5. In vitro vascularization improves myobundle calcium handling following implantation into mouse dorsal skinfold window chambers

Engineered muscle tissues have long held promise as a therapeutic approach to regenerate skeletal muscle following large injuries or genetic diseases [66]. We have previously implanted myobundles transduced with MHCK7-GCaMP lentivirus into athymic nude mouse dorsal skinfold window chambers, to allow live, minimally invasive measurements of implant vascularization and Ca²⁺ handling [18, 19, 46]. Building on this work, we utilized the window chamber implantation model to assess the longevity of engineered myovascular bundles in vivo and study how the presence of in vitro engineered vasculature affects myobundle vascularization/perfusion and Ca²⁺ handling post-implantation (Fig 5a,b). Specifically, by intravital in situ imaging, we found progressive increase in perfused blood vessel density (BVD) within both avascular and pre-vacularized myobundles over 14 days post-implantation (Fig 5c,d). Interestingly, in vitro vascularization of myobundles showed no significant effects on perfused vessel ingrowth at any point post-implantation compared to vessel ingrowth in avascular implants (Fig 5c,d). By specifically tracking the implanted vasculature via mCherry fluorescence, we further found that implanted mCherry⁺ vessels remained stable after implantation (Fig 5e,f). After two weeks, implanted myobundles were explanted and electrically stimulated to assess potential effects on Ca²⁺ handling properties. Despite no differences in time course of in vivo vascularization, Ca²⁺ transient amplitude was significantly increased in myovascular vs. avascular myobundle explants (Fig 5g-i) and associated with a faster Ca²⁺ transient rise (Fig 5j). Together, these studies showed that compared to muscle-only implants, in vitro pre-vascularization of engineered muscle tissues can result in improved muscle Ca²⁺ handling upon implantation in vivo.

Figure 5: Improved functionality of myovascular bundles implanted in mouse dorsal window chambers.

a, Schematic of dorsal skinfold window chamber experimental design, timeline, and end point. b, Photograph of myobundles implanted into a dorsal window chamber in a nude mouse. c-d, Representative intravital images of blood vessel ingrowth into initially avascular and pre-vascularized myobundles on day 4 and 14 of post-implantation (c) and corresponding quantification of blood vessel density (BVD, n=18-28 myobundles, significant increasing trend, p<0.0001, d). e-f, Representative intravital images of mCherry fluorescence at post-implantation D4, 7, 11, and 14 (e), and corresponding quantification of % mCherry⁺ area within myobundle (n=24-26 myobundles, ns: not significant, f). g, Representative ex vivo recorded snapshots of baseline and peak GCaMP6 fluorescence during twitch (1Hz stimulation) and tetanic (40 Hz stimulation) contraction in Mu and Mu+EC myobundle explants. h-j, Representative traces of twitch and tetanic Ca²⁺ transients (shown as relative GCaMP6 fluorescence, ΔF/F₀, h) and corresponding quantifications of twitch and tetanic Ca²⁺ transient amplitudes shown relative to Mu group (n=17-18 myobundles, **p<0.01, i) and time to peak twitch (n=17-18 myobundles, *p<0.05, j).

4. Discussion

In this study, we have developed an in vitro model of vascularized skeletal muscle with similar contractile function and increased SC density compared to avascular, muscle-only tissues. In vitro engineering of functional vascularized muscle tissues has been hindered by incompatibility between muscle and EC culture media, compromising both tissue myogenesis and vasculogenesis [31, 36, 65]. In agreement, use of EC growth media (EGM-2) in our study, even when diluted, inhibited muscle differentiation and fusion (Sup Fig 2). We thus optimized a custom-made serum-free HPLM differentiation media which when supplemented with VEGF and other pro-myoangiogenic factors (e.g. bFGF and EGF) supported both muscle (Sup Fig 2, Fig 2) and EPC viability and capillary network formation (Sup Fig 1, Fig 1). Previous studies with HPLM media reported more clinically predictive in vitro drug responses and altered cellular metabolism, including rewiring of urea metabolism in cancer cells [32-35]. HPLM has also been shown to drastically alter T lymphocyte transcriptome, cytokine secretion, and improve their activation response [34]. To our knowledge, this is the first report where HPLM was used for co-culture of cell types with distinct media requirements. Underlying mechanisms of how HPLM supports successful engineering of myovascular bundles remain to be explored. Importantly, the use of customized HPLM for the first time allowed generation of highly functional vascularized muscle tissues with similar levels of muscle differentiation and contractile strength compared to muscle-only controls (Fig 2).

In particular, the specific forces of myovascular bundles (~10mN/mm²) are orders of magnitude higher than previously reported (10-40 μN/mm²) for engineered vascularized skeletal muscle [30, 36, 67]. Previous compartmentalized muscle-endothelium models have demonstrated that human umbilical vein ECs (HUVECs) can increase contractile force of C2C12 [65] and primary human [30, 36] engineered muscle, presumably via secretion of trophic paracrine factors. In contrast, we found that media conditioned by myovascular bundles had no effect on force generation of avascular myobundles (Fig 4). The reason for this difference may be distinct secretomes of different EC sources used, different culture media, high myobundle functionality limiting potential for additional force increase, or the much lower endothelial:myogenic cell ratio (1:20) used in this compared to other conditioned media studies [65]. Nevertheless, the myovascular bundles developed here permit simple 3D co-culture of human myogenic and endothelial cells characterized by robust muscle function, native vascular density, and no need for use of complex compartmentalized fluid systems.

Previously, our group and others have shown that engineered 3D muscle tissues can form and maintain a niche-like environment in which Pax7⁺ SC-like cells attain a non-cycling state and approximate in vivo SC transcriptome and injury response [18, 25, 68-70]. In vivo, the function of SC niches is influenced by adjacent capillaries whereby SCs secrete VEGFA to attract ECs, while ECs maintain SC quiescence though expression of a Notch ligand DLL4 [10]. Here we show, for the first time in engineered skeletal muscles, that the presence of EC capillary structures increases SC numbers (Fig 3b). Furthermore, we find that SCs are preferentially located within 20 μm of formed vessels (Fig 3e), where SCs adopt a less proliferative state (Fig 3f), as observed in vivo [10]. These findings support reciprocal paracrine [7, 9, 10] and juxtacrine [10] interactions suggested to direct SC fate and EC patterning. Our conditioned media studies (Fig 4) and similar rates of SC proliferation at D0 and D7 (Fig 3) further suggest that increased SC numbers are likely not due to secreted ECs paracrine factors promoting SC proliferation [9, 17]. However, as SC numbers are increased within 20μm of capillaries, the potential of local secreted paracrine factors increasing SC numbers cannot be ruled out. Alternatively, the main factor leading to increased SC numbers in myovascular bundles may be EC juxtacrine signaling which may act in part through Notch signaling [10] to increase cell survival [71], inhibit fusion to myofibers [72], and support SC quiescence [73, 74] and self-renewal [75].

In addition to robust contractile function, lumen density in vascularized myobundles (~24 lumens/mm²) was higher than previously reported for in vitro engineered muscles (~7.5 lumens/mm²) [31], but still significantly lower than in native muscle [62, 63]. Interestingly, while the human vascular networks in myobundles were robustly maintained for at least two additional weeks after implantation in vivo (Fig 5f), this was not the case beyond two weeks of 3D co-culture in vitro. Timed addition of growth factors, such as angiopoietin [76] and platelet derived growth factor-BB [77], may further improve the density, lumenization, and longevity of myobundle vessels in vitro. Furthermore, incorporation of additional cell types, such as pericytes, to support vessel stability [78], macrophages, to complement myovascular interactions and improve muscle regenerative capacity [11, 19], fibro-adipogenic progenitors (FAPs), as a source of VEGFA [10, 79], or functional motoneurons [80, 81], to model the intricate neuromyovascular milieu of skeletal muscle in vivo, could all improve the physiological relevance of myovascular bundles which currently lack distinct fiber types [82] and only approximate the native extracellular matrix [83, 84] and SC niche [4, 85]. Beyond the studies of homeostatic roles of ECs in healthy skeletal muscle, myovascular bundles could also serve to investigate muscle-EC crosstalk in the context of muscle development [86, 87], exercise [15, 88], regeneration [19, 25], and various myovascular pathologies, including Duchenne’s muscular dystrophy [89, 90], myositis [91, 92], and peripheral vascular disease [93, 94]. Moreover, as incorporation of ECs in 3D engineered tissues [95, 96] and use of HPLM to culture cancer cells [32, 33, 35] alter in vitro drug responses, use of myovascular bundles may allow for more clinically predictive drug screening in the future.

Previous studies have shown that engineered C2C12 [28, 97] and human [26] muscle tissues pre-vascularized in vitro exhibit enhanced vascularization, perfusion, and survival in vivo. In our 2-week implantation study, in vitro pre-vascularization of human myobundles did not accelerate blood vessel ingrowth compared to that of avascular implants (Fig 5d), potentially due to use of a different scaffold, cell types, or an implantation model compared to previous studies. For example, we utilized blood-derived EPCs as opposed to HUVECs, commonly used as a model of mature ECs in vascularized engineered tissues [28, 98-100]. While EPCs show significant expansion potential [101], network forming capacity [102], and pro-angiogenic factor secretion [103] in vitro, there have been conflicting reports about their ability to maintain network stability post-implantation [104, 105]. In our studies, mCherry-labelled EPC networks within implanted myobundles remained stable for two weeks in vivo, but their ability to stimulate ingrowth of host capillaries and form anastomoses relative to other available EC sources remains to be studied within the same in vivo setting. Moreover, while EPCs are commonly utilized in various in vitro applications [106, 107], the EC function is known to be organ-specific [108, 109], thus warranting the future studies with muscle-resident ECs [110], ideally using methods that also enable perfusion of engineered vessels [111] in myobundles.

Notably, we for the first time demonstrate that in vitro pre-vascularization of engineered muscle tissues improves their Ca²⁺ transient amplitude and kinetics after implantation (Fig 5g-j). This improved Ca²⁺ handling without improved vascularization in vivo, may suggest an increase in regenerative capacity due to increased SC numbers or the presence of ECs. We have previously shown that upon implantation in window chambers, myobundles undergo a transient ischemic insult, likely due to no mass transfer through the window chamber glass coverslip and the switch from dynamic environment in vitro to static environment in vivo. In such conditions, increased SC density and myobundle regenerative capacity may eventually improve implanted muscle Ca²⁺ handling in vivo [18, 19]. Simultaneously, in transient hypoxic conditions post-implantation, EPCs could secrete factors that promote myofiber survival, SC proliferation, and/or myogenesis, similar to their responses to native muscle injury [9, 11, 17, 112]. Together, these findings warrant future mechanistic studies of how in vitro vascularization yields improved Ca²⁺ handling of implanted myobundles. Additionally, in vivo Ca²⁺ transient studies remain to be complemented with measurements of implant contractile force, resistance to fatigue, and neuromuscular integration to fully elucidate the effects of in vitro vascularization on muscle functionality post-implantation.

In summary, we have developed a biomimetic in vitro model of human vascularized skeletal muscle tissue that maintains robust functional metrics of avascular myobundles, exhibits increased SC density, mimics the native proximity between ECs and SCs, and results in improved Ca²⁺ handling compared to muscle-only tissues after implantation in vivo. In the future, this 3D cell culture platform could be used to model and study roles of myovascular crosstalk in muscle growth and disease/injury response, and, ultimately, enable the development of engineered skeletal muscle tissues with enhanced therapeutic potential.

Statement of Significance.

In native skeletal muscle, intricate relationships between vascular cells and muscle stem cells (“satellite cells”) play critical roles in muscle growth and regeneration. Current methods for in vitro engineering of contractile skeletal muscle do not recreate capillary networks present in vivo. Our study for the first time generates in vitro robustly vascularized, highly functional engineered human skeletal muscle tissues. Within these tissues, satellite cells are more abundant and, similar as in vivo, they are more dense and less proliferative proximal to endothelial cells. Upon implantation in mice, vascularized engineered muscles show improved calcium handling compared to muscle-only implants. We expect that this versatile in vitro system will enable studies of muscle-vasculature crosstalk in human development and disease.

Data Availability

The raw/processed data required to reproduce these findings are available upon request.