Human tendon-on-a-chip for modeling vascular inflammatory fibrosis

Understanding vascular inflammation and myofibroblast crosstalk is critical to developing therapies for fibrotic diseases. Here we report the development of a novel human Tendon-on-a-Chip (hToC) to model this crosstalk in peritendinous adhesions, a debilitating fibrotic condition affecting flexor tendon, which currently lacks biological therapies. The hToC enables cellular and paracrine interactions between a vascular compartment harboring endothelial cells and monocytes with a tissue hydrogel compartment containing tendon fibroblasts and macrophages. We find that the hToC replicates in vivo inflammatory and fibrotic phenotypes in preclinical and clinical samples, including myofibroblast differentiation and tissue contraction, excessive ECM deposition, and inflammatory cytokines secretion. We further show that the fibrotic phenotypes are driven by the transmigration of monocytes from the vascular to the tissue compartments of the chip. We demonstrate significant overlap in fibrotic transcriptional signatures in the hToC with human tenolysis samples, including mTOR signaling, a regulatory nexus of fibrosis across various organs. Treatment with rapamycin suppressed the fibrotic phenotype on the hToC, which validates the hToC as a preclinical alternative for investigating fibrosis and testing therapeutics.

brotic microenvironment 12,13 .Recent advances in liver, gut, and kidney microphysiological systems (MPS), to name a few examples, highlight the modernization of in vitro models for drug testing of complex diseases 14 .Of particular value are vitro models of blood-tissue interfaces that capture the dynamic vascular and immune responses driving in ammatory diseases such as brosis 15 .In this context, we seek a model of the injured human tendon, including its interaction with the vasculature and immune system, as a platform for discovering therapies to treat tendinopathies.
In this study, we develop and evaluate a human Tendon-on-a-Chip (hToC) model of in ammatory vascular brosis.The hToC platform is a modular device featuring a 'vascular compartment' with endothelial cells and monocytes and a 'tissue compartment' featuring a collagen-based ECM with tenocytes and resident macrophages.The hToC modularity enables the compartments to be prepared separately and then brought together to study the crosstalk between vascular endothelial cells, leukocytes, and tissue broblasts in response to pro-in ammatory and reparative factors such as TGF-β1.
Our results demonstrate that the brotic and in ammatory phenotypes observed in vivo also develop in the hToC, similar to tendon injury responses observed in mouse and human tissues.Further, we validate the transcriptional pro le in the hToC against data derived from human tenolysis tissues 16 .Notably, gene expression analysis reveals the activation of mTOR signaling as a consensus transcriptional signature of brosis in the hToC and the human tenolysis tissues.Given that mTOR signaling has recently emerged as a druggable target in the treatment of major organ brosis, which was also observed in the human tenolysis tissues and the hToC, we tested the therapeutic e cacy of mTOR inhibition in mitigating the brotic traits in the hToC.We found that a single dose of rapamycin suppressed myo broblast activation and attenuated the mTOR activation in the vascular compartment, although it did not fully resolve the in ammatory secretome.These ndings underscore the potential of the hToC as a model of the tendon injury microenvironment and, more broadly, as a platform for investigating the vascular-in ammatory crosstalk in brosis.

A Modular Tendon-Vascular Interface Model
While a healthy tendon is hypovascular, neovascularization is frequently observed in tendinopathy and scar tissue following acute injury and is considered a hallmark of chronic pathology.Traditionally, the role of vascularization in acute tendon injury and brotic peritendinous adhesions has been underemphasized, with the assumption that its primary purpose is to facilitate the arrival of immune cells and in ammatory factors to the injury site 17,18 .In human tenolysis samples, we have observed notable spatial colocalization of neovascularization with circulating and tissue-in ltrating leukocytes in proximity to myo broblast.Furthermore, this spatial localization was associated with increased activation of downstream mTOR effector proteins S6 and 4EBP1, which suggests a potentially aberrant cellular and molecular crosstalk involving mTOR signaling in the injury milieu (Fig. 1a).We postulated that these interactions at the tendon-vascular interface play a pivotal role in the pathogenesis of brosis and might offer novel therapeutic targets to attenuate brotic progression.
To model the tendon-vascular interface (Fig. 1b), we customized a two-compartment vascular barrier model 19 to be integrated with a tendon tissue hydrogel (Fig. 1c,d).
Brie y, the hToC is composed of two custom-fabricated components: Component 1 is an acrylic block component featuring a 100 µL well with two full-thickness uidic ports to access the bottom component and a bottom ledge lined with a pressure-sensitive adhesive (PSA) to seal to a membrane chip (Fig. 1e).
The membrane chip contains an ultrathin (~ 100 nm thick) optically clear, porous silicon nitride membrane patterned in a 700 µm x 2 mm window 20,21 .The membrane features dual-scale porosity with 5 µm pores superimposed on a nanoporous silicon nitride (NPN) background (Fig. 1e).The 5 µm pores provide portals for cell transmigration from the vascular to the tendon hydrogel compartment, while the nanopores (~ 60 nm diameter; ~ 15% porosity) allow unhindered paracrine signaling by small molecule exchange throughout the membrane interface between compartments.Component 2 features a uidic channel into which a broblast-laden collagen hydrogel (~ 65 µL) can be cast, with horizontal anchors suspended on either end of the channel (Fig. 1c,f) to constrain the broblast-mediated hydrogel axial contraction while permitting lateral contraction.A double-sided PSA lining on the top surface of Component 2 facilitates bonding to Component 1 (Fig. 1g) during the assembly of the full hToC (Fig. 1d).
Human umbilical vein endothelial cells (EC) were seeded onto the porous membrane to form the vascular interface monolayer (Fig. 1e).Human primary tendon broblasts (TC), isolated from tissues obtained from hand surgery were embedded in a TeleCol ® -3 collagen hydrogel (500,000 cells/ml) and pipetted into Component 2 channel (Fig. 1f).The addition of tissue-resident macrophages (tMϕ) into the tissue hydrogel and circulating monocytes (cMϕ) into the vascular well simulates the leukocyte interactions in the vascularized injury milieu as observed in Fig. 1b.
The modular design of the hToC enables the independent culture of the vascular and tissue compartments before their integration as a composite tissue (Fig. 2a).This feature is critical given the different culture conditions and times required for the EC monolayers to achieve con uent barrier function (24 hours) and for the differentiation and maturation of monocytes to tMϕ in the hydrogel (6 days).By accommodating cell-speci c timelines and requirements, the hToC can be used to study different cell combinations, including monocultures (TC only), cocultures (TC/EC or TC/tMϕ), tricultures (TC/EC/CMϕ or TC/tMϕ/EC), or quad cultures (TC/tMϕ/EC/cMϕ) (Extended Fig. 1).Careful selection of device materials and layer thicknesses enables live microscopy and multiplex uorescent imaging of cellular and molecular interactions in situ.The uidic ports enable media sampling for cytokine analysis.

Modeling In ammatory-Fibrovascular Scar Tissue
Using a quad-culture protocol in the hToC, we modeled the injured tendon microenvironment involving endothelial cells, leukocytes, and tendon broblasts.Monocytes isolated from human peripheral blood donors were embedded and co-cultured with tenocytes (TC) in TeleCol®-3 hydrogel for a total cell concentration of 500 cells/µl at a broblast-to-monocyte ratio of 7:1 22 .We then differentiated the hydrogel-embedded monocytes into macrophages (tMϕ) in Lonza's XVIVO-10™ serum-free media supplemented with monocyte-colony stimulating factor (M-CSF) over six days.In the TC-laden hydrogel monoculture, treatment with TGF-β1 over seven days in culture increased hydrogel contraction to ~ 50% of its original area (Fig. 2a-c).Interestingly, hydrogel containing TC and tMϕ also contracted to ~ 50% in the absence of TGF-β1, which was accompanied by the activation -SMA + myo broblasts (Fig. 2d).The role of the tenocytes/myo broblasts and leukocytes crosstalk in effecting the TC/ tMϕ hydrogel contraction was veri ed by the lack of contraction in the tMϕ hydrogel monocultures.Together, this data suggests that even in the absence of TGF-β, tMϕ secrete factors that lead to the activation of tenocytes into myo broblasts, which subsequently contract the collagen hydrogel.
To create a vascular barrier, endothelial cells (EC) were plated on the membrane chip in the vascular compartment at 40,000 cells/mm 2 and allowed to adhere to the membrane and reach con uence over 24 hours.To mimic circulating monocytes (cMϕ) in ltration to the injury site through the vasculature, freshly isolated monocytes were introduced into the vascular compartment at 10,000 cells per 100 µL of XVIVO-10™ serum-free media (day 0).The two compartments were then assembled to create a quad culture system (TC/tMϕ/EC/cMϕ), which was treated with TGF-β1 (10 ng/ml) or vehicle, directly administered to the tendon hydrogel compartment for 24-72 hours (Fig. 3a).
When cultured separately, the vascular compartment exhibited robust activation of in amed endothelium at 72 hours independent of the luminal exposure to TGF-β1 (Fig. 3b).This activation was manifested by the positive staining of intercellular adhesion molecule-1 (ICAM-1) and vascular cell adhesion molecule-1 (VCAM-1) and noticeable gaps resulting in a leaky vascular monolayer.Introducing TGF-β1 from the EC abluminal side in quad culture conditions resulted in comparable, robust vascular activation.Furthermore, quanti cation of the transmigration of uorescently labeled cMϕ through the vascular endothelium into the tissue hydrogel demonstrated the critical role of tMϕ in stimulating the transmigration behavior, even in the absence of TGF-β1 treatment (Fig. 3c).Interestingly, there were no signi cant increases in transmigrated cMϕ after 24 hours in quad cultures without TGF-β1 treatment (Fig. 3d).Finally, treatment with TGF-β1 in the quad cultures did not signi cantly increase the number of transmigrated cMϕ after 24 hours (Fig. 3e).
Under the quad culture conditions, treatment with TGF-β1 led to an increased abundance of markers of brosis, including α-smooth muscle actin (α-SMA + ) myo broblasts, with characteristic contractile morphology, HSP47, a collagen-binding glycoprotein chaperone in the endoplasmic reticulum (ER), and Ki67, a marker of proliferation (Fig. 3f).Furthermore, treatment with TGF-β1 increased activation of the mammalian target of rapamycin (mTOR) signaling, evidenced by increased abundance of phosphorylated 4EBP1 and P70S6 (Fig. 3f), consistent with the observed upregulation of mTOR effector proteins p4EBP1 and pS6 in the vicinity of myo broblasts in tenolysis tissue (Fig. 1a).
Previous experiments in mice have shown that several cytokines are systemically upregulated early and sustained over time following acute tendon injury 7 .We probed for eight of these secreted cytokines in the hToC culture supernatant using a Luminex detection assay at 24 and 72 hrs (Extended Fig. 2).MCP-1, IL-1β, IL-6, IL-10, and TNF-were signi cantly increased in quad-cultures at 24 and 72 hours compared to the monoculture of tenocyte-laden hydrogel (Fig. 3g).However, TGF-β1 treatment did not signi cantly affect their levels in the quad culture hToC.These results underscore the signi cance of cellular crosstalk of TC, tMϕ, cMϕ, and EC in driving the in ammation and TGF-β1 induced myo broblast activation and mTOR signaling within the quad culture model, which could serve as markers of tendon brosis in the hToC.

Validating the hToC as a Peritendinous Fibrosis Model
To validate the hToC quad culture as an in vitro model of vascular-in ammatory brosis, we performed bulk RNA-sequencing and benchmarked the results against published gene expression data from human tenolysis tissue obtained 2-3 weeks after tendon injury and healthy exor tendon tissue (GSE108933) 16 .We rst assessed the effect of TGF-β1 by analyzing the differential gene expression in TGF-β1 treated (+ TGF-β1) hToC compared to control, untreated hToC (-TGF-β1).Based on a twofold change (± 2FC) and adjusted p-value < 0.05 differential expression criteria, TGF-β1 treatment resulted in 462 downregulated genes and 304 upregulated genes (Fig. 4a).Notably, the vascular component exhibited fewer than 70 differentially expressed genes (DEGs), indicating a more signi cant effect of TGF-β1 on the tendon hydrogel.Subsequently, we compared the gene expression pro le of the in vitro disease model (TGF-β1 + hToC) to healthy human tendon, which identi ed 1,011 downregulated genes and 597 upregulated genes (Fig. 4b).In contrast, a total of 1,668 genes were downregulated and 1,101 were upregulated in human tenolysis tissue relative to healthy tendon tissue (Fig. 4c).
We performed pathway enrichment analyses based on the DEGs in these comparisons.To evaluate the delity of the hToC disease model in simulating vascular-in ammatory brosis, we identi ed the top signi cantly enriched Reactome pathways in the hToC and demonstrated signi cant overlap with the pathways enriched in the human tenolysis tissue, which encompassed biological processes related to ECM deposition and remodeling and in ammatory signaling (Fig. 4d).To further assess the congruence of gene expression in the overlapping Reactome pathways, a Jaccard dissimilarity heatmap was plotted based on Euclidian distance of the gene sets between the hToC and tenolysis tissue (Fig. 4e).Despite observing high dissimilarity indices over many of the shared pathways, we observed signi cant congruence in key ECM, in ammation, interleukin signaling, including the PI3K/AKT (mTOR) signaling pathways.Additionally, gene set enrichment analysis (GSEA) revealed that the mTOR signaling pathway was signi cantly enriched in the hToC upon treatment with TGF-β1 (Fig. 4f).

Rapamycin Effects on mTOR Signaling and Fibrotic Responses in hToC
Given the observed involvement of mTOR in tendon adhesions in vivo, which is also recapitulated in the hToC, we then investigated the effects of rapamycin, an mTOR kinase inhibitor and FDA-approved drug 23 , as a treatment to mitigate the brosis phenotype in the hToC.In these experiments, we added 10 ng/ml of rapamycin to the culture media at the time of TGF-β1 and cMϕ introduction (D 0 ) to the quad culture hToC (Fig. 5a).The dose was selected based on measured plasma drug concentration in clinical studies 23 .We evaluated rapamycin's effects on mTOR signaling, vascular in ammation, and the activation of myo broblasts in the hToC.While treatment with rapamycin exerted negligible effects on pAKT and p4EBP1 activation in the endothelial cells, it signi cantly inhibited pS6 activation in the vascular endothelium (Fig. 5b).Rapamycin induced a uniform and more evenly distributed expression of junctional VE-cadherin and mitigated the noticeable gaps in the leaky vascular monolayer that are prominent in the TGF-β1 control (Fig. 5c).
In the tissue compartment of the TGF-β1 control, p4EBP1 expression was not limited to broblastic cells but was also observed in leukocytes (tMφ and in ltrating cMφ) (Fig. 5d).Upon treatment with rapamycin, the expression of p4EBP1 decreased in tendon broblasts but not in leukocytes.Treatment with rapamycin showed no discernable effects on pS6 in the tendon broblasts (Fig. 4d).However, the expression of α-SMA in the tendon broblasts signi cantly decreased and was accompanied by remarkable morphological changes such that the number of enlarged and irregularly shaped myo broblasts with prominent stress bers was reduced (Fig. 5e).These results collectively demonstrate that rapamycin treatment in the hToC model had signi cant effects in both the vascular and tissue compartments, attenuating the effects of TGF-β1 on the vascular endothelium and the myo broblast phenotype.
Three monocyte donors were used to provide tMφ and cMφ to account for patient variability in the experimental design.To monitor changes in the in ammatory and brotic response to rapamycin, secreted in ammatory cytokines were evaluated after 24 hours of treatment with a Luminex ® assay for each distinct donor.While no signi cant rapamycin-induced changes were observed from controls on average, the secreted in ammatory cytokine levels varied signi cantly between donors (Fig. 5f).This nding underscores the importance of using multiple donors in the assessment of drug safety and e cacy in pre-clinical studies with MPS platforms and demonstrates the hToC's sensitivity to capture the expected patient variability.
To gain further insights into the effects of rapamycin on the regulation of the in ammatory and brotic response, bulk RNA-seq analysis of the hToC was carried out for each of the three different M donors (rapamycin compared to control) using the integrated Differential Expression & Pathway (iDEP1.12)web application 24 .A representative heatmap of the top 2000 differentially expressed genes in the hToC (Fig. 6a) shows distinct transcriptional clustering and the associated enrichment of Gene Ontology (GO) Biological Processes (BP) (Fig. 6b) terms.In general, GO terms related to in ammation, ECM remodeling, cell cycle, cell adhesion and transmigration, and various signaling pathways, including the PI3K-Akt pathway, were enriched for all donors.Differential gene expression analysis revealed that depending on the donor, rapamycin treatment resulted in 238-452 downregulated genes and 86-248 upregulated genes in the vascular (EC/cMϕ) compartment (Fig. 6c).In contrast, 15-35 genes were downregulated and 11-35 genes were upregulated in the tissue hydrogel (TC/tMϕ) compartment (Fig. 6f).This underscores a predominant in uence of rapamycin on the in amed vascular endothelium, likely associated with the enhanced vascular barrier integrity observed post-rapamycin treatment (Fig. 5c), which is consistent with published reports of rapamycin's cytoprotective effects against endothelial cell damage 25 .The GO BP terms enriched by DEGs in the vascular and tissue hydrogel compartments were then examined.In the vascular compartment, downregulated genes in rapamycin-treated devices enriched immune and defense responses pathways, including the regulation of cytokines and responses to external stimuli (Fig. 6d), and the upregulated genes enriched pathways regulating metal ion levels, detoxi cation, and stress responses (Fig. 6e).In the tissue hydrogel compartment, the downregulated genes in rapamycin-treated devices enriched recruitment of immune cells, chemokine signaling, and various immune responses (Fig. 6g), and upregulated genes enriched tissue damage and repair, including cell death and survival processes and responses to growth factors (Fig. 6h).

Discussion
In this study, we engineered and characterized a novel human Tendon-on-a-Chip (hToC) platform to study in ammatory brosis.The hToC model recapitulates in ammatory and brotic characteristics observed in preclinical and clinical specimens.These characteristics include key markers of in ammatory brosis, such as myo broblast differentiation, tissue contraction, excessive extracellular matrix (ECM) deposition, and the secretion of in ammatory cytokines.
Our data suggest that interactions between leukocytes and endothelial cells in the quad cultures are critical for promoting elevated in ammatory secretome and activating adhesion molecules that mediate leukocyte transendothelial migration, regardless of exogenous TGF-β1 treatment, which is consistent with established paradigms 18,26 .Tendon broblasts, in turn, are affected by the secreted cytokines, which activate their differentiation into myo broblasts, characterized by elevated collagen production and a proliferative and contractile state.The activation of myo broblasts is induced by TGF-β1, a major in ammatory growth factor released during healing by neutrophils, macrophages, and T cells 27 .Macrophages are innate immune cells that accumulate at in ammatory brosis sites.The in ammatory signals from macrophages amplify brosis through various mechanisms, such as attracting more immune cells, inducing myo broblast differentiation and cellular senescence, and impairing tissue remodeling 28 through secreting a multitude of growth factors and cytokines, including TGF-β1 29 , leading to the persistence of an in ammatory response, ultimately culminating in chronic scar tissue.However, the role of angiogenesis in the development of tendon adhesions has been less appreciated until recently.Our research and others have shown increased angiogenesis in the formation of brotic adhesions in tendons 7,9,30 .Yet, the endothelial vasculature's roles in brosis remain to be fully elucidated.
The hToC provides the rst in vitro model of the myo broblast microenvironment (MME), which we de ne to include the myo broblast and the surrounding blood vessels, immune cells, tenocytes, signaling molecules, and the extracellular matrix (ECM).This is analogous to the tumor microenvironment (TME), de ned as the cellular and molecular environment in which cancer cells reside and interact with surrounding non-cancerous cells, extracellular matrix, and signaling molecules 31 .The invasiveness of cancer cells, including their propensity to metastasize and their resistance to cytotoxic drugs, is regulated by the TME.By analogy, we postulate that disrupting the MME therapeutically could resolve brotic pathologies.Therefore, the hToC will be a valuable new tool for delineating novel cellular interactions and signaling in the MME, which could lead to therapies to resolve peritendinous adhesions and, more broadly, chronic brosis.
We identi ed a substantial convergence in the transcriptional signatures associated with brosis between the hToC model and human tenolysis samples 16 .Speci cally, we observed commonalities in the activation of the mTOR pathway.While a speci c role for mTOR signaling in tendon injury and peritendinous adhesions has not been previously reported, it serves as a regulatory hub for brosis 32 .
Tendon-speci c Raptor/mTOR knockout mouse models have decreased tendon strength and elasticity, collagen ber diameter, and paw grip strength, suggesting that mTORC1 signaling is a critical regulator of postnatal tendon development 33,34 .A recent genome-wide association study (GWAS) of susceptibility to idiopathic pulmonary brosis strongly implicated mTORC1 signaling as a causal driver of brosis pathobiology in humans 35 .Recent clinical trials demonstrated that short-term treatment with sirolimus, an mTOR inhibitor, led to a decrease in the levels of circulating brocytes in individuals with IPF, accompanied by an acceptable safety pro le 36 .
Having identi ed mTOR activity as a driver of the brotic phenotype in the hToC, we then used the platform to explore the therapeutic potential of rapamycin.Rapamycin substantially impacted the vascular endothelium, as evidenced by the signi cant inhibition of pS6 activation and the promotion of a more uniform expression of junctional VE-cadherin.Moreover, it mitigated the distinctive gaps between endothelial cells in the vascular compartment following TGF-β1 treatment.The ability of rapamycin to improve the barrier function of the vasculature has been previously reported in neuroin ammatory diseases, in part by attenuating the in ammatory response 37,38 .In the tissue hydrogel, rapamycin reduced the expression of p4EBP1 in tendon broblasts and a notable decrease in α-SMA expression, indicating a modulation of the myo broblast phenotype.
Additionally, bulk RNA-seq analysis provided insights into the transcriptional changes induced by rapamycin.Notably, rapamycin predominantly in uenced the vascular endothelium, with many downregulated genes related to immune and defense responses and leukocyte transendothelial migration, corroborating in vivo observations 39 .Furthermore, rapamycin downregulated genes in the tissue hydrogel enriched immune cell recruitment and chemokine signaling while upregulated genes enriched tissue repair and cell survival mechanisms, consistent with previous ndings 40 .In contrast, the single dose of rapamycin had a minimal impact on the in ammatory cytokine pro le in the hToC, although the results depended on the donor used to populate resident and circulating monocytes in the model.Thus, while circulating and resident immune cells still produced an in ammatory secretome despite treatment with rapamycin, the endothelial cells, tenocytes, and myo broblasts became more resistant to the in ammatory microenvironment.This work is not without limitations.The interactions between monocytes and the vascular endothelium were evaluated under static (no ow) conditions.Introducing physiological uid ow would allow for clearance of secreted cytokines and enable ECs to undergo shear priming and physiological conditioning 41 .Integrating uid ow would also facilitate interactions between ECs and circulating leukocytes, which will be evaluated in future studies.Furthermore, while using primary cells is practical, it introduces donor variability and heterogeneity into the model, highlighted here by the variable cytokine secretion response with different monocyte donors.Such variable, di cult-to-control effects make primary cells undesirable for screening.To overcome this challenge, improved models of peritendinous injury could leverage induced pluripotent stem cells (iPSCs), which offer the advantage of enabling isogenic models and well-de ned patient cohorts for simulated clinical trials.
In conclusion, this study introduces and characterizes a novel human Tendon-on-a-Chip (hToC), revealing its ability to recapitulate key features of vascular in ammatory brosis in injured tendons.Our results highlight previously underestimated contributions of vascular endothelial cells in the myo broblast microenvironment and identify the mTOR pathway as a critical regulator of brosis in this context.Rapamycin shows promise in mitigating vascular endothelial dysfunction and myo broblast activation in the hToC model, though its impact on in ammatory cytokines is modest and varies between donors.Thus, while our ndings shed light on novel aspects of brovascular in ammatory myo broblast microenvironment and demonstrate the potential of the hToC to evaluate potential therapeutic interventions, we acknowledge the need for future studies incorporating physiological uid ow and the use of induced pluripotent stem cells to re ne our understanding and personalize therapeutic screening.

Methods
Silicon Dual-Scale Porous Membranes 'Dual-scale' silicon nitride membranes were purchased from SiMPore Inc. (Rochester, NY).These are commercial versions of membranes rst described by Salminen et al. (2019) 21 .The membrane is created by patterning a microporous array atop a nanoporous silicon nitride (NPN) background.The nanoporous background has an average nanopore size of 44.6 nm, a porosity of ~ 15%, and a membrane thickness of ~ 100 nm.We patterned micropores at 5 µm in diameter to enable monocyte transmigration from the vascular to tissue sides of the hToC.We limited the microporous density to produce a ~ 1% porosity (superimposed upon the nanoporous background to give a total of ~ 16% porosity).This density of micropores is well below the 10% value shown to give monolayer stability 30 and does not hinder imaging due to light scattering by micropores.

hToC Components
The top well component (top component) and bottom channel component (bottom component) of the hToC were manufactured at ALine Inc. (Signal Hill, CA) using laser cutting and lamination processes that are compatible with mid-volume production (hundreds to tens of thousands) of micro uidic components in a single production run.Parts were produced using a batch process and diced after nal lamination for more reliable handling in the laboratory.The top well component contains uidic access ports to the bottom channel that create sealed ts against P20/P200 pipette tips (VWR, 76323-390).Using pressuresensitive adhesive (PSA), the top and bottom components were reversibly adhered to form the hToC device.Devices are free of any molded PDMS, and although they contain silicone as part of the adhesive layers, the material accounts for < 5% of the uid-exposed surface.Thus, concerns about the loss of small molecules via absorptive loss 42 are minimized in the hToC.The external surfaces of the shipped components include an additional protective layer (masking material) to maintain cleanliness and sterility during shipment and storage.The masking material is removed by the user before assembly of the components in a laminar oor hood as described in McCloskey et al. ( 2022) 19 .

Tissue Formation and Maturation
Human Tenocyte Isolation & Culture Tenocytes were isolated from tendon tissue fragments retrieved from hand surgery procedures at the University of Rochester Medical Center under an approved Institutional Review Board (IRB) protocol (STUDY00004840).The isolated tendon tissues were immersed in alpha minimum essential medium (αMEM) supplemented with 10% Pen-Strep (10,000 U/mL penicillin/10 mg/mL streptomycin) and cut into 1-mm 3 pieces using ophthalmic scissors and a gentleMACS™ Dissociator (Miltenyl Biotec, 130-093-235).
After mechanical digestion, the tissue was transferred to an enzyme solution consisting of 2.5mg/ml of Collagenase D (Millipore Sigma, 11088858001), 3mg/ml of Dispase II (Millipore Sigma, D4693), and 1mg/ml of DNase (New England Biolabs, M0303) dissolved in αMEM.Tissues in enzyme solution were placed in a Roto-Therm™ Plus Incubated Rotator (Benchmark Scienti c, H2024) at 37°C while rotating in combination with oscillations for an hour.αMEM supplemented with 10% Fetal Bovine Serum (FBS) and 1% Pen-Strep was added to the enzyme solution at a 2:1 ratio to inactivate the proteases in the digestive solution.The complete tissue solution was strained through a 70-micron lter, centrifuged, and resuspended in αMEM supplemented with 10% FBS, 1% Pen-Strep, and 55 µM 2-Mercaptoethanol (Thermo Fisher Scienti c, 21985023).The isolated tenocytes were seeded in a T-75 ask (Corning) coated with 1 µg/cm 2 Fibronectin (Sigma-Aldrich, F1056).
Cell culture was performed under standard conditions (37°C, 5% CO 2 , 95% humidity) with a media change after 4 days post-plating and then every other day until sub-con uence was achieved.Cells were passaged at a 1:4 split ratio using 0.25% trypsin/0.02%EDTA solution (Sigma-Aldrich, 25200056).After passage 3, the cells were cryopreserved in Recovery™ Cell Culture Freezing Medium (Thermo Fisher Scienti c, 12648010) for upcoming experiments.At that point, cells were thawed and cultured in a T-175 ask under standard conditions in αMEM supplemented with 20% FBS for 24 hrs.Media was changed to αMEM supplemented with 10% FBS for 3 additional days, after which the media was changed to αMEM supplemented with 5% FBS.Tenocytes were trypsinized at day 5 after reaching ~ 80% con uency for suspension in the collagen hydrogels.

Monocytes/Macrophages
All blood samples were collected in EDTA pre-coated tubes from healthy volunteers under an IRBapproved protocol (STUDY0004777) and processed immediately after collection.PBMCs were isolated from whole blood by density separation over a solution of 1-Step™ Polymorph (Accurate Chemical & Scienti c Co., AN221725) at 500×g for 30 min without brakes.The white buffy PBMC layer was then washed twice in Wash buffer (HBSS without Ca 2+ and Mg 2+ with 10mM HEPES and 5mg/ml BSA) and centrifuged at 350×g for 7 min without brakes to remove platelets.Red blood cells were lysed by exposing the washed PBMC layer to 1/6x PBS for 1 min, followed by 4x PBS for 1 min, at a 3:1 ratio, and spinning at 350×g for 7 min.PBMCs were resuspended in Wash buffer and spun down at 350×g for 7 min to remove any remaining PBS.PBMCs were resuspended in an Isolation buffer (1XDPBS without Ca 2+ and Mg 2+ with 2mM EDTA and 1 mg/ml BSA).CD14 + Monocytes were isolated from PBMCs by positive magnetic enrichment using the QuadroMACS Starting Kit (LS) (Miltenyi Cat# 130-091-051) using the manufacturer's protocol.
For TC/tMϕ co-cultures, freshly isolated monocytes were suspended in TeleCol®-3 with tenocytes at a ratio of 1:7.The hydrogel was pipetted into the bottom channel and 200 µl of X-VIVO™ 10 media supplemented with 20 ng/ml M-CSF (PeproTech, 300 − 25) was added to induce monocyte polarization into naïve macrophages.M-CSF supplemented media was entirely replaced on D − 4 and D − 2 and ½ replaced on D − 5 and D − 3 .

Endothelial Cells
Pooled Human Umbilical Vein Endothelial Cells (HUVECs) were purchased from LONZA (C2519A) and maintained in a T-75 ask at 37°C, 5% CO 2 , 95% humidity in EGM-2™ media (LONZA, CC-3162).HUVECs were used between passage number 3-7 as recommended by the supplier and cultured according to the supplier's protocol.

Quad-Culture hToC Assembly
The hToC top component was assembled as described above.To prevent bubble formation at the interface of the trench of the porous membrane and the tissue hydrogel in the bottom channel, the membrane trench was back lled with collagen solution and was then placed in the incubator at 37°C, 5% CO 2 , 95% humidity for 20 min to allow the collagen hydrogel to crosslink.After 20 min, the top components were removed from the incubator, and the reservoirs were attached and lled with 225 µl of EGM TM -2 media.The back-lled top components were placed onto the media-lled reservoirs to prevent the collagen hydrogel in the trench from drying out.The well-side of the porous membrane was then coated with 0.17 mg/ml bronectin (Sigma, F1141) for 1 hr at room temperature to facilitate cell adhesion.Expanded HUVECs were plated into the 100 µl top component well at a 40,000 cell/cm2 density.Cells were allowed to settle for 3 hours before rinsing with EGM TM -2 media to remove nonadherent cells.HUVECs were incubated for 24 hours in the top component alone before combining with the bottom component containing the TC/ tMϕ hydrogel.EGM TM -2 media was replaced with X-VIVO™ 10 media in the top component well after assembly to establish serum-free conditions during the experimental timeline.
For tri-and quad-culture assembly, the top component is adhered to the bottom channel to form a tendonendothelial barrier.Prior to assembly, the bottom channel cultures must be ready for assembly (24-hr culture for TC only and 6-day culture for TC/tMϕ co-cultures), the ECs in the top component must have formed a con uent monolayer (expected after 24-hrs), and CD14 + monocytes freshly isolated.After full assembly of the hToC, the bottom channel was lled with 100 µl of X-VIVO™ 10 media ± TGF-β1.Freshly isolated CD14 + monocytes suspended at 100,000 cells/ml in 100 µl of X-VIVO™ 10 media were added to the top component well containing the EC monolayer.For all rapamycin experiments in this study, a 10 ng/ml rapamycin solution was added to the EC well at D0.The fully assembled hToC was then placed in an incubator and cultured at standard conditions.Culture supernatants from the top well and bottom channel were sampled from 24-and 72-hour devices for cytokine analysis.At the end of the experiments, EC monolayers and hydrogels were xed or processed for immuno uorescence or alternative downstream assays at 24 and 72 hours, respectively.
Live-Stain Imaging and Quanti cation of Macrophage Transmigration in hToC TC, tMϕ, and cMϕ were live-stained using the ViaFluor SE Cell Proliferation Kit (Biotium, 30068-T) before seeding into the hToC and imaged at 3-, 24-or 72-hrs, respectively.For each image, a z-stack was taken every 10 minutes over 45 minutes, starting ~ 50 µm on top of the porous membrane and 600 µm under the porous membrane to capture cMϕ migration through the EC monolayer and into the bottom channel compartment.The hToC was placed on an incubation stage at standard conditions for the entire imaging session to prevent cell death during imaging.

Luminex Multiplex Immunoassay
Human Luminex Discovery Assay (R&D Systems) was used for the quanti cation of selected human cytokines and chemokines: MCP-1, CCL3, CXCL10, IL-1β, IL-6, IL-10, IL-17, and TNF-α in cell supernatants collected from the top well and bottom channel (~ 110µl) and frozen immediately at -80°C.The measurements were performed according to the manufacturer's speci cations using a Luminex 200 Instrument.The experiments were performed with at least three biological replicates per condition, each performed in duplicate.

Immunostaining and Fluorescence Microscopy
Tendon constructs were rinsed with 1× PBS, xed in 4% Paraformaldehyde (PFA) for 10 min, and then washed 2 × 5 min in PBS.Cells were then permeabilized with 0.5% Triton X-100 (Amresco, 0694) for 10 min and rinsed.The constructs were blocked in 1% Bovine Serum Albumin (BSA) (Cell Signaling Technology, 9998) solution for 30 min at Room Temperature (RT) and incubated with the primary antibody solutions (Supplemental Table) diluted in 0.1% BSA or PBS overnight at 4°C.The respective secondary antibody (Supplemental Table 2) used at the manufacturer's recommended dilution was diluted in 0.1% BSA or PBS and incubated for 1 hr at RT. Hoechst stain was then diluted in PBS and incubated for 10 min at RT.Alternatively, hydrogels were xed in 4% PFA, dehydrated, removed from the channel and embedded in para n.The blocks were sectioned at 10 µm thickness.For IHC, depara nized and rehydrated sections were incubated in 10 mM sodium Citrate (pH 6.0) for 1 hr at 65°C for antigen retrieval.Slides were rinsed with ddH 2 O for 5 min followed by 3 × 4 min washes in PBS and 2 × 4 min washes in PBST (PBS + 0.1% Tween-20).Slides were blocked for 1 hr in 5% BSA diluted in PBST and incubated with primary antibodies (Supplemental Table) at 4°C overnight.After primary antibody incubation, the slides were washed 3 × 10 min in PBS and then incubated with respective secondary antibodies diluted in 5% BSA diluted in PBST.Slides were then rinsed 3 × 5 min in PBST, followed by a 2 × 5 min wash in PBS.Sections were cured with ProLong™ Diamond Antifade Mountant with DAPI (Thermo Fisher, P36962) overnight at RT in the dark.Clear nail polish was used as a sealant the following day, and slides were stored at 4°C until imaged.
For ICAM-1 and VCAM-1 expression in EC monolayers, live stains were used (Supplemental Table 1).This required adding the primary antibodies diluted in X-VIVO™ 10 media and incubating at standard conditions for 15 min before xing.Cells were then blocked with blocking buffer (5% BSA + 0.1% Triton X-100) and incubated at RT for 10 min.For ICAM-1 and VCAM-1, secondary antibodies and Hoechst stain were added in blocking buffer and incubated for 1 hr at RT.In the case of VE-Cadherin and PECAM-1 staining, primary antibodies were diluted in blocking buffer and incubated for 1 hr at RT after the blocking step.After rinsing the well, the secondary antibody + Hoechst stain diluted in blocking buffer were added and incubated for 1 hr at RT.All samples were stored in a humidi ed chamber at 4°C and protected from light until imaging.All images were acquired using a Dragon y Spinning Disk Confocal System (Andor, Belfast, UK) at the University of Rochester High Content Imaging Core.

RNA Extraction from Collagen Construct and Bulk RNA-Seq
The cell-laden hydrogels (n = 6 samples per group, with each sample consisting of 2 pooled hydrogels) were harvested from the hToC bottom compartment immediately following the experiment.The hydrogel was digested to release the cells by adding 500 µl of 1 mg/ml (215 units/mg) of collagenase I (Thermo Fisher, 17018029) diluted in X-VIVO™ 10 medium and stored at standard conditions for 45 min.Total RNA was isolated from released cells using the RNeasy Plus Micro Kit (Qiagen, 74134) per manufacturer recommendations.To isolate RNA from the EC monolayer, 100 µl of Lysis Buffer was added to the top component well for 10 min.The EC lysate was immediately processed following the manufacturer's protocol.The RNA concentration was determined via NanoDrop 1000 spectrophotometer (NanoDrop, Wilmington, DE), and RNA quality was assessed with the Agilent Bioanalyzer (Agilent, Santa Clara, CA).
Messenger RNA isolation and next-generation RNA sequencing analysis were performed by the University of Rochester Genomics Core.

Bioinformatics and statistical analysis of DEGs
All bioinformatic analysis was performed with original RStudio scripts (R Core Team (2022).R: A language and environment for statistical computing.R Foundation for Statistical Computing, Vienna, Austria.URL https://www.R-project.org/).Using DESeq2, gene expressions were compared between the human (uninjured and tenolysis from Zheng et al 16 ) and hToC samples (vascular and tissue separately).To correct for differences attributed to batch effects, all count data was batch adjusted using ComBatseq.The Wald test was used to generate P values and log 2 (fold changes).Genes with P < 0.05 and absolute log 2 (fold changes) > 1 were selected as differentially expressed genes for each comparison.
Heatmaps, dot plots, and enrichment maps were performed using the Complex Heatmap Visualization and Enrichplot packages.Pathway analysis for Gene Ontology (GO), Kyoto encyclopedia of genes and genomes (KEGG) and Reactome were performed using their respective homo sapien database in RStudio.
The differentially expressed genes obtained from RNA sequencing were ranked from highest to lowest expression and uploaded to GSEA software.

Declarations
Research Fellowship Award from the National Science Foundation.RGA is funded by NIAMS training award T32AR076950.The content is the sole responsibility of the authors and does not represent the o cial views of the National Institutes of Health or the National Science Foundation.for the hToC quad-culture assembly.b, The effect of the exogenous TGF-β1 treatment was examined on the EC mono culture stained for leukocyte adhesion and activation markers VE-cadherin (green, row 1), ICAM-1(purple, row and VCAM-1(yellow, row 3) at 24-and 72-hours.Exogenous TGF-β1 was added directly to the luminal EC surface, incubated for 72-hours, and assessed with the same markers (b, column 3).These results were compared to the EC monolayer removed from a 72-hour quad culture (b, column 4).Scale bar = 50 mm.The transmigration of uorescently labeled circulating monocytes (cM from the vascular compartment to the tissue hydrogel was quanti ed under various conditions, including c, tri cultures (without tM ) or quad cultures (with tM ), d, quad cultures without TGF-β1 over time, and e, quad cultures without or with exogenous TGF-β1 at 72 hours.f, Representative immuno uorescence sections of the tissue hydrogel in quad-cultures show broblastic morphology in tendon construct in the absence or presence of exogenous TGF-β1 treatment.Sections were stained for markers for collagenspeci c molecular chaperone (HSP47), cellular proliferation (Ki67), myo broblast activation (α-SMA), and mTOR pathway expression (pAKT, p4EBP1, pS6).g, The quad culture supernatants were tested from the combined vascular and tissue sides and sampled using a multiplex Luminex® assay.The in ammatory cytokine secretory protein pro le at 24 hours in the presence and absence of exogenous TGF-β1 (N=3-4) is displayed, where the dashed blue line represents secretion levels in monocultures (tenocytes-laden hydrogels) treated with TGF-β1.Scattered data presented with mean and standard deviation marks (n=3 devices).One-way ANOVA was used for comparison (*p < 0.05).

Figure 1 Design
Figure 1