Decoding Viral Transmission Dynamics in Structured Populations Using a Microfluidic Herd‐Immunity‐on‐a‐Chip Platform Authors: Jiande Zhang, Narina Jung, Min‐Hyeok Kim, Wanyoung Lim, Zhenzhong Chen, Byung Mook Weon, Sungsu Park Published: 2026 ABSTRACT ABSTRACTUnderstanding how population structure shapes viral transmission remains challenging because most in vitro models assume well‐mixed conditions and lack spatial structure. We present a Herd‐Immunity‐on‐a‐Chip (HIC) platform—a 444‐chamber microfluidic network that recreates structured populations—and apply it to human coronavirus 229E infecting MRC‐5 fibroblasts. By varying initial inoculum (I 0), susceptible density (S 0), cell motility, and the fraction of non‐susceptible cells (U 0), we decode how seeding, density, immunity and movement reshape outbreak trajectories. Increasing S 0 or I 0 raised local contact rates, reduced effective intercellular spacing, and increased the reproduction number (R 0), accelerating spread. In contrast, higher U 0 suppressed transmission; with U 0 ≥ 80% of the fixed uninfected population (S 0 + U 0), outbreaks collapsed (herd immunity). Independently, lowering S 0 slowed early spread but did not, alone, prevent eventual transmission. We integrate the data with mathematical modelling of spatial, contact‐structured transmission to quantify changes in R 0 and apparent herd‐immunity thresholds. HIC offers a generalizable, bench‐top framework for outbreak forecasting and intervention testing. Introduction The coronavirus disease 2019 (COVID‐19) pandemic constitutes a major global health crisis, with high morbidity and mortality. Growing evidence links transmission rates to societal factors, including population density, adherence to social distancing, and vaccination coverage. Densely populated regions—such as Hubei Province in China and major U.S. metropolitan areas—exhibited faster spread than less‐populated settings [1], whereas effective distancing measures and vaccination campaigns consistently attenuated transmission [2, 3]. These patterns underscore the interaction between human behavior, public‐health policy, and transmission dynamics, and motivate the development of targeted, evidence‐based interventions. Traditional compartmental models—such as Susceptible–Infected–Recovered (SIR) [4, 5] and Susceptible‐Exposed‐Infected‐Recovered (SEIR) [6, 7] —have yielded important insights into epidemic dynamics by enabling simulation of epidemiological parameters and evaluation of candidate interventions. However, their standard formulations typically assume homogeneous mixing and time‐invariant transmission rates, assumptions that neglect substantial variability arising from epidemiological, socioeconomic, and behavioral factors [8]. For example, transient surges in effective population density in high‐density settings (e.g., New York City) can amplify contact rates and accelerate spread [9], while early, concentrated outbreaks (e.g., Daegu, Korea) illustrate how the size of the initially infected population shapes subsequent transmission trajectories [10]. Consequently, a gap remains between theoretical predictions from classical models and the context‐dependent realities of transmission in complex human populations. Conventional cell culture and animal models have advanced our understanding of viral mechanisms but cannot recapitulate the heterogeneous, network‐like interactions that govern transmission in human populations. Simplified biology and interspecies differences limit their translational fidelity, obscuring the roles of contact structure, mobility, and context‐dependent susceptibility in shaping epidemic trajectories [11, 12]. These limitations motivate experimental platforms that capture societal structure and enable quantitative testing of transmission drivers under controlled, reproducible conditions. Microfluidic technologies provide rigorously controlled, physiologically relevant microenvironments and have illuminated microbial interaction networks, drug‐resistance evolution, and directed cell migration [13, 14], yet applications to population‐level viral transmission are scarce. We therefore developed Herd‐Immunity‐on‐a‐Chip (HIC), a 444‐chamber hexagonal network with a central epicentre that enforces a structured contact topology, adapting a validated microchamber architecture [15] not previously applied to viral transmission. Conceptually, percolation on this lattice persists at 70% non‐susceptible coverage but fragments at 80%, yielding a herd‐immunity–like collapse (Figure 1a). HIC operationalises this logic by relaying infection locally from the epicentre while non‐susceptible cells reduce effective contacts (Figure 1b). Seven‐day imaging with daily counts produces chamber‐resolved transmission trajectories (Figure 1c). Guided by this framework, we hypothesised that higher S 0 density and larger I 0 accelerate spread, whereas increased U 0 and constrained contacts slow it. We used GFP‐labelled MRC‐5 (S 0), human coronavirus‐229E (HCoV‐229E)–infected lung fibroblast MRC‐5 (I 0), and aminopeptidase N (APN)‐knockdown MRC‐5 (U 0), treating U 0 as a vaccination‐like reduction in susceptibility [16]. Transmission was quantified in both the HIC and 96‐well controls, which respectively realise the structured population and its homogeneous‐mixing limit, allowing the effect of contact topology to be isolated. The reproduction number R0 was estimated directly from the experimental S0 depletion curves, which were reproduced by a Susceptible–Exposed–Non‐susceptible–Infected‐Dead (SEUID) model. The close experiment–simulation [17] agreement establishes HIC as a quantitative platform to probe transmission drivers and herd‐immunity thresholds. Herd‐Immunity‐on‐a‐Chip (HIC): structured transmission and threshold behavior. (a) Conceptual percolation. In a hexagonal contact lattice, 70% U 0 (yellow) permits infection (red) to percolate through remaining susceptible hosts (green), whereas 80% U 0 fragments connectivity and halts long‐range spread (herd‐immunity–like collapse). (b) Device and relay mechanism. HIC comprises 444 interconnected hexagonal microchambers arranged around a 700 µm‐diameter epicentre (seeding port). Infection relays locally via contacts between initially infected and susceptible cells; U 0 cells provide physical/functional shielding that reduces effective contacts. (c) Readout. Daily snapshots (D0–D7) yield chamber‐resolved trajectories. Enumeration of viable S 0 provides quantitative transmission metrics (e.g., R 0) and enables detection of a herd‐immunity threshold in the structured network. U 0 represents non‐susceptible cells (APN‐knockdown) used as a phenomenological proxy for vaccination‐induced reductions in susceptibility. Results and Discussion Susceptible‐Cell Density Governs Transmission in Structured Networks Urban centres during the early COVID‐19 outbreak highlighted how local susceptible‐host density shapes transmission trajectories [18]. To recapitulate density effects, we first established baseline behavior in homogeneous 96‐well plates by varying S 0 = 102, 103, 104 at fixed I 0 = 103 (Figure 2a), and in parallel applied the same conditions to HIC by seeding the epicentre and allowing spread across the compartmentalised network (Figure 2c). Seven‐day readouts quantified propagation in both formats. Effect of susceptible‐cell density (S 0) on viral transmission in homogeneous wells and HIC. (a) 96‐well plates. Confocal images at day 0 and day 7 for S 0 = 102, 103, 104 at fixed I 0 = 103 (n = 3 wells per condition). Blue: initially infected cells (I 0). Green: susceptible cells (S 0). Red: dead cells. Cartoon icons indicate relative susceptible population size. (b) Quantification in wells. Time courses of remaining S 0 over 7 days for each condition. R 0 estimated by fitting S 0‐depletion curves with a susceptible–infected–dead (SID) model [20]. (c) HIC spatial dynamics. Left: hexagonal‐lattice maps at days 0, 4, 7 showing chamberwise counts of S 0 (green scale) and I 0 (red scale); dead cells are indicated in dark red. Right: corresponding confocal images from the boxed lower‐right region. Blue: initially infected cells (I 0). Green: susceptible cells (S 0). Red: dead cells. (d) Schematic of contact‐mediated propagation in the HIC under low and high S 0. Low S 0 yields sparse encounters between initially infected (red) and susceptible (green) cells and slow front propagation, whereas high S 0 increases contact opportunities and accelerates spread. (e) Quantification in HIC. S 0 time courses for S 0 = 102, 103, 104 over 7 days. Solid lines, experiments; dotted lines, SEUID simulations. R 0 values estimated from S 0‐depletion fits. Scale bar, 500 µm. Note: The reference condition S 0 = 103 / I 0 = 103 in both the 96‐well plate and HIC experiments is shared with Figure 3 because it served as a common baseline condition in the same experimental run. In homogeneous wells, infection expanded rapidly and nearly all cells died by day 7 regardless of the initial S 0 (Figure 2a). Time courses of remaining susceptible cells yielded R 0 values of 2.66, 2.76, and 5.42 for S 0 = 102, 103, 104, respectively (Figure 2b), indicating that well‐mixed conditions amplify transmission and largely obscure any protective effect of low density. By contrast, HIC revealed strong S 0 dependence: early spread was constrained by local contacts, higher S 0 (103–104) accelerated propagation after a delay, and low S 0 (102) remained relatively protected. By day 4, devices at S 0 = 104 were predominantly infected, whereas S 0 = 102–103 showed substantially fewer infected chambers; by day 7, widespread infection persisted at higher densities while many low‐density chambers remained uninfected (Figure 2c; Movies S1–S3). Robustness was observed with MRC‐5/HCoV‐OC43 and Huh‐7 (human hepatoma)/HCoV‐229E (Figure S1). Comparison with SEUID simulations showed close agreement on HIC at S 0 = 102–104, with R 0 ≈ 1.72, 2.17, and 2.49 (Figure 2e). These data support a contact‐centred mechanism whereby increasing S 0 increases the probability of infectious cell‐cell encounters (effective contact rate, β). Consistent with this interpretation, reducing cell motility by focal adhesion kinase (FAK) knockdown [19] without changing S 0 delayed outward spread and attenuated transmission on HIC, while producing only modest mitigation in homogeneous wells (Figure S2, S3; Movies S4–S6). Together, the results indicate that spatial compartmentalisation exposes density‐ and contact‐limited dynamics that are masked in well‐mixed environments, with S 0 and contact opportunities jointly determining effective transmission. We acknowledge, however, that cell proliferation during the 7‐day culture period may also contribute to local cell redistribution and cannot be completely distinguished from infection‐driven spread in this in vitro model. Initial Infection Density (I 0) Tunes Outbreak Onset in Structured Networks We modelled scenarios from small clusters to large outbreaks by varying the initial infected population size (I 0), reflecting early COVID‐19 superspreading linked to a large religious gathering in South Korea [10]. We first examined how I 0 influences spread in a homogeneous setting by inoculating 96‐well plates with I 0 = 10, 103, or 105 at fixed S 0 = 103. Across all I 0 values, wells exhibited rapid spread: substantial involvement by day 4 and near‐complete depletion of S 0 by day 7 (Figure 3a). Quantifying S 0 over time and fitting a SID model yielded R 0 = 2.42, 2.76, and 2.81 for I 0 = 10, 103, and 105, respectively (Figure 3b), indicating that the well‐mixed environment strongly amplifies transmission and is only weakly sensitive to initial infection density. Initial infection density (I 0) tunes outbreak onset in homogeneous wells and HIC. (a) 96‐well plates. Confocal images at day 0 and day 7 for fixed S 0 = 103 and varying I 0 = 10, 103, and 105 (n = 3 wells per condition). Blue: initially infected cells (I 0); green: susceptible cells (S 0); red: dead cells. Cartoon icons indicate relative I 0. (b) Quantification in wells. Time courses of remaining S 0 over 7 days; R 0 is estimated from the experimentally obtained S 0‐depletion curves. (c) HIC spatial dynamics. Left: hexagonal‐lattice maps at days 0, 4, 7 showing chamberwise counts of S 0 (green scale) and I 0 (red scale); dead cells indicated in dark red. Right: matching confocal images from the boxed lower‐right region (n = 3 devices per condition). Blue: initially infected cells (I 0); green: susceptible cells (S 0); red: dead cells. (d) Contact‐based interpretation. Low I 0 seeds few early foci and lengthens the lag; high I 0 seeds many foci, shortening the lag and advancing onset. (e) Quantification in HIC. Time courses of remaining S0 for I 0 = 10, 103, and 105, with experimental data (solid lines) and SEUID simulations (dotted lines). R0 values were derived from the experimentally obtained S0 ‐depletion curves. Scale bar, 500 µm. Note: The reference condition S 0 = 103 / I 0 = 103 in both the 96‐well plate and HIC experiments is shared with Figure 2 because it was obtained from the same experimental run and served as a common baseline condition. We next tested the same seeding conditions in the HIC with I 0 = 10, 103, 105 (Figure 3c). Low I 0 produced slow, localised spread with no detectable infection until day 7, whereas higher I 0 increased early exposure frequency and accelerated dissemination, yielding extensive spread by day 4 and near‐complete loss of S 0 by day 7. Once initiated, the I 0 = 10 case propagated at a rate comparable to higher inocula, producing a similar final attack size (Movies S7, S8). The trend was robust across MRC‐5/HCoV‐OC43 and Huh‐7/HCoV‐229E experiments (Figure S4). Contact‐based reasoning (Figure 3d): low I 0 starts few clusters and lengthens the delay before spread takes off; high I 0 starts many clusters, shortening the delay and bringing earlier onset. SEUID simulations recapitulated the HIC data, with R 0 declining from 2.34 to 2 as I 0 decreased (Figure 3e). In contrast, well‐mixed controls produced higher R 0 (2.42, 2.76, 2.81) with weak sensitivity to I 0 (Figure 3b), masking the early‐phase delay revealed by the structured HIC network. Thus, while higher I 0 accelerates early spread, it chiefly tunes outbreak onset (lag duration), with structured networks exposing early‐phase delays that are obscured in well‐mixed settings. To further evaluate the physiological relevance and application potential of HIC, we performed additional I 0‐focused validation experiments. First, we repeated the I 0‐density transmission experiment using primary human lung fibroblasts. Similar to the results obtained with established cell lines, primary cells showed rapid transmission in homogeneous 96‐well plates, whereas HIC resolved I 0‐dependent differences in outbreak onset and spatial propagation (Figure S13). This result indicates that the structured‐transmission behavior captured by HIC is not restricted to established cell lines, but can also be reproduced in primary human lung cells, supporting the physiological relevance of the platform. In addition, to explore the potential use of HIC for early intervention testing, we pretreated I0 cells with imipramine, a reported inhibitor of coronavirus entry and intracellular trafficking, before loading them into the HIC epicentre, while the pre‐seeded susceptible population was maintained in normal medium. This targeted treatment delayed viral spread in HIC and reduced R0 compared with untreated and dimethyl sulfoxide (DMSO) vehicle controls, whereas 96‐well controls showed rapid cell death from day 4 and retained high R0 values regardless of treatment (Figure S14). These findings suggest that HIC can reveal intervention‐dependent delays in spatial transmission that are masked in well‐mixed culture. Non‐Susceptible Fraction (U 0) Establishes Herd‐Immunity Thresholds in Structured Networks Herd immunity depends on the fraction of immune (or otherwise protected) individuals and, in practice, on vaccine coverage and effectiveness [21, 22]. In our framework, this protection is operationalised as a non‐susceptible fraction (U 0) generated by APN knockdown in MRC‐5, rendering them refractory to infection by HCoV‐229E [16]. The remaining APN‐intact cells constitute S 0. As a homogeneous‐mixing control, we first benchmarked transmission in conventional 96‐well plates, which eliminate spatial compartmentalization while maintaining the same virus–host interactions as the HIC, using U 0 = 60%, 70%, 80% at fixed S 0 + U 0 = 103 and I 0 = 103. Across all U 0 levels, wells showed rapid spread: S 0 was substantially infected by day 4 and fully depleted by day 7 (Figure 4a). The experimentally obtained S 0 depletion curves yielded R 0 = 3.17, 2.97, 2.48 for U 0 = 60%, 70%, 80%, respectively (Figure 4b), indicating that well‐mixed environments fail to establish herd immunity even at high U 0. Effect of the non‐susceptible fraction (U 0) on virus transmission in homogeneous wells and HIC. (a) 96‐well plates. Confocal images at days 0 and 7 for fixed S 0 + U 0 = 103 and I 0 = 103, with U 0 = 60%, 70%, or 80% of the 103 uninfected cells (n = 3 wells per condition). Blue: initially infected (I 0); green: susceptible (S 0); yellow: non‐susceptible (U 0); red: dead cells. Cartoon icons indicate the U 0 proportion. (b) Quantification in wells. Time courses of remaining S 0 over 7 days; R 0 was estimated from the experimentally obtained S 0‐depletion curves. (c) HIC spatial dynamics. Left: hexagonal‐lattice maps at days 0, 4, 7 showing chamberwise counts of S 0 (green scale), I 0 (red scale), and U 0 (yellow). Right: corresponding confocal images from the boxed lower‐right region (n = 3 devices per condition). Blue: initially infected (I 0); green: susceptible (S 0); yellow: non‐susceptible (U 0); red: dead cells. (d) Contact‐based interpretation. Low U 0 leaves many permissive contact paths and permits extensive spread; high U 0 sparsifies the contact graph, shielding susceptible clusters and breaking inter‐chamber transmission. (e) Quantification in HIC. Time courses of remaining S 0 for U 0 = 60%, 70%, and 80% with experimental (solid) and SEUID simulation (dotted) fits; R 0 values were derived from S 0‐depletion curves. Scale bar: 500 µm. We then applied the same U 0 fractions to HIC (Figure 4c). In HIC, U 0 = 60–70% produced widespread infection by day 7, whereas U 0 = 80% left more than half of S 0 uninfected, with spread largely confined to the high‐contact epicentre and strongly attenuated peripherally (Figure 4c, Movies S9–S11). Similar trends were observed when CASD1 (CAS1 domain‐containing protein 1)‐knockdown MRC‐5 cells served as U 0 (Figure S5a,b). CASD1 is a sialate O‐acetyltransferase that generates 9‐O‐acetylated sialic acid, the attachment moiety for HCoV‐OC43. Similar results were also obtained when APN‐knockdown Huh‐7 cells were used for HCoV‐229E (Figure S5c,d). Mechanistically (Figure 4d), low U 0 leaves many permissive contact paths, whereas high U 0 sparsifies the contact graph, shielding susceptible clusters and breaking inter‐chamber relay. Quantitatively, increasing U 0 from 60% → 80% reduced R 0 from 1.55 → 0.98 in HIC, closely matching SEUID predictions; at 80% U 0 infection of S 0 was effectively halted by day 7 (Figure 4e). By contrast, well‐mixed plates retained R 0 > 2.48 even at 80% U 0 (Figure 4b), precluding herd immunity. Thus, herd‐immunity thresholds depend on both the protected fraction and contact topology: spatial compartmentalisation amplifies the protective effect of U 0, enabling outbreak suppression at lower immune fractions than in well‐mixed settings. For modelling, treating U 0 as a parameter that sparsifies the contact kernel provides a principled route to predicting thresholds in structured populations. The herd‐immunity threshold identified in the HIC is specific to the present experimental configuration and should not be interpreted as a universal threshold for human populations. Rather, the platform is intended to experimentally demonstrate how spatial contact topology and the proportion of protected units jointly influence transmission dynamics under controlled conditions. Future studies incorporating more complex and heterogeneous contact networks may further improve its translational relevance. Conclusion We established a structure‐aware microfluidic platform—HIC—that quantitatively links population composition and contact mechanics to coronavirus transmission in a 444‐chamber hexagonal network. Increasing S0 or I0 elevated effective encounter rates and accelerated spread, whereas increasing U0 progressively attenuated transmission and, at 80% of the fixed uninfected pool, produced a herd‐immunity–like collapse that prevented network‐wide percolation. Reducing cell motility via FAK knockdown further confirmed that the frequency of physical cell–cell contacts is a direct driver of transmission, and that spatial isolation is a key regulator of intervention efficacy—dependencies that are largely masked in well‐mixed culture wells. By fitting susceptible‐cell depletion trajectories to estimate R 0 and comparing them with SEUID simulations, HIC couples directly observable infection fronts to mechanism‐aware modelling, improving the interpretability of density, seeding, immunity, and motility effects in structured populations. While the present chip uses a limited set of host cell types and a diffusion‐dominated 2D geometry, the framework is readily extensible to multi‐cell and immune co‐cultures, in situ sensing, and models with contact kernels that explicitly depend on density, motility, and spatial topology. Collectively, these results position HIC as a low‐barrier, scalable testbed for structure‐aware forecasting and for scenario testing of antivirals, vaccination coverage, and contact‐modulating strategies in preparation for future outbreaks. Experimental Section Cell Culture and Viruses The MRC‐5 lung fibroblast and Huh‐7 human hepatoma cells were purchased from the Korean Cell Line Bank (Seoul, Korea). Green fluorescent protein (GFP)‐labelled MRC‐5 (GFP‐MRC‐5) and Huh‐7 (GFP‐Huh‐7) cells were purchased from Lugen Sci Co., Ltd. (Bucheon, Korea). The cells were maintained in minimum essential medium (MEM) (HyClone Laboratories, Logan, UT, USA) and Roswell Park Memorial Institute (RPMI) 1640 medium (HyClone Laboratories) supplemented with 10% (v/v) foetal bovine serum (FBS) (HyClone Laboratories), 100 units penicillin mL−1 (Life Technologies, Carlsbad, CA, USA), 100 µg streptomycin mL−1 (Life Technologies), 25 mM sodium bicarbonate (Sigma‐Aldrich), and 25 mM hydroxyethyl piperazine ethane sulfonic acid (HEPES) (Life Technologies) at 37°C with 5% CO2. The HCoV strains (229E and OC43) were provided by the Korea Research Institute of Bioscience and Biotechnology (KRIBB) (Daejeon, Korea). Virus Propagation and Measurement of the 50% Tissue Culture Infective Dose (TCID50) Virus propagation and measurement of TCID50 values were performed as described previously [23]. Cytopathic Effect (CPE) MRC‐5 cells (104 cells per well) were seeded in 96‐well plates and incubated overnight at 37°C with 5% CO2. The cells were washed with phosphate‐buffered saline (PBS) and inoculated with HCoV‐229E viruses at 10−3 to 103 TCID50 mL−1 diluted in modified MEM with 2% FBS at 33°C in a 5% CO2 incubator [24, 25]. After 3 days of incubation, CPE, such as rounding and aggregation of cells, was monitored (Figure S6). Preparation of Infected Cells (I0 ) To generate infected cells, host (MRC‐5) cells (5 × 105 cells) were seeded in a T‐75 flask (SPL Life Sciences, Pocheon, Korea) and incubated overnight at 37°C with 5% CO2. The cells were washed with PBS and inoculated with HCoV‐229E at 100 TCID50 mL−1 in MEM at 33°C with 5% CO2. On the next day, infected cells were labelled with Hoechst 33342 (2’‐[4‐ethoxyphenyl]‐5‐[4‐methyl‐1‐piperazinyl]‐2.5‐bi‐1H‐benzimidazole trihydrochloride trihydrate) (Sigma‐Aldrich), which stains the nuclei of living cells, before seeding into the epicentre (Figure S7). Preparation of U0 and Low‐Motility S0 cells GFP‐MRC‐5 cells were seeded in 6‐well plates at 2 × 105 cells/well and cultured overnight in MEM containing 10% FBS. To knock down APN, CASD1, and FAK, the respective cells were transfected with 100 nM AccuTarget siRNA (Table 1) using Lipofectamine RNAiMAX transfection reagent (Invitrogen) according to the manufacturer's protocol. As a negative control, parallel wells received 100 nM AccuTarget negative control siRNA (Bioneer Co., Daejeon, Korea). After 48 h of transfection, KD efficiency was verified using quantitative reverse transcription‐PCR (qRT‐PCR) and western blot before seeding these APN or CASD1‐KD cells as U 0 and FAK‐KD cells as low‐motility S 0 in the HIC experiments. TABLE 1siRNA sequences.SenseAntisense FAK siRNA 5’‐ AAC CAC CUG GGC CAG UAU UAU ‐3’5’‐ AUA AUA CUG GCC CAG GUG GUU ‐3’ APN siRNA 5’‐GCA GCA GAU CUG UAU AUU U‐3’5’‐AAA UAU ACA GAU CUG C‐3’ CASD1 siRNA 5’‐GGA UAU GCC CGU UCA GUU U‐3’5’‐ AAA CUG AAC GGG CAU AUC C ‐3’John Wiley & Sons, Ltd. siRNA sequences. Evaluation of the Transfection Efficiency Total RNA was extracted and purified from the cells using a spin column kit (Zymo Research, Irvine, CA, USA). Complementary DNA (cDNA) fragments were synthesised from the extracted RNA using a first‐strand cDNA synthesis system (LeGene Biosciences, San Diego, CA, USA). The primer sequences used were designed by Bioneer Co. (Table 2). The increase in amplified DNA was measured using iQ SYBR Green premix (Bio‐Rad, Hercules, CA, USA). TABLE 2Primer sequences.SenseAntisense FAK 5’‐GGC TGT CAT CGA GAT GTC CA ‐3’5’‐TCT CAT CCA CCG TGG CTA GT‐3’ APN 5’‐GTG ATG GCA GTG GAT GCA CAG CTT‐3’5’‐CCT GTC CGA GGA CTG TA‐3’ CASD1 5’‐GTG GAT TTT CTG TGG CAT CC‐3’5’‐AAG CGC TTC ACT GCT ACC AT‐3’ GAPDH 5’‐ AGG TGG TGA AGC AGG CGT CGG AGG G‐3’5’‐CAA AGT GGT CGT TGA GGG‐3’John Wiley & Sons, Ltd. Primer sequences. Western blotting was performed to confirm transfection efficiency. Cells were lysed using radioimmunoprecipitation assay (RIPA) buffer (Bio‐solution, Suwon, Korea), and proteins were extracted. The protein concentration in RIPA buffer was measured using Bradford reagent (Sigma). Proteins (30 µg) were separated using 10% SDS‐PAGE and then probed with antibodies (Figure S8). For the first antibody treatment, anti‐CD13 (APN) antibody (ab108382; Abcam, Cambridge, UK) at a 1:5,000 dilution, CASD1 antibody (PA5‐69624) from Invitrogen (Waltham, MA, USA) at a 1:1,000 dilution, FAK antibody (AHO1272) from Invitrogen at a 1:500 dilution, and GAPDH antibody (Santa Cruz Biotechnology) at a 1:8,000 dilution were used. For the secondary antibody treatment, anti‐rabbit‐HRP antibody (Thermo Fisher) at a 1:10,000 dilution and anti‐mouse HRP antibody (ab205719; Abcam, Cambridge, UK) at a 1:5,000 dilution were used. Evaluation of Virus Infectivity MRC‐5 and APN‐KD MRC‐5 cells were seeded in 96‐well plates (Corning Inc., Corning, NY, USA) (104 cells/well) and incubated overnight at 37°C with 5% CO2. The cells were washed with PBS and inoculated with HCoV‐229E at 10 to 103 TCID50 mL−1 diluted in MEM with 2% FBS at 33°C with 5% CO2 for 3 days (Figure S9). Measurement of Cell Migration Ability MRC‐5 (Con) and FAK‐KD MRC‐5 cells (104 cells/well) were seeded on a 6‐well plate (Corning Inc.). To observe the migration distance of cells, images of displacement with cells were obtained at 1 and 24 h in a live cell chamber (Live Cell Instrument, Seoul, Korea) at 37°C with 5% CO2 using a DeltaVision microscope (GE Healthcare) equipped with a CoolSnap HQ2 camera (Photometrics, Tucson, AZ, USA). Cell movement tracks were labelled, and total displacement (Figure S10; Movies S12–S15) was measured using ImageJ (NIH, Bethesda, MD, USA). Fabrication of HIC A previously developed chip [15, 26] was used to prepare the HIC (Figure S11). Briefly, a mould of HIC was first fabricated using a negative photoresist SU–8 (MicroChem Co., Westborough, MA, USA) spin‐coated on a 4‐inch wafer for 30 s (SU–8 2025, 2000 rpm, 40 µm thickness). The HIC was cast into polydimethylsiloxane (PDMS) (Sylgard 184, Dow Corning Co.) by soft lithography [27]. The HIC comprised a 444‐chamber microfluidic network (200 µm long and 40 µm deep) created using microchannels that were 30 µm wide. 5 µm‐high sidewalls were present at the edges of the microchambers to allow nutrient flow into the interior of the microchambers. Nutrients were perfused from the four reservoirs through two microchannels (150 µm wide) surrounding the chambers. A 700 µm‐diameter seeding port (‘epicentre’) was punched. The process of sterilising and coating the chips has been described previously [26]. Cell Loading Strategy for Establishing the Initial Spatial Distribution in HIC To establish the initial spatial organization of the HIC, S 0/U 0 and I 0 cells were loaded using different procedures. For S 0 seeding, or S 0/U 0 seeding in the herd‐immunity experiments, gentle manual flow was applied through the epicentre inlet to distribute the cell suspension throughout the interconnected microchamber network. After loading, the HIC was incubated overnight to allow cell attachment and stabilization across the network. In contrast, I 0 cells were intended to serve as a localized initial infection source. Therefore, I 0 cells were loaded without applying manual flow; the infected‐cell suspension was gently placed in the epicentre and allowed to settle by gravity. This procedure maintained I 0 cells in the central epicentre region at day 0, while S 0 and U 0 cells were distributed throughout the surrounding microchamber network. The initial distribution of each fluorescently labeled cell population was confirmed from day 0 stitched fluorescence images before longitudinal transmission analysis. Effect of Susceptible Cell (S0 ) Density on Virus Transmission in HIC and a 96‐Well Plate Before observing the virus transmission in the HIC, totals of 103, 104, and 105 GFP‐MRC‐5 cells were suspended in 10 µL culture medium and 1 µL of cell suspension was injected via the epicentre using a 200 µL pipette tip (T‐200, Axygen; Union City, CA, USA), so that the final number of initial S 0 was 102, 103, 104 cells/HIC, respectively. Considering that the area of the well of a 96‐well plate is 1.6 times that of a chamber of the HIC, the final number of initial S 0 in the 96‐well plate was set as 1.6 × 102, 1.6 ×103, 1.6 ×104 cells/well to normalize the same cell density as that in the HIC. The HIC and a 96‐well plate were incubated overnight at 37°C with 5% CO2. On the next day, I 0 at 104 cells/10 µL in a 10‐µL pipette tip (T‐300) were injected and flowed into the epicentre by gravity for virus transmission. The final number of I 0 cells was 103 cells/HIC and 1.6 × 103 cells/well. The HIC and 96‐well plate containing both S 0 and I 0 cells were incubated at 33°C with 5% CO2 for 7 days. Each day, the medium was replaced with 200 µL of fresh culture medium. To generate the cell number data, optical and fluorescence images were obtained daily using a K1‐Fluo confocal microscope (Nanoscope Systems, Inc., Daejeon, Korea). We employed the K1‐array viewer to acquire stitched images across various channels at a 4× magnification. Images were processed using ImageJ. The number of live susceptible cells was manually counted from days 0 to 7. Effect of I0 Density on Virus Transmission in the HIC and a 96‐Well Plate GFP‐MRC‐5 cells at 103 cells/HIC and 1.6 × 103 cells/well were seeded in the HIC and 96‐well plate and incubated overnight at 37°C with 5% CO2. The next day, I 0 at 101, 103, and 105 cells/HIC and 1.6 × 101, 1.6 × 103, 1.6 × 105 cells/well were flowed into the epicentre by gravity. The HIC containing both S 0 and I 0 cells was incubated at 33°C with 5% CO2 for 7 days. The subsequent steps were the same as those described in the ‘Effect of S 0 density on virus transmission in HIC’. For primary‐cell validation, the same I 0‐density transmission experiment was repeated using primary human lung fibroblasts purchased from Lonza (Walkersville, MD, USA) and maintained in fibroblast basal medium (FBM) from Lonza, with the same seeding, infection, imaging, and 7‐day culture procedures described above. Effect of U0 Density on Virus Transmission in the HIC and a 96‐Well Plate For herd immunity experiments, S 0 (GFP‐MRC‐5) and U 0 (APN‐KD MRC‐5) were mixed at a ratio of 4:6 to 2:8 with a total of 103 cells/HIC and 1.6 × 103 cells/well seeded into the HIC and 96‐well plate, respectively. To visualise U 0, cells were labelled with the CellMask Deep Red Plasma Membrane Stain (Invitrogen, Waltham, MA, USA) according to the manufacturer's instructions before seeding. After 1 day of incubation, I 0 at 104 cells/10 µL in a 10‐µL pipet tip were injected and flowed into the epicentre by gravity for virus transmission, and 1.6 × 103 cells/well were added as the I 0 in the 96‐well plate. Effect of Imipramine‐Treated I0 Cells on Virus Transmission in the HIC and a 96‐Well Plate To evaluate the applicability of HIC for early intervention testing, imipramine treatment was applied only to I 0 before loading. S 0 cells were seeded in the HIC and 96‐well plates using the same procedure described above and maintained overnight in normal culture medium without drug treatment. On the next day, HCoV‐229E‐infected I 0 cells were prepared as described above and resuspended in culture medium containing imipramine. For vehicle controls, I 0 cells were resuspended in culture medium containing 0.1% (v/v) of DMSO. Untreated control I 0 cells were resuspended in normal culture medium. The I 0 suspensions were then loaded into the HIC epicentre or added to 96‐well plates using the same loading procedure as described for the I 0‐density experiment. The pre‐seeded S 0 cells in the HIC were not directly exposed to imipramine or DMSO before I 0 loading. After I 0 loading, the HICs and 96‐well plates were incubated at 33°C with 5% CO2 for 7 days. Daily imaging, medium replacement, live/dead staining, S 0 counting, and R 0 estimation were performed as described above. SEUID‐Based Computational Model of Virus Transmission To describe virus transmission on the HIC platform, we developed a spatial model incorporating five cell states: susceptible (S), exposed (E; infected, but not yet infectious), non‐susceptible (U), infected (I), and dead (D). This model accounts for spatial confinement by indexing each microchamber as I, with virus transmission driven by local interactions and constrained interchamber migration. S and U cells undergo division at rate γ, while I cells do not divide and are assumed to die (death rate µ), precluding recovery. The system is governed by the following equations: (1) dSidt=−βwSiIi−βn∑jn.n.SiIj+γSi (2) dEidt=βwSiIi+βn∑jn.n.SiIj−αEi (3) dIidt=αEi−μIi (4) dUidt=γUi (5) dDidt=μIi here, βw and βn represent the intra‐chamber and nearest‐neighbour inter‐chamber contact rates, respectively. To reflect the suppressive effect of immune cells on contact probability in confined domains, these rates are modulated as: β w = β0(1 − c(Ui − Ucr ) m ) and β n = β1(1 − c(Ui − Ucr ) m ). The model was fitted to the experimental data using nonlinear least squares regression. Calculation of R0 In the standard SID framework [17, 20], the basic reproduction number R 0 is defined as R0=βμ,where β is the transmission rate and μ is the death rate. In the present analysis, rather than deriving R 0 from the microscopic parameters β and μ of a SID or SEUID model, we estimated it as an effective reproduction number from operational quantities obtained directly from the experimental data: R0eff=βeffμeff=βeffτinfec, where β eff is the effective transmission rate and τinfec=1μeff is the infectious period. The effective transmission rate β eff was operationalised as the average daily fractional decrease of S(t) over the first two days of spread: βeff=−1S0S(2ndday)−S02days. The infectious period τ infec was approximated as the time required for a detectable cytopathic effect (CPE) to appear at the microchamber level, including cell rounding, aggregation, loss of GFP‐positive cells, and the appearance of dead‐cell staining. We used τ infec = 2 days, based on the time required for HCoV‐229E infection to produce an observable CPE. Statistical Analysis Statistical significance was determined using Student's t‐test. The results are presented as the mean ± standard deviation (S.D.). All quoted p‐values were two‐tailed, and differences were considered statistically significant at *p < 0.05, **p < 0.01, and ***p < 0.001. Author Contributions Zhenzhong Chen: data curation, methodology, validation. Narina Jung: data curation, writing – original draft, software, methodology, investigation, formal analysis, writing – review and editing. Sungsu Park: writing – original draft, supervision, investigation, conceptualization, methodology, funding acquisition, validation, writing – review and editing, project administration, data curation. Wanyoung Lim: conceptualization, methodology, data curation, visualization, investigation. Min‐Hyeok Kim: data curation, validation, visualization, methodology, writing – review and editing, investigation. Byung Mook Weon: conceptualization, supervision, investigation. Jiande Zhang: methodology, data curation, writing – original draft, investigation, formal analysis, visualization. Funding This work was supported by National Research Foundation of Korea (NRF) grants funded by the Ministry of Science and ICT (RS‐2023‐00234581) and the Ministry of Education (RS‐2019‐NR040077), and by KIAS individual grant (PG085701) at Korea Institute for Advanced Study. Conflicts of Interest The authors declare no conflicts of interest. Supporting information Supporting File 1: advs77383‐sup‐0001‐SuppMat.docx. Supporting File 2: advs77383‐sup‐0002‐SuppMat.zip.