3D microfluidic gradient generator for combination antimicrobial susceptibility testing

The study targets the challenge of rapidly determining effective antibiotic therapies for antimicrobial-resistant (AMR) infections, which pose major health and economic burdens. Standard AST methods (MIC and CDS) are accurate but slow and labor-intensive, often taking days, which can drive empiric use of broad-spectrum antibiotics and further resistance. While microfluidic µ-CGGs can accelerate AST by generating discrete concentration gradients, conventional planar (2D) designs are inherently limited to symmetric gradients of only two inputs, constraining combination testing to pairs of antibiotics. The research question is whether a truly three-dimensional microfluidic network can generate symmetric concentration gradients for three inputs to support higher-throughput combination AST. The purpose is to design, model, fabricate, and validate a 3D-printed µ-CGG that overcomes 2D routing limitations, enabling accurate, symmetric three-drug gradients for MIC and CDS with clinically relevant antibiotics against resistant E. coli.

Conventional AST (MIC and CDS) requires multiple manual fluid handling steps and lengthy incubation (often 2–7 days), hindering timely clinical decision-making. Microfluidic AST platforms, particularly µ-CGGs, have improved throughput and reduced reagent volumes by producing discrete gradients between two inputs (e.g., antibiotic and buffer), enabling rapid on-chip or off-chip assays. However, planar MEMS/PDMS fabrication yields essentially 2D channel networks, which cannot generate symmetric gradients for more than two fluids; prior multi-input planar µ-CGGs are nonsymmetric and omit combinations of nonadjacent inputs. Attempts at quasi-3D structures via multilayer PDMS alignment are complex and limited in geometric complexity and function. High-resolution 3D printing has emerged for microfluidics, yet prior 3D-printed devices have not demonstrated discrete gradients for more than two antibiotics in AST. Thus, there is a gap for a truly 3D µ-CGG capable of symmetric three-input gradient generation for combination antibiotic testing.

Design and analysis: The team designed a 3D microchannel network featuring a tetrahedral arrangement of nodal combination–mixing–splitting units to achieve symmetric routing in all three spatial dimensions. The device has three inlets and 13 discrete outlets, generating outputs that capture combinations of inputs 1&2, 2&3, 1&3, and 1&2&3, which is not achievable in planar devices. A nodal analytical method (extending conventional 2D CGG analysis) was used to compute expected outlet flow rates (Qi) and species concentrations (Ci) under pressure-driven flow, assuming complete mixing within each nodal unit.

3D mixing structures and CFD: Because outlet accuracy depends on intra-node mixing, three variants were studied in COMSOL CFD simulations: (i) smooth-walled vertical channels (reference), (ii) vertical channels with repeating hollow spherical sidewall indentations (bulbous µ-mixer), and (iii) vertical channels with embedded 3D rifled sidewall structures (3D rifled µ-mixer). Simulations evaluated normalized concentration distributions across 13 outlets over symmetric input flow rates from 0 to 4000 µL/min and compared to analytical predictions.

Device geometry and features: The network comprises three layers of spherical bulbs (≈1.25 mm diameter) connected by vertical channels integrating micromixers. Representative dimensions include: overall vertical channel length ≈5 mm, micromixer pitch ≈1 mm, feature depth d ≈130 µm, inter-layer microchannels and 60°/120° routing to maintain tetrahedral symmetry. The final selected design incorporated the 3D rifled µ-mixer due to superior mixing.

Fabrication and postprocessing: Devices were fabricated via Multijet 3D printing as a single solid body with embedded hollow channels and bulbs. Postprocessing removed internal support material by immersion in mineral oil at 60 °C for ~10 min, followed by cleaning to reveal clear internal structures.

Experimental validation—gradient generation: Fluorescent tracer experiments validated concentration outputs. Analytical and CFD predictions guided operation at flow rates where concentration became flow-independent (~1000 µL/min). Gradient uniformity and symmetry were visualized by imaging the positive solids model surfaces.

Biological assays—MIC and CDS: Discrete outlet solutions (µ-drug cocktails) were collected using a pressure-driven setup (compressed air and pressure controller) delivering three antibiotics (tetracycline, ciprofloxacin, amikacin). For AST, collected gradients were combined with antibiotic-resistant E. coli, growth media, and resazurin viability indicator. Cultures were incubated at 37 °C, and fluorescence imaging (green LED 525 nm excitation, 585 nm emission filter) in a light-insulating box with a DSLR camera quantified bacterial proliferation. The platform supported single-antibiotic MIC and pairwise and three-way CDS assays.

• The analytical normalized outlet concentrations are intrinsically independent of flow rate; CFD simulations for all designs converged to flow-rate-independent behavior around ~1000 µL/min.
• Mixing strongly affected accuracy relative to analytical predictions: • Smooth-walled vertical channel design: average errors ~24% at outlets 2 & 3 and ~57% at outlets 6 & 8. • Bulbous µ-mixer design: average errors ~26% at outlets 2 & 3 and ~60% at outlets 6 & 8. • 3D rifled µ-mixer design: all outlets within ≤10% error versus analytical predictions, meeting the commonly accepted ≤10% accuracy criterion for AST CGGs.
• The selected 3D rifled µ-mixer-integrated µ-CGG produced discrete outputs spanning useful proportions of each input (approximately 1.0, ~0.7, 0.5, ~0.3, ~0, plus a near-equal 1:1:1 mixture), delivering a symmetric three-input gradient not possible with planar routing.
• Proof-of-concept biological testing demonstrated feasibility for MIC and pairwise/three-antibiotic CDS against antibiotic-resistant E. coli using tetracycline, ciprofloxacin, and amikacin.

The results demonstrate that a truly three-dimensional microchannel network with tetrahedral nodal routing overcomes the inherent two-dimensional constraint of conventional µ-CGGs, enabling symmetric gradients among three inputs and thereby supporting higher-throughput combination AST. CFD-guided evaluation of integrated micromixer geometries identified that 3D rifled sidewall structures achieve near-complete mixing within nodal units, yielding outlet concentrations that match analytical targets within the ≤10% accuracy required for reliable AST. This fidelity allows the device’s 13 outputs to represent a comprehensive set of discrete three-drug combinations, facilitating efficient MIC and CDS testing with minimal manual handling. The ability to collect discrete µ-drug cocktails and perform downstream bacterial assays indicates practical applicability to rapid AST workflows, potentially reducing time-to-result and mitigating empiric broad-spectrum antibiotic use.

This work introduces and validates a 3D-printed µ-CGG featuring a tetrahedral nodal network capable of generating symmetric three-input concentration gradients, a functionality unattainable in planar microfluidics. Analytical modeling, COMSOL CFD, and experiments show that integrating 3D rifled micromixers enables accurate (≤10% error) discrete outputs across 13 outlets. Proof-of-concept MIC and combination (pairwise and three-way) CDS with tetracycline, ciprofloxacin, and amikacin against resistant E. coli demonstrate utility for AST. Future directions include expanding to more than three inputs, integrating on-chip culture and detection to further shorten assays, optimizing operation across broader flow regimes, and validating with diverse pathogens and clinical samples.

Analytical predictions assume complete mixing within nodal units; accuracy depends on micromixer performance and may degrade with insufficient mixing (as seen in smooth and bulbous designs). Device performance becomes flow-rate-independent only beyond a threshold (~1000 µL/min), suggesting sensitivity to operating conditions at lower flows. Additive manufacturing imposes constraints on feature resolution and uniform support removal, which can impact channel fidelity and mixing structures. Biological demonstration was proof-of-concept; detailed MIC values and comprehensive clinical validation were not provided in the excerpt.