Acoustofluidic system for targeted antibody removal in transplantation: Enabling small-volume therapeutic apheresis

Abstract

Antibody-mediated rejection is a leading cause of allograft failure and mortality in pediatric solid organ transplant recipients. Current apheresis systems require large blood volumes and are primarily designed for adults, making them unsuitable for children and small animals. These systems often indiscriminately remove both harmful and protective antibodies, increasing the risk of complications such as life-threatening infections. To address this critical need, we developed acoustofluidic-based system for targeted antibody removal in transplantation (A-START). A-START is engineered to handle small blood volumes and has demonstrated efficacy in preclinical small animal trials. In a sensitized rodent skin transplantation model, A-START only demands 240 microliters of blood, selectively removing donor-specific alloantibodies (DSAs) while preserving protective antibodies, such as tetanus antibodies. A-START retains ~95% of beneficial antibodies and achieves a 60% improvement in DSA removal compared to conventional methods. These findings highlight the transformative potential of A-START as a promising, reliable, scalable solution for improving outcomes in pediatric transplantation and treating antibody-mediated diseases.

A-START is a therapeutic apheresis system that can remove harmful antibodies from blood samples of children and small animals.

INTRODUCTION

Solid organ transplantation is a life-saving intervention that offers children with end-stage organ failure a substantially improved quality of life (1). Despite this, long-term graft survival remains a pressing challenge, with more than half of all transplanted organs failing within a decade due to antibody-mediated rejection (2–4). Antibody-mediated rejection occurs when the recipient’s immune system overwhelms the body with donor-specific alloantibodies (DSAs) generated against major histocompatibility complex (MHC) molecules located on the endothelial surface of the transplanted organ (5). This immune response triggers a variety of pathological events, including vascular injury, thrombosis, arteriosclerosis, and ischemia (6–8). These pathological events are often fatal, with antibody-mediated rejection carrying an 80% higher risk of mortality. Of the transplanted children who survive, 50% will require a second transplant by age 25 (9). In an era of limited organ availability, children who are broadly sensitized and have antibodies to their prospective donors are more likely to die while awaiting a new organ (10, 11). These challenges underscore the urgent need for innovative therapeutic strategies to mitigate antibody-mediated rejection and improve outcomes for pediatric transplant recipients.

Hence, strategies to decrease DSA production during solid organ transplantation are essential to prolonging the lifespan of the transplanted organ and improving patient outcomes. A cornerstone method for DSA reduction is apheresis, in which blood is passed through a centrifuge (12), column (13), or filter (14) to remove antibodies and subsequently returned to the patient’s circulation (15, 16). However, existing apheresis devices face substantial limitations when used with pediatric populations, including infants and small children. These limitations include (i) designs that are optimized for adults; (ii) exposure to synthetic and allergenic materials; (iii) nonspecific removal of both harmful and protective antibodies, increasing the risk of life-threatening infections; and (iv) an extracorporeal volume requirement ranging from 165 to 280 ml (averaging 170 ml) (17). While this volume represents a negligible ~6% of total blood volume for adults, it represents more than 50% of the total blood volume of an infant or small child (18, 19). Hence, when conventional apheresis devices are applied toward children, they must be primed with blood, which often increases antibody production and sensitization (20). These challenges highlight the urgent need for pediatric-specific apheresis technologies that address the specific physiological and immunological requirements of children.

Contemporary plasmapheresis methods not only achieve antibody removal rates of ~20 to 50% but also indiscriminately deplete beneficial antibodies, essential proteins, and other critical components such as exosomes from the blood (21). A device capable of removing DSA without requiring a blood prime or depleting beneficial antibodies would represent a transformative advancement in pediatric organ transplantation. This device would specifically lower DSA levels, reduce antibody-mediated rejection in patients, and improve overall patient outcomes. Moreover, this device could advance the field of apheresis by facilitating preclinical studies in animal models of antibody-mediated rejection, which are currently limited by the large blood volumes required for conventional apheresis systems (22).

Acoustofluidic devices (23–34) are excellent candidates for addressing these challenges for several reasons. First, acoustofluidic devices are highly biocompatible (35–45), operating at power intensities and frequencies comparable to ultrasonic imaging, a widely accepted and safe medical technology. Second, acoustofluidic devices excel at separating particles suspended within fluids with high precision and specificity. This enables the integration of antigen-coated beads into an acoustofluidic system that can facilitate the selective capture and removal of DSA from the whole blood. Third, acoustofluidic devices are versatile, capable of handling both large volumes and small volumes of fluids, including microliter-scale samples for small-volume therapeutic apheresis. Last, acoustofluidic solutions eliminate the need for filters, columns, or costly reagents, reducing both material expenses and the risk of allergic reactions associated with traditional separation methods.

Therefore, we developed and evaluated an acoustofluidic apheresis device that uses antigen-coated beads to selectively remove DSA in a sensitized rodent skin transplantation model. To our knowledge, the device, called the acoustofluidic-based system for targeted antibody removal in transplantation (A-START), represents the first apheresis technology capable of selective removal of harmful antibodies and is suitable for preclinical animal studies with potential scalability in pediatric care. By selectively removing DSA while maintaining endogenous immunity, A-START can reduce the risk of life-threatening infections, a major complication with conventional apheresis methods, especially in children with naïve immune systems (46–48). With its innovative capabilities, A-START may offer a promising platform for advancing research into pediatric sensitization, antibody-mediated rejection, and other antibody-related disorders.

RESULTS

Schematic of targeted DSA removal with A-START compared to conventional apheresis

Figure 1 illustrates the apheresis strategy for selective DSA removal using A-START. A fully mismatched MHC I and MHC II skin transplantation was performed in a rodent model to mirror the process of sensitization in which a recipient develops high levels of DSA (Fig. 1A). These DSAs circulate in the recipient’s bloodstream, bind to the skin graft endothelium, and mediate rejection, ultimately leading to allograft failure (Fig. 1A). Unlike conventional apheresis systems, A-START incorporates donor-specific antibody–targeting cell-bead complexes (DSA-trapping beads) to selectively remove DSA from recipient blood while preserving protective antibodies against infections. The system uses an A-START chip designed for precise and stable handling of whole blood and buffer fluids, enabling efficient blood component separation.

Fig. 1. Schematic illustration of mechanisms associated with the development of DSA and selective DSA reduction with the A-START.

(A) Four primary events including pregnancy, recurrent infection, prior transplantation, and blood transfusions can contribute to the development of DSA, which binds to the graft endothelium and triggers immune-mediated endothelial injury. The complement system is activated, and antibody-dependent cell-mediated cytotoxicity involving macrophages, natural killer (NK) cells, and neutrophils occurs, leading to thrombosis, vasoconstriction, and ischemia of the endothelium. (B) Conventional commercial apheresis machines nonselectively remove plasma and its components, including harmful DSAs as well as beneficial immunoglobulin G (IgG), IgM, and exosomes. Furthermore, their operating volume is ~170 ml, making them unsuitable for small animals. In contrast, the A-START system can selectively remove DSA from protective antibodies in whole blood, requiring an extracorporeal volume of only 240 μl. (B) Created in BioRender. Ma, Z. (2025) https://BioRender.com/xchp4a2.

Overall, the A-START consists of two main components: a portable instrument and a modular acoustofluidic chip that are detailed in fig. S1. The portable instrument integrates peristaltic pumps, a power supply, and a temperature controller to support stable extracorporeal circulation and on-chip thermal regulation. The acoustofluidic chip is composed of three key units: (i) a fluid stabilizer unit, (ii) a blood separation unit, and (iii) a targeted antibody removal unit. Peristaltic pumps initially facilitate extracorporeal blood circulation (49–50). To maintain a stable input flow rate, we integrate fluid stabilizers into the circulation system, each containing four cavities connected to the main fluid microchannel (fig. S2). Within the four cavities, a dynamic equilibrium between fluid flow and air compression is established. As the fluid’s flow rate and pressure increase, the liquid level in the overlying channel rises, compressing the air within the microchannel. Conversely, when the flow rate and pressure decrease, the liquid level falls, and the air decompresses. In this context, the fluid stabilizers act as a capacitor, storing excess blood flow during periods of high pressures while providing excess blood flow during periods of low pressure. This mechanism ensures a consistent flow rate within the microfluidic channel, even in the presence of systolic pressure fluctuations or sudden pressure changes.

The blood separation unit of A-START uses surface acoustic waves (SAWs) generated by interdigitated transducers (IDTs) to create an acoustic radiation force that selectively separates red blood cells (RBCs), white blood cells (WBCs), and platelets, which are then returned to the body. The remaining plasma flows into the targeted antibody removal unit, where specially engineered DSA-trapping beads—constructed using B and T cells isolated from donor rat spleens—capture and remove DSA with high specificity. Unlike conventional apheresis techniques that indiscriminately remove all antibodies (Fig. 1B), this approach preserves non-DSA antibodies. The DSA-trapping beads, along with the captured DSA, are then separated from the plasma using SAWs, effectively removing DSA while ensuring that the purified plasma can be collected for characterization and returned to the recipient.

The A-START system (Fig. 1B) integrates these three units to enable efficient, selective blood component separation and targeted antibody removal. In particular, it uses ladder-shaped IDTs (LSIDTs) that uniquely enable dual–standing wave generation, efficiently reflecting acoustic waves from the central IDT to the side IDTs. This design ensures that the microfluidic channel experiences four successive standing wave fields, substantially enhancing blood separation efficiency and throughput.

Rat sensitization, tetanus toxoid inoculation, and DSA-trapping bead optimization

A rat model of sensitization with DSA (Fig. 2A) was established by transplanting fully MHC I– and MHC II–mismatched dorsal skin allografts from a dark agouti (DA) donor rat onto a Lewis recipient rat. We performed two sequential skin grafts on postoperative days (POD) 0 and 56, respectively (Fig. 2, A and B). Protective antibodies were modeled in the same system by inoculating the recipient rates with tetanus toxoid (TT) to generate TT antibodies (Fig. 2A). Successful engraftment status post–skin transplantation was shown on POD 5 (Fig. 2B). TT antibodies were stimulated in the Lewis recipient rat by injection of TT on POD 56 (Fig. 2A). To demonstrate the persistent elevation of MHC I and MHC II DSA in untreated, sensitized recipient rat blood, we measured DSA concentrations with flow cytometry at PODs 70, 84, 98, 112, and 126. After 56 days, the concentration of DSA in the blood of Lewis recipient rats in the absence of treatment remained consistently high (Fig. 2C). As a negative control to confirm the specificity of the DSA-trapping beads, the concentration of anti-TT immunoglobulin G (IgG) was also measured and found to remain unaffected by treatment with the DSA-trapping beads (Fig. 2, D and E).

Fig. 2. Creation of T and B cell capture beads and successful removal of DSA but not TT antibody from a sensitized rat.

(A) Schematic describing our sensitized rat model with elevated DSA and TT antibody in vivo. (B) Photographic series of a rat following skin transplantation surgery: The left image shows the graft on the day of the transplant. In contrast, the right image shows the graft 5 days postoperatively, after the rejection of the graft has begun. (C) All rats were successfully sensitized by day 70, with persistently elevated DSA levels without A-START (n = 3). (D) Recipient plasma (400 μl) from naïve and sensitized rats is mixed with DSA-trapping beads. These are specific for DSA and do not bind TT antibodies. (E) Anti-TT IgG levels are unaffected by the addition of DSA-trapping beads, and high levels persist despite the addition of DSA-trapping beads (n = 3). (F and G) Dose optimization experiments for (F) MHC I and (G) MHC II show that a plasma concentration of 10⁷ beads/100 μl is most effective in removing DSA. (H) The addition of plasma with 10⁷ DSA-trapping beads/100 μl reduces MHC I DSA by 95%, from 25,000 to 3000 mean fluorescence intensity (MFI), and reduces MHC II DSA by 80%, from 20,000 to 5000 MFI, in sensitized rat recipient plasma ex vivo after 30 min. [(A) and (B)] Created in BioRender. Ma, Z. (2025) https://BioRender.com/xchp4a2.

DSA-trapping beads were generated by isolating B and T cells from DA splenocytes using anti-CD45RA and anti-CD3 antibodies and affixing them to streptavidin-labeled Dynabeads (Fig. 2D). B cells have MHC I and MHC II antigens on their surface, while T cells have only MHC I antigens. Bead concentration was titrated (1 × 10⁶ to 1 × 10⁷) against rat plasma volume to optimize for DSA removal. To confirm the specificity of DSA-trapping beads to the donor cells, we also used splenocytes and streptavidin-labeled Dynabeads as capture antigens (fig. S3). We confirmed that DA splenocytes and DSA-trapping beads reduced DSA in plasma similar to levels without changing TT antibody levels (fig. S4). We demonstrated that treating plasma with 1 × 10⁷ DSA-trapping beads/100 μl was most effective in reducing DSA concentrations, with an optimal incubation time of 30 min for maximal reduction (Fig. 2, F and G). After 30 min of treatment with 1 × 10⁷ DSA-trapping beads/100 μl, MHC I and MHC II DSA levels decreased by ~95 and 80%, respectively (Fig. 2H).

Safety, efficacy, and biocompatibility of blood separation via A-START

The A-START system integrates the peristaltic pump, polystyrene tubing (inside diameter, 0.28 mm), cooling plate, temperature controller, and a 90 mm–by–35 mm A-START chip into a 28 cm–by–22 cm desktop unit. In total, the device requires an extracorporeal volume of just 240 μl, a substantial reduction from conventional apheresis machines. The functionality of the A-START chip was demonstrated through precise particle manipulation in x-y planes under acoustic activation, as evidenced by the controlled displacement and pressure fields (fig. S5).

To evaluate the performance of the blood separation unit in the A-START system, we introduced the blood from a sensitized rat into the A-START chip (Fig. 3A). The soluble antibodies present in plasma were quantitatively assessed using enzyme-linked immunosorbent assay (ELISA). RBC, WBC, and platelet counts were concurrently quantified using flow cytometry. Figure 3 (B and C) depicts a simulation of two IDTs positioned on opposite sides of the dual-channel region generating SAWs at 39.8 MHz. The LSIDTs uniquely enable bidirectional acoustic wave radiation from the central IDT to the side IDTs, ensuring that standing waves form on both sides. With the microfluidic channel precisely routed around the central IDT, the sample undergoes four consecutive standing wave interactions, markedly improving separation efficiency and throughput compared to conventional IDT designs. This standing wave field on the substrate leaks into the fluid within the dual channels, creating a three-dimensional (3D) acoustic standing wave pattern. This pattern exerts acoustic radiation forces that can trap particles and blood cells, enabling the separation of blood cells from flowing plasma. Bright-field imaging of the blood separation unit’s outlet (Fig. 3D) shows differential cell displacement toward the top outlet, both in the absence and presence of an acoustic field. Meanwhile, the plasma (here containing 100-nm nanoparticles to aid in imaging) is unaffected by the acoustic field (fluorescent images).

Fig. 3. Safety, efficacy, and biocompatibility of plasma separation from cellular blood components using the A-START.

(A) Schematic of the A-START separation process with an emphasis on the measurements obtained to characterize the performance of the blood separation unit. (B) Schematic and simulation of the LSIDT system generating 39.8-MHz SAWs. (C) Visualization of the z-directional displacement, acoustic pressure field, and distribution of 10-μm particles in the y-z plane, shown with acoustics off (top) and on (bottom). (D) Comparison of plasma and cell flow in A-START under acoustic off and on conditions. (E) Cell recovery is significantly higher from the cell outlet postpurification (n = 6). (F) Comparison of live-cell fluorescence values after A-START processing at various power levels versus control and centrifugation (n = 6). (G to I) Distribution of cellular and soluble components in the samples collected from cell outlet and plasma outlet of the A-START when acoustics are on (n = 3). (J to L) DSA (MHC I) and DSA (MHC I and MHC II) level testing for the sample collected from the plasma outlet and cell outlet of the blood processor portion of A-START. (M and N) Performance of separation at varied conditions regarding hemoglobin (HGB) and P-selection level of the outputs from cell outlet comparison with centrifugation (n = 3). (O) D-dimer levels after A-START processing at different power levels, control, and centrifugation showed no significant differences, indicating stable coagulation status post–A-START (n = 3). (P and Q) Characterization of the RBC and platelet (PLT) recovery rate after processing by the A-START under varied conditions of flow rate and different input voltages (n = 3). (R to U) Flow cytometry data show the number of blood cells and platelets in the sample collected from the A-START’s cell outlet (left) and plasma outlet (right). (C) Created in BioRender. Ma, Z. (2025) https://BioRender.com/xchp4a2.

To test the A-START system’s enrichment performance for both micro- and nanosized particles, we injected three types of polystyrene particles (d = 5.0 μm, 1.9 μm, and 100 nm) into the microchannel. As shown in fig. S6, the A-START system has a substantial enrichment effect on the 5.0- and 1.9-μm particles, but no enrichment effect on the 100-nm particles. Our experimental results indicate that LSIDTs can effectively aggregate large particles such as RBCs (~5 μm) while having a minimal effect on small moieties (<100 nm), including antibodies. Therefore, the A-START system can efficiently separate RBCs, WBCs, and platelets from plasma. Cellular analysis showed that 87% of RBCs were sorted into the cell outlet, with only 13% remaining in the plasma component (P = 0.0001) (Fig. 3E). Figure 3F indicates that the A-START system maintains high cell viability across various power levels, significantly higher viability than the control and centrifugation group, demonstrating the biocompatibility of the A-START system. Meanwhile, most (72, 77, and 64%, respectively) of IgG, IgM, and TT antibodies were retained in the plasma component (P = 0.0018, P = 0.001, and P = 0.0001) (Fig. 3, G to I). In addition, blood gas parameters remained stable after A-START processing, demonstrating its biocompatibility when compared to methods such as centrifugation (fig. S7).

Soluble component analysis showed that DSAs were sorted into the plasma outlet at a higher concentration than into the cell outlet, with 63.2% of MHC I and MHC II DSA recovered in the plasma outlet (P = 0.05) (Fig. 3, J to L). Regarding safety analysis, the concentration of free hemoglobin, which reflects RBC hemolysis, was similar between blood samples that had undergone A-START and unprocessed samples of whole blood, reflecting minimal RBC breakdown (P = 0.131). In contrast, free hemoglobin concentration was increased in samples separated by centrifugation compared to unprocessed whole blood (P = 0.013) (Fig. 3M). P-selectin concentration, a reflection of platelet activation (51), was lowest in whole blood preseparation and blood separated by A-START (P = 0.243) and significantly higher in blood separated by centrifugation (P = 0.012), suggesting significantly less platelet activation with the A-START technique (Fig. 3N). The D-dimer assay results show no significant changes in coagulation indices after A-START processing, confirming that the A-START process does not affect blood coagulation status (Fig. 3O). RBC and platelet recovery varied inversely with flow rate and directly with input voltage to the A-START chip but remained >80% for flow rates as high as 80 μl/min at 48.7 dBm (Fig. 3, P and Q). Overall, blood separated by A-START had a similar safety profile (minimal RBC hemolysis and platelet activation) compared to unprocessed blood, while blood separated by centrifugation showed an increase in both RBC hemolysis and platelet activation. A comprehensive flow cytometry profile is presented in Fig. 3 (R to U). A comparison of the plasma and cell fractions shows that RBCs and WBCs are predominantly located in the cell fraction (Fig. 3, R and S). A CD61 fluorescence assay confirmed that most of the constituents within the platelet group were intact platelets and captured in the plasma fraction (Fig. 3, T and U) (SSC-A, side scatter area; FSC-A, forward scatter area; PE-A, phycoerythrin area). These data demonstrate that the A-START chip effectively separates cellular components from plasma while minimizing damage to cells and other bioparticles.

Targeted DSA removal from sensitized rat plasma via A-START

Figure 4A illustrates the superior performance of the A-START system across three key metrics: DSA removal rate, recovery rate for beneficial antibodies, and required blood volume in extracorporeal circulation. Unlike conventional apheresis methods including those by Cervantes et al. (52), Hirata et al. (53), Francey and Schweighauser (54), and Tagawa et al. (21), which require substantial blood volumes (71 to 180 ml) and often compromise recovery rate for beneficial antibodies, A-START achieves near-complete protective and endogenous antibody recovery (~95%) with an extracorporeal volume of only 240 μl, while accomplishing a >50% reduction in DSA mean fluorescence intensity (MFI). These results underscore A-START’s potential for safer and efficient antibody removal, particularly suitable for pediatric and small-volume applications where minimizing blood loss and preserving immunity are paramount.

Fig. 4. Targeted DSA removal from sensitized rat plasma via donor-specific antibody–targeting cell-bead complexes (DSA-trapping beads) incorporated within the A-START.

(A) 3D plot illustrating A-START’s superior DSA removal, high antibody recovery, and minimal blood volume requirement compared to conventional plasmapheresis techniques, along with a comparative analysis of A-START against other apparatus reported in the literature. (i) (52), (ii) (53), (iii) (54), and (iv) (21). (B and C) Overview of targeted MHC I/II DSA removal using DSA-trapping beads in the A-START unit. T and B cells were isolated from DA rat spleens by mechanical filtration. Biotinylated anti-CD45RA and anti-CD3 were conjugated to streptavidin nanobeads to generate DSA-trapping beads, which selectively bound and removed DSA from circulation. (D) Simulation results show distinct displacement patterns for proteins, platelets, RBCs, and DSA-trapping (DSAT) beads, highlighting the varying effects of acoustic radiation force and fluid flow on particles of different sizes. (E to G) The concentration of IgG, IgM, and TT was measured using ELISA both before acoustofluidic apheresis and in the plasma and cell fractions after DSA separation (n = 3). (H) As measured by NTA, particle size distribution shows that after the removal of DSA-trapping bead/DSA complexes, the waste contains minimal other plasma substances (n = 3). (I) DSA-trapping beads concentration was measured before A-START apheresis and for samples collected from the plasma and cell outlets of the A-START (n = 3). (J and K) The concentration of MHC I and MHC II for samples collected from preapheresis, the return outlet, and the waste outlet (n = 3). (L) The A-START system achieves high recovery rates for IgG, IgM, and TT antibodies while selectively removing MHC antibodies (n = 3). (B and C) Created in BioRender. Ma, Z. (2025) https://BioRender.com/xchp4a2.

Figure 4B demonstrates the manufacturing process of the DSA-trapping beads and their mechanism of binding with DSAs. After separation from whole blood within the A-START blood separation unit, plasma from sensitized Lewis rats—containing both DSA and TT antibodies—entered the target antibody removal unit module, where DSA-trapping beads are introduced (Fig. 4C). Simulations reveal that different particle types experience unique displacement patterns within the A-START channel, influenced by their size and response to acoustic radiation force. Proteins (diameter used in simulation, 100 nm) remain largely unaffected, maintaining a stable y-axis position, while RBCs (diameter used in simulation, 6 μm) and DSA-trapping beads (diameter used in simulation, 2 μm) exhibit greater displacement along the x axis, indicating effective separation under acoustic forces. These results confirm the A-START system’s capability to selectively manipulate particles, critical for targeted antibody removal in biomedical applications (Fig. 4D). When compared to baseline (sensitized plasma before treatment), plasma processed by the A-START had a significantly higher concentration of native rat IgG, IgM, and anti-TT antibodies in the return fraction (P = 0.563, P = 0.28, and P = 0.58 respectively) (Fig. 4, E to G). Conversely, these antibodies were rarely found in the waste output—the concentrations of IgG, IgM, and anti-TT antibodies were significantly lower in the waste fraction compared to untreated plasma (P = 0.0026, P < 0.0001, and P = 0.0018 respectively) (Fig. 4, E to G). Overall, the targeted antibody removal unit with incorporated DSA-trapping beads effectively separated DSA from plasma and returned protective IgG, IgM, and anti-TT antibodies to the subject.

The size distribution of the plasma before A-START process, the waste fraction, and the return fraction were evaluated using nanoparticle tracking analysis (NTA) (Fig. 4H). Following the separation of DSA-trapping bead/DSA complexes from plasma, the NTA profile of the return fraction closely matched that of the preapheresis sample, indicating minimal disruption to plasma composition. In contrast, the waste fraction exhibited a marked absence of plasma components, confirming the selective removal of DSA-trapping bead/DSA complexes.

A-START was also used to separate DSA-trapping bead/DSA complexes from DSA-free plasma. After processing, we found that the DSA-trapping beads were primarily found in the waste fraction versus the return (plasma) fraction (P = 0.001; Fig. 4I); a gating strategy for three different DSA-trapping beads samples is shown in fig. S8. The scanning electron microscopy (SEM) images of DSA-trapping bead/DSA complexes are found in fig. S9.

Overall, treatment with A-START reduced both MHC I and MHC II DSA, as demonstrated in Fig. 4 (J and K), as DSA concentration in the return fraction was significantly lower compared to the preseparation sample. When compared to both baseline (sensitized plasma before treatment) and pretreated plasma, plasma treated by A-START had markedly lower MHC I and MHC II DSA in the return fraction, with a clinically meaningful reduction in MFI by >50% (P = 0.045 and P = 0.025) (Fig. 4K). Figure 4L highlights the selective antibody processing capability of the A-START system, which effectively recovers beneficial antibodies (IgG, IgM, and TT) (~93.8, 92.7, and 85.3%, respectively) while selectively removing MHC antibodies (~32.1%). These results underscore the system’s ability to preserve protective immunity while effectively eliminating harmful DSAs. Furthermore, transmission electron microscopy (TEM) images of plasma pre- and postpurification confirmed the preservation of intact exosomes (fig. S10), demonstrating the biocompatibility of A-START with plasma nanovesicles.

Collectively, these findings indicate that the composition of the purified (return) plasma fraction closely mirrors that of the original (unpurified) plasma, while the waste fraction contains significantly fewer beneficial native protein and nanovesicles. This selective processing capability highlights A-START’s potential to minimize immune compromise and maintain plasma integrity during antibody removal.

DISCUSSION

In this work, we developed A-START, an apheresis system designed for small animal models with potential translation toward pediatric applications. Leveraging microfabrication and nanofabrication methods, we downsized apheresis systems into a compact chip capable of processing blood volumes as low as 240 μl with high precision and efficiency. The integration of an advanced acoustic transducer (i.e., LSIDT), which generates enhanced acoustic radiation forces, enables rapid and efficient separation of blood components. To demonstrate its effectiveness, we used a sensitized rat skin transplantation model to selectively remove DSA while preserving beneficial native immunity, as evidenced by the retention of anti-TT antibodies. A-START represents a substantial advancement in solid organ transplantation by providing targeted removal of DSA and can be the foundation of standardizing methods for selective antibody elimination under other childhood conditions such as autoimmune antibody–mediated diseases. This device has broad and transformative potential for both biological research and clinical applications, allowing researchers to explore apheresis systems using small animal models that can ultimately mature into life-saving pediatric interventions.

Currently, traditional plasmapheresis devices suffer from critical limitations, including the nonspecific removal of harmful and protective antibodies from blood (55, 56). This indiscriminate removal means that these conventional techniques can often further compromise immunity, particularly in immunosuppressed recipients. In addition, because of its sizeable extracorporeal volume, traditional plasmapheresis equipment is unsuitable for pediatric populations without notable modifications (57). For pediatric patients, the extracorporeal volume can represent a substantial proportion of their total blood volume, leading to hemodynamic instability and intolerance to the procedure (58). Unlike conventional systems, A-START is uniquely engineered to selectively remove donor-specific antibodies while retaining a broad spectrum of beneficial antibodies, proteins, exosomes, and vital small molecules in whole blood. This selective removal critically preserves key immune components, minimizing the risk of immune compromise reducing the risks associated with traditional apheresis. Furthermore, this device requires just 240 μl of extracorporeal volume, a stark contrast to the much larger volumes (~170 ml) needed for conventional plasma exchange. This substantial volume reduction minimizes the procedure’s invasiveness for small animal models and pediatric patients and reduces associated risks such as hypotension and the need for blood transfusion, which can induce further sensitization in transplant recipients.

Future work will focus on optimizing the blood separation unit to enhance the separation efficiency between antibodies and cells. This involves improving the chip design, working conditions, and acoustic transducer design to generate stronger acoustic radiation forces. In addition, we will continue verification studies to understand the effects of repeated in vivo acoustofluidic treatments using the A-START device. Last, further tests will demonstrate the efficacy of this device in large animals such as piglets and nonhuman primates to demonstrate the scalability and precision of A-START, paving the way for clinical translation of A-START toward pediatric populations.

This work provides a compelling proof of concept for the use of acoustofluidics in the precise removal of DSA from small volumes of blood, offering a previously unexplored approach to addressing antibody-mediated diseases in pediatric populations. We have used a sensitized rat model with elevated DSA levels to mirror the sensitization observed in humans and shown that A-START is a potentially viable bedside acoustofluidic apheresis device. This system not only holds promise for reducing rejection and maintaining protective immunity in pediatric transplant recipients but also provides a robust platform for preclinical testing of apheresis adjuncts. Thus, a scaled version of this technology could substantially improve patient outcomes for children suffering from antibody-mediated rejection and autoimmune diseases, while also enabling further research to expand its therapeutic applications.

MATERIALS AND METHODS

Chip fabrication

The A-START chip is composed of three units. The first is a fluid stabilizer unit that converts the cyclic flow driven by the peristaltic pump into stable laminar flow within the device. The second is a blood separation unit that separates and returns the cellular components of whole blood. Last, the targeted antibody removal unit selectively isolates and removes DSA. The fluid stabilizer unit was fabricated using soft lithography techniques and comprises four polydimethylsiloxane (PDMS; Dow Corning, USA) sheets: The top and bottom sheets are unaltered, while the middle two sheets contain microfluidic channels for fluid flow and through-layer cavities that maintain a relatively stable output pressure (fig. S2). Briefly, the microchannels were patterned with a positive SU8 photoresist mold. Inlet and outlet holes were added using a hole punch (69039-10, Electron Microscopy Sciences), and the PDMS sheets were plasma cleaned and assembled. The blood separation unit and targeted antibody removal unit were fabricated using similar techniques. Each unit contains a piezoelectric substrate (Y + 128° X-propagation LiNbO₃) and a PDMS microfluidic channel. Briefly, electrodes were patterned on the piezoelectric subcomponent via electron beam evaporation, photolithography, and liftoff. The substrate and PDMS subcomponents were then plasma cleaned and bonded. The fabrication process and final device schematic are shown in fig. S11.

Configuration of the A-START

Our A-START system contains peristaltic pumps, a power supply, a temperature controller, a cooling plate, and other electronic accessories. The P625 peristaltic pumps (Instech Laboratories Inc., Plymouth Meeting, PA, USA) convey samples to the corresponding chip inlets via plastic tubing (Smith Medical International, USA). To prevent overheating, the A-START chip rests on a CP-031 cold plate (TE Technology Inc., Traverse City, MI, USA) maintained at 10°C with a TC-48-20 pulse-width modulation temperature controller (TE Technology Inc., Traverse City, MI, USA). The electronic components are powered at 60 peak-to-peak voltage with a PS-12-8.4A dc power supply (TE Technology Inc., Traverse City, MI, USA). A 39.8-MHz acoustic signal was generated using a function generator (E4422B, Agilent, USA) and amplifier (100A250A, Amplifier Research, USA).

Rat sensitization

Lewis rats (Charles River Laboratory, Wilmington, MA) were sensitized through sequential skin allograft transplantation from DA donors (Charles River Laboratory, Wilmington, MA) to generate sustained high titer DSA levels to simulate a highly sensitized environment (59–60). All rats’ experiments were performed in accordance with the guidelines approved by the Institutional Animal Care and Use Committee of Duke University, in compliance with all relevant ethical regulations. Briefly, a 4 inch–by–4 inch (10.16 cm–by–10.16 cm) skin graft was procured from the posterior skin of each DA rat. The skin flap was then divided into four 2 inch–by–2 inch (5.08 cm–by–5.08 cm) segments, and each segment of the divided donor graft was transplanted onto one Lewis recipient rat. Four weeks posttransplantation, each Lewis rat was injected with 125 μg of TT subcutaneously. To allow recipient sensitization, rats were euthanized, and blood was collected 56 days after tetanus injection.

A-START separation of whole blood

Whole blood was collected humanely from rats under general anesthesia with isoflurane. Briefly, a midline incision was made, and the intrahepatic inferior vena cava (IHIVC) was dissected and exposed. Using a 16-gauge needle, 10 ml of blood was collected from the IHIVC and immediately transferred into a 15-ml tube containing 0.5 ml of heparin. The blood sample was then immediately run through the A-START device, and 2 ml of fluid was collected at each outlet (cellular fraction and plasma fraction) of the device. RBCs in the cellular fraction were quantified using a TC-20 automatic cell counter (Bio-Rad, USA). Meanwhile, the plasma fraction was stored at −80°C for further analysis. Figure S12 explains the principles behind the proposed LSIDTs in the A-START chip. In the spiral channel, inertial focusing will affect the distribution of blood cells within the channel (61). Figure S13A demonstrates the impact of inertial focus on particles in microchannels. Our simulation results, as shown in fig. S13 (B and C), indicate that the lift force substantially exceeds the drag force in the channel of the microparticle separator. Figure S14 shows the distribution of particles of different sizes at various flow rates in the microchannels, indicating that the spiral channel has minimal impact on the distribution of particles. Therefore, inertial focusing does not adversely affect the separation efficiency of our A-START chip.

Synthesis of DSA-trapping beads

To demonstrate the isolation of DSA from plasma using DSA-trapping beads, the spleen was first isolated from a naïve DA rat. After induction of general anesthesia, a midline incision was made, and the intestines were retracted to expose the spleen. The splenic artery and vein were cut to remove the spleen and exsanguinate the animal. The spleen was then mechanically digested through a 100-μm cell filter in 30 ml of high-yield fixative-free lysis buffer (Thermo Fisher Scientific, Waltham, MA, USA) for 10 min to lyse RBCs. The digest was then centrifuged at 300g for 5 min at room temperature. The supernatant was removed, and the pellet was resuspended in phosphate-buffered saline (PBS) at a concentration of 10⁷ cells/100 μl. Next, 2 μl each of biotin-labeled anti-rat CD45RA and biotin-labeled anti-rat CD3 were incubated with 100 μl of the splenocyte cell suspension for 15 min at 4°C. This suspension was then incubated with 25 μl of M-270 streptavidin-labeled Dynabeads (Thermo Fisher Scientific, Waltham, MA, USA) for 15 min to generate bead-cell complexes (DSA-trapping beads; Fig. 4A). The DSA-trapping beads were then washed with 2 ml of Mojo sort buffer (BioLegend Inc., San Diego, CA, USA) and magnetically isolated for 5 min. Last, the supernatant was removed, and the DSA-trapping beads were resuspended in 100 μl of PBS.

Characterization of DSA-trapping beads binding to DSA in plasma

Figure S15 is the gating strategy for three different samples. To understand the time required for DSA-trapping beads binding, whole blood was collected from sensitized and naïve Lewis rats as described above (Fig. 2C) in EDTA purple top tubes. The samples were then centrifuged at 3500 rpm for 10 min at 4°C, and the supernatant (plasma) recovered. DSA-trapping beads were added to the sample, and aliquots were serially collected using a magnetic stand at t = 0, 5, 15, 30, and 60 min. Flow cytometry was then used to measure the DSA concentration in each 100-μl aliquot. To evaluate how bead concentration influences binding efficiency, plasma collected from sensitized rats was incubated with different amounts of DSA-trapping beads (20 μl per sample) for 1 hour. This allowed assessment of the beads’ capacity to bind MHC I and MHC II DSA. To evaluate the specificity of DSA-trapping beads, anti-TT concentration was measured in treated (DSA-trapping bead+) samples and untreated (bead-free) controls by ELISA.

A-START plasma separation using DSA-trapping beads

Assay by flow cytometry demonstrates specific removal of DSA from plasma via DSA-trapping beads binding and acoustic separation (fig. S16). The whole blood was collected from sensitized Lewis rats as described above in EDTA purple top tubes and centrifuged at 3500 rpm for 10 min at 4°C. The supernatant (plasma) was recovered and combined with 100 μl of 10⁸/ml of DSA-trapping beads per 4000 μl of sensitized plasma (n = 3) for 1 hour. The DSA-trapping bead–treated plasma was then fractionated using the A-START device, and 2 ml of fluid was collected at each output, corresponding to the return and waste fractions (fig. S17).

Antibody quantification

IgG, IgM, and TT were quantified after separation of plasma from whole blood using the blood separation unit and after purification of DSA from the plasma using DSA-trapping beads with the target antibody removal unit. IgM was quantified using an IgM rat ELISA kit (Abcam Plc., Boston, MA, USA); IgG was quantified using an IgG rat ELISA kit (Abcam Plc., Boston, MA, USA); and anti-TT antibody was quantified using an IgG rat anti-TT kit (Innovative Research Inc., Novi, MI, USA). The concentration of each antibody in plasma was determined using optical density measured by Synergy H1 Microplate Reader (BioTek Instruments, Winooski, VT, USA). Because of the difference in dilution factors before and after treatment in Fig. 4, when calculating the dilution factor for ELISA, the pretreatment samples are multiplied by 1.5 times compared to the posttreatment samples.

Donor-specific antibody quantification

B and T cell flow cytometry cross-matches were performed using plasma samples and donor splenocytes. Fluorescently labeled anti-CD3 allophycocyanin (Thermo Fisher Scientific, Waltham, MA, USA), anti-CD45R/B220 Cy7 phycoerythrin (BD Biosciences), and anti-rat IgG fluorescein isothiocyanate (BioLegend Inc., San Diego, CA, USA) were used to stain and quantify T lymphocytes, B lymphocytes, and MHC interactions, respectively. The viability of T and B lymphocytes was determined by live/dead staining using the Fixable Blue Dead Cell Stain Kit (Thermo Fisher Scientific, Waltham, MA, USA). A FACSCaliber Cytometer (BD Biosciences, San Jose, CA, USA) was used to acquire 10,000 events from each sample. Data analysis was performed with FlowJo software (BD Life Sciences, Franklin Lakes, NJ, USA).

Electron microscopy analysis

To examine the morphology of the separated components, samples were collected from two different parts of the targeted antibody removal unit: the return fraction and waste fraction. These samples were then analyzed using SEM (Thermo Fisher Scientific, Apreo S, Waltham, MA, USA) and TEM (FEI Tecnai G² Twin). The samples collected from the beads outlet and plasma outlet were prepared by initially coating them on a glass cover for SEM analysis and on a grid for TEM analysis. The samples were fixed in 4% glutaraldehyde [prepared in 10 mM PBS (pH 7.4)] for 1 hour at 4°C. After fixation, they were washed three times with PBS (10 mM) (pH 7.4), followed by another three washes with distilled water. Each washing step lasted ~15 min. Dehydration was performed through a graded ethanol series (70, 80, 90, and 100%), with each step maintained for 15 min. After dehydration, the samples were left in a fume hood overnight to ensure complete drying. For SEM analysis, the final step involved coating the sample with a 5-nm-thick layer of gold. This coating was applied using a Denton Desk V sputter coater to enhance electron conductivity and imaging quality.

Numerical simulation

Finite element simulations were performed using COMSOL Multiphysics to investigate the 3D acoustic field distributions and the resulting acoustic radiation forces in fluid channels. The pressure acoustics, solid mechanics, and electrostatics modules were used to model the 3D acoustic field within the fluid domain, induced by SAWs propagating on a lithium niobate substrate. These SAWs were generated by applying voltages to the IDTs. To quantify the forces acting on particles, the acoustic Gor’kov potential was first computed from the simulated pressure field p and acoustic velocity field $\mathbf{v}$ ( $v_x$ , $v_y$ ) via

$$U=V_{\mathrm{p}}(\frac{f_1}{2{\mathrm{\rho }}_0{c}_0^2}\langle p·p^{\ast }\rangle -\frac{3}{4}{\mathrm{\rho }}_0{f}_2\langle \mathbf{v}·{\mathbf{v}}^{\ast }\rangle )$$

where $f_1=1-{\mathrm{\rho }}_0{c}_0^2/{\mathrm{\rho }}_{\mathrm{p}}{c}_{\mathrm{p}}^2$ and $f_2=2({\mathrm{\rho }}_{\mathrm{p}}-{\mathrm{\rho }}_0)/(2{\mathrm{\rho }}_{\mathrm{p}}+{\mathrm{\rho }}_0)$ . The parameters used were ${\mathrm{\rho }}_0=1$ g/cm³ and $c_0$ = 1500 m/s for water, and ${\mathrm{\rho }}_{\mathrm{p}}$ = 1.04 g/cm³ and $c_{\mathrm{p}}$ = 2320 m/s for particles. Here, V_p represents the volume of the particle considered in the simulation, and ⟨·⟩ denotes the time-averaging operator. Subsequently, the laminar flow module, along with a stationary solver, was used to compute the resultant acoustic streaming field based on the simulated pressure and velocity distributions. Last, the particle tracing module, in conjunction with a time-dependent solver, was used to track particle trajectories over time. A total of 50,000 particles were randomly released into the channel, and their motions were governed by the computed drag and acoustic radiation forces obtained from the Gor’kov potential. To ensure numerical stability and convergence, a time step of 1 ms and a total simulation time of 2 s were used, with a relative tolerance of 0.001.

Image recording

Fluorescence and bright-field images were acquired using an upright microscope (Olympus Corporation, BX51WI, Hachioji, Tokyo, Japan) and a charge-coupled device camera (Photometrics, CoolSNAP HQ2, Tucson, AZ, USA). Images were then processed using ImageJ (National Institutes of Health, USA).

Statistical analysis

Statistical comparisons were conducted using unpaired t tests or one-way analysis of variance (ANOVA), followed by Tukey’s multiple comparison test in GraphPad Prism version 10.0. Results are expressed as means ± SD. Significance was determined at *P < 0.05, **P < 0.01, and ***P < 0.001.

Use of artificial intelligence

Portions of this manuscript were refined with assistance from OpenAI’s ChatGPT (GPT-4). Prompts used included “Improve text conciseness” and “Enhance clarity.” All AI-assisted edits were critically reviewed, revised, and approved by the authors.

Supplementary Materials

This PDF file includes:

Figs. S1 to S17