Mechanically mediated cargo delivery to cells using microfluidic devices

Drug delivery technologies aim to actively or passively transport or release drugs to the target cells, thereby minimizing off-target effects to improve drug efficiency.¹ Since their discovery, research into cell biology and ways in which cells can be modified have been ceaseless. Mechanisms by which drugs can be delivered across cell membranes into the intracellular environment have always been a significant area of interest in drug delivery technology research. In particular, with the latest advances in different therapeutic modalities in recent years, including, but not limited to, small molecules,² proteins, and biologics³ (e.g., monoclonal antibodies and peptides^4,5), as well as nucleic acids, have necessitated the development of novel drug delivery technologies.⁶ Therefore, there is a need to optimize and develop new drug delivery strategies to meet these requirements.

In drug delivery systems, the process of delivering exogenous substances (e.g., a drug or gene) into living cells is called intracellular cargo delivery, which can also be called cargo delivery into cells. This complex technology is widely applicable and is emerging as a highly promising tool in cell biology, research as well as therapeutic development in gene therapy, gene editing, regenerative medicine, and cell therapy.^3,7–11 In recent years, the development of micro/nanotechnology has promoted breakthroughs in intracellular delivery research based on physical methods. Mechanical methods, especially, have shown great potential in various aspects.

Generally, due to the different mechanisms of transmembrane regulation, cargo delivery technologies can be divided into three groups: biological carrier-mediated [Fig. 1(a)], chemical carrier or nanoparticle-mediated [Fig. 1(b)],^12,13 and physical-mediated [Fig. 1(c)].¹⁴ Among the three categories of cargo delivery methods, virus-mediated biological vectors [Fig. 1(a)(i)], which are associated with high delivery efficiency and high cell viability, are the most widely used ones.^15–18 In 2006, the Japanese scientist, Shinya Yamanaka,¹⁹ used viral transfection technology for the first time to introduce the Oct4, Sox2, Klf4, and c-Myc (OSKM) gene combination, encoding four transcription factors, into mouse fibroblasts. The introduction and expression of these transfection factors led to the successful reprogramming of these somatic cells into induced pluripotent stem cells (iPSCs). This groundbreaking achievement has revolutionized somatic cell reprogramming technology and led to Yamanaka receiving the Nobel Prize in Medicine or Physiology in 2012. However, general concerns surrounding the safety of using viral vectors persist, specifically in cell and gene therapy.^20,21 However, the safety of using viruses as vectors has always been questioned, especially in gene and cell therapy. This is mainly because viral transfection and gene insertion will greatly increase the risk of cell carcinogenesis, especially the reactivation of a c-myc retrovirus, which can lead to tumor formation, cytotoxicity, as well as unnecessary immune and inflammatory responses.^22–24 Chemical carrier-mediated approaches include a variety of options, such as liposomes²⁵ and liposome complexes,^26–29 among others. This approach allows for transport and delivery of loaded cargo through the cell membrane and into the cell by membrane fusion [Fig. 1(b)(i)]. For example, as gene carriers, they deliver DNA or RNA for gene therapy and gene editing. Because conventional liposome transfection technology is often associated with the issue of low transfection efficiency, it can be combined with physical-mediated methods, such as electroporation, to further improve efficiency.

The phenomenon of electroporation³⁰ as a potential physical-mediated method of targeted cargo delivery was observed as early as 1898 when researchers found that the addition of an exogenous electric field could change the permeability of the cell membrane, allowing dyes to enter the cell.³¹ Modern electroporation technology³² is characterized by high transfection efficiency and mainly uses short and high voltage electrical pulses to temporarily create small pores in the cell membrane [Fig. 1(c)(i)], allowing targeted delivery of drugs, nucleic acids, or other molecules to the cell. However, when delivering biomacromolecules, which are relatively larger in size compared to other substance types (e.g., small molecules and nucleic acids), it is difficult to ensure sufficient cell viability because the strength of the electrical field needs to be increased to create larger pores in the cell membrane. Voltages above a certain threshold can create irreversible damage to cell membranes leading to cell death.^33,34 Recently, Mukherjee et al. proposed the local electroporation (LEP) technology based on bulk electroporation (BEP) and successfully delivered the largest proteins, protein–nucleic acid conjugates, and Cas9-ribonucleoprotein complexes to date using a nanochannel-based local electroporation device (LEPD) while ensuring good cell activity.^35,36 Microinjection [Fig. 1(c)(ii)], another form of physical-mediated drug delivery technology,^37–39 mainly operates on single cells and is often used in reproductive assisted technology. Although this technology boasts a high degree of accuracy, it is also associated with a low throughput and low efficiency, making it unsuitable for use in cell transfection, gene therapy, and cell therapy. Importantly, each method has its own unique characteristics [Fig. 1(d)]. In addition to electroporation and microinjection, new physically mediated delivery platforms are in continuous development owing to advances in micro- and nanoprocessing technology.

As the term suggests, mechanical-mediated cargo delivery strategies utilize external mechanical means to temporarily damage the integrity of the cell membrane to transiently increase cell membrane permeability by also introducing tiny pores [Fig. 1(c)(iii)], thereby facilitating entry of external cargo into the intracellular environment. Compared with electroporation, mechanical transfection allows for efficient delivery of macromolecules, does not require an additional energy field, and also avoids biologically mediated adverse reactions and immunogenic reactions.⁴⁰ Moreover, under high-throughput conditions, mechanical transfection shows a higher transfection efficiency. In addition, due to its favorable biocompatibility and low cost, mechanical transfection is emerging as an increasingly promising technology for future development. Herein, in this review, we will summarize the existing microfluidic-based mechanically mediated delivery technologies, including filtroporation, microchannel constriction, obstacle array extrusion, point constriction, fluid shear, hydrodynamic stretching, and spiral hydroporation strategies. Moreover, we discuss the future applications of artificial intelligence (AI)-enabling technologies in novel mechanical delivery platforms.

Mechanically mediated delivery technologies induce cell deformation by applying external force to cells, thereby opening force-sensitive channels⁴¹ to generate transient pores in the cell membrane, thus facilitating the entry of exogenous substances. Mechanical cargo delivery was first reported by McNeil and colleagues in the 1980s, including scrape-⁴² and bead-loading.⁴³ With the former, fluorescently labeled macromolecules are used as a substitute for culture medium and attached to the periphery of adherent cells. Subsequently, a rubber scraper is used to repeatedly scrape the adherent cells off the matrix to resuspend them in solution [Fig. 2(a)]. Since the scraping force is difficult to control to ensure uniformity, the pressure on each cell is different, causing some cells to die immediately while others are barely affected. Only in cells that are moderately stressed do exogenous molecules diffuse into the cell interior through transient membrane damage. With bead loading, cells in medium containing beads (75–500 μm) and target molecules are shaken so that the beads and adherent cells collide with each other [Fig. 2(b)], resulting in damage to the plasma membrane and subsequent delivery of the target molecule. However, because the degree of plasma membrane damage varies considerably between cells, the delivery efficiency of the target molecules and cell viability are highly inconsistent. Additionally, these methods generate a large amount of cell debris, further affecting the stability of the experimental results. Moreover, these methods have some limitations in terms of the use of expensive reagents due to the large amount of target molecules required during the delivery process. Compared with viruses, liposomes, and electroporation methods, these two mechanical delivery methods have lower throughput, efficiency, and cell viability. As such, these limitations have hindered their widespread application.

In recent years, with the continuous development of micro- and nanoengineering technologies, microfluidics technology has provided new opportunities for precisely controlling the plasma membrane deformation process by specifically manipulating liquid flow at the microscale. Specifically, microfluidic platforms allow for the regulation of fluid dynamic parameters at the micro- and nanometer scale, including flow rates, shear force, and fluid mixing. These precise controls enable researchers to simulate and manipulate cell responses to different environmental conditions. Importantly, the controlled disruption of the plasma membrane is essential for efficient and safe introduction of exogenous molecules. In addition, the highly integrated design of microfluidic systems allows multiple delivery parameters to be simultaneously optimized on a single chip, thereby permitting large-scale high-throughput cell processing. This integration and automation improve the repeatability and reliability of experiments and also greatly reduces the operation time and associated costs. In Secs. II A and II B, we will introduce microfluidic-based methods of cargo delivery, which are associated with advantages that may solve the limitations of the traditional delivery methods described above. These mechanical approaches can be divided into two categories: constriction and fluid shear-induced cargo delivery.

In 1999, Williams et al.⁴⁴ reported a method based on filter constriction, which forces cell suspensions through a 12 μm thick polycarbonate filter with pore sizes ranging from 5 to 18 μm [Fig. 2(c)]. In this method, cells are subjected to mechanical pressure from the filter pores, and due to the fluidity and flexibility of the plasma membrane, the squeezing force separates the phospholipid molecules in the membrane structure from each other. This results in temporary micro-pores in the cell membrane, allowing substances to be transported into the cell by diffusion, allowing for delivery of large fluorescent glucan molecules (~500 kDa). Furthermore, studies have shown that increasing the driving pressure or reducing the micropore size leads to an increase in the macromolecular delivery. Based on these studies, Yen et al.⁴⁵ developed a microfluidic integrated filter constriction array system known as TRansmembrane Internalization Assisted by Membrane Filtration (TRIAMF). Consisting of 24 highly integrated split-flow filter constriction modules, this system innovatively combines microfluidic and filter constriction technologies. The integrated microfluidic design of the TRIAMF system significantly improves the throughput and efficiency of cargo delivery and has been used to successfully deliver Cas9-ribonucleoprotein (RNP) to human hematopoietic stem cells, achieving a 63.1% knockout efficiency of β2-microglobulin (B2M), demonstrating the applicability of this method in primary cells.

Inspired by the mechanical pressure of cells passing through the pores, other researchers have utilized microchannels to squeeze the cells and aid drug delivery. In 2013, Sharei and co-workers.^46–48 developed a microfluidic platform for cargo delivery by rapidly contracting and squeezing cells through parallel microchannels, also known as the SQZ platform [Fig. 2(d)]. The SQZ platform is designed with 45 parallel channels to avoid equipment failure due to blockage of any single channel, thus ensuring consistent cell processing and high throughput. The base material of this microfluidic device is a wafer on which photolithography and deep reactive ion etching technologies are used to introduce the microchannels. Subsequently, the device is then sealed with Pyrex layers. Cell suspensions are then injected into the channels under pressure, and since the narrowest part of the microchannel is smaller than the cell diameter, the cell membrane is squeezed and deformed, creating transient pores that permit the diffusion of exogenous substances into the cytoplasm. In addition, the research group that developed this SQZ platform also studied the associated plasma membrane recovery kinetics of this cargo delivery platform and found that the Ca²⁺ concentration had a significant impact on membrane repair. Their data show that membrane recovery kinetics and cell viability can be successfully fine-tuned using buffers with different Ca²⁺ concentrations. In the absence of Ca²⁺, membrane repair time can range from 30 s to a few minutes, further improving transfection efficiency.⁴⁹ 

In addition to the above silicon-based microfluidic chips, Han et al.⁵⁰ proposed a method for cargo delivery using an array of obstacles with an extrusion effect in 2015 [Fig. 2(e)]. The chips used in this method are fabricated by standard polydimethylsiloxane (PDMS) microfluidic technology. Compared with the silicon chip discussed above, the PDMS-glass microfluidic structure is cheaper and less difficult to process during manufacture. The microchannel design of this technology consists of an array of ten diamond-shaped barrier columns with a gap of 4 μm, and cells are repeatedly squeezed and deformed when passing through these arrays. Single-stranded DNA (ssDNA), small interfering RNA (siRNA), and large-sized plasmids have been successfully delivered into different types of cells, including adherent and non-adherent cells, difficult-transfected lymphoma cells, and embryonic stem cells using this method. Furthermore, the same research group improved the constriction structure by modifying the diamond-shaped obstacle column array to make it star-shaped⁵¹ and fishbone-shaped arrays⁵² for a CRISPR–Cas9 RNP delivery and siRNA internalization, respectively. These optimized designs effectively improved delivery efficiency and broadened the application of the PDMS microfluidic technology in different cell types.

In addition to the use of contraction structures to deform cells, researchers have also developed other constriction strategies. In 2018, Liu et al.⁵³ discovered a novel cell volume convective exchange transfer (VECT) phenomenon. Cells are extruded through ridges with rectangular cross sections, which deform significantly at ultrafast time scales (10 μs) to produce transient and significant volume changes (up to 30%) without compromising cell viability. In that study, the authors successfully achieved rapid delivery of molecules of various types and sizes [e.g., dextran (4–2000 kDa), plasmids, mRNA, and nanoparticles (100 nm)] to human cells. Furthermore, the results indicated that as the number of consecutive constrictions increased, the time for cell relaxation decreased, leading to increased volume exchange. In addition, Liu et al.⁵³ also showed that more than 80% of the molecular delivery occurred during the brief cell compression stage (<0.1 s). In comparison, when fluorescent molecules were provided immediately after the cell had exited the channel, transfection efficiency was reduced to 33%, thereby confirming that efficient delivery of large molecules is achieved through fluid exchange during compression and relaxation processes in cell-based VECT. In order to enhance the convective exchange of cells, in 2021, Chung et al.⁵⁴ proposed drop-based constriction technology for efficient transfection of human primary T lymphocytes. This method cleverly combines microfluidic droplet technology with cell constriction technology, encapsulating the cell with external cargo (such as mRNA or plasmid DNA) in droplets. When these droplets pass through narrow channels, the cell membrane is mechanically compressed to create pores, and the surrounding cargo is exchanged into the cell interior through convective solution. The recirculation flow within the droplet further promotes the convective exchange and significantly improves the delivery efficiency. In addition, droplet packaging reduces the consumption of delivered reagents, thereby reducing the requirements for expensive reagents. Microchannel- and filter-based constriction may cause blockages, resulting in large back pressure. To solve this problem, Xing et al.⁵⁵ studied low back pressure single cell spot contraction technology based on a microfluidic platform for substance delivery [Fig. 2(f)]. Importantly, this study demonstrates that the novel point contractile structure enables efficient molecule delivery under low pressure, while maintaining high cell viability. Unlike conventional constriction channels, the point constriction structure compresses cells in multiple dimensions and imposes minimal back pressure on the system due to its unique semicircular geometry. Using this point constriction structure, the delivery efficiency of fluorescein-conjugated dextran (70 kDa) was up to ~86% under 4 bar, and the cell viability was also significantly high at ~85%.

Unlike constriction, fluid shear depends on the action of fluid on the cell membrane. When water molecules move at high speed in parallel to the membrane surface, they cause lipid heads to tilt in the direction of shear, resulting in membrane instability, which may eventually trigger plasma membrane rupture. Additionally, jets of water molecules that impact the membrane vertically may puncture the cell membrane. Although fluid shear forces are less invasive to cell membranes and will not directly disrupt membrane structures, controlling such shear forces in a liquid environment is generally more complex and challenging. One of the simplest ways to generate regions of high fluid shear stress is to alter the fluid flow by sudden contraction that is wider than the cell diameter, thereby rapidly increasing the fluid shear stress, causing the cell membrane to open and allowing for target cargo delivery. In 1992, Clarke and McNeil⁵⁶ used syringe loading to achieve cargo delivery. This method relies on mixing the cell suspension with the target cargo by repeated aspiration through a needle syringe [Fig. 2(g)]. During this process, the diameter of the needle is greatly reduced compared to the syringe tube and the cell container, resulting in a sudden increase in the fluid shear force for transient penetration of the cell membrane and cargo delivery. Furthermore, the flow rate of the repeated aspirations through the needle also determines the speed at which cells pass through the contraction zone, thereby determining the shear force on the cells. However, since the fluid flow is manually controlled, considerable experience and skills are required to ensure a consist flow rate for optimal transfection.⁵⁷ To improve accuracy and repeatability, in 2008, Prausnitz and co-workers⁵⁸ and colleagues designed a simple flow-through microfluidic device with parallel contractions to generate consistent fluid shear regions [Fig. 2(h)]. They used lasers to drill 50–300 μm conical microchannels in 100–250 μm thick mylar films, and syringe pumps were used to flow cell suspensions through the channels at controlled flow rates, thus subjecting the cells to constant shear forces.

Compared with shear deformation, fluid extensional deformation has higher analytical throughput and induces more significant cell deformation at high strain rates. In 2013, Dudani et al.⁵⁹ proposed a new hydrodynamic stretching strategy known as hydropipetting. This method uses inertial focusing to accurately position individual cells, uniformly transport cells to the extension flow in the cross-region, and stretch and deform them through the lateral shear force of the high-speed fluid, while recording the deformation process with a high-speed camera.^60,61 Through high resolution microscopic imaging and automated image analysis, researchers extracted mechanical parameters, such as cell strain and apparent viscosity, thus providing a new tool for disease diagnosis and cell state characterization. Additionally, Kwon et al.⁶² introduced a nonlinear microfluidic cell stretcher platform, known as a μ-cell stretcher, in 2023, which combines contraction and fluid stretching methods to achieve intracellular mRNA delivery. Here, methylcellulose (MC) solution was used as a viscoelastic fluid to apply high shear forces to cells through a single contraction zone in a microfluidic channel, thereby generating transient nanopores to promote the effective entry of macromolecules (e.g., mRNA) into cells. This platform has a delivery efficiency of ~97% and can operate at a high throughput of ~3.5 × 10⁵ cells/min with almost no clogging. In 2019, Kizer et al.⁶³ developed a cargo delivery platform called a hydroporator [Fig. 2(i)], which does not require the use of viscoelastic fluids, but only uses fluid inertia to focus, guide, and stretch cells to perforate cell membranes. In this method, the cell suspension is mixed with the target cargo and injected into the microchannel at a moderate Reynolds number. The inertial effect in the microchannel is used to exert strong deformation at the confluence point. The stretching flow elongates the cell and creates discontinuities in the membrane, allowing for the rapid delivery of external material into the cell. In addition to relying on passive diffusion, this platform also generates rapid solution exchange across the cell membrane through the deformation process when achieving the cargo delivery of exogenous substances, significantly improving the delivery efficiency. Moreover, additional work also investigated the relationship between delivery efficiency and intrinsic mechanical properties of cells, namely, deformability, which showed a strong correlation. Cells with lower stiffness were shown to be more likely to deform under constant fluid shear, allowing for increased stretching and an increased number of nanopores in the cell membrane. In comparison, cells with increased stiffness require a larger shear force to form transient pores. In a subsequent study using a similar channel layout, Hur et al.⁶⁴ proposed a T-shaped cell stretching platform with a cavity, which significantly stretches the cell through an elongated circulation flow in the channel, creating discontinuities on the cell membrane and, thus, enabling efficient internalization of nanomaterials. Materials, such as plasmid DNA (7.9 kbp), mRNA, siRNA, quantum dots, and large nanoparticles (300 nm), were successfully delivered to different cell types, including hard-to-transfect primary stem cells and immune cells, with delivery rates as high as 98%. This technology is emerging as a potentially promising tool for gene therapy.

Another strategy used to generate fluid shear in microchannels is to exploit the vortex shedding phenomenon, widely studied flow oscillation behavior that occurs behind a blunt body. In 2019, Jarrell et al.⁶⁵ placed a series of equidistant columns in a microfluidic device to induce vortexes when fluid struck the cylinders for cargo delivery. The device, termed a microfluidic vortex shear (μVS) device, was fabricated on a silicon wafer using a deep reactive ion etching technique. When the fluid flows through the array column, vortexes are induced under conditions with Re values greater than 40, resulting in membrane rupture, allowing exogenous molecule delivery to the cell. Studies have shown that mRNA concentrations >80 μg/ml result in maximal transfection efficiencies and cell viabilities of 63.6% and 81%, respectively, in human T cells. Furthermore, these parameters improve with increasing mRNA concentrations.

In 2020, Kang et al.⁶⁶ designed a spiral vortex delivery platform, which includes a cross junction and two bifurcated T-junctions, designed to achieve efficient delivery of large nanomaterials by utilizing spiral vortexes and vortex rupture in the cross junction at medium Re values. Based on this platform, cell suspensions mixed with target nanomaterials were pumped into two opposing channels at different flow rates (Re = 30.3, 37.9, 122, 203, 285, or 366). This study showed that when the cells flowed and approached the intersection, the fluid flow interface remained sharp and symmetrical at low Re values. Moreover, at Re values >37.9, vortex motion was the dominant force and the cells elongated asymmetrically, resulting in transient pore formation, followed by the influx of nanomaterials from the extracellular environment. As the cell leaves the stagnation region, inertia guides it to the center of the channel to the T-junction and the cell collides with the channel wall again, further causing cell deformation. This repeated process of cell deformation and repair produces more nanopores in the cell membrane, thereby improving the diffusion and transport of target substances into the cytoplasm. Kang et al.⁶⁶ successfully achieved delivery efficiency of up to 96.5% with a throughput of 1 × 10⁶ cells/min and a cell viability of 94%, while being able to rapidly deliver large-sized nanoparticles (200 nm gold nanoparticles) and macromolecules (>2000 kDa dextran) into cells in approximately 1 min.

These delivery methods enable controlled membrane deformation through precise control of the constriction and fluid shear forces. Although all of the above microfluidic delivery systems require certain experimental skills when preparing the chip, except for syringe loading, unless an integrated chip is customized to simplify the preliminary preparation work. In practical applications, constriction delivery often faces the common problem of channel blockage when injecting cells, while fluid shear delivery does not require contraction smaller than the cell diameter, which enables it to significantly reduce the common blockage problem in constriction delivery. In general, the highly integrated design of microfluidic platforms reduces cell damage and improves throughput and consistency, simplifies the operating procedures of traditional biological and chemical methods, improves the repeatability and reliability of the experiment, and overcomes the defects of traditional mechanical delivery methods, thereby providing more efficient and reliable solutions for current and future cargo delivery technologies.

In this work, we briefly review the widely used biological, chemical, and traditional physical methods for intracellular cargo delivery and mainly discuss a series of microfluidic technologies based on mechanical delivery that have emerged in recent years (Table I). These technologies use mechanical or fluid dynamic effects to create transient pores in the cell membrane, thereby achieving efficient delivery of macromolecules, such as proteins, nucleic acids, and nanomaterials. Each technology is associated with its own advantages, for example, filter constriction regulates delivery efficiency by controlling pore size and pressure, while the SQZ platform improves throughput and consistency with parallel channel design. Furthermore, obstacle arrays are suitable for a wider range of cell types through multiple extrusion deformations, and fluid shear and stretching technologies exploit fluid properties to achieve efficient delivery while maintaining relatively high cell viability. Continued advances in these delivery technologies, along with the expansion of basic research into these methods, will allow for even more precise regulation of cell function and targeted delivery, thereby driving innovation in gene therapy, regenerative medicine, cell therapy, and bioengineering.

However, as with many scientific advances, progress will no doubt result in many challenges along with promising opportunities. First, the complexity of microfluidic systems makes it difficult for non-specialists (e.g., clinicians or biologists) to adopt such technologies immediately, which has hindered their application in clinical practice. Therefore, simplifying the operation of microfluidic systems, improving their ease of use, and reducing human intervention through automation will be critical for future developments. Although microfluidic technology has performed well in terms of delivery efficiency and cell viability, its versatility still requires further investigation. This technology should be able to deliver fluorescent materials as well as cargo that can alter cell function, and it should be suitable for difficult to transfect primary cells. In addition, good manufacturing practice compliance is also an important goal of new cargo delivery platforms, especially in clinical applications. Ensuring the repeatability, aseptic operation, and process standardization of cargo delivery process is essential if we are to realize its potential in therapeutic applications.

With the increasing complexity of delivery technology and the increasing demand for high-throughput experiments, the ways in which big data can be processed and analyzed, as well as process automation, are urgent issues that need to be addressed. The advent of AI has brought new opportunities for the development of microfluidic technology. High-throughput microfluidic systems can generate a large amount of data in a single experiment. Traditional manual analyses methods cannot cope with the complexity and scale of these data. AI, especially deep learning technology, can quickly extract meaningful information for classification, prediction, and optimization through automated data processing and pattern recognition. For example, researchers have proposed a fully automated nano-fountain probe electroporation (NFP-E) system that does not require any additional nuclear labeling but uses a fully convolutional network (FCN) to achieve cell segmentation and cell localization.⁶⁷ During cargo delivery, AI can monitor cell status in real time, process cell image data, calculate Young's modulus, and dynamically adjust delivery parameters according to cell stiffness to ensure experimental consistency. Especially in large-scale experiments, the application of AI significantly reduces manual intervention. This intelligent and automated data analysis and feedback mechanism enables microfluidic system to operate efficiently in complex experimental environments. In addition, AI plays a crucial role in the design and optimization of systems. By using AI-driven optimization algorithms, researchers can automatically generate optimal system configurations, thereby reducing design iteration time and costs.⁶⁸ In the future, the deep integration of AI and microfluidic technology is expected to promote the development of personalized medicine. Through analysis and AI prediction models, researchers can design personalized intracellular cargo delivery solutions for different patient conditions to achieve precision treatments. AI can also predict and prevent possible issues in experiments, such as equipment failure or cell damage, thereby improving the success rate and safety of experiments. In the field of drug screening and gene therapy, AI-driven microfluidics platforms will greatly shorten the new drug development cycle and reduce R&D costs. AI technology will play a vital role in future advances, especially when dealing with complex diseases and personalized treatment needs. AI has shown great potential in the field of intracellular cargo delivery, but its application is not without challenges. The “black box” nature of AI models limits their interpretability in medical and scientific research, making it difficult for researchers to understand their decision-making processes, particularly regarding delivery parameter optimization and cell response prediction. Furthermore, protecting the privacy of sensitive data in personalized medicine is crucial, necessitating robust measures for data security and ethical compliance. Therefore, multidisciplinary collaboration will be essential in the future to comprehensively address the reliability and operability of this technology, promoting the effective application of AI-assisted cell delivery in clinical and scientific research.

This work was funded by the Key Program of the National Natural Science Foundation of China (No. 12332016), the National Key Research and Development Program of China (No. 2021Yfb3200804), and the Foundation of National Center for Translational Medicine (Shanghai) SHU Branch (No. SUITM-202416).

The authors have no conflicts to disclose.

Zhiyu Mao: Conceptualization (equal); Investigation (equal); Writing – original draft (lead). Bori Shi: Investigation (equal); Visualization (lead). Jinbo Wu: Conceptualization (equal); Project administration (equal). Xinghua Gao: Funding acquisition (equal); Project administration (equal); Writing – review & editing (lead).

Data sharing is not applicable to this article as no new data were created or analyzed in this study.