The past two decades have seen a paradigm shift from manual, time-intensive SELEX to automated, high-throughput platforms. Traditional SELEX suffers from a high number of selection rounds and long processing times^[8,9], as shown in Figure 3A. Integration of SELEX with microfluidics enables efficient aptamer separation, significantly shortening the selection time while minimizing the consumption of biological materials^[29,30], as shown in Figure 3B. Capillary electrophoresis SELEX eliminated target immobilization, a critical advantage for preserving native EV epitope conformations, and yielded high-affinity aptamers (Kd = 27.6 nM) in as few as five rounds^[28].

Most recently, fully automated platforms integrating microfluidics, robotic liquid handling, real-time quantitative polymerase chain reaction (qPCR) monitoring, and machine learning have achieved the automated, hands-free selection of intact EVs within days^[30]. These advances have transformed aptamer discovery from an artisanal craft into an engineering discipline. However, these platforms, alongside existing hybrid approaches that combine robotic handling with microfluidic modalities such as droplet microfluidics or acoustofluidics, still fall short of addressing three essential prerequisites for automated multi-analyte liquid biopsy applications in a single instrument: (1) supporting the processing of large sample volumes; (2) enabling the processing of more than one sample at a time; and (3) providing orthogonal extraction mechanisms as required^[31]. Figure 3 presents a comparison of conventional versus advanced SELEX strategies for aptamer selection targeting EVs.

The shift from soluble proteins to intact EVs represents a fundamental expansion in target complexity. Early efforts focused on individual EV surface markers such as epithelial cell adhesion molecule (EpCAM), mucin 1 (MUC1), and CD63^[21]; however, such predefined targets cannot resolve EV subset heterogeneity. The breakthrough came with whole-EV SELEX, which treats EVs as complex molecular assemblies^[26]. Cell-SELEX against cancer cells followed by counter-selection against normal cells enabled the discovery of aptamers recognizing unknown, disease-specific surface features^[32,33]. EDGE-SELEX applied to intact exosomes, together with immobilization-free SELEX that avoids conformational artifacts^[26], has further pushed the boundary. Most recently, AI-driven SELEX design has optimized library complexity for multi-epitope recognition on EV surfaces^[33], while fully automated platforms now enable programmable, high-throughput selection against intact EVs^[29]. Collectively, these advances have shifted the field from targeting purified molecules to confronting physiologically relevant, complex targets.

Initially regarded as antibody substitutes for detection, aptamers have undergone a functional redefinition: selection criteria now extend beyond binding affinity to encompass biological activity^[34,35]. Emerging strategies prioritize aptamers that not only bind with high affinity but block protein-protein interactions, mediate drug internalization, or activate immune signaling^[36,37]. Examples include aptamers that block heat shock protein 70 (HSP70)-mediated immunosuppression^[38], as well as aptamers that recognize membrane protein targets on exosomes derived from highly metastatic colorectal cancer and directly suppress exosome-induced pro-metastatic activities, such as cell migration, invasion, and angiogenesis^[39]. This functional turn shifts the paradigm from binding to intervention, positioning aptamers as therapeutic effectors rather than passive targeting ligands.

In some reported cases, functional selection may involve a trade-off with binding affinity^[34]. For instance, a study using the DNA aptamer GS24 targeting the transferrin receptor revealed that this aptamer exists in two conformations; however, only one of them (A-2F) possesses biological activity. Engineering modifications, such as truncation or mutation, were employed to stabilize the active conformation and enhance its biological activity; nevertheless, the affinity of the truncated variant GS24min was approximately five times lower than that of the parental GS24^[40]. However, this trade-off is not a universal rule; whether it occurs depends critically on whether the aptamer’s epitope overlaps with a functional site (e.g., a receptor-binding domain) or a non-functional structural region^[34]. Several examples demonstrate that aptamers can maintain both high affinity and full biological activity when the epitope is appropriately positioned^[41].

The observed trade-off in EV applications is primarily based on a limited number of systems, such as HSP70^[38], and should therefore be considered a cautionary observation rather than an established limitation. Nevertheless, for EV applications where aptamers must simultaneously recognize surface epitopes with high specificity and exert downstream functions (e.g., inhibiting EV-mediated signaling or directing cargo internalization), balancing these two properties remains an important design consideration.

Second, the inherent heterogeneity of EVs, both across and within populations, severely compromises inter-laboratory reproducibility. EVs derived from the same cell type can vary significantly in size, surface marker density, and cargo composition depending on culture conditions, isolation methods, and storage protocols^[5]. This batch-to-batch variability means that aptamers selected against one EV preparation may perform poorly against another, even from the same nominal source^[23,24]. To address this, the field urgently needs standardized biological and physical reference materials to benchmark aptamer binding affinity and specificity across different laboratories and platforms. Promising candidates include recombinant EVs (rEVs) engineered to mimic native biochemical and biophysical properties for quantitative data normalization, nanoerythrosomes, and engineered vesicles or synthetic liposomes with predefined sizes, refractive indices, and marker/epitope abundances.

Third, concurrent target discovery and functional aptamer selection without prior knowledge of EV surface features remains a technical blind spot. Most current selection strategies either begin with a known target [e.g., CD63, programmed death-ligand 1 (PD-L1)]^[21] or use whole-EV selection to generate aptamers without immediately knowing their binding targets^[32].

However, when an aptamer selected against intact EVs elicits a desired functional effect, such as blocking EV uptake or neutralizing a pathological activity, identifying its cognate surface epitope becomes a major bottleneck. Reverse engineering the target from a functional aptamer is technically demanding and time-consuming, typically requiring immunoprecipitation, mass spectrometry, or clustered regularly interspaced short palindromic repeats (CRISPR)-based validation. Until this blind spot is resolved, the systematic discovery of novel EV surface biomarkers that are both therapeutically actionable and diagnostically useful will remain constrained.

Magnetic separation is the most established aptamer-based platform for EV isolation. Biotin-labeled aptamers are immobilized on streptavidin-coated magnetic beads^[47,48], enabling rapid EV capture under an external magnetic field. At the optimal aptamer concentration of 1 μM, the capture efficiency reaches 80%, significantly higher than the 65% achieved by conventional antibody-functionalized Fe₃O₄ beads^[48]. Single-target magnetic separation has since been applied to various disease-associated EVs^[49].

Multitarget magnetic separation further improves specificity and coverage^[50,51]: dual-aptamer systems integrated with graphene field-effect transistors permit label-free detection of cancer-derived vesicles^[42], while triple-aptamer electrochemical approaches simultaneously recognize multiple EV surface epitopes^[52]. Critically, complementary strand displacement and deoxyribonuclease I cleavage have been optimized for non-destructive elution, both of which preserve EV morphology and membrane protein integrity^[53,54].

Microfluidics integrates EV isolation into miniaturized chip formats, substantially reducing sample consumption and processing time^[55]. Aptamer-based microfluidic devices enable target-specific EV capture with efficiencies of 10⁷-10⁸ particles/mL^[56], while select filtration-based platforms enable direct EV isolation from whole blood without pre-processing^[57]. Magnetic microfluidics further enhances isolation efficiency by combining magnetic separation with on-chip flow control, completing isolation within 1 h^[58,59].

However, substantial challenges remain. Wide variability in chip geometry, flow rate, and surface chemistry across laboratories hinders cross-study comparison, and the lack of standardized reference samples for calibration impedes clinical translation.

Smart responsive materials enable dynamic control over EV capture and release. Thermosensitive aptamer-conjugated materials achieve > 60% recovery rate through temperature switching^[21]. Photoactivatable magnetic beads offer spatiotemporal precision^[60]. DNA hydrogels incorporating CRISPR/CRISPR-associated protein 12a (Cas12a) respond to EV recognition through conformational changes, enabling signal-triggered release. Moreover, these hydrogels possess a high encapsulation capacity, with a maximum of 7.6 × 10⁷ exosomes per 50 μL of hydrogel, which facilitates efficient EV loading for downstream analysis^[61,62]; in this system, EV binding induces a conformational change in the hydrogel, activating Cas12a’s trans-cleavage activity, which then releases a reporter or triggers cargo liberation. Molecularly imprinted polymers combined with aptamers can recognize glycosylation patterns, a capability beyond conventional aptamers^[63].

Although these platforms have expanded the toolbox from passive adsorption to active regulation, most remain confined to benchtop operation^[64]; the lack of integrated automated liquid handling and external temperature/light control modules continues to impede their routine use.

Integrated platforms that combine isolation with downstream analysis enable sample-to-answer workflows. The CLCtips platform, for instance, integrates aptamer-mediated capture and on-chip enzymatic digestion for trace EV proteomics, identifying over 500 proteins from minimal sample volumes^[65]. Microfluidic chips coupled with mass spectrometry or next-generation sequencing permit multi-omics profiling from single isolation runs^[66-68], while proximity hybridization-mediated FRET (Fluorescence Resonance Energy Transfer) assays achieve high-specificity EV detection without separate isolation steps^[69].

Key bottlenecks for integrated platforms include multi-step compatibility (capture, lysis, digestion, analysis), the purity-sensitivity trade-off, sample throughput, and automation costs.

Target selection for aptamer-mediated EV delivery generally adheres to two criteria: (i) overexpressed surface biomarkers unique to pathological tissues [e.g., EpCAM, human epidermal growth factor receptor 2 (HER2), nucleolin, CD20]; and (ii) receptors capable of mediating robust internalization (e.g., transferrin receptor). Whole-cell SELEX offers an alternative route, enabling isolation of functional aptamers without prior target identification, provided they support intracellular uptake.

Nucleic acid therapeutics - small interfering RNA (siRNA), messenger RNA (mRNA), circular RNA (circRNA), and CRISPR-Cas9 - all require cytosolic delivery for activity. Aptamers serve a dual role, directing cell-specific uptake and, in certain designs, facilitating endosomal escape. For RNA interference, aptamer-displayed EVs have been validated for circRNA delivery to cancer cells^[95], and CL4 aptamer-modified EVs delivering long non-coding RNA (lncRNA) DARS-AS1 siRNA have been shown to reverse chemoresistance in triple-negative breast cancer^[96]. For mRNA delivery, hepatocyte-targeted systems have been developed for familial hypercholesterolemia^[97]; circular RNA and microRNA (miRNA) sponge delivery via aptamer-EV platforms has also been demonstrated^[98,99].

The most advanced application is CRISPR-Cas9 delivery. Elsharkasy et al. devised a modular strategy in which aptamers mediate ribonucleoprotein (RNP) loading into EVs, with ultraviolet (UV)-triggered release upon endocytosis^[100]; Hegeman et al. further refined aptamer affinity to enhance delivery precision^[101]; and Xu et al. engineered a reactive oxygen species (ROS)-responsive, aptamer-conjugated EV system for combined oncolytic virus OH2 and anti-PD-L1 therapy in prostate cancer^[102].

The endosomal escape bottleneck is most severe for these cargoes. Whereas a few siRNA copies per cell can achieve potent gene silencing, mRNA and CRISPR-Cas9 demand substantially higher cytosolic concentrations^[99,100]. Current efforts to optimize escape through aptamer design - surface charge modulation or minor conformational adjustments - remain largely empirical and insufficient. Transformative progress will likely depend on equipping EVs with viral fusion machinery.

Protein therapeutics (molecular weight > 10 kDa, fragile tertiary structures) impose additional demands: protection during delivery^[103,104], controlled release, and evasion of ubiquitin-proteasome degradation. Aptamer-EV systems have addressed these challenges through three functional paradigms.

Functional antagonism: the aptamer itself acts as the effector. Gobbo et al. designed a peptide aptamer targeting HSP70 on tumor EVs, thereby blocking Toll-like receptor 2 (TLR2)/TLR4-mediated dendritic cell suppression^[38]; Quintavalle et al. reported an aptamer against gremlin-1 that disrupts tumor-stroma crosstalk in breast cancer^[105].

Growth factor delivery: Luo et al. conjugated aptamers to bone marrow stromal cell EVs - directing them to hydroxyapatite or bone sialoprotein - while encapsulating bone morphogenetic protein 2 (BMP-2), thus shielding the growth factor from circulating proteases and preserving its bioactivity^[106].

Gene editor delivery: as described above, aptamer-EV systems for CRISPR-Cas9 RNP delivery represent the leading edge of this field^[101].

Three strategies targeting different nodes of the tumor-immune axis have emerged. The first strategy focuses on blocking immunosuppressive signals. Aptamers against HSP70 or PD-L1 on tumor EVs prevent these molecules from engaging immune receptors^[38,107], thereby restoring dendritic cell maturation and T cell activation. The second strategy involves delivering immunostimulatory agents. EpCAM aptamer-modified bacterial EVs serve as tumor vaccines, combining targeted delivery with natural adjuvant activity to enhance antitumor immunity^[108]. The third strategy aims at eliminating suppressive EVs themselves. Aptamer-guided capture removes pro-inflammatory or immunosuppressive EV subsets, as demonstrated in allergic airway inflammation, and this approach holds promise for reshaping the tumor microenvironment^[109]. Together, these three strategies illustrate the versatility of aptamer-EV platforms in reprogramming immune responses against cancer.

Immune cell-derived EVs naturally carry major histocompatibility complex (MHC) molecules and co-stimulatory signals, making them attractive platforms for immunomodulation. However, their therapeutic utility is limited by poor targeting efficiency and heterogeneous production. Aptamer modification addresses the first limitation by enhancing the tumor-homing capacity of immune cell-derived EVs, enabling their accumulation at disease sites with higher precision^[109,110]. To tackle the second, microfluidic strategies enable size-based separation of T cell receptor (TCR)-CD3 EV subsets, ensuring manufacturing consistency and reproducible therapeutic outcomes^[23]. Beyond oncology^[111], this engineering approach has shown promise in inflammatory diseases: aptamer-functionalized EVs loaded on biomimetic periosteum can recognize nerve growth factor (NGF) released from injured neurons, delivering anti-inflammatory cytokines and neurotrophic factors to promote tissue repair^[112]. Collectively, these strategies demonstrate that aptamer-mediated engineering can simultaneously improve targeting accuracy and manufacturing consistency of immune cell-derived EVs, addressing two key barriers to their clinical translation.

Real-time monitoring of immune checkpoint expression on circulating EVs is essential for evaluating patient responses to immunotherapy and guiding treatment decisions. Aptamers meet this need by enabling specific capture and high-sensitivity quantification of immune checkpoint molecules on circulating EVs^[113]. For example, He et al. developed a wash-free, homogeneous assay based on an aptamer-cholesterol-anchored DNAzyme assembly, simplifying the workflow without sacrificing accuracy^[114]. In another example, Cong et al. constructed a microfluidic device-based extracorporeal circulation system that combines CD63 aptamer-modified magnetic nanospheres for small extracellular vesicle (sEV) capture and a PD-L1-specific trigger aptamer for immune checkpoint recognition, coupled with hybridization chain reaction (HCR) for fluorescence signal amplification; this platform enables continuous in vivo monitoring of PD-L1⁺ sEVs during tumor progression, reflecting tumor burden and distinguishing early-stage from advanced disease, thus providing real-time feedback for immunotherapy management^[55]. In addition, a range of signal transduction platforms have been adapted for aptamer-based EV detection, including SERS (Surface-Enhanced Raman Spectroscopy) probes^[115], quantum dot nanospheres integrated with machine learning^[116], ellipsometry, and thermophoretic assays^[117]. All these platforms share a common three-step cascade - aptamer capture, signal amplification, and quantitative readout - with the principal differences residing in the signal transduction module and achievable sensitivity, entailing trade-offs among complexity, cost, and detection limits. A detailed comparison of these immune monitoring platforms is provided in Supplementary Table 3.

Aptamer-EV systems have been applied to bone, nerve, and vascular regeneration. Luo et al. pioneered bone targeting by functionalizing bone marrow stromal cell (BMSC)-derived EVs with hydroxyapatite-binding aptamers and loading them with BMP-2^[106], while Su et al. achieved combined neural, vascular, and bone repair via NGF targeting and c-Jun N-terminal kinase 3 (JNK3) mitogen-activated protein kinase (MAPK) pathway activation^[112].

In neurodegenerative diseases, Ren et al. designed DNA aptamers against α-synuclein aggregates, enabling EV-mediated delivery of chaperones or autophagy inducers to attenuate pathology in Parkinson’s disease models^[118]; Du et al. constructed multi-target engineered hybrid EVs bearing amyloid beta (Aβ)-specific aptamers for both oligomer neutralization and anti-inflammatory miRNA delivery^[119].

For anti-infection applications, Zhang et al. employed marine-derived EVs with bacterial-targeting aptamers for infected bone defect repair^[120], and Cui et al. developed a glycosylated extracellular vesicle-like receptor (termed GlycoEVLR) that enables rapid detection of viral antigens^[121].

Collectively, tissue regeneration and anti-infection applications constitute emerging frontiers where aptamer-EV systems offer distinct advantages, notably the capacity to target specific cell types or anatomical sites without requiring endosomal escape for therapeutic activity. Unlike intracellular drug delivery, where cytosolic cargo release is essential, tissue regeneration frequently benefits from sustained surface signaling or paracrine effects^[122], thereby mitigating the endosomal escape bottleneck. This distinction indicates that aptamer-EV technologies may achieve earlier clinical translation in regenerative medicine than in intracellular drug delivery.