MAXTIA: A high‐throughput platform for rapid functional epitope mapping by kinetic screening of mutant libraries Authors: Kihoon Kim, Ryo Matsunaga, Takanori Yokoo, Makoto Nakakido, Kouhei Tsumoto Published: 2026 ABSTRACT AbstractAccurate identification of epitope residues is essential for developing biopharmaceuticals and understanding the mechanisms of immune recognition. However, experimental approaches for residue‐level epitope mapping remain time‐consuming and labor‐intensive, while accurate computational prediction of protein–protein interfaces remains challenging. Here, we present MAXTIA, a high‐throughput kinetic screening platform that integrates cell‐free protein synthesis with high‐throughput surface plasmon resonance and demonstrate its application to alanine scanning‐based functional epitope mapping. This workflow enables the rapid preparation and kinetic characterization of up to 384 protein variants, allowing the identification of functional epitope residues within 3 days while simultaneously providing binding affinity and kinetic parameters (K D, k on, and k off). We applied MAXTIA to map the epitope of the single‐domain antibody (VHH) N1 against the pentraxin domain of neuronal pentraxin‐2 (NP2 PTX). Alanine substitutions that cause substantial affinity losses clustered within a localized region on the AlphaFold3‐predicted NP2 PTX structure, defining a functional epitope site. These residues closely matched the interface observed in the NP2 PTX–VHH N1 crystal structure, validating the accuracy of MAXTIA. Beyond epitope identification, MAXTIA provides a simple and versatile platform for the quantitative analysis of protein–protein interactions, including high‐throughput screening of antibody variants for affinity optimization. This approach should accelerate biopharmaceutical development and facilitate mechanistic studies of molecular recognition.Keywords: cell‐free protein synthesis, epitope mapping, high‐throughput experimental system, interaction analysis, kinetic screening, protein engineering INTRODUCTION An epitope, also known as an antigenic determinant, is a specific region of an antigen that is recognized and bound by the antigen binding site of an antibody. Epitope mapping is the experimental process of identifying and characterizing specific binding sites at the molecular or amino acid residue level (Sharon et al., 2014; Talucci et al., 2024), linking molecular recognition to function (Sela‐Culang et al., 2013). Precise epitope mapping is critical because it enables the rational design and optimization of therapeutic antibodies, vaccines, and diagnostic reagents by facilitating the engineering of molecules with improved specificity and reduced off‐target effects (Ehlers et al., 2019; Grewal et al., 2024; Potocnakova et al., 2016). Despite their utility, current epitope mapping methods are constrained by inherent methodological limitations that create specific trade‐offs among applicability, resolution, and throughput (Abbott et al., 2014; Jethva & Gross, 2023; Kumar et al., 2023; Vengesai et al., 2022). Peptide‐scanning approaches enable rapid and scalable epitope mapping (Hansen et al., 2013; Hilpert et al., 2007) but exhibit an intrinsic bias toward linear epitopes (Vengesai et al., 2022). These methods often fail to detect discontinuous or conformational epitopes, which predominate in the recognition of native protein structures by B cells (Ansari & Raghava, 2010; Stephenson et al., 2015; Vengesai et al., 2022). Hydrogen‐deuterium exchange mass spectrometry (HDX‐MS) can detect conformational epitopes (Sevy et al., 2013), but its spatial resolution is limited by proteolytic peptide coverage, which generally restricts analysis to the peptide level rather than to individual residues (Konermann & Scrosati, 2024; Stofella et al., 2024). Furthermore, developing of broadly applicable approaches for the atomistic interpretation of HDX patterns remains a significant methodological challenge (Stofella et al., 2024). High‐resolution structural techniques, such as x‐ray crystallography and cryo‐electron microscopy, provide a near‐atomic view of antibody–antigen interfaces (Abbott et al., 2014) but remain labor‐intensive and limited in throughput. These approaches often require well‐ordered crystals (Liljeroos et al., 2015) or highly homogeneous and stable complexes (Skiniotis & Southworth, 2015), conditions that are particularly challenging for flexible proteins. Moreover, even if contacting residues are identified, they provide no direct information on per‐residue energetic contributions without complementary mutational or thermodynamic analyses (Bogan & Thorn, 1998; Cunningham & Wells, 1993). Mutagenesis‐based epitope mapping, as exemplified by alanine scanning (Gershoni et al., 2007), can quantify the energetic contribution of individual residues within an antigen–antibody interface (Davidson & Doranz, 2014). By systematically replacing residues with alanine, which removes side‐chain interactions while minimizing structural disruption (Cunningham & Wells, 1993), and measuring the resulting changes in binding affinity or kinetics, these approaches identify “hotspot” residues that disproportionately drive binding free energy (Pons et al., 1999). This level of resolution provides insight not only into which residues contact the antibody but also into which residues are critical for maintaining high‐affinity recognition (Van Petegem et al., 2008). However, mutagenesis‐based methods face significant bottlenecks, as they require sequential gene construction, expression, and characterization of individual variants, which imposes a substantial technical burden and limits throughput (Ahmad et al., 2016; Dang et al., 2023; Keen et al., 2025). Alanine scanning coupled with display platforms, such as mammalian or yeast display, can dramatically increase throughput (Potocnakova et al., 2016; ThaléN et al., 2024). However, these approaches generally do not directly measure binding affinity derived from association and dissociation rate constants, thereby limiting mechanistic insights into the distinct contributions of individual residues to binding kinetics (D'arrigo et al., 2024; Hunter & Cochran, 2016; Liu et al., 2021). Cell‐free protein synthesis (CFPS) systems provide versatile platforms for the rapid expression of target proteins (Dondapati et al., 2020). By bypassing the time‐consuming steps inherent in cell‐based expression systems, CFPS enables protein production directly from DNA or PCR templates within a few hours, thereby facilitating the high‐throughput generation of protein variants required for mutant screening (McSweeney & Styczynski, 2021; Sawasaki et al., 2002). More recently, CFPS has been coupled with label‐free interaction analysis platforms, including bio‐layer interferometry (BLI) and high‐throughput surface plasmon resonance (SPR), offering a powerful approach for the large‐scale quantitative characterization of protein–protein interactions (Agu et al., 2025; Matsunaga & Tsumoto, 2025; Predeina et al., 2025; Warfel et al., 2023). This integrated strategy enables the rapid generation and kinetic profiling of extensive mutant libraries, thereby addressing the major limitations of mutagenesis‐based epitope mapping. In this study, we developed Multiplexed Assay using CFPS and Kinetic Screening Technology for Interaction Analysis (MAXTIA), a novel high‐throughput kinetic screening platform, and applied it to alanine scanning‐based functional epitope mapping. This system integrates colony PCR with a CFPS platform for the rapid generation of mutant antigen libraries, followed by large‐scale interaction screening using high‐throughput SPR. This system enables the precise identification of functional epitope residues that contribute to binding free energy within 3 days by simultaneously providing comprehensive kinetic profiles for 384 variants, along with parallel sequence verification of each variant using nanopore sequencing (Figure 1). We applied this approach to alanine‐scanning‐based epitope identification using a single‐domain antibody (VHH). The functional epitope residues identified through the analysis of affinity changes closely corresponded to those defined by crystal structures. Overall, this system represents a promising platform for the comprehensive kinetic analysis of protein–protein interactions and substantially accelerates the kinetic screening of large mutant libraries. Overview of the MAXTIA system. Created in BioRender. Kihoon, K. (2026) https://BioRender.com/07ut4m9. RESULTS Model antigen and antibody pair We selected the neuronal pentraxin‐2 (NP2) pentraxin (PTX) domain and the anti‐NP2 VHH, designated N1, as the model system. The complex structure of N1 bound to NP2 PTX has been previously resolved (PDBID 9LAD) (Yokoo et al., 2026). The NP2 PTX domain corresponds to the C‐terminal module of NP2 and adopts a conserved PTX fold that encompasses calcium‐binding sites. This domain plays a pivotal role in modulating NP2 interactions and oligomerization at synapses, though it remains monomeric in isolation. The anti‐NP2 PTX VHH N1 was generated by immunizing alpacas with an NP2 construct, followed by phage display selection. Biophysical and structural characterization demonstrated that N1 binds the PTX domain with high specificity and affinity, inhibiting NP2 receptor engagement and function (Yokoo et al., 2026). Importantly, N1 was shown to bind the NP2 PTX domain even in the absence of calcium; accordingly, all interaction analyses in this study were performed under calcium‐free conditions. NP2 PTX variant gene library preparation To identify the epitope of VHH N1 on NP2 PTX, we selected 77 solvent‐exposed residues for alanine‐scanning mutagenesis based on a structural model predicted by AlphaFold 3. Residues were chosen using a criterion of solvent‐accessible surface area greater than 20% and a low predicted propensity for substantial backbone rearrangement upon mutation (Gilis & Rooman, 1996). These positions were individually substituted with alanine via PCR‐based site‐directed mutagenesis, and the resulting variants were produced using the PUREfrex cell‐free protein synthesis system (Shimizu et al., 2001). For cell‐free expression, a plasmid encoding NP2 PTX under the control of the T7 promoter was constructed with a C‐terminal 10 × His‐tag. This plasmid served as a template for site‐directed mutagenesis to introduce alanine substitutions at the 77 target positions. Following PCR amplification, the amplicons were pooled, purified, and digested with DpnI to remove residual parental plasmid DNA. The resulting linear DNA fragments were reassembled into circular plasmids using NEBuilder HiFi DNA Assembly and transformed into Escherichia coli. Individual transformants were arrayed into 384‐well plates and subjected to colony PCR with primers containing specific barcode sequences, amplifying the NP2 PTX variant‐coding region (from the T7 promoter to the 10 × His tag) for each clone. The resulting amplicons from each well were used directly as template DNA for cell‐free protein synthesis. In parallel, aliquots of the barcoded colony PCR products from all wells were pooled into a single tube, purified, and analyzed by nanopore sequencing to determine the sequence identity of each NP2 PTX variant (Currin et al., 2019; Ito et al., 2024). Cell‐free synthesis of an NP2 PTX variant library and interaction analysis Colony PCR products generated in a 384‐well format were used directly as DNA templates for CFPS using the PURE system without purification. The feasibility of protein expression under these conditions was verified via SDS‐PAGE and Western blot analysis (Figure S1). Cell‐free protein synthesis was carried out in a 384‐well plate at 30°C; subsequently, the reaction mixtures were diluted in assay buffer for high‐throughput SPR measurement. The diluted mixtures were applied to a nickel‐chelating NTA sensor chip to capture the His‐tagged NP2 PTX variants as ligands for interaction analysis. Kinetic measurements were performed using multiple concentrations of the analyte (VHH N1), and binding parameters were determined from the acquired sensorgrams. Kinetic and energetic interpretation of alanine scanning data for epitope identification Kinetic parameters (K D, k on, and k off) were successfully determined for 74 of the 77 alanine mutants by performing two independent experimental cycles using the MAXTIA system under identical conditions. One variant (L429A) was not recovered during colony PCR, while two variants (P301A and D380A) yielded single clones that were excluded from further analysis due to poor fitting quality based on empirically defined criteria. The distributions of K D values for the remaining 74 variants are summarized as box plots in Figure 2; the corresponding distributions of k on and k off are shown in Figures S2 and S3, respectively. For measurements with no detectable binding, placeholder values (k on = 104 M−1 s−1, k off = 10−1 s−1, R max = 100 RU) were assigned to enable visualization and summary‐level comparison. These placeholder values do not represent fitted kinetic constants and were interpreted only as indicators of undetectable binding. Median values were used to represent each mutant. Box plots showing the K D values of NP2 PTX variants binding to VHH N1. For D328A, R381A, and W419A, K D values could not be determined for all data points because of their low affinity. Because unpurified cell‐free synthesized products were directly used for Ni‐NTA capture in the MAXTIA workflow, the observed capture response may include background contributions from the crude reaction mixture and nonspecifically captured components. Therefore, the capture level was not used as a quantitative measure of active ligand density, and theoretical Rmax values or occupancy values based on capture level were not calculated. The fitted Rmax values shown in Table 1 were used as empirical parameters obtained from kinetic fitting. TABLE 1Kinetic parameters of variants exhibiting substantial alterations in binding affinity following alanine substitution.Construct n R max (RU) k on (× 105 M−1 s−1) k off (× 10−3 s−1) K D (nM) K D,mut/K D,wt ΔΔG (kcal mol−1)WT19224.26.53.44.81.00Y235A15118.6142.31.60.3−0.66V412A474.784.27.6183.70.78G415A252.933.14.3142.90.64S417A4179.15.18.0163.40.73K418A449.573.725173.60.76P420A7175.54.68.8183.80.78 Note: Median value for each parameter was calculated independently. Consequently, discrepancies may arise between the reported K D values and the corresponding k off/k on ratios in this table. Fitted R max values are shown as empirical parameters obtained from kinetic fitting. Theoretical R max and occupancy values were not calculated because the ligand capture level in the crude CFPS‐based Ni‐NTA capture format does not quantitatively represent the amount of active target protein on the sensor surface. Kinetic parameters of variants exhibiting substantial alterations in binding affinity following alanine substitution. Note: Median value for each parameter was calculated independently. Consequently, discrepancies may arise between the reported K D values and the corresponding k off/k on ratios in this table. Fitted R max values are shown as empirical parameters obtained from kinetic fitting. Theoretical R max and occupancy values were not calculated because the ligand capture level in the crude CFPS‐based Ni‐NTA capture format does not quantitatively represent the amount of active target protein on the sensor surface. Comprehensive kinetic analysis revealed that W419A, R381A, and D328A exhibited no measurable binding across all detected clones despite confirmed expression via Western blotting (Figure S1). In addition, V412A, G415A, S417A, K418A, and P420A showed a nearly threefold or greater reduction in binding affinity, while Y235A showed a more than threefold increase in affinity (Table 1). Representative sensorgrams for variants displaying pronounced affinity changes are shown in Figure 3, and the corresponding kinetic parameters are summarized in Table 1. Changes in binding free energy were calculated as ΔΔG using the relationship ΔΔG = RT ln(K D,mut/K D,wt), with the wild‐type value set to zero. Excluding variants with no measurable binding, the observed ΔΔG values ranged from approximately −0.66 to 0.78 kcal mol−1. These values were mapped onto the AlphaFold3‐predicted structure of NP2 PTX using a color scale (Figure 4a). Representative sensorgrams of alanine‐substituted NP2 PTX variants exhibiting either pronounced alterations in binding affinity or complete loss of detectable binding. NP2 PTX variants were His‐captured on a sensor chip, and different concentrations of VHH N1 (0.32, 1.6, 8, 40, 200, 1000 nM) were injected as analytes. AlphaFold3‐based structural mapping of binding free‐energy changes in NP2 PTX epitope residues. Residues of NP2 PTX are colored using AlphaFold3‐predicted structure according to the relative change in binding free energy (ΔΔG) calculated from surface plasmon resonance (SPR) kinetics (ΔΔG = RT ln(K D,mut/K D,wt)). Blue indicates residues whose alanine substitution yielded the smallest ΔΔG (least disruptive or affinity‐enhancing), while red indicates residues with the largest ΔΔG among the variants that retained measurable binding. Residues shown in magenta correspond to alanine mutants for which no measurable binding was detected by SPR, indicating a loss of binding beyond the detection limit, generating an effectively infinite ΔΔG. Residues with ΔΔG = 0 are shown in white. Residues not included among the mutation targets are colored in gray. (a) Surface of AlphaFold3‐predicted structure of NP2 PTX is shown from two orthogonal views. (b) AlphaFold3‐predicted structure of NP2 PTX colored by affinity changes aligned to the NP2 PTX‐N1 crystal complex (PDBID 9LAD). (c) Close‐up view of the interaction interface shown in (b). Residues showing a mutation‐induced binding free‐energy change (ΔΔG) >0.5 kcal mol−1 upon alanine substitution were operationally classified as candidate epitope‐associated residues (Nam et al., 2020). Since alanine substitution can also perturb local stability, folding, or ligand presentation, this classification does not necessarily indicate direct antibody contact. Notably, a localized region contained a cluster of candidate epitope‐associated residues. Within this region, W419A did not exhibit measurable binding. Surrounding residues V412, G415, S417, K418, and P420 exhibited ΔΔG values of 0.78, 0.64, 0.73, 0.76, and 0.78 kcal mol−1, respectively, following alanine substitution. Together, these residues form a contiguous surface patch consistent with a functional epitope site. In contrast, alanine substitution of Y235, located proximal to this putative epitope cluster, produced a favorable affinity change (ΔΔG = −0.66 kcal mol−1). Detailed kinetic analysis further revealed that the affinity reductions observed for V412A, S417A, K418A, and P420A were primarily driven by an increase in k off (Table 1). In contrast, the affinity decrease observed for G415A was mainly attributed to a reduction in k on (Table 1). Conversely, Y235A exhibited both an increased k on and decreased k off, resulting in a more favorable binding profile relative to the wild‐type. Interestingly, R381A and D328A, which are distal to this cluster, also showed no measurable binding, classifying them as candidate epitope‐associated residues (Figure 4a). However, unlike with W419A, neighboring residues did not display significant affinity changes. Structural validation of the mapped epitope using the NP2 PTX‐VHH N1 crystal structure To validate the epitope mapping results obtained using MAXTIA, we compared the candidate epitope‐associated residues identified by alanine scanning with the residues in structural contact with VHH N1 in the antibody–antigen crystal complex (Figure 4b,c). Although the distal candidate epitope‐associated residues (R381 and D328) were far from the interaction interface, all residues in the putative epitope cluster identified by MAXTIA (V412, G415, S417, K418, W419, and P420) were located within the binding interface. Consistent with this observation, the crystal structure revealed that W419, identified as a binding hotspot by MAXTIA, inserts into a cavity on VHH N1 and interacts with N98 and R100 in CDR3, as well as F37 and F47 in framework region 2 (FR2), forming a core interaction interface (Yokoo et al., 2026) (Figure 4b,c). This structural arrangement supports the conclusion that W419 is an energetic hotspot residue essential for binding. Residues V412 and K418 were oriented inward within the β‐strand and faced each other, forming a hydrophobic packing that stabilized the local binding‐site conformation. Simultaneously, both residues formed hydrogen bonds with VHH N1 (Table S1). Alanine substitution at these positions resulted in decreased k on and increased k off, indicating impaired association efficiency and reduced complex stability, consistent with the disruption of the optimal binding geometry. G415 is located within a loop region connecting the β‐strands; given the conformational flexibility conferred by glycine residues, substitution with alanine likely alters the local geometry or restricts the loop flexibility required for productive binding. This interpretation is supported by the observation that the affinity decrease for G415A was primarily driven by a reduction in k on, in contrast to other epitope mutants (V412A, S417A, K418A, and P420A), for which affinity losses were mainly driven by increased k off (Table 1). Interestingly, despite its proximity to G415, alanine substitution of G414 resulted in increased affinity (ΔΔG = −0.46 kcal mol−1), suggesting that G414 is a permissive residue whose substitution improves binding‐site geometry or conformational preorganization. In addition, the median R max values for V412A (74.78 RU), G415A (52.93 RU), and K418A (49.57 RU) were substantially lower than that of the wild‐type (224.15 RU), suggesting reduced expression levels or partial misfolding. Given that V412 and K418 are inward‐facing β‐strand residues and G415 resides in a loop region, these residues likely contribute not only to binding but also to the structural stability of NP2 PTX. Notably, Y235, which forms close contact with E44 in the FR2 of VHH N1 in the crystal structure (Figure 4c), exhibited a more than threefold increase in affinity upon alanine substitution (Table 1, Figure 4). This enhancement was primarily driven by an increase in k on accompanied by a decrease in k off (Table 1). These results suggest that despite mediating close intermolecular contacts, the bulky tyrosine side chain imposes steric constraints; substitution with alanine alleviates this hindrance, leading to more favorable binding kinetics. Finally, we compared the buried surface area (BSA) of individual residues derived from the crystal structure with the corresponding ΔΔG values obtained from MAXTIA for residues with solvent exposure greater than 20% (Figures 5 and 6). All residues in the putative epitope cluster were buried at the interface (BSA ≥25 Å2). However, several residues, including D407 and D411, showed negligible ΔΔG values despite having considerable BSA values (Figure 6), indicating that structural burial alone does not necessarily correspond to energetic importance. Furthermore, some buried residues, including Y235, K241, and G414, functioned as “cold spots,” where alanine substitution enhanced binding affinity (Figure 6). Comparison between the residue‐wise buried surface area in the crystal complex and ΔΔG values obtained by MAXTIA. Scatter plot illustrating the relationship between ΔΔG and buried surface area for interface residues of NP2 PTX with solvent accessibility greater than 20%, as analyzed in this study. DISCUSSION In this study, we developed a high‐throughput system termed MAXTIA, which enables rapid kinetic screening of mutant libraries, and applied it to alanine scanning‐based epitope mapping. Using this platform, functional epitope residues contributing to the binding free energy (ΔΔG) were accurately identified within 3 days, accompanied by detailed kinetic characterization of each mutant. MAXTIA allows the direct measurement of K D, k on, and k off for up to 384 variants using high‐throughput SPR, without the need for labor‐intensive protein expression and purification. The overall workflow is simple, bench‐scale, and rapid. Gene construction and E. coli transformation are performed on day one, followed by 384‐well colony PCR, sequence analysis, and cell‐free protein synthesis on day two, and high‐throughput SPR analysis on day three. Although commercially available CFPS kits are readily accessible tools that enable the rapid and straightforward production of large numbers of protein variants (given sufficient resources) (Chong, 2014), the primary bottleneck in high‐throughput kinetic screening remains the construction of mutant gene libraries. Conventionally, each mutant DNA construct is generated individually through site‐directed mutagenesis, assembled by ligation, transformed into E. coli, amplified by plasmid preparation, and sequence‐verified separately, resulting in a substantial technical burden and limited throughput. To overcome this limitation, we implemented a colony‐PCR‐based strategy in which E. coli colonies served directly as DNA templates using primers containing specific barcode sequences, enabling pooled sequencing and demultiplexing by nanopore sequencing. Importantly, residual contaminants present in the crude colony PCR products did not interfere with CFPS because they were sufficiently diluted, allowing direct use without purification. This approach enabled the high‐throughput preparation of up to 384 mutant DNA variants in quantities sufficient for CFPS, thereby substantially increasing the scalability of mutant library generation. Kinetic screening using a high‐throughput SPR system enabled the rapid characterization of up to 384 cell‐free synthesized ligand samples in a single experiment, yielding clear interaction profiles even when diluted crude CFPS products were used in a 384‐well format. The ability of high‐throughput SPR to capture His‐tagged proteins directly from crude solutions was exploited in our previous study. However, those results indicated that efficient His‐tag capture of proteins expressed in bacteria‐based systems requires ammonium sulfate precipitation to remove low‐molecular‐weight components derived from the culture medium (Matsunaga et al., 2023). In contrast, in the present study, robust interaction profiles were obtained by directly capturing the diluted crude CFPS products generated using the PUREfrex system via their His‐tags, without any additional precipitation or purification steps, which substantially simplified the workflow and increased the overall throughput of the MAXTIA platform. This improvement is attributed to the composition of the PUREfrex system, which is reconstituted from individually purified E. coli translation components, including initiation, elongation, and termination factors; a ribosome recycling factor; aminoacyl‐tRNA synthetases; methionyl‐tRNA formyltransferase; ribosomes; T7 RNA polymerase; and enzymes for energy regeneration (Shimizu et al., 2001). Unlike cell‐extract‐based systems, the use of purified components stabilizes the background, enabling direct compatibility with high‐throughput SPR analysis. The results obtained using the MAXTIA system highlight the distinction between structural interface residues and functional epitope residues, which directly contribute to binding energetics. Residues V412, G415, S417, K418, W419, and P420 formed a localized cluster on the surface of NP2 PTX and exhibited substantial increases in ΔΔG upon alanine substitution, indicating that they collectively constituted the functional epitope recognized by VHH N1. Among these residues, W419 emerged as a critical hotspot because its substitution with alanine abolished detectable binding across all clones. Structural analysis of the NP2 PTX–VHH N1 crystal complex revealed that W419 is inserted into a cavity on the antibody surface and forms a central interaction network, consistent with its dominant energetic contribution. Furthermore, kinetic decomposition of the binding affinity revealed distinct mechanistic roles for individual residues. Mutations such as V412A, S417A, K418A, and P420A primarily increased the dissociation rate, whereas G415A predominantly affected the association rate, suggesting that functional epitope residues contribute not only to complex stability but also to the efficient formation of the binding interface. Comparison of the NP2 PTX‐VHH N1 crystal complex with alanine scanning results obtained using the MAXTIA system (Figures 4b,c, 5, and 6) also revealed several “cold‐spot” residues whose alanine substitution led to increased binding affinity. Notably, residue R257 of NP2 PTX contacts R103 within CDR3 of VHH N1, and the R257A mutation resulted in a modest affinity enhancement (ΔΔG = −0.19 kcal mol−1). Similarly, Y235 of NP2 PTX forms close contact with E44 in FR2 of VHH N1, and its alanine substitution yields a larger affinity gain (ΔΔG = −0.66 kcal mol−1). In addition, K241 of NP2 PTX interacts with S63 in FR3 of VHH N1, with the K241A mutation producing an affinity increase (ΔΔG = −0.30 kcal mol−1). These observations indicate that interactions involving cold‐spot residues on NP2 PTX and their corresponding partners on VHH N1 are not optimized for maximal binding affinity. Importantly, such unfavorable energetic contributions cannot be inferred from static structural analysis of the crystal complex alone, underscoring the value of quantitative mutational approaches for elucidating epitope functionality. Furthermore, these results provide actionable insights into antibody affinity optimization, suggesting that rational modification of interaction partners corresponding to cold‐spot residues could further enhance binding. Variants incorporating these mutations can be efficiently generated and screened in a high‐throughput manner using the MAXTIA system. In the configuration used here, antigen variants were His‐captured on the sensor surface and the antibody was injected as the analyte. By reversing this assay orientation, the same workflow can be used to generate and screen mutant antibody libraries against the antigen, thereby enabling high‐throughput antibody optimization and epitope‐focused analysis. This highlights the synergistic integration of MAXTIA with structural analyses and underscores its versatility as a general platform for kinetic screening of protein variants. Although the MAXTIA system enables high‐throughput kinetic screening and the rapid identification of functional epitope residues, it is not without limitations. First, the PUREfrex‐based CFPS system lacks eukaryotic post‐translational modification machinery and therefore cannot support glycosylation, which restricts its ability to produce native glycoproteins (Matsumoto et al., 2023). Consequently, glycan‐dependent protein folding, epitope presentation, and antibody recognition cannot be faithfully reproduced, thereby limiting the applicability of MAXTIA to the analysis of glycosylation‐dependent epitopes. Second, the epitope mapping strategy employed in this study relied on alanine scanning, which can perturb the local or global protein structure and, in some cases, induce misfolding (Zoete & Meuwly, 2006). Because Western blot detection of His‐tagged expression confirms protein production but does not directly demonstrate proper folding, additional folding controls, such as binding analysis using an independent conformational antibody recognizing a different epitope or direct biophysical characterization of representative variants, would further strengthen the interpretation of binding‐loss mutants. Consistent with this limitation, alanine substitution at D328 and R381 resulted in no measurable binding in SPR analysis (Figures 2, 3, and 4a), suggesting that these residues play a critical role in maintaining structural integrity rather than directly contributing to binding interactions. This interpretation is supported by the absence of substantial affinity changes in neighboring residues surrounding D328 and R381, in contrast to the case of W419, which also exhibited no measurable binding but was flanked by residues showing pronounced affinity changes, indicative of a binding hotspot (Figure 4a). Computational studies of protein–protein interfaces indicate that binding hotspots are embedded within clusters of neighboring residues that collectively shape the binding energetics, and that alanine substitution of residues in such hotspot‐centered clusters can perturb binding affinity, thereby helping to distinguish structurally essential residues from those directly mediating molecular recognition (Koes & Camacho, 2011; Ozbek et al., 2013). However, when structurally essential residues are located near epitope regions or when epitope residues themselves contribute to structural stability, the interpretation of alanine scanning data becomes more complex. Furthermore, the criterion applied to select residues for alanine scanning in this study (specifically, solvent exposure greater than 20%) may not have identified all interface residues, as illustrated by residues F225 and S228 (Tables S1 and S3). Despite these limitations, this approach successfully revealed energetic contributions for most interface residues, even when target residues were selected based on predicted structures, thereby demonstrating its potential to accurately identify critical epitope residues in cases where experimentally determined structures are not available. (Figures 4, 5, 6). CONCLUSIONS The demand for the rapid identification of functional epitopes has increased substantially with the growing emphasis on antibody therapeutics and vaccine development. However, existing approaches remain labor‐intensive, time‐consuming, and insufficiently quantitative. In this study, we developed MAXTIA, a high‐throughput platform for the kinetic screening of hundreds of protein variants applied to alanine scanning‐based epitope mapping, which integrates colony PCR, nanopore sequencing, CFPS, and high‐throughput SPR. MAXTIA provides a simple and rapid workflow that enables the identification of functional epitope residues that contribute to binding free energy (ΔΔG) within 3 days. Beyond rapid epitope identification, this system delivers quantitative kinetic insights into antibody–antigen interactions, thereby facilitating a mechanistic understanding of protein–protein interactions that remain difficult to predict accurately using computational approaches. Collectively, MAXTIA establishes a practical platform for rapid functional epitope mapping and quantitative kinetic analysis of protein–protein interactions, with potential applications in antibody engineering and related biopharmaceutical research. MATERIALS AND METHODS Expression and purification of VHH N1 The DNA sequence encoding VHH N1, including a TEV protease cleavage site and a C‐terminal His‐tag, was cloned into the pRA2 vector (Makabe et al., 2005). The construct was expressed in E. coli BL21 (DE3) cells. Protein expression was induced by adding 0.5 mM isopropyl β‐D‐1‐thiogalactopyranoside (IPTG) when the culture reached an optical density (OD600) of 0.8–1.0, followed by overnight incubation. Cells were harvested by centrifugation, resuspended in lysis buffer (20 mM Tris–HCl, 500 mM NaCl, and 5 mM imidazole; pH 8.0), and lysed by sonication. The clarified lysate was subjected to immobilized metal affinity chromatography (IMAC) using Ni‐NTA agarose (30,230; Qiagen, Hilden, Germany). Bound proteins were eluted with elution buffer (20 mM Tris–HCl, 500 mM NaCl, and 200 mM imidazole; pH 8.0). The eluted protein was treated with TEV protease to remove the 6 × His‐tag, followed by a second IMAC purification step under the same buffer conditions to separate the cleaved protein from His‐tagged contaminants. After TEV protease digestion, the target protein was recovered from the flow‐through fraction. The protein was subsequently dialyzed against 10 mM HEPES–NaOH and 150 mM NaCl (pH 7.4). Final purification was performed by size‐exclusion chromatography (SEC) using a HiLoad 16/60 Superdex 75 pg column (28989333; Cytiva) at 4°C, equilibrated with the dialysis buffer. Protein purity was confirmed by SDS–PAGE. Plasmid construction and DNA preparation The gene encoding the NP2 PTX domain was cloned into a pET28b‐based expression vector under the control of the T7 promoter, incorporating a C‐terminal 10 × His tag to enable detection and His‐tag capture on the NiHC200M biosensor surface (Carterra). The resulting plasmid served as a template to construct the NP2 PTX variant genes. AlphaFold prediction of protein structure The amino acid sequence corresponding to the NP2 PTX domain (residues 223–431) was submitted to the AlphaFold server (Abramson et al., 2024) for structure prediction to facilitate epitope mapping in the absence of experimentally determined structures. Among the five models generated from a random seed, the top‐ranked model (model #0) was selected for subsequent analysis. Generation of NP2 PTX variant plasmids and E. coli transformation A panel of 77 alanine mutants targeting residues with solvent‐accessible surface areas greater than 20% was designed based on the AlphaFold3‐predicted structure visualized using Swiss‐PDB Viewer (Guex & Peitsch, 1997). For each mutant, forward and reverse PCR primers containing the desired mutations and overlapping sequences were designed for assembly using the AAscan software (Sun et al., 2013). PCR amplification was performed using KOD One DNA Polymerase (TOYOBO). The resulting linear DNA fragments were pooled and purified using a FastGene Gel/PCR Extraction Kit (Nippon Genetics). The purified DNA mixture was then treated with DpnI to remove residual template plasmid DNA. The DpnI‐digested linear fragments containing the desired mutations were assembled into circular plasmids using the NEBuilder HiFi DNA Assembly Kit (New England Biolabs), with the assembly reaction performed at 50°C for 1 h. The assembled plasmid mixture was then used to transform competent E. coli JM109 cells. Colony PCR and Nanopore sequencing A total of 384 colonies grown on kanamycin‐containing agar plates were individually picked and transferred to a 384‐well plate containing KAPA2G Robust HotStart ReadyMix with dye (Kapa Biosystems), using sterile toothpicks to provide the DNA templates for colony PCR. PCR amplification was performed to generate DNA fragments corresponding to the target gene variants. For sequence identification, a nanopore sequencing workflow based on a multiplexed barcoding strategy (Currin et al., 2019; Ito et al., 2024) was employed. Barcoded primer sequences used in this study are listed in Table S2. Forward primer positions were identical across each set of four sections within a 384‐well plate, whereas reverse primer positions differed, enabling multiplexed sample identification. This design allowed the simultaneous sequencing of up to three 384‐well plates. Following PCR amplification, the products from all wells were pooled and purified using the NucleoSpin Gel and PCR Clean‐up Kit (Macherey‐Nagel). Sequencing libraries were prepared from the purified DNA using the Ligation Sequencing Kit V14 (Oxford Nanopore Technologies). The prepared library was loaded into a Flow Cell (R10.4.1) and sequenced using a MinION Mk1B device (Oxford Nanopore Technologies). Data acquisition and base calling were performed using MinKNOW software (Oxford Nanopore Technologies). The resulting FASTQ files were demultiplexed by barcode using a custom Python script. Sequence clustering and consensus sequence determination were performed using SeekDeep (Hathaway et al., 2017), with the clustering threshold set to 98% sequence identity. The monoclonality of each sample was evaluated using both inter‐ and intra‐cluster mixture scores. The inter‐cluster mixing score was calculated as the ratio of reads in the second‐largest cluster to those in the largest cluster, with a threshold value of 0.1. When no reads were present in the second cluster, a value of one read was assumed for calculation. The intra‐cluster mixing score was determined by identifying the highest mutation ratio within the consensus sequence of the dominant cluster using SNP information generated with the “writeOutFinalInternalSnps” option, with a threshold value of 0.2. Only samples classified as monoclonal according to these criteria were included in the final dataset (Ito et al., 2024). Cell‐free protein synthesis Cell‐free protein synthesis was performed using the PUREfrex 2.1 system supplemented with DnaK Mix and DsbC Set (GeneFrontier), according to the manufacturer's instructions. Each 2.86‐μL reaction mixture contained 4 mM reduced glutathione, 3 mM oxidized glutathione, 5 μM DnaK, 1 μM DnaJ, 1 μM GrpE, and 1 μM DsbC, and 0.5 μL of a 40‐fold diluted crude colony PCR product as the DNA template. The reactions were incubated at 30°C overnight (Matsunaga & Tsumoto, 2025). Interaction analysis using a high‐throughput SPR instrument Cell‐free‐synthesized NP2 PTX variants obtained from the 384‐well plate were diluted in HBS‐T buffer (10 mM HEPES, 150 mM NaCl, 50 μM EDTA, and 0.05% Tween 20; pH 7.4) supplemented with 20 mM MgCl2 to a final volume of 250 μL. Association and dissociation kinetics were measured using an LSA high‐throughput surface plasmon resonance instrument (Carterra). Each NP2 PTX variant was His‐captured for 10 min on the surface of a NiHC200M biosensor chip (Carterra). Measurements were performed at 25°C using HBS‐T as the running buffer. VHH N1 was injected as the analyte using a fivefold dilution series (0.32–1000 nM) without regeneration between injections. The association and dissociation phases were monitored for 5 min each. After each experimental cycle, the sensor chip was regenerated using 0.5 M EDTA followed by 0.1 M NaOH. Kinetic parameters were determined using two analytical approaches, selecting the model that provided the best fit. For interactions exhibiting slow dissociation, the starting response at each analyte concentration was not corrected, and the entire dissociation phase was analyzed. Sensorgrams were fitted using the k d–k a model with the “Float t0” and “Float Rb (Bulk Shift)” options enabled. For interactions exhibiting fast dissociation, the starting response at each analyte concentration was normalized to 0 RU, and the dissociation phase was analyzed over a 1‐min interval. In this case, sensorgrams were fitted using the k d–k a model with the “Float t0” and “Float Rb (Bulk Shift)” options disabled (Matsunaga et al., 2023). Dataset construction Variants with poor fitting quality, defined by an empirical threshold (20