Single-residue engineering of lambda (λ) antibody light chains reduces conformational flexibility and enhances thermal stability

Abstract

Monoclonal antibodies with lambda (λ) light chains are less commonly used in therapeutics due to their lower biophysical stability compared to kappa (κ) variants. Here, we identify a conserved glycine residue (Gly111) in the λ light chain hinge as a driver of large-scale Fab elbow-angle transitions. Using microsecond-scale molecular dynamics simulations of the EBV-neutralizing Fab AMMO1, we show that substituting Gly111 with threonine (G111T) increases the free energy barrier between conformational states, effectively arresting these transitions. Structural and biophysical analyses—including crystallography, differential scanning fluorimetry, and surface plasmon resonance—confirm that the mutation maintains Fab architecture and antigen binding while increasing thermal stability by up to 2.5 °C. The same mutation applied to a second λ-Fab yielded similar stabilization, and simulations of three clinical λ-Fabs revealed consistent reductions in elbow-angle flexibility. These results demonstrate a generalizable, single-residue engineering strategy to enhance the stability of λ-based Fabs without compromising function, with direct implications for therapeutic antibody development and manufacturability.

Graphical Abstract

1. Introduction

Over the past 30 years, antibodies have become the fastest-growing class of therapeutics for a broad range of diseases [1], [2], [3], [4]. Monoclonal antibodies (mAbs) are widely used in cancer therapy, immune system modulation, and treatment of chronic and infectious diseases [5], [6]. Immunotherapy with mAbs has become a cornerstone of cancer treatment, with continued efforts to improve efficacy, pharmacokinetics, pharmacodynamics, and manufacturability. As such, the generation and stabilization of multispecific mAbs remains a significant challenge and an active area of investigation [7], [8], [9].

mAbs are composed of two identical light chains (κ or λ isotypes) and two identical heavy chains, forming a characteristic Y-shaped structure. This includes two Fab (antigen-binding) regions—the antibody arms—and one Fc (fragment crystallizable) region. Each Fab consists of one heavy and one light chain [10]. While κ and λ light chains occur at comparable frequencies in humans, most biophysical studies have focused on κ-based Fabs [11]. These typically exhibit moderate thermal stability, unfolding cooperatively with melting temperatures (T_m) in the range of 70–80 °C [11], [12], [13], [14]. In contrast, the thermodynamic properties of λ-based Fabs remain less well-characterized. Prior studies suggest that λ light chain-containing Fabs are significantly less stable than their κ counterparts [15].

The increased flexibility of λ Fabs is thought to stem from a single-residue insertion in the hinge region connecting the variable light (VL) and constant light (CL) domains [16]. This hinge is one residue longer in λ light chains than in κ light chains. The Fab elbow angle—defined as the angle between the variable (FV) and constant (CH1/CL) domains—is influenced by this hinge. In λ-based Fabs, elbow angles range from 112° to 227°, while in κ-based Fabs they typically range from 125° to 195°. One study demonstrated that mutations near the elbow region in broadly neutralizing HIV-1 antibodies can affect both interdomain flexibility and paratope plasticity [17]. For example, somatic mutations such as P14S (near the elbow) in the VH domain reduced thermal stability and altered Fab orientation. However, the direct effects of hinge residue mutations on structural stability and antigen binding have not been systematically investigated.

Efforts to improve broadly neutralizing antibodies against viruses such as Epstein-Barr Virus (EBV) have explored structural engineering approaches [17]. The Bjorkman group, for instance, used DNA and protein linkers to couple Fabs to GP120, enhancing affinity through avidity [18]. Our lab developed a method using meditope technology, offering an alternative approach to achieve affinity through avidity. This technology enables covalent attachment of additional binding moieties through cysteine conjugation [19], [20]. However, as this technology requires access of a cyclic 12 amino acid peptide to the “hole” within the Fab region (created by the heavy and light chains) we suspect the extreme elbow angles and flexibility inherent in λ-based Fabs could impact application of this technology. To better understand this flexibility, we turned to molecular dynamics (MD) simulations.

MD simulations have become essential for probing atomic-level conformational changes, providing insights into antibody flexibility and stability [21], [22], [23]. These tools have been extensively applied to study elbow angle and interface dynamics in Fab fragments [24], [25]. For example, Fernández-Quintero and colleagues examined the impact of affinity maturation, humanization, and light-chain isotype on Fab dynamics, revealing rapid interdomain fluctuations on nanosecond timescales and increased rigidity upon maturation.

In this study, we applied extensive MD simulations to analyze elbow-angle transitions in the EBV-neutralizing λ-based antibody AMMO1 (PDB code: 6bla [26]). We identified specific hinge residues, particularly gly111, that underwent significant dihedral changes associated with the conformational transition. Based on these observations, we hypothesized that introducing a mutation at this position could stabilize λ light chain Fabs by restricting conformational mobility. Threonine, a β-branched amino acid, was selected to replace glycine due to its compatibility with Ramachandran space, its ability to reduce local backbone flexibility, and its minimal impact on structural packing. We generated both parental and mutant versions of the antibody. Surface plasmon resonance (SPR) measurements confirmed that the substitution did not affect antigen-binding affinity, and X-ray crystallography supported the structural neutrality of the mutation. Thermal stabilization was demonstrated using differential scanning fluorimetry (DSF). Collectively, this single-residue substitution increased thermal stability without perturbing the overall structure.

We further extended this approach to the anti-NANP λ antibody (PDB code: 6wfy [27]) involved in malaria immunity and to three clinical λ antibodies Guselkumab (PDB code: 4m6m [28]), Erenumab (PDB code: 6umg [29]), and Avelumab (PDB code: 4nki [30]) to assess generalizability. In all cases, we observed significant elbow-angle fluctuations driven by dihedral flexibility within the hinge region.

These findings highlight a rational, single-point mutation strategy that may be broadly applicable to stabilize λ light chain Fabs and potentially improve manufacturability and shelf-life in therapeutic antibody development.

2. Materials and methods

2.1. Simulation methods

All-atom MD simulations were performed using GROMACS [31] version 2019.3 on the Gemini supercomputer (TGen, Arizona). The CHARMM36 force field was applied to the system, and the TIP3P model was chosen for water. Each native simulation was initiated from the PDB coordinates. To generate the starting structure of the mutated Fabs, MODELLER [32] version 9.3 was used, which provided the initial coordinates for those simulations. For the AMMO1 second conformation simulation, we used the final snapshot (∠_elb = 138°) from the previously completed 1 µs native simulation (simulation number 1), applied the mutation, and then continued the simulation. CHARMM-GUI [33] was used to generate simulation models. The simulation box sizes were approximately 10 × 10 × 10 nm³ , providing a 10 Å padding on the longitudinal sides of the Fab in all simulations. Each system was first neutralized by adding counterions (Na⁺ or Cl⁻) to achieve zero net charge. Additional NaCl ion pairs were then added to reach ∼150 mM bulk salt. Periodic boundary conditions were applied in all three dimensions. Long-range electrostatic interactions were treated using the Particle Mesh Ewald (PME) method, and van der Waals interactions were calculated using Lennard-Jones potentials with a cutoff of 10 Å.

After energy minimization, a short 10 ns MD simulation using the NVT ensemble was performed, followed by system equilibration at constant temperature (303 K) and constant pressure (1 bar) using a Berendsen thermostat and a Parrinello-Rahman barostat, respectively. Production simulations were then carried out for approximately 1 µs using the NPT ensemble with timesteps of 2 fs. Elbow angles were calculated using a Python script inside PyMOL [34] graphical tool. Principal Component Analysis (PCA) and Normal Mode Analysis (NMA) were performed using the ProDy [35] tool. For each trajectory, we uniformly subsampled 400 frames across the full 1-µs simulation; with three replicates per condition, this yielded 1200 frames per condition. To address comparability within a figure, PCA was computed on the concatenated frames for the conditions shown in that figure (e.g., WT vs. mutant), and all datasets in that panel were projected into the same PC space. VMD Tcl scripts were executed in the VMD TkConsole [36] to calculate root mean square fluctuations (RMSF), root mean square deviation (RMSD) and to visualize domain motions as vectors derived NMA. Residue numbering in the text and figures follows the Chothia scheme; simulations and analyses used Fab backbone heavy atoms (N, Cα, C, O).

2.2. Biophysical methods

All Fabs were expressed in Chinese hamster ovary cells (ExpiCHO, Thermo Fisher Scientific) and purified using a modified protein G column (KanCapG, KANEKA Corporation), followed by size exclusion chromatography (Superdex 75 and Superdex 200; Cytiva Lifesciences). The mammalian expression vector harboring Fabs or Fab with mutations were transfected to ExpiCHO cells. Cells were incubated at 32 °C for 7 days. The synthetic (NPNA)₄ peptide used in this study was purchased from GenScript. Surface plasmon resonance (SPR) experiments were performed on a Biacore T200 instrument (Cytiva Lifesciences) to assess Fab–peptide interactions. Different Fabs were immobilized on a CM5 sensor chip (Cytiva Lifesciences) using EDC/NHS coupling chemistry. Various concentrations (0, 3, 10 and 30 nM) of the (NPNA)₄ peptide and a negative control peptide were flowed over the sensor surface in HBS-EP+ buffer (Cytiva Lifesciences). Kinetic parameters were determined using Biacore Evaluation Software, version 3.0. Protein thermal stability was assessed with 5 μg of Fab by differential scanning fluorimetry using SYPRO Orange dye (Thermo Fisher Scientific) on a QuantStudio 3 Real-Time PCR system (Thermo Fisher Scientific).

2.3. Protein crystallization

The native anti-NANP Fab and threonine mutant anti-NANP and AMMO1 Fabs were crystallized using sitting-drop vapor diffusion. Fab proteins were prepared in crystal buffer (10 mM Tris-HCl, pH 8.0, 50 mM NaCl) at concentrations of 12.5, 10, and 12 mg/mL, respectively, and mixed with reservoir solution at a 1:1 ratio. Native and threonine mutant anti-NANP Fabs were optimized in 2 µL drops, while the threonine mutant AMMO1 Fab was optimized in 0.2 µL drops.

Native anti-NANP Fab crystals were grown in 100 mM Tris-HCl, pH 7.0, 20 % (w/v) PEG 1000 and formed overnight. Crystals were harvested after 4 days and required no additional cryoprotection before flash-freezing in liquid nitrogen.

Threonine mutant anti-NANP Fab crystals were grown in 200 mM ammonium acetate, 100 mM Bis-Tris-HCl, pH 5.5, 25 % (w/v) PEG 3350 and also formed overnight. Crystals were harvested after 2 days and flash-frozen without additional cryoprotection.

Threonine mutant AMMO1 Fab crystals were grown in 170 mM ammonium acetate, 85 mM sodium citrate-HCl, pH 5.6, 25.5 % (w/v) PEG 4000, and 15 % (v/v) glycerol. Crystals formed within 2 days, were harvested after 4 days, and required no additional cryoprotection before flash-freezing in liquid nitrogen.

2.4. Data collection and structure determination

Data for the native anti-NANP Fab were collected at the Cornell High Energy Synchrotron Source (CHESS) at Cornell University on beamline MX 7B2. Data for the both threonine mutant anti-NANP Fab and AMMO1 Fab were collected at the Stanford Synchrotron Radiation Lightsource (SSRL) at the SLAC National Accelerator Laboratory on beamline BL12–2. Diffraction data were processed using XDS [37], and molecular replacement phasing was performed using Phenix Phaser-MR [38]. Models were refined through iterative rounds of refinement in Phenix and manual model building in COOT [39].

3. Results

3.1. Molecular dynamics of AMMO1

Initially, we performed an extended (Δt = 1 µs) all-atom MD simulation of the native AMMO1 Fab based on the X-ray structure (Figure S1a). At approximately 450 ns into the simulation, we observed a large conformational transition occurring between two distinct conformational states (Fig. 1a – Native), consistent with a shift in the elbow angle of the Fab (Fig. 1b, dashed orange trace). The elbow angle (∠_elb) at the beginning of the simulation was 218°, corresponding to the conformation observed in the crystal structure (defined as Conformation 1). The alternate conformation (Conformation 2) was centered around an elbow angle of ∠_elb = 138° ± 7°. To validate this substantial ∼70° conformational shift (Δ∠_elb), we performed two additional independent MD simulations. In both cases, a similar transition was observed, occurring earlier at 90 ns and 120 ns, respectively (elbow angles shown in Fig. 1b, solid orange lines).

Fig. 1.

Impact of G111T Mutation on AMMO1 Fab Conformation and Dynamics. (a) Crystal structure of native AMMO1 (PDB ID: 6bla) colored orange (left). The elbow angle measured from the native Fab (∠_elb = 218°) is indicated. After a 100 ns MD simulation, a new conformation with (∠_elb = 138° was identified (orange shadow)). Upon introducing the G111T mutation into the native Fab, a distinct conformation (∠_elb = 203°) emerged, shown in teal (middle); this simulation started from the native AMMO1 conformation (Conf1, ∠_elb = 218°). A second conformation resulting from simulations with the G111T mutation (∠_elb = 142°) is depicted in silver (right); here, simulations were initiated from Conf2 (∠_elb = 138°). Structures are aligned via constant domains. (b) Elbow angles from three independent 1 µs simulations of native AMMO1 (orange). In simulation 1 (dashed orange), the elbow angle transitions from 218° to 138° after 450 ns. This transition occurs after ∼100 ns in the other two simulations. G111T mutant simulations starting from the native structure (Conf1) are shown in teal, and simulations initiated from Conf2 (∠_elb = 138°) are shown in silver. (c) Ramachandran plots for Gly111 in the native Fab (orange dots), Thr111 in G111T mutant simulations initiated from Conf1 (teal dots), and Conf2 (silver dots). (d) Principal Component Analysis (PCA) plots (PC1 vs. PC2) for native AMMO1 (orange), G111T Conf1 (teal), and G111T Conf2 (silver). (e) Normal mode analysis of native AMMO1 (left), G111T Conf1 (middle), and G111T Conf2 (right). Arrows represent direction and magnitude of domain movements.

Given that the transition is largely driven by changes in the elbow angle, we sought to determine whether one or more residues in the hinge region participate in or gate this transition. To investigate this, we analyzed the dihedral angles of residues connecting the variable and constant regions of the light and heavy chains. Specifically, we focused on residues 108–114 of the light chain and residues 121–126 of the heavy chain, which we define as the hinge region (Figure S1b, shown in red and green licorice representation). Surprisingly, we observed a significant change in the dihedral angles of a single residue—Gly111 in the light chain (highlighted in green in Figure S1b; backbone dihedral angles shown as orange points in Fig. 1c)—rather than small, distributed changes across multiple hinge residues (Figure S1c, Native).

As glycine is the least conformationally restricted amino acid in terms of dihedral angles [40], we substituted Gly111 with threonine (Thr), a hydrophilic, β-branched amino acid, and tested whether this substitution would inhibit the transition between conformational states. To address this, we performed six additional 1 µs MD simulations with the Thr substitution: three initiated from Conformation 1 and three from Conformation 2. In all cases, the Thr substitution prevented the conformational transition over the same simulation timescale (Δt = 1 µs). Simulations starting in Conformation 1 (initial ∠_elb = 218°) fluctuated within a local energy well centered at an elbow angle of 203° ± 11°. The average structure and elbow angle trajectory for this condition are shown in teal in Fig. 1a and b. Similarly, simulations of the Thr-mutated Fab starting in Conformation 2 did not transition to the alternate state. These remained in a local well centered around an elbow angle of 142° ± 7°, with average structure and elbow angle trace shown in grey in Fig. 1a and b.

The dihedral angles of residue 111 in the light chain hinge region (between the Fab Fv and CH1/CL; Figure S1b, green) for native and Thr-substituted Fabs, plotted in Fig. 1c, show that the G111T substitution restricts the residue to specific regions of Ramachandran space. The dihedral angles associated with Conformation 1 are clustered within the α-helical region, whereas those associated with Conformation 2 are clustered within the β-sheet region (Fig. 1c, gray points). In comparison, the dihedral angles of the native Fab are more broadly distributed between the α-helical and β-sheet regions (Fig. 1c, orange points).

To determine whether the G111T substitution affected other regions of the Fab, we performed additional analyses. First, we conducted principal component analysis (PCA) using snapshots from three independent simulations of each system. The first two principal components (PC1 and PC2) revealed that the threonine-substituted Fab adopted a single conformational population (Fig. 1d, teal and gray points), whereas the native Fab exhibited two distinct populations (Fig. 1d, orange points). Normal mode analysis (NMA) further confirmed that the dominant mode of motion corresponds to the transition between the two conformations (Fig. 1e). The arrows represent the direction and amplitude of motion for each Cα atom in the Fab. A marked reduction in amplitude was observed when comparing the parental (orange arrows) and G111T-substituted Fabs (teal and gray arrows, corresponding to Conformations 1 and 2, respectively).

In simulations of the native AMMO1 Fab, the dihedral angles of residues within the heavy chain hinge region (residues 121–126) remained clustered within a narrow region of Ramachandran space (Figure S1c). In contrast, simulations of the G111T mutant starting from Conformation 1 revealed the emergence of a second population of dihedral angles in Ser125 and Thr126 of the heavy chain. This second population was not observed in G111T simulations initiated from Conformation 2. To test whether these residues facilitate the conformational transition observed in the native Fab, we conducted a new set of simulations in which Ser125 of the heavy chain was substituted with proline, the most conformationally restricted amino acid. Two independent simulations were performed for each of the native and G111T-mutated Fabs, starting from Conformation 1 using the same simulation parameters.

The S125P substitution in the heavy chain did not block the conformational transition in the native Gly111 Fab, whereas no transition was observed in the G111T/S125P double mutant (Fig. 2a, purple trace). Fig. 2b illustrates the distribution of backbone dihedral angles for residue Gly111 in the S125P and G111T/S125P variants. The conformational distribution of backbone dihedral angles in these two variants remained nearly identical. PCA of the S125P simulations starting from Conformation 1 revealed at least two distinct populations, but the addition of the G111T mutation collapsed this second population, as seen by a shift toward zero along the PC1 axis (Fig. 2c, green and purple points). Furthermore, NMA indicated that the S125P mutation alone had little effect on Fab amplitude (Fig. 2d, green arrows), while a substantial reduction in amplitude was observed for the G111T/S125P double mutant (Fig. 2d, purple arrows). These findings, together with the dihedral angle analyses of the light chain hinge (Figure S1c, S2a), support the conclusion that the conformational transition between the two elbow states is largely driven by the low energy barrier associated with Gly111.

Fig. 2.

Structural and Dynamic Effects of S125P and G111T/S125P Mutations on AMMO1 Fab, fluctuations of G111T simulations, and thermal stability experiments. (a) Elbow angles from two independent 1 µs simulations of the S125P mutant (green). In both cases, elbow angles change from 218° to ∼143° within 400 ns. G111T/S125P simulations (purple) also started from Conf1. (b) Ramachandran plot for Pro125 in S125P (green dots) and G111T/S125P mutants (purple dots). (c) PCA plots of S125P (green) and G111T/S125P (purple). (d) Normal mode analysis for S125P (green arrows) and G111T/S125P (purple arrows). (e) Per-residue RMSF plots for native AMMO1 (orange), G111T Conf1 (teal), and G111T Conf2 (silver), with light-chain variable shown at top, heavy-chain shown at bottom. Standard deviations represent variations from triplicate simulations. (f) Differential scanning fluorimetry (DSF) curves for native AMMO1 (orange) and G111T mutant (teal).

In addition to threonine, serine at position 111, a hydrophilic, non-β-branched residue, was also evaluated. Three independent 1-µs simulations of the G111S variant in the AMMO1 background were performed; elbow angles from each trajectory are shown in Supplementary Figure S2b. In each simulation, the Fab transitioned from Conformation 1 to Conformation 2, most clearly in replicate 2 (green trace). Thus, serine at position 111 permits the elbow transition and, unlike threonine, is not as restrictive.

To examine other potential differences, we calculated the root mean square fluctuations (RMSF) of the Cα atoms over the final 500 ns of each 1 µs simulation (Fig. 2e). The terminal 500 ns were selected as piecewise RMSF profiles had converged by ∼500 ns. RMSF values for the light chain variable domain are shown in the top panel, and those for the heavy chain variable domain in the bottom panel. Standard deviations were calculated from three independent runs. Regardless of the starting conformation (Conformation 1 or 2), the RMSF values of the Thr-substituted Fabs remained within one standard deviation of those from the native Fab across both CDR and non-CDR regions, further suggesting that the conformational transition is largely localized to the hinge region.

3.2. Biophysical characterization of AMMO1

To validate the conclusions drawn from the MD simulations, we produced and purified Fabs of both the native and G111T mutant AMMO1 variants. Each construct expressed as soluble protein (yields: 100–200 mg/L) and showed a symmetric elution profile by size exclusion chromatography, indicating that the threonine substitution was well tolerated.

We also set up crystallization trials for the G111T mutant, obtained and optimized crystals, and collected a complete diffraction dataset to 2.08 Å resolution. The structure was determined by molecular replacement and refined using Phenix and Coot; refinement statistics are listed in Table S1. The overall structure of the mutant is shown in Figure S2c. Notably, the elbow angle of the mutant Fab (225°) was nearly identical to that of the native AMMO1 (218°), suggesting that crystal packing forces may stabilize the open conformation. Root mean square deviation (RMSD) calculations over various domains of the G111T and native Fabs (Table S2) revealed that the RMSD between the variable (light and heavy) and constant (CH1 and CL) domains was 0.46 Å and 0.74 Å, respectively, indicating that the G111T substitution did not significantly alter the overall domain structure.

Given prior reports that limiting conformational flexibility can improve thermal stability, we performed differential scanning fluorimetry (DSF) to assess thermal transitions. Both native and G111T Fabs showed well-defined melting transitions. The native AMMO1 Fab melted at 68.5 °C (Fig. 2f, orange trace), while the G111T variant melted at 71.0 °C (Fig. 2f, teal trace), indicating an increase in thermal stability of 2.5 °C (n = 3).

As the S125P substitution alone did not block the conformational transition in simulation, we also tested its effect on thermal stability using DSF. The S125P mutant melted at 68.5 °C, while the G111T/S125P double mutant melted at 70.5 °C (Figure S2d, green and purple traces, respectively), consistent with the stabilizing effect of the G111T substitution.

3.3. Generalization of Gly substitution - anti-NANP Fab

Given these results, we sought to determine whether the observed effects were generalizable. Searching the Protein Data Bank (PDB) for human Fabs with a lambda light chain, we identified the anti-NANP Fab. This Fab shares an identical hinge sequence with AMMO1. Additionally, its antigen is a small peptide, (NPNA)₄, which allowed us to experimentally quantify the effect of the equivalent G111T substitution on antigen binding. As with AMMO1, three independent MD simulations were performed under identical conditions. The glycine-to-threonine mutation (Gly112 in anti-NANP Fab) was introduced using MODELLER, and subsequent simulations were conducted on the mutated structure.

As shown in Fig. 3a, the elbow angle of the wild-type anti-NANP Fab fluctuated between ∠_elb = 243° and ∠_elb = 137° (solid orange lines). Unlike AMMO1, this Fab frequently switched between the two conformations in one simulation, showed fewer transitions in the second, and rapidly transitioned to a new conformation early in the third simulation (dashed orange trace). In simulations of the G112T mutant, transitions were reduced in one case (dashed teal trace) and occurred once in the other two simulations (solid teal traces). In contrast to AMMO1, the mutation did not fully lock in a single conformation. However, Ramachandran plots of ϕ/ψ angles for the light chain hinge region showed tighter clustering in the threonine-substituted Fab, consistent with reduced flexibility (Figure S3a). No significant differences were observed in the heavy chain.

Fig. 3.

Structural Dynamics and Stability of Native and G112T Mutant anti-NANP Fab (PDB code: 6wfy). (a) Elbow angles from three independent 1 µs simulations of native anti-NANP (orange) and G112T mutant (teal). (b) PCA plot of native (orange) and G112T mutant (teal) anti-NANP Fab. (c) Normal mode analysis of native (left) and G112T mutant (right) anti-NANP. Arrows illustrate domain movements. (d) DSF data for native anti-NANP (orange) and G112T variant (teal). (e) Surface plasmon resonance (SPR) sensorgrans of native anti-NANP Fab with peptide antigen (left), G112T variant (middle), and control Fab (right).

As with the AMMO1 simulations, we performed PCA, NMA, and RMSF analyses. The PC1/PC2 projection (Fig. 3b) showed that the threonine substitution reduced the magnitude of PC1 relative to the native Fab. Mapping of normal modes onto the average structures (Fig. 3c, blue ribbons) revealed large-scale motion in the native Fab (orange arrows), which was markedly reduced in the threonine mutant (teal arrows). RMSF calculations over the final 500 ns of each simulation showed that the G112T variant remained within one standard deviation of the native Fab across both CDR and non-CDR regions of the variable domains (Figure S3b). While the substitution did not prohibit transitions between elbow-angle-defined conformations, the data again indicate that substitution of glycine with threonine in the hinge region substantially alters elbow dynamics.

To experimentally validate these findings, we produced and purified both the wild-type and G112T mutant anti-NANP Fabs. Both variants expressed as soluble protein (wild-type: 100 mg/L; G112T: 60–150 mg/L) and showed symmetric elution profiles by size exclusion chromatography (Figure S3c), indicating proper folding. Differential scanning fluorimetry (DSF) revealed a melting temperature of 68.3 °C for the wild-type Fab (Fig. 3d, orange trace), compared to 70.8 °C for the G112T variant (Fig. 3d, teal trace), reflecting a 2.5 °C increase in thermal stability (n = 3).

We also used surface plasmon resonance (SPR) to measure antigen-binding kinetics. The Fab was immobilized on a CM5 chip, and increasing concentrations of the (NPNA)₄ peptide were passed over the surface. Both the wild-type and G112T variants showed comparable binding affinities (Fig. 3e; left: native; center: mutant; right: control), indicating that the substitution did not impair antigen recognition. The fit kinetic parameters and affinity are presented in supplementary Table S3. No binding was observed using a control Fab.

Finally, we determined the crystal structures of the native and G112T mutant. As observed in AMMO1, the elbow angle of the mutant remained similar to that of the native Fab (∠_elb = 222° for the mutant, ∠_elb = 219° for native). RMSD analysis (Table S4) indicated no significant structural deviations, with an overall backbone RMSD of 0.82 Å between the two Fabs.

3.4. Clinical lambda-based fabs

As the experimental data were consistent with our MD simulations, we extended our MD analysis to the Fab regions of three clinical mAbs. Given the extensive optimization typically performed on clinical-grade antibodies, we asked whether the corresponding glycine-to-threonine substitution would further reduce conformational dynamics and, by extension, enhance Fab stability. For each Fab (Guselkumab, Erenumab, and Avelumab) we conducted three independent MD simulations (Δt = 1 µs) of both the wild-type and Gly-to-Thr variants under the same conditions as in previous simulations.

In Guselkumab and Erenumab, the threonine substitution blocked transitions between the two major conformations observed in the native Fabs (Fig. 4a and b). For Avelumab, a brief transition to an extreme elbow angle occurred in one wild-type simulation but quickly reverted to the original conformation (Fig. 4c, orange traces). In all three cases, the Gly-to-Thr variant exhibited reduced elbow-angle fluctuations relative to the native Fabs.

Fig. 4.

Comparative Dynamics of Native and Mutant Therapeutic Antibody Fabs (Guselkumab, Erenumab, and Avelumab). (a-c) Elbow angles from three independent 1 µs simulations of native (orange) and mutant (teal) variants of: (a) Guselkumab (PDB code: 4m6m), (b) Erenumab (PDB code: 6umg), and (c) Avelumab (PDB code: 4nki). (d-f) PCA plots for native (orange) and mutant (teal): (d) Guselkumab, (e) Erenumab, and (f) Avelumab. (g-i) Normal mode analyses of native (orange arrows) and mutant (teal arrows): (g) Guselkumab, (h) Erenumab (native left, mutant right), and (i) Avelumab. Arrows indicate domain movements.

RMSF analysis showed that fluctuations in the variable domains of the Gly-to-Thr mutants remained within one standard deviation of the native Fab values for all three antibodies (Figure S4a for Guselkumab, Figure S5a for Erenumab, and Figure S6a for Avelumab). In Guselkumab, the threonine substitution led to a tighter clustering of ϕ/ψ dihedral angles for residues in the light chain hinge region (Figure S4b), but no significant change was observed in the heavy chain (Figure S4c). Similarly, in Erenumab, the mutation produced tighter dihedral clustering in both the light chain hinge (Figure S5b) and the heavy chain elbow region (Figure S5c). For Avelumab, tighter clustering of ϕ/ψ angles was again observed for both the light (Figure S6b) and heavy chain hinge regions (Figure S6c).

Principal component analysis showed a consistent reduction in PC1 magnitude for all three Gly-to-Thr mutants, with points clustering closer to zero compared to the wild-type Fabs (Fig. 4d–f; native in orange, mutant in teal). This trend was further supported by normal mode analysis, which showed reduced overall motion in the threonine-substituted Fabs (Fig. 4g–i; native in orange, mutant in teal).

3.5. Aggregate statistics across simulations

In our simulations, we calculated the SD of elbow angles (Supplementary Table S5) and defined flexibility as SD > 10°, because at the 10° threshold the mutant showed a highly significant shift (two-sided Fisher’s exact test, P = 0.0006; Supplementary Table S6). Using this cutoff, 2/15 native Fabs (13 %) were non-flexible (SD ≤ 10°), compared with 15/25 mutant Fabs (60 %; Supplementary Fig. S7a).

We also estimated the probability of remaining in the starting conformation over time from the elbow-angle trajectories. Native simulations gradually jumped to the new over the 1 µs imulation, whereas threonine-mutant simulations largely remained confined to the initial conformation (Supplementary Fig. S7b).

4. Discussion

Our study explores the structural dynamics and stability of λ-isotype antibodies, with a specific focus on the impact of hinge-region mutations on the Fab elbow angle and overall thermal stability. Through extensive MD simulations and biophysical characterization, we provide insights into conformational transitions and the stability enhancements achievable through a single amino-acid substitution. A survey of human germline sequences confirms invariance of Gly at the Fab elbow of λ antibodies, with no observed polymorphism; this Gly is absent at the equivalent position in κ (supplementary Table S7). Its invariance arises because hinge residues are specified by framework and constant domains, not by V(D)J diversification.

The MD simulations revealed significant conformational changes in the elbow angle of the Fab region, predominantly driven by the dihedral angles of specific hinge residues. Substituting glycine with threonine at residue 111 in the light-chain hinge (G111T) restricted conformational flexibility and stabilized the Fab structure. This restriction was evident from tighter clustering of dihedral angles within defined regions of the Ramachandran plot and from the stabilized elbow angles observed across simulations. Principal component analysis (PCA), along with reduced amplitudes of motion in normal-mode analysis (NMA), further supported this finding. The threonine substitution resulted in a single dominant conformational population, in contrast to the two distinct populations observed in the native Fab. These results suggest that the G111T mutation imposes a higher free-energy barrier, limiting conformational transitions and enhancing structural stability.

Experimental validation using differential scanning fluorimetry (DSF) demonstrated a measurable increase in the melting temperature of the threonine-substituted Fab compared to the native Fab, indicating improved thermal stability. Crystallographic analysis further corroborated the simulation results, showing that the elbow angle remained consistent between native and mutant Fabs. This suggests that crystallization conditions, including crystal contacts, likely stabilize the observed conformation. Surface plasmon resonance (SPR) measurements using the anti-NANP system confirmed that the G111T substitution had no adverse effect on antigen binding. The kinetic profiles and affinities of the wild-type and mutant Fabs were comparable, indicating that the enhanced stability does not compromise antigen recognition or functional efficacy.

To assess whether these findings could be generalized, the analysis was extended to additional λ light-chain antibodies, including the anti-NANP Fab and three clinically relevant Fabs (Guselkumab, Erenumab, and Avelumab) whose sequences have been extensively validated for stability, expression, and developability. This provides a consistent experimental baseline and avoids confounding effects arising from poorly folded or aggregation-prone antibodies. Across all systems, the glycine-to-threonine substitution consistently reduced conformational fluctuations and, in the case of the anti-NANP Fab, improved thermal stability, in agreement with the results from AMMO1.

Variation in the elbow angle could, in principle, alter the VH–VL relative orientation and modulate the paratope geometry. Indeed, a single mutation in the Fab region may hinder paratope binding in solution [41]. Additionally, molecular dynamics (MD) simulations suggest that framework mutations can significantly shift the distribution of VH–VL interface angles, leading to local changes in antibody flexibility through alterations in the solvent-accessible surface area [42]. Importantly, our experimental affinity data indicate that such structural fluctuations do not measurably affect antigen binding. Due to the complexity and cost of antigen preparation for SPR experiments, the binding study was limited to the anti-NANP Fab, which binds a small peptide. While a detailed analysis of paratope distortion along the trajectories would be valuable, it is beyond the scope of the present study.

5. Conclusion

Monoclonal antibodies are a major therapeutic platform, and increasing their stability remains a core engineering goal. Here, a single hinge substitution (G111T) in λ-type antibodies improved thermal stability without compromising antigen binding. Molecular dynamics and biophysical data indicate that G111T restricts Fab elbow flexibility and enhances structural resilience. Together, these results provide a practical, generalizable strategy for stabilizing λ light-chain Fabs with minimal structural disruption and potential benefits for development, formulation, and long-term storage.

Authorship contribution statement

J.C.W. conceived, designed, and secured funding for the study. Y.J. performed molecular dynamics simulations. M.P. oversaw Fab production, as well as SPR and DSF experiments. K.L. and A.G. produced Fabs and conducted SPR and DSF experiments. T.Y. contributed to structure determination. Y.J., J.C.W., M.P., T.Y., and C.M. contributed to manuscript preparation, figure generation, and revisions. All authors reviewed and approved the final submitted manuscript.

CRediT authorship contribution statement

Williams John Charles: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Resources, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis. Miso Park: Writing – review & editing, Visualization, Validation, Supervision, Methodology, Formal analysis. Kevin Ly: Methodology, Data curation. Gonzalez Asual: Methodology, Data curation. Mallet T Conn: Writing – review & editing, Writing – original draft, Visualization, Validation, Formal analysis, Data curation. Yead Jewel: Writing – review & editing, Writing – original draft, Visualization, Validation, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Tynan Young: Writing – review & editing, Writing – original draft, Visualization, Validation, Investigation, Formal analysis, Data curation.

Declaration of Competing Interest

The authors declare that this research was conducted without any commercial or financial relationships that could be construed as a potential conflict of interest.

Appendix A. Supplementary material

Supplementary material

Supplementary material

Data availability

The raw data supporting the conclusion of this article will be made available by the authors, without undue reservation. The atomic coordinates of AMMO1 G111T, anti-NANP Fab & anti-NANP G112T were deposited in the Protein Data Bank under the accession codes 9Q0P, 9PZ5 and 9PZ6, respectively.