All tests and the biodemographic/individual information have been approved by the Lehigh University Institutional Review Board. Recruitment was done widely, with the majority of participants being Lehigh University students, but also included students, faculty and administrators at Lehigh and other institutions. All subjects were adults over the age of 18. We collected background information from our participants, which included the following: sex, age, professional/career goals, hobbies/hobby frequency, mental health diagnoses, highest degree attained, medications, smoking habits, alterations related to identity, and previous experience with cognitive/anxiety experiences. This comprehensive data helped us gain insights into the diverse experiences and backgrounds of our participants, allowing us to consider other factors that may contribute to the relevance of our study (Supplementary Table S1). We collected from 624 participants who completed the anxiety questionnaires (STICSA and/or STAI), 179 male participants and 416 female participants, the balance being participants who answered “other” or did not answer that question. The ratio of male and female participants reflected the demographics of the recruitment pool, who were largely Lehigh University natural science students. The average age was 20.7 +/− 0.28 years and ranged from 18 to 63 with a median age of 20. All data underwent a double coding system to ensure anonymity, confidentiality and security of data. Collection of data and samples occurred in batches of roughly 50–100 participants, and samples from each batch were analyzed by different sets of researchers in different locations to reduce the likelihood that a detected SNP allele was a result of a spurious event such as contamination.

The State–Trait Inventory for Cognitive and Somatic Anxiety (STICSA) (Ree et al., 2008) and State–Trait Anxiety Inventory (STAI) (Spielberger et al., 1983) were used to examine human anxiety levels. These tests assess subjects’ chronic and current anxiety levels based on self-reports. The STICSA takes cognitive and somatic anxiety dimensions into account. Both tests provide summarized state and trait anxiety scores calculated from frequency ratings that subjects provide for a variety of anxiety-related symptoms. Participants are asked to report on the anxiety symptoms they are experiencing at the time of the test (state) as well as the anxiety symptoms they experience more generally, outside the bounds of the test and throughout their daily lives (trait). The STAI furthermore allows for a standardized interpretation taking into account inherent anxiety level differences in different populations (based on age, student status, etc.). In addition to anxiety tests, a battery of creativity tests which measure convergent and divergent thinking, including the Creativity Behavioral Inventory (CBI), Remote Associates Test (RAT), and Alternative Uses Task (AUT), were used to examine human creativity due to previous reports of the correlation between creativity and mental illness (Andreasen, 1987; Runco and Richards, 1997; Kyaga et al., 2011; Kyaga et al., 2012). The researchers analyzing the psychology data were different from those performing DNA analysis, to keep the information about SNP carriers blind to those analyzing psychological data.

Genomic DNA was extracted from saliva samples collected from each participant during psychological testing. Saliva and cheek cells were deposited into conical tubes by the participants after rinsing with a 0.9% saline solution for 60 s. The saline and saliva samples were kept on ice at 4°C for up to one week. A portion of the saliva samples were used for genomic DNA extraction with Chelex resin (BioRad 1421253) following the manufacturer’s recommendations within a week of collection. The remainder of the saliva-saline samples were moved into long-term storage at −20° or − 80°C. Exons 3–4 of the LYNX2 gene representing the mature protein coding sequence were amplified with PCR primers (Table 1) 100–200 bp outside of the exon/intron boundaries. Primers were identified by comparing isoform cDNA sequences from NCBI (GeneBank name: LYPD1, chromosomal location 2q21.2). Successful PCR reactions were purified using ExoSap (Affymetrix cat no.: 78200) or by gel excision of visualized bands and gel extraction (GeneJET Gel Extraction Kit, ThermoFisher, K0691). Purified PCR products were sent to GenScript USA for Sanger sequencing. Chromatogram analyses of sequence files were performed using DNA Baser Sequence Assembler software to generate contigs of overlapping sequences under the high-quality option. If a contig did not form under high quality, the results were discarded, and PCR was performed again for high-quality sequence results.

Alternative allele detection was conducted using manual identification of the chromatograms by more than one individual to verify alternative allele t identification. Alignment to the reference genome (GRCh38) was done using Molecular Evolutionary Genetics Analysis (MEGA) software. We performed 2x sequence coverage on all samples using both a forward and reverse sequencing primer, analyzing in large batches of samples, with further in-depth (4–6x) sequence coverage for select samples. Samples from individuals with variant alleles and a representative number of reference samples were further confirmed through extra rounds of amplification and sequencing, as well as re-extraction of DNA, each performed by unique researchers. Eight of the nine identified SNPs were in dbSNP.¹ The ninth SNP was not previously reported in dbSNP, Gnomad², or Bravo³ at the time of publication.

The structure of two adjacent α7 subunits (α7:α7) was taken from the Protein Data Bank (PDB) entry 3SQ6 (Li et al., 2011). The α7 subunits have a closed state conformation of loop C. We obtained 350 α7:α7 snapshots with apo-state loop C conformations from a 10 ns molecular dynamics (MD) simulation by applying a weak positional restraint to glutamate185 Cα atom in the loop C.

The three-dimensional (3D) structure of LYNX2 was predicted from the sequence using the standalone I-TASSER suite (Yang et al., 2015) by specifying a LYNX1 PDB structure (PDB entry 2 L03) (Lyukmanova et al., 2011) as a template. We obtained 100 energy-minimized snapshots from a 5 ns MD simulation of the LYNX2 model and then identified the best snapshot among them by evaluating the quality of each structure using ModFOLD (Maghrabi and McGuffin, 2017). For the best snapshot, we sampled thermally acceptable 50 conformations for LYNX2 loop I using the Backrub application in Rosetta (Smith and Kortemme, 2008).

A low-resolution structure model of α7 nAChR in complex with LYNX2 was generated based on our LYNX1/α4:α4 complex model (Nissen et al., 2018). The α7 subtype was chosen because a homomeric subtype was more tractable to model than heteromeric nAChRs α4β2 and α4β2 subtypes, to which LYNX2 also binds. A set of LYNX2/α7:α7 complex models was built by mapping each α7 subunit (in the ensemble of 350 α7:α7 conformations) and each LYNX2 (in the ensemble of 50 LYNX2 conformations) onto α4:α4 and LYNX1 in the LYNX1/α4:α4 model, respectively, thus generating 350 × 50 combinations. For each complex, LYNX2 was moved outwards from α7:α7 by ≤2 Å using an interval of 1 Å along a vector defining the geometric centers of two protein binding surfaces to find a complex structure with the minimum number of bad (energetically unfavorable) contacts between α7:α7 and LYNX2, which was determined as the final model of α7 nAChR in complex with LYNX2 (Supplementary Figure S1).

The LYNX2/α7:α7 complex model was subjected to 135 ns MD simulations to obtain a refined complex model. 5 ns simulations with or without weak positional restraints were iterated during the first 35 ns simulations. The positional restraint was applied to the Q39 Cα atoms of LYNX2, aiming at maintaining LYNX2 loop II in the ligand-binding pocket under α7:α7 loop C by gradually reducing the force constant. We also applied positional restraints to prevent α7:α7 from dissociating; no restraints were applied to loop B, loop C and β8–β9 loop in α7:α7, allowing for conformational flexibility, as they involve protein–protein interactions. After the 35 ns simulation, we removed the positional restraints to LYNX2 and performed additional 100 ns simulations for equilibration.

The Q61H mutation of the full-length LYNX2 gene (the nascent polypeptide before cleavage of the signal sequence) is at amino acid 39 in the fully-processed mature LYNX2 protein and will be referred to as Q39H. To generate the LYNX2 Q39H alternative allele, we computationally replaced LYNX2 Q39 with histidine in the α7:α7/ LYNX2 complex structure refined by MD simulations (Supplementary Figure S1). We carried out 400 ns MD simulation without any positional restraints to LYNX2 for three independent replicates (MD-1, MD-2, and MD-3) of α7:α7 in complex with native LYNX2 and its Q39H alternative allele. The convergence of the simulations was assessed by measuring root-mean-square deviation (RMSD) of LYNX2 in terms of simulation time, indicating that MD-3 s of each system show good convergence (Supplementary Figure S1C). The trajectories of these well-converged replicates were then analyzed to calculate nonbonded interaction (van der Waals and electrostatic) potential energies between the entire α7:α7 interface and LYNX2. Note that, to calculate the RMSD, we first aligned α7:α7 of each MD snapshot to the initial structure (i.e., the structure before the MD simulations) and then calculated the RMSD of the entire LYNX2 Cα atoms relative to its initial structure. Therefore, the calculated RMSD includes both the LYNX2 conformational changes and the changes of LYNX2 positions relative to α7:α7. Given the flexibility of α7:α7/LYNX2 interactions, RMSD values above 5 Å appear to be reasonable and α7:α7 and LYNX2 interact strongly as shown in Supplementary Figure S1C (the interaction energy time series) (Dong et al., 2020).

All MD simulations were prepared through CHARMM-GUI (Jo et al., 2008; Lee et al., 2016) and performed with explicit water molecules and 150 mM NaCl using CHARMM36m force field (Huang et al., 2017) and NAMD software (Phillips et al., 2005) at 300 K.

Recombinant hexahistidine-tagged human LYNX2 gene was made in the pcDNA3.1 plasmid, expressed in HEK 293 T cells and purified from culture supernatant using Ni-NTA chromatography. The expression and purification of the protein were confirmed by SDS-PAGE and Western blot using an anti-his-HRP antibody.

Purified recombinant LYNX2 was attached to an atomic force microscopy (AFM) cantilever by covalently cross-linking the two using a heterobifunctional PEG linker (Ebner et al., 2007). The AFM cantilevers (MLCT: Bruker Nano, Camarillo, CA) were first silanized with 3-aminopropyltriethoxysilane (APTES) using a gas phase coating method (Wildling et al., 2011). An acetal-PEG27-NHS linker was then attached to the silanized cantilevers in chloroform via its N-Hydroxysuccnimide (NHS) group and then rinsed and dried under nitrogen gas. Next, the acetal group of the PEG linker was converted to an aldehyde group using a 1% citric acid solution vol/vol, and 1 μM of LYNX2 was then added to the solution, which coupled to the activated PEG linker. After quenching the free aldehyde group with ethanolamine, the cantilevers were stored in TBS buffer (50 mM Tris, 0.15 M NaCl, pH 7.4) until experimental use.

All single-molecule force measurements were conducted using a custom-designed AFM that employs a single-axis piezoelectric translator equipped with a strain gauge (Physik Instrumente, Waldbronn, Germany) to control the absolute position of the AFM cantilever (Chen and Moy, 2002; Zhang et al., 2002; Zhang et al., 2004; Fu et al., 2015). The deflection of the cantilever was monitored optically using an inverted optical system attached to the AFM. A focused laser spot from a fiber-coupled diode laser was reflected off the back of the cantilever onto a two-segment photodiode to monitor the cantilever deflection. The photodiode signal was then pre-amplified, digitized and processed by a computer. The force apparatus was suspended inside a refrigerator housing to reduce both mechanical and thermal instabilities.

The human α7 nAChR receptor was expressed on SH-EP1 cells by transfection of pCEP4 vector containing the cDNA of the α7 receptor. The stably transfected SH-EP1-hα7-nAChR cell line was previously engineered and characterized as described in Zhao et al. (2003). The transfected cells were maintained in media supplemented with 0.13 mg/mL hygromycin B. The cells were plated on 35 mm cell culture dish 24 h prior to AFM force measurements.

The AFM force measurements were performed on an apparatus designed for operation in the force spectroscopy mode (Zhang et al., 2002; Zhang et al., 2004; Zhang et al., 2009; Fu et al., 2015). Using a piezoelectric translator, the LYNX2-functionalized cantilever tip was lowered onto a live SH-EP cell expressing α7 nAChRs (Figure 2A). Tip-cell contact may (or may not) lead to LYNX2/α7 nAChR interactions. The binding interaction was then determined from the deflection of the cantilever via a position-sensitive two-segment photodiode. To calibrate the cantilever (320 μm long by 22 μm wide triangle), the spring constant at the tip was characterized via thermally induced fluctuations (Hutter and Bechhoefer, 1993). The spring constants (9.7 ± 1.9 pN/nm) of the calibrated cantilevers agreed with the values specified by the manufacturer.

Except for the adhesion specificity test (Supplementary Figure S1D), the contact time and indentation force between the cantilever and the sample were minimized to obtain measurements of the unitary unbinding force. An adhesion frequency of ~30% in the force measurements ensured that there was a > 83% probability that the adhesion event was mediated by a single LYNX2/α7 nAChR bond (Evans, 2001). AFM measurements were collected at cantilever retraction speeds ranging from 0.1 to 10 μm/s in order to achieve the desired loading rate (~30–4,000 pN/s). All measurements were conducted at 25°C in the cell culture medium. Loading rates were determined directly from the force-extension data by multiplying the system spring constant (Figure 2A) of the unbinding pulling trace by and the retraction speed of the cantilever.

The Bell-Evans model, a theory to determine energy landscape properties, describes the influence of an external force on the rate of bond dissociation (Bell, 1978; Evans and Ritchie, 1997). According to this model, a pulling force, f, distorts the intermolecular potential of a receptor-ligand complex, leading to a lower activation energy and an increase in the dissociation rate k(f) or a decrease in the lifetime t(f) as follows:

where k⁰ is the dissociation rate constant in the absence of a pulling force, γ is the width of the activation barrier, or the position along the reaction coordinate relative to the bound state, T is the absolute temperature, and k_B is the Boltzmann constant. For a constant loading rate r_f, the probability density for the unbinding of the complex as a function of the pulling force f is given by:

with the most probable unbinding force f*.

Hence, the Bell-Evans model predicts that the most probable unbinding force f* is a linear function of the logarithm of the loading rate. Experimentally, f* was determined from the mode of the unbinding force histograms. The Bell-Evans model parameters were determined by fitting Equation 3 to the plot of f* versus ln(r_f).

The STAI (State–Trait Anxiety Inventory) was administered in this study because it distinguishes clearly between “state” anxiety, which refers to anxiety being experienced at the time of assessment, and “trait” anxiety, which predicts the degree to which a person is prone to generalized anxiety. To assess participants’ (n = 624) state and trait anxieties, the complete STAI test was administered to two cohorts, which we will refer to as the “discovery cohort” and the “replication cohort.”

Half of the STAI addresses state anxiety, and participants are asked to answer assessment items intuitively and in accordance with their present feelings. This part of the test includes items such as “I feel calm,” “I am tense,” and “I feel satisfied,” and participants are asked to characterize their agreement with the statement on a scale of 1–4. 1 corresponds to the answer “Not at all,” and 4 to the answer “Very much so.”

The other half addresses trait anxiety, in which participants are asked to answer in accordance with their feelings in general, and beyond the scope of the time in which the test was taken and into their daily lives. This part of the test includes items such as “I feel pleasant,” “I feel satisfied with myself,” and “I lack self-confidence,” and participants are asked to characterize their agreement with the statement on the same scale of 1–4.

To score the STAI, we added the scoring weights of each question in accordance with the official manual. For the statements which indicate more anxiety (e.g., “I feel tense”), a participant scoring of 4 corresponds to an item weight of 4, and a participant scoring of 1 corresponds to an item weight of 1. For the statements that, inversely, indicate less anxiety (e.g., “I feel calm”), a participant scoring of 4 corresponds to an item weight of 1, and a participant scoring of 1 indicates an item weight of 4, such that the increased weight of an item corresponds to more anxiety-expressing responses.

The scoring of the STICSA was comparable to that of the STAI. The STICSA also makes the distinction between state and trait anxieties, and total scores were calculated by the summation of the items’ weights, in accordance with the STICSA manual. STAI responses were analyzed according to the STAI manual.

Statistical analysis involved comparing scores of human LYNX2 mutant populations to scores of the LYNX2 WT, or reference, population. This was done with a two-tailed Fisher’s Exact Test, or an independent two-tailed t-test, to determine whether there existed a statistically significant difference between the means of the groups. In some tests, we combined state and trait results of the STICSA and STAI to produce a “total score,” and in others, we made the distinction between state and trait in order to test them separately. In performing a Fisher’s Exact Test to compare scores between human participants with and without the LYNX2 Q39H mutation, we chose a threshold score of 43 for the STICSA test based on literature determining that it was the score that could predict incidence of an anxiety disorder in individuals (Van Dam et al., 2013). For the STAI, we chose a threshold score of 45 as a score that has been used in literature to represent “high anxiety” (Kayikcioglu et al., 2017).

The Creative Behavioral Inventory, CBI, was analyzed with a Fisher Exact Test comparing the average total CBI score of those with the Q39H mutation and the reference group. The CBI total score is determined by summing the individual scores of its components, which are Literature, Music, Crafts, Art, Math & Science, and Performing Arts. Average scores for mutant and reference groups were found and compared, according to each population’s size.

For each pulling speed, over 1,000 force curves were recorded, which yielded 80 to 350 unbinding forces. Curve fitting was performed using IGOR Pro or Origin software by minimizing the chi-square statistic for the optimal fit. Unless otherwise stated, the data is reported as the mean and the standard error of the estimate. Statistical analyses between groups were performed using an unpaired t-test or ANOVA, with a p-value less than 0.05 considered to be statistically significant.

To assess if LYNX2 has an impact on human anxiety, we investigated whether there is a correlation between variation in the LYNX2 gene and self-reported anxiety in a population of 624 participants. The average age of our available population (M = 20 +/− 0.28 years) was below the average age of clinical diagnosis of anxiety disorders at the time of testing (M = 30, Gonçalves and Byrne, 2012; Vaingankar et al., 2013; Lijster et al., 2017). We used two established tests of self-reported anxiety for phenotyping: the State–Trait Inventory for Cognitive and Somatic Anxiety (STICSA) and the State–Trait Anxiety Inventory (STAI). The STICSA was chosen because it is designed to discriminate between symptoms of anxiety and depression. These two disorders are often comorbid, which can make interpretation of genetic risk for anxiety difficult. The STAI was chosen because it can discriminate between the state and trait components of anxiety (Oei et al., 1990; Grӧs et al., 2007; Elwood et al., 2012; Van Dam et al., 2013).

After extracting DNA and performing targeted Sanger sequencing, we detected SNPs in nine different positions across the mature protein coding sequence of LYNX2 (Figure 3A). Eight of the nine SNPs were previously reported in dbSNP, and consisted of a mix of synonymous, missense, and frameshift mutations (Table 2). We did not find enough participants harboring these eight mutations to draw conclusive results about how they function in anxiety. One SNP, however, was not previously reported: (Tables 2, 3) a C → T missense variant resulting in a glutamine (Q) to histidine (H) substitution at amino acid 61 of the full-length protein, and amino acid 39 of the mature protein. We will refer to this variant as Q39H. This was the most common SNP variant in our population, occurring in 14/624 (2.2%) of participants, and was exclusively heterozygous. The Q39H variant was found across different batches of samples throughout the course of the study (Supplementary Table S1). None of the nine variants had been previously associated with anxiety or had pre-existing reports of clinical significance.

As a group, individuals harboring any alternative SNP allele in LYNX2 reported significantly higher STICSA and STAI scores than those with a reference allele (p = 0.005; Cohen’s d = 1.34, Table 3) (Figure 1B). The individuals with the Q39H mutation, in particular, were associated with significantly increased anxiety scores comparable to those clinically diagnosed with an anxiety disorder. Including anxiety, panic attacks, and PTSD. In the STICSA test (Figure 3), the Q39H group was significantly different from the reference group in all measured anxiety tests, with a general value (representative of trait anxiety) of 48.43 +/− 2.11, and the reference group 36.60 +/− 0.40 (p < 1.0 × 10⁻⁴), and “at the moment” value of 45.86 +/− 2.50 compared to the reference group of 33.92 +/− 0.40 (p < 1.0 × 10⁻⁴). In the STAI test, the Q39H group was also significantly different from the reference group, with a “general” value (representative of trait anxiety) of 54.07 +/−2.79, and the reference group 43.28 +/− 0.49 (p = 4.0 × 10⁻⁴), and “at the moment” value (representative of state anxiety) of 46.93 +/− 2.79 compared to the reference group of 37.94 +/− 0.49 (p = 0.0281) (Table 2; Figure 3). We found two individuals harboring a synonymous mutation at the same location (nucleotides encoding amino acid 39 of the mature protein and amino acid 61 of full length LYNX2) with anxiety levels similar to the reference group. This suggests that only the nonsynonymous mutations confer a functional consequence.

We further analyzed the anxiety test results by comparing answers from questions that matched the Diagnostic and Statistical Manual of Mental Disorders (DSM-5) criteria for Generalized Anxiety Disorder (GAD), an anxiety disorder reflective of increased trait anxiety (Chambers et al., 2004). We identified 11 questions that matched diagnosis symptoms. In these DSM-aligned questions, the Q39H group answers reflected significantly higher anxiety scores (Figure 3D). These data suggest a relationship between the LYNX2 mutation Q39H and heightened anxiety levels. The individuals with the Q39H mutation in LYNX2 also had anxiety scores comparable to those of individuals with diagnosed panic disorder (Van dam et al., 2013). 61.5% of individuals harboring the Q39H had high anxiety levels, above the levels for GAD, whereas in the reference group, 15.9% of respondents exhibited such levels. There was good concordance in the entire dataset between the STICSA and STAI values including the reference group and the Q39H carriers (Figure 4). This suggests a functional consequence of the mutation in elevating anxiety levels, indicating its potential role in predisposing individuals to heightened anxiety traits.

To explore a previously discovered link between creativity and heightened trait anxiety levels, we performed and analyzed the results of the Creative Behavioral Inventory (CBI), taking interest in the comparison of scores between those harboring the Q39H mutation and those not (Figure 5). We found there to be a significant difference between the average scores of the groups, with a Q39H group score of 30.35 +/− 4.79 and a reference group score of 43.46 +/− 1.62 (p = 0.45). This provides evidence for a link between the Q39H mutation and creativity.

To study the possible effect of the most common Q39H variant on nAChR binding, we performed computational modeling and binding studies with the human LYNX2 protein and a nAChR subtype. The Q39H mutation is at amino acid 39 in the mature protein (Figure 2A). In the mature protein the 39th amino acid mutation site is at the tip of loop II (Figure 2A), a region associated with nAChR binding similar to three-looped nAChR-binding proteins like α-bungarotoxin (Tsetlin, 2014) and LYNX1 (Lyukmanova et al., 2013).

To examine the possible effect of the mutation at the molecular level, we performed computational modeling and simulations of LYNX2 binding to the α7:α7 interfaces of α7 nAChR (Figure 2), a common nAChR subtype expressed in anxiety pathways of rodents and shown to be critical in regulating amygdala excitability (Klein and Yakel, 2006; Pidoplichko et al., 2013; Almeida-Suhett et al., 2014; Jiang et al., 2016). We chose the α7 subtype to model because a homomeric subtype was more tractable to model than heteromeric nAChRs α4β2 and α4β2 subtypes. A low-resolution structural model of the α7 nAChR in complex with LYNX2 was generated based on our LYNX1/α4:α4 complex model (Nissen et al., 2018). A set of LYNX2/α7:α7 complex models was built by mapping each α7 subunit (in the ensemble of 350 α7:α7 conformations) and each LYNX2 (in the ensemble of 50 LYNX2 conformations) onto α4:α4 and LYNX1 in the LYNX1/α4:α4 model, respectively, thus generating 350 × 50 combinations. From this we selected the complex structure with the minimal number of energetically unfavorable contacts as the final model (Supplementary Figure S1).

The modeling indicates that the wild-type LYNX2 forms a complex with the α7:α7 subunit interface through multiple salt bridges with E8, R46 and K70 of LYNX2 (Figure 2B). However, in the Q39H LYNX2 variant, the LYNX2 and α7:α7 subunit interface salt bridges are established only by R46 of LYNX2 (Figure 2B). Further, three residues in the loop II of WT LYNX2, including Q39, form additional hydrogen bonds with α7:α7. These hydrogen bonds are, however, not observed in the α7:α7 complex with the Q39H variant LYNX2. We measured intermolecular interaction energy between α7:α7 and LYNX2 during the 400 ns molecular dynamics (MD) simulations (Figure 2C). The results indicate a more favorable interaction of WT LYNX2 with α7:α7 after 200 ns than the Q39H variant. Therefore, our modeling and simulations suggest that the Q39H LYNX2 variant would have a lower binding affinity to α7-nAChRs and thus suppresses the function of α7-nAChRs less than a WT protein.

To further gain insight into the molecular consequences of naturally occurring variants underlying the LYNX2/nAChR interactions, we applied single-molecule atomic force spectroscopy (AFM) to study the reaction kinetics between LYNX2 and α7 nAChRs at the single-molecule level. For these studies, purified recombinant LYNX2 was covalently attached to an AFM cantilever and a SH-EP1 mammalian cell line stably expressing concatemeric α7 nAChRs (Pisapati et al., 2021) were used (Figure 6A). LYNX2/α7 nAChR interactions were analyzed by dynamic force spectroscopy (DFS), the plot of most probable unbinding force (Figure 6B, Equations 1–3). The unbinding force of the LYNX2/α7 nAChR interaction increased with the logarithm of the loading rate, ranging from 60 to 90 pN over a loading rate of 30–4,000 pN/s, respectively (Figure 6B, Equations 1–3). The LYNX2 variant with the Q39H mutation found in the human population resulted in significantly decreased unbinding force under all tested loading rates tested. The mutation shortened the lifetime of the interaction by 180-fold (from 1,470 to 8.1 s). This result indicates that this human mutation in LYNX2 weakens the physical association with α7 nAChRs in vitro.