PaperlayerView exact revision on arXiv

Source-linked AI summary

Measuring the sequence-affinity landscape of antibodies with massively parallel titration curves

Rhys M. Adams, Thierry Mora, Aleksandra M. Walczak, Justin B. Kinney

arXiv:1601.02160v2q-bio.QMq-bio.BM

TL;DR

Quantitative links between antibody sequence and antigen affinity remain poorly characterized. Tite-Seq measures binding titration curves for thousands of antibody variants in parallel, and its application to CDR1H and CDR3H variants revealed distinct affinity and expression patterns alongside structural associations. The authors caution that these observations come from a single antibody system, while presenting Tite-Seq as adaptable to other compatible binding assays.

  • Problem

    Quantitative sequence–affinity relationships remain unclear, despite their relevance to understanding adaptive immunity and interpreting antibody repertoires.

    Source: §I · p. 1

  • Method

    Tite-Seq applies yeast-display Sort-Seq across antigen concentrations to infer binding curves and absolute KD values for many variants.

    Source: §I · p. 2

  • Results

    In the studied scFv, CDR3H variation affected affinity more than CDR1H variation, while CDR1H was more optimized for expression; affinity effects also correlated with intraprotein molecular contacts.

    Source: §III · p. 8

  • Takeaways & Limitations

    Tite-Seq can be applied to protein-binding assays compatible with yeast display and FACS and may be adapted to other expression platforms.

    Source: §III · p. 8

    The CDR-region observations are limited in scope, and the authors caution against generalizing them to arbitrary antibody–antigen interactions.

    Source: §III · p. 8

Abstract

from arXiv · show

Despite the central role that antibodies play in the adaptive immune system and in biotechnology, much remains unknown about the quantitative relationship between an antibody's amino acid sequence and its antigen binding affinity. Here we describe a new experimental approach, called Tite-Seq, that is capable of measuring binding titration curves and corresponding affinities for thousands of variant antibodies in parallel. The measurement of titration curves eliminates the confounding effects of antibody expression and stability that arise in standard deep mutational scanning assays. We demonstrate Tite-Seq on the CDR1H and CDR3H regions of a well-studied scFv antibody. Our data shed light on the structural basis for antigen binding affinity and suggests a role for secondary CDR loops in establishing antibody stability. Tite-Seq fills a large gap in the ability to measure critical aspects of the adaptive immune system, and can be readily used for studying sequence-affinity landscapes in other protein systems.

I. INTRODUCTION

Antibody CDRs shape antigen recognition, but quantitative links between sequence and affinity remain poorly understood. Tite-Seq measures binding curves and absolute affinities in parallel, and the study applies it to CDR1H and CDR3H variants of an anti-fluorescein scFv.

I. INTRODUCTION

  • CDR3H is highly variable and typically contributes most to antigen specificity, while the roles of other CDRs are less clear.The six CDRs are each roughly 10 amino acids long; some often do not contact the target directly.Source: §I · p. 1
  • Quantitative sequence–affinity relationships remain an open question, including how many antibodies bind at a specified affinity and how CDR positions interact.These questions matter for understanding adaptive immunity and using repertoire sequencing to diagnose and monitor disease.Source: §I · p. 1
  • Existing DMS approaches had not established quantitative dissociation constants in molar units, and SORTCERY required separate low-throughput calibration.For folded proteins, it was also unclear whether SORTCERY could separate affinity effects from sequence-dependent expression and stability effects.Source: §I · p. 2
  • Tite-Seq applies Sort-Seq to yeast-displayed antibodies across antigen concentrations to infer thousands of titration curves and absolute KD values without expression or stability confounding.The assay measures affinities over many orders of magnitude.Source: §I · p. 2
  • The study tested CDR1H and CDR3H variants of a fluorescein-specific scFv and validated measurements against low-throughput binding curves.The measurements reveal differences between the regions’ mutation effects and inform structural aspects of antigen recognition independent of antibody stability.Source: §I · p. 2

A. Overview of Tite-Seq

Tite-Seq estimates antibody affinities by measuring binding across antigen concentrations rather than relying on a single fluorescence readout. Repeated sorting and sequencing yield per-variant mean-bin measurements that can be fit to binding curves.

A. Overview of Tite-Seq

  • Tite-Seq repeats Sort-Seq at multiple antigen concentrations and fits sigmoidal curves to each variant’s mean bin number to infer KD.The mean bin number is used as a proxy for mean cellular fluorescence and bound antigen.Source: A. Overview of Tite-Seq · p. 2
  • Figure 1 depicts sorting and sequencing variant antibodies into fluorescence bins, then using mean-bin measurements across concentrations to infer curves and KD.Its panel E is simulated under idealized conditions; inference from real data must account for multiple sources of experimental noise.Source: Figure 1 · p. 3
  • Affinity is quantitatively readable only when labeling antigen concentration is comparable in magnitude to KD, so measurements must span the variation caused by mutations.The binding relationship is sigmoidal, and single mutations can shift KD by multiple orders of magnitude.Source: A. Overview of Tite-Seq · p. 3
  • Single-concentration sorting can misrank affinity because a high-affinity, low-expression variant may bind less antigen than a low-affinity, high-expression variant.Full titration curves are needed to separate sequence effects on affinity from effects on functional protein expression.Source: A. Overview of Tite-Seq · p. 3

B. Proof-of-principle Tite-Seq experiments

The proof-of-principle experiments assayed pooled CDR1H and CDR3H scFv libraries alongside control clones. Tite-Seq used antigen titrations, fluorescence sorting, sequencing, and curve fitting to infer variant KD values.

B. Proof-of-principle Tite-Seq experiments

  • The pooled 1H and 3H libraries were assayed with WT, OPT, and nonfunctional Δ scFvs as controls.Each library varied a 10-aa region containing CDR1H or CDR3H.Source: B. Proof-of-principle Tite-Seq… · p. 4
  • Cells were labeled at eleven fluorescein-biotin concentrations and sorted into four PE-fluorescence bins, with a separate BV sort measuring scFv expression.The experiment used three independent replicates performed on different days.Source: B. Proof-of-principle Tite-Seq… · p. 4, Figure 3 · p. 4
  • The sequencing dataset averaged 2.6 × 10^6 reads per bin across all 48 bins.Sorted bins were regrown, DNA-extracted, and the 1H and 3H regions were PCR-amplified and sequenced.Source: B. Proof-of-principle Tite-Seq… · p. 4
  • Variant KD values were inferred by independently fitting a sigmoidal mean-bin binding curve to each Tite-Seq replicate.The fitting procedure accounts for multiple sources of experimental noise beyond the schematic illustration.Source: B. Proof-of-principle Tite-Seq… · p. 4
  • The study separately determined each scFv’s expression level from BV-labeled Sort-Seq data, scaling synonymous WT variants to a mean expression value of 1.0.Expression was represented by the mean bin E for each scFv gene.Source: B. Proof-of-principle Tite-Seq… · p. 5

C. Low-throughput validation experiments

Low-throughput flow cytometry binding curves provided a clone-level validation comparison for Tite-Seq measurements.

C. Low-throughput validation experiments

D. Tite-Seq can measure dissociation constants

Tite-Seq affinity measurements agreed with low-throughput flow-cytometry measurements and were reproducible across replicates. Single-concentration enrichment poorly captured full-curve affinities, while simulations supported the analysis pipeline.

D. Tite-Seq can measure dissociation constants

  • Tite-Seq measured WT scFv KD at 10^-8.87±0.02 M, consistent with 10^-8.61±0.07 M by low-throughput flow cytometry and 1.2 nM = 10^-8.9 M reported previously.The comparison averaged 33 synonymous variants across three Tite-Seq replicates and 10 flow-cytometry replicates.Source: D. Tite-Seq can measure dissociation… · p. 5
  • Across replicates, Tite-Seq KD measurements had Pearson coefficients of r = 0.82 to r = 0.89.Fit variability error bars in Fig. 4C also support reproducibility.Source: D. Tite-Seq can measure dissociation… · p. 5
  • At each tested antigen concentration, high-PE-bin enrichment correlated poorly with full-curve KD, and detectable correlations occurred only for scFvs with KD near that concentration.A single concentration also cannot separate affinity differences from differences in the fraction of displayed scFvs competent to bind antigen.Source: D. Tite-Seq can measure dissociation… · pp. 5–6
  • Simulated Tite-Seq data yielded KD values that agreed accurately with simulation inputs, supporting the analysis pipeline..Source: D. Tite-Seq can measure dissociation… · p. 6

E. Properties of the affinity and expression landscapes

Most substitutions in CDR1H and CDR3H weakened binding, with CDR3H more sensitive to mutation, while expression effects showed a different pattern. An additive matrix model correctly ranked OPT above WT but underestimated OPT’s affinity, consistent with assay limits and ignored epistasis.

E. Properties of the affinity and expression landscapes

  • 88% of CDR1H and 93% of CDR3H substitutions weakened binding; the median KD was 10^-6.84 in 1H versus ≳10^-5.0 in 3H (P = 4.7 × 10^-4).The majority increased KD above the 10^-5 M detection threshold (36% in 1H; 52% in 3H), and multiple random mutations tended to further deteriorate affinity.Source: E. Properties of the affinity and… · p. 6, Figure 5 · p. 6
  • The authors interpret the mutation patterns as consistent with WT CDR1H and CDR3H having been selected for high-affinity fluorescein binding.They describe this as a validation of the Tite-Seq measurements.Source: E. Properties of the affinity and… · p. 6
  • Five of six OPT-specific mutations reduced KD or were nearly neutral; D106E alone increased KD, though prior structural analysis suggests it may increase affinity with S101A.The comparison concerns OPT mutations within the assayed CDR1H and CDR3H regions.Source: E. Properties of the affinity and… · p. 6
  • The additive matrix model predicted OPT KD = 10^-9.16 M and correctly ranked it above WT, but the known OPT KD is 10^-12.6 M.The model covered only 6 of OPT’s 14 differences, censored W108L at 10^-9.5 M, and ignored epistasis; only 4.7 × 10^-5 of simulated six-mutation variants matched or exceeded its predicted affinity.Source: E. Properties of the affinity and… · pp. 6–7
  • Median expression effects did not differ between 1H (E = 0.826) and 3H (E = 0.822; P = 0.96), but effect variance was larger in 3H (P = 9.9 × 10^-16).Double or triple mutations further reduced expression in both CDRs; the authors infer that WT is more expression-optimized in CDR1H than CDR3H.Source: E. Properties of the affinity and… · p. 7

F. Structural correlates of the sequence-affinity landscape

The authors quantified how substitutions affect affinity and expression at each position, then related those sensitivities to structural features. Affinity sensitivity tracked intraprotein contacts but not distance to antigen; expression sensitivity tracked neither feature.

F. Structural correlates of the sequence-affinity landscape

  • At each position, SK and SE quantify sensitivity of KD and expression to substitutions, using WT measurements and an average over the 19 non-WT amino acids.The quantities compare WT KD and expression with corresponding values for single-substitution variants.Source: F. Structural correlates of the… · pp. 7–8
  • Affinity sensitivity SK strongly correlated with the WT residue’s number of intraprotein contacts, but not its distance to antigen.The structural comparison used the WT scFv in complex with fluorescein.Source: F. Structural correlates of the… · p. 8
  • Expression sensitivity SE did not significantly correlate with either intraprotein contact count or distance to antigen.Source: F. Structural correlates of the… · p. 8

III. DISCUSSION

Tite-Seq measures antibody sequence–affinity landscapes with full binding curves, yielding absolute affinities while separating affinity from expression-related effects. In the studied anti-fluorescein scFv, CDR3H variation had larger affinity effects, while CDR1H was more optimized for expression; structural patterns suggest possible roles for interacting residues and secondary CDRs.

III. DISCUSSION

  • The measured affinity range, 10^-9.5 M to 10^-5.0 M, includes a large fraction of the physiological range relevant to affinity maturation, 10^-10 M to approximately 10^-6 M [44–46].The authors note that extending measurements below 10^-9.5 M might require larger-volume labeling reactions.Source: §III · p. 8
  • Tite-Seq uses full binding titration curves to determine absolute KD values in molar units and disentangle affinity from expression and functional-display effects.Unlike single-concentration DMS, curves span concentrations and can capture a wider range of KDs.Source: §III · p. 8
  • Tite-Seq is presented as a general solution for separating protein-ligand binding effects from other sequence-dependent biochemical effects, and may work with yeast-display/FACS-compatible assays or be adapted to mammalian display.The paper says the procedure can be applied to other protein binding assays compatible with yeast display and FACS.Source: §III · p. 8
  • In the anti-fluorescein scFv, CDR3H variation had larger effects on affinity, while CDR1H was more optimized for expression; the authors caution against generalizing beyond this system.The authors suggest secondary CDRs may help ensure antibody stability, potentially freeing CDR3H to encode antigen specificity.Source: §III · p. 8
  • Across the studied scFv, residue-position effects on affinity correlated strongly with antibody contacts in the co-crystal structure, unlike expression effects; the authors caution that this is one antibody.If observed more broadly, the authors suggest paratopes could be tightly interacting residue networks governing specificity, but not stability.Source: §III · p. 8

IV. METHODS

The methods generate CDR1H and CDR3H variant libraries, display and label yeast-bound scFvs, sort cells across antigen concentrations, and sequence the sorted populations. Simulations and replicate experiments support measurement and processing of the resulting data.

IV. METHODS

  • Variant 1H and 3H regions were inserted into the 4-4-20 scFv; yeast-display experiments were sorted, regrown, and sequenced, with three replicate experiments performed on different days.Clones were also measured by low-throughput flow cytometry for validation.Source: §IV · pp. 8–9

Appendix A: Schematic simulations

  • Simulated sorts used two hypothetical scFvs with different KD and fluorescence parameters, eleven antigen concentrations, noisy cell fluorescence, and three fluorescence bins to illustrate curve fitting.The simulated fluorescence included multiplicative log-normal noise, and fitted curves were compared with mean fluorescence or mean bin values.Source: Appendix A · p. 9

Appendix B: Yeast display

  • Yeast cultures were induced for scFv expression and fluorescently labeled with biotinylated fluorescein across antigen concentrations; 0.1% glucose in induction cultures was essential for reliable expression in many cells.A separate anti-c-Myc labeling reaction measured expression.Source: Appendix B · p. 9

Appendix C: Cloning strategy

  • The separately cloned 1H and 3H libraries varied one variable region at a time and yielded ≥10^5 yeast transformants; after filtering, 144 clones were picked randomly, and 19 were chosen based on preliminary Tite-Seq for KD measurements.Source: Appendix C · p. 10

Appendix D: Tite-Seq procedure

  • The mixed yeast libraries, including positive and negative controls, were labeled at eleven antigen concentrations, sorted by fluorescence into four bins, and sequenced with bin-specific barcodes.Under-sampled bins from all three replicates were prepared with different barcodes and sequenced again.Source: Appendix D · pp. 10–11

Appendix E: Inference of KD from Tite-Seq data

The inference procedure fits each sequence’s binding titration curve from sorting-bin read counts rather than direct fluorescence measurements. It estimates the curve parameters and bin-assignment probabilities jointly using a likelihood model.

Appendix E: Inference of KD from Tite-Seq data

  • Each sequence’s mean cellular fluorescence across antigen concentrations is modeled with a non-cooperative Hill function, with background fluorescence, saturation increase, and sequence-specific KD as parameters.The background B is set to mean fluorescence at zero antigen, while A_s is the increase at saturation.Source: Appendix E · p. 11
  • Because Tite-Seq does not directly measure mean fluorescence, the procedure estimates it from sorting-bin data, relating sequence read counts to bin-assignment probabilities while accounting for sequencing-depth differences and sorted-cell counts.The probabilities are inferred alongside the Hill-function parameters rather than modeled explicitly as functions of mean fluorescence.Source: Appendix E · p. 11
  • A Poisson model relates observed sequence reads to expected reads, and maximum-likelihood optimization estimates KD, saturation increase, library frequency, and bin probabilities subject to the stated constraint.The fitting procedure scans candidate KD and saturation values and selects parameters maximizing the likelihood.Source: Appendix E · p. 11

Appendix F: Inference of KD from flow cytometry experiments on individual clones

The flow-cytometry procedure fits individual clone titration curves to infer KD and fluorescence parameters. A separate realistic-data simulation tests the Tite-Seq analysis pipeline.

Appendix F: Inference of KD from flow cytometry experiments on individual clones

  • For individual clones, measured mean fluorescence across antigen concentrations is fit to a non-cooperative Hill function to infer KD and saturation increase, with background estimated from zero-antigen measurements.A grid search restricts KD to 10^-10 M through 10^-3 M and saturation increase to a range based on the session’s measured fluorescence.Source: Appendix F · pp. 11–12

Appendix G: Realistic Tite-Seq simulations

  • Realistic Tite-Seq simulations test the analysis scripts using randomly drawn KD values from 10^-10 M to 10^-4 M and fluorescence distributions modeled with a Gaussian mixture in log space.Simulated saturation amplitudes span the bulk of experimentally observed values.Source: Appendix G · p. 12
Loading 1601.02160v2…