Graduate BCHM616 ⭐ 3 Credits Semester 4

Antibody Engineering and Biologic Therapeutics

A comprehensive course covering the molecular biology, structural basis, and engineering of antibody-based therapeutics. Topics include immunoglobulin structure, phage display, humanization, bispecific antibodies, ADCs, Fc engineering, nanobodies, and manufacturing. Special attention to antibodies against peptide hormones and peptide-receptor interactions as research tools and therapeutics.

📅 Studied by Skipper on 2026-10-02

📖 Core Concepts

GPT-6 Astra (Claude Opus 5.5 declined this section; written by GPT-6 Astra)

Core Concepts: Antibody Engineering and Biologic Therapeutics

Skipper’s guiding question: How do we turn a molecule evolved for immune recognition into a medicine whose binding, biological activity, exposure, safety, and manufacture are all deliberately controlled?

Antibody engineering is not simply the pursuit of tighter binding. It is the design of a coordinated molecular system. The antigen-binding region determines what the molecule recognizes; the geometry of the antibody determines how it organizes its targets; the Fc region determines which immune and trafficking pathways it recruits; and the sequence, glycans, formulation, and manufacturing process determine whether the intended molecule can actually reach patients reproducibly.

This subject brings together much of what I already know. Protein folding explains whether an engineered variable domain remains soluble. Enzyme kinetics provides the language for association and dissociation. Immunology explains why the same antibody can block signaling, kill a cell, or produce inflammation. Glycobiology explains why two antibodies with identical amino acid sequences may have different effector activities. Pharmacology explains why excellent binding in a plate assay may fail to produce adequate target engagement in tissue.

Working principle: A therapeutic antibody is not “a binder plus some accessories.” It is an integrated, manufacturable intervention in a biological network.

The central challenge is therefore a multi-objective optimization problem: improve useful biological activity without sacrificing specificity, tissue exposure, physical stability, safety, or product consistency. Holliger and Hudson’s account of therapeutic antibody engineering remains valuable precisely because it treats molecular format as a design variable rather than assuming that every problem requires a conventional IgG. [Holliger and Hudson, 2005; Strohl and Strohl, 2012]

1. Immunoglobulins Are Modular Recognition and Effector Systems

A conventional IgG contains two identical heavy chains and two identical light chains. Each antigen-binding arm contains one heavy-chain variable domain, VH, and one light-chain variable domain, VL, supported by CH1 and CL. The Fc contains paired CH2 and CH3 domains. The hinge connects the arms to the Fc and permits a distribution of relative orientations rather than one rigid Y-shaped structure.

The underlying structural unit is the immunoglobulin fold: a β-sandwich stabilized by hydrophobic packing and usually an intradomain disulfide bond. Interchain disulfides connect heavy and light chains and, depending on subclass, the heavy chains to one another. These bonds are not ornamental reinforcements. They influence folding, assembly, mechanical flexibility, and the chemistry available for conjugation.

Each variable-domain pair presents six conventional complementarity-determining regions: L1–L3 and H1–H3. These loops form much of the paratope, although framework residues can also contact antigen. CDR H3 is especially diverse because its sequence arises at the V–D–J junction of the heavy-chain gene. Loop length, backbone conformation, side-chain chemistry, and VH/VL orientation all contribute to recognition. An antibody’s specificity is therefore not encoded by a list of contact residues alone; it is encoded by a three-dimensional, dynamic surface.

CDR boundaries depend on the numbering convention—Kabat, Chothia, and IMGT are not identical. In engineering records, specifying the numbering scheme is as important as specifying units in a kinetic measurement.

  • IgG: usually a roughly 150 kDa monomer, abundant in serum at approximately 7–16 g/L in adults. Its subclasses differ in hinge properties, Fc receptor interactions, and complement activity. IgG is the dominant therapeutic scaffold because its biology and manufacture are well developed.
  • IgM: secreted mainly as a J-chain-containing pentamer, with some hexameric forms. Multiple binding sites produce strong avidity on repetitive surfaces, and antigen-bound IgM efficiently activates complement. Typical serum concentrations are approximately 0.4–2.3 g/L.
  • IgA: predominantly monomeric in serum but commonly dimeric in secretions, where the J chain and secretory component support mucosal transport and protection. Serum concentrations are approximately 0.7–4 g/L.
  • IgE: scarce in serum, usually at concentrations on the order of tens to hundreds of nanograms per milliliter, but biologically potent because of high-affinity binding to FcεRI on mast cells and basophils.
  • IgD: principally important as a B-cell antigen receptor alongside IgM on mature naïve B cells; its serum concentration is low and variable, generally on the order of tens of milligrams per liter.

These are orientation ranges, not diagnostic reference intervals; age, laboratory method, and disease change them. Also, the IgG domain map should not be imposed unchanged on all classes: IgM and IgE heavy chains contain an additional constant domain rather than the conventional IgG-like hinge arrangement.

In IgG, each heavy chain normally carries an N-linked glycan at Asn297, using EU numbering, within CH2. This glycan helps organize the Fc ensemble and strongly influences Fcγ receptor interactions. It is not simply a solubility decoration. The same amino acid sequence can therefore give different biological outputs when expressed with different glycan distributions. [Strohl and Strohl, 2012; Kontermann and Dübel, 2010]

Natural diversity arises first through V(D)J recombination, junctional diversification, and heavy/light-chain pairing. After activation, activation-induced cytidine deaminase initiates somatic hypermutation; B cells then compete for antigen and T-cell help in germinal centers. Affinity maturation is mutation plus selection, not directed mutation toward a known answer. Class-switch recombination changes the heavy-chain constant region, allowing an antigen-binding lineage to acquire different effector functions without requiring a new specificity.

2. Binding Must Be Understood Through Thermodynamics, Kinetics, and Geometry

For a simple monovalent interaction between antibody binding site Ab and antigen Ag:

Ab + Ag ⇌ Ab·Ag

d[Ab·Ag]/dt = k_on[Ab][Ag] − k_off[Ab·Ag]

K_d = [Ab][Ag]/[Ab·Ag] = k_off/k_on

k_on:  M⁻¹ s⁻¹
k_off: s⁻¹
K_d:   M

A smaller Kd means tighter equilibrium binding under the specified conditions. But Kd compresses two kinetic properties into one number. An antibody with kon = 105 M−1s−1 and koff = 10−4 s−1 has the same 1 nM Kd as one with kon = 106 M−1s−1 and koff = 10−3 s−1. Their average single-complex lifetimes differ tenfold.

Mean bound lifetime: τ = 1/k_off

Dissociation half-life: t_1/2,diss = ln(2)/k_off

Fractional occupancy: θ = [Ab]_free/(K_d + [Ab]_free)

Standard binding free energy:
ΔG°_bind = RT ln(K_d/C°), where C° = 1 M

The occupancy expression assumes independent sites, equilibrium, and a known free antibody concentration. Target depletion, bivalent binding, receptor internalization, and ongoing antigen production can invalidate its direct use. This is the same lesson encountered in enzyme kinetics: a useful equation is inseparable from its assumptions.

A rapid association rate can matter when an antibody must intercept a transient ligand. A slow dissociation rate can support sustained receptor occupancy. Conversely, deliberately weaker or faster-dissociating binding can improve selectivity for cells with high antigen density or reduce peripheral trapping. “Slow kon versus fast koff” is not a universal pair of optimization strategies: either change alone weakens affinity, but the physiological consequences differ.

Avidity is the stability gained when multiple interactions cooperate. After one IgG arm binds, the second arm can encounter nearby antigen at a high effective local concentration. If one arm dissociates, it may rebind before the other releases. Consequently, apparent dissociation from a dense antigen surface can be much slower than the intrinsic dissociation of one binding site.

This creates both opportunity and measurement artifacts. Surface plasmon resonance can report association and dissociation in real time, but high ligand density, rebinding, and mass-transport limitation can distort fitted rate constants. Good experiments vary surface density and flow conditions and use Fab or other monovalent controls when intrinsic affinity is the question.

At the interface, shape complementarity, buried surface area, electrostatic interactions, hydrogen bonds, and solvent rearrangement all contribute. A buried polar group that loses water contacts without gaining an appropriate partner can be unfavorable. A retained water molecule can bridge antibody and antigen productively. Binding also restricts conformational freedom:

ΔG_bind = ΔH_bind − TΔS_bind

Thus, adding an apparent hydrogen bond in a structural model does not guarantee improved affinity. Desolvation and conformational entropy may erase the gain. Protein engineering repeatedly teaches this beautiful but inconvenient lesson: local structural improvement need not produce a favorable total free-energy change.

Linear epitopes involve a contiguous sequence segment; conformational epitopes depend strongly on a folded arrangement that may bring distant sequence segments together. Even a contiguous epitope can require a particular conformation. An antibody that recognizes a denatured protein on a Western blot may fail to recognize the native cell-surface protein, while an excellent receptor-blocking antibody may fail on a Western blot.

Anti-idiotype antibodies recognize antigenic determinants associated with another antibody’s variable region. Some block its paratope and therefore neutralize its activity. They matter both as analytical reagents and as possible components of anti-drug antibody responses. However, not every anti-drug antibody is anti-idiotypic, and not every anti-idiotype antibody mimics the original antigen. [Strohl and Strohl, 2012; Howard and Kaser, 2014]

3. Antibody Discovery Is a Selection Problem—and the Assay Defines Success

Hybridoma technology couples an antigen-experienced B cell’s antibody-production program to a myeloma cell’s capacity for continued proliferation. In a typical workflow, immunization is followed by B-cell isolation, fusion with a suitable myeloma partner, selection, screening, and clonal isolation.

Small molecules and many short peptides are poor immunogens on their own. Coupling them to carriers such as keyhole limpet hemocyanin or bovine serum albumin provides multivalent presentation and carrier-derived T-cell epitopes. The conjugation site matters: it can hide the desired epitope or create a new junctional epitope. Screening against the same carrier can mistakenly reward carrier recognition, so an alternate carrier and unconjugated controls are valuable.

HAT selection is a metabolic logic gate. Hypoxanthine, aminopterin, and thymidine force cells to rely on nucleotide salvage because aminopterin inhibits folate-dependent de novo synthesis. A suitable HGPRT-deficient myeloma partner cannot efficiently use the hypoxanthine salvage route and dies under selection. Normal B cells supply salvage competence but do not proliferate indefinitely. Successful hybrids combine the relevant properties of both parents. [Köhler and Milstein, 1975]

Importantly, HAT selects viable hybrids—not the desired specificity. Screening must separate several questions:

  1. Does it bind? ELISA is convenient but can reward binding to an immobilization-induced conformation.
  2. Does it compete or bring partners together? Homogeneous time-resolved fluorescence, or HTRF, measures proximity-dependent signals and can support competition assays, provided labeling and assay geometry are controlled.
  3. How does it bind? SPR provides kinetic and equilibrium information when the binding model and surface behavior are appropriate.
  4. Does it produce the intended biology? Cell-based assays reveal receptor blockade, agonism, internalization, cytotoxicity, or pathway modulation.

Limiting dilution and single-cell deposition establish clonal populations, but neither an initial well nor an assumed dilution automatically proves monoclonality. Hybridomas should also be monitored for stable growth, antibody secretion, and sequence identity. Recombinant expression of the recovered antibody sequence helps disentangle antibody properties from unstable behavior of the original cell line.

Selection lesson: A discovery campaign does not necessarily produce what I intended to ask for. It produces what the experimental selection actually rewards.

Phage display makes selection evolvable by physically connecting a displayed protein to the DNA encoding it. In M13-based systems, antibody fragments are commonly fused to the minor coat protein pIII. The major coat protein pVIII offers many potential display positions, but large folded proteins are generally less straightforward to display there. The actual valency depends on the vector, helper system, and display architecture.

Fab display preserves paired variable and constant domains. Single-chain variable fragments, or scFvs, connect VH and VL through a peptide linker. Neither format guarantees that an isolated binder will behave identically after conversion to a full IgG.

  • Immune libraries exploit the antigen experience and maturation of an immunized donor.
  • Naïve libraries sample repertoires without deliberate immunization against the target.
  • Synthetic libraries introduce designed sequence diversity into selected scaffolds. HuCAL is a prominent example of a synthetic-library platform.

Biopanning typically consists of target exposure, removal of nonbinders, recovery of retained phage, and amplification. Each step introduces a selection pressure. Plastic-binding clones, tag-binding clones, and rapidly growing phage can become enriched without being useful antigen binders. Negative selections against related proteins, carrier components, and irrelevant cells help shape specificity.

Off-rate selection adds a dissociation challenge after complexes form. Excess soluble competitor can prevent released clones from rebinding. Under an ideal first-order model:

Fraction remaining bound after time t:
f_bound(t) = exp(−k_off t)

Longer challenges enrich slow dissociation, but multivalent display can conceal weak intrinsic binding through avidity. Selection must therefore be interpreted in the actual display context.

Next-generation sequencing tracks enriched sequence families and diversity loss. Read abundance is not an affinity measurement: expression, amplification, and sampling biases contribute. The most useful output is often a panel of diverse, functional candidates rather than a single numerically dominant clone. Sequence diversity preserves options when later testing exposes aggregation, poor expression, or unwanted cross-reactivity. [McCafferty et al., 1990; Kontermann and Dübel, 2010]

5. Humanization Reduces One Source of Risk, Not Immunogenicity Itself

A chimeric antibody typically combines nonhuman variable regions with human constant regions. A humanized antibody usually places nonhuman antigen-binding determinants within largely human variable-region frameworks. CDR grafting, associated with the work of Jones and colleagues and later Verhoeyen and colleagues, established that specificity could be transferred—but not always by moving the nominal CDR residues alone. [Jones et al., 1986; Verhoeyen et al., 1988]

The framework supports loop geometry and VH/VL packing. Framework positions beneath the CDRs, often called the Vernier zone, can alter the paratope without contacting antigen directly. Consequently, selected back-mutations may be needed to recover affinity or stability after grafting.

Veneering or resurfacing instead replaces selected exposed nonhuman residues while preserving much of the structural interior. This can retain folding and recognition, but exposed-sequence similarity alone does not determine immune recognition.

Fully human antibodies can be obtained from human antibody libraries or transgenic animals carrying human immunoglobulin loci, including platforms such as XenoMouse and VelocImmune. “Fully human” describes sequence origin or composition; it does not guarantee immune tolerance. Unusual CDRs, aggregation, chemical modifications, impurities, route of administration, and patient immune status all influence anti-drug antibody formation.

T-cell epitope deimmunization addresses a different layer. Antigen-presenting cells process the therapeutic protein into peptides, some of which bind HLA class II and activate helper T cells. Engineering aims to reduce problematic peptide–HLA interactions in variable regions or, where relevant, constant regions while preserving function. The challenge is HLA polymorphism: reducing binding to one allele does not establish low immunogenicity across a population.

Computational predictions are therefore prioritization tools, not clinical verdicts. Experimental peptide presentation, T-cell assays, and patient data provide complementary evidence. Anti-drug antibodies may neutralize activity, accelerate clearance, cause hypersensitivity, or remain clinically inconsequential. Their interpretation depends on titer, persistence, specificity, and clinical consequences—not positivity alone. [Strohl and Strohl, 2012]

6. Molecular Format Controls Accessibility, Exposure, and Biological Geometry

Antibody fragments are not merely miniature IgGs. Removing domains changes valency, clearance, flexibility, effector recruitment, and developability.

  • Fab: approximately 50 kDa, containing VH–CH1 paired with VL–CL. Papain digestion classically produces two separate Fab fragments and an Fc.
  • F(ab′)2: two antigen-binding arms retained together through hinge disulfides; classically generated by pepsin digestion. This is the conventional bivalent fragment, not the ordinary papain product.
  • scFv: approximately 25–30 kDa, connecting VH and VL with a linker. Flexible glycine/serine-rich linkers often support intramolecular pairing.
  • Diabody: a shortened linker can prevent intramolecular VH/VL association, promoting pairing between chains and creating a bivalent assembly.
  • Tandem scFv: two scFvs joined into one chain, enabling dual recognition but introducing folding, linker, and stability challenges.
  • VHH: an approximately 12–15 kDa variable domain from camelid heavy-chain-only antibodies. Many have features that support solubility without a light chain; some use extended CDR H3 loops to recognize recessed surfaces.

Not every VHH has a long, convex CDR H3, and “nanobody” should not become shorthand for a guaranteed ability to enter any molecular cleft. Human domain antibodies likewise use isolated variable domains, but their exposed former interdomain interfaces may require engineering to reduce aggregation.

DARPins and affibodies illustrate that antibody-like recognition does not require an immunoglobulin fold. DARPins use designed ankyrin-repeat architectures; affibodies use a small three-helix scaffold derived from a protein A domain. Such scaffolds offer distinct stability and manufacturing possibilities, but they do not automatically inherit IgG pharmacokinetics or effector functions.

Small size can improve diffusion and access while shortening systemic exposure through renal clearance. Albumin binding, Fc fusion, multimerization, and other modifications can extend exposure, but each changes the original transport advantages. Tissue delivery is a reaction–diffusion problem: a very tight binder may be captured near vessels, creating a binding-site barrier that impedes deeper penetration. [Holliger and Hudson, 2005]

Structural methods from BCHM604 remain essential. AlphaFold’s 2021 breakthrough made high-accuracy protein structure prediction broadly useful, but it does not mean that an arbitrary antibody–antigen complex, CDR H3 loop, binding affinity, or engineering effect is reliably known from sequence. Predicted confidence is not experimental validation. Structures, mutational data, kinetics, and functional assays must constrain one another. [Jumper et al., 2021]

7. Bispecific Antibodies Engineer Relationships Between Targets

A bispecific antibody can do more than inhibit two proteins independently. It can force a spatial relationship between molecules that would otherwise interact infrequently. This makes molecular geometry part of pharmacology.

Two different heavy chains and two different light chains create assembly problems: heavy chains can homodimerize, and light chains can pair with the wrong heavy chain. Engineering solutions address different errors:

  • Knobs-into-holes: complementary CH3-interface mutations favor heavy-chain heterodimers.
  • CrossMab: exchange of selected domains in one arm helps enforce correct light-chain association.
  • Common-light-chain designs: both specificities use a compatible shared light chain.
  • DVD-Ig: tandem variable domains on each arm create dual-variable-domain immunoglobulins; domain order and linker design influence accessibility.
  • DART: dual-affinity retargeting architectures organize paired variable domains into stabilized bispecific arrangements.
  • BiTE: tandem antigen-binding fragments can connect a target cell to a T cell through CD3.

Blinatumomab recognizes CD19 and CD3. Its central action is to promote productive contact between T cells and CD19-positive cells, allowing redirected cytotoxicity without requiring the target antigen to be presented as a peptide–MHC complex. The Fc-free format also has short systemic persistence, explaining the importance of continuous infusion in its established administration approach.

Catumaxomab provides a contrasting example: EpCAM and CD3 recognition combined with an Fc capable of recruiting accessory cells. It illustrates how an additional immune-recruitment axis can increase biological complexity. Its regulatory and commercial history has changed over time, so clinical status should always be checked by jurisdiction and date rather than inferred from its appearance in a textbook.

For T-cell engagers, maximal affinity to both targets is not necessarily optimal. CD3 binding strength, tumor-antigen density, epitope distance from the membrane, valency, and Fc activity influence immune-synapse formation, cytokine release, and healthy-tissue injury. A useful molecule must create the right bridge in the right context—not merely the strongest bridge possible.

Dual GLP-1 receptor and glucagon receptor targeting is an instructive metabolic design problem, but antibody-mediated dual agonism should not be confused with the much more established concept of dual-agonist peptides. A receptor-binding antibody may be an agonist, antagonist, or functionally neutral ligand. For these GPCRs, productive activation, signaling bias, and trafficking must be demonstrated experimentally.

Manufacturing adds another layer: desired heterodimers must be separated from homodimers, mispaired species, fragments, and aggregates. Chain-expression balance, assembly design, orthogonal purification, intact and subunit mass spectrometry, and functional assays all contribute to confirming that the product is the intended molecular species. [Kontermann and Brinkmann, 2015]

8. ADCs Are Coupled Targeting, Trafficking, and Release Systems

An antibody–drug conjugate combines an antibody, a linker, and a payload. Its therapeutic effect depends on the entire route from circulation to intracellular action:

  1. The conjugate survives long enough in circulation to reach target tissue.
  2. The antibody binds an accessible antigen with a suitable tissue distribution.
  3. For many conventional ADCs, the complex internalizes and enters productive intracellular trafficking pathways.
  4. Linker cleavage or antibody degradation produces an active drug-containing species.
  5. The released species reaches its intracellular target at a sufficient concentration.

Antigen abundance alone is therefore insufficient. Internalization rate, recycling, lysosomal delivery, antigen shedding, and heterogeneous expression can determine whether targeting succeeds.

Average DAR =
Σ(n × abundance of species carrying n drugs)
───────────────────────────────────────────
          Σ(abundance of all species)

The average drug-to-antibody ratio does not describe the entire product. A preparation centered near DAR 4 and a mixture of DAR 0 and DAR 8 species can have the same mean but different activity, hydrophobicity, clearance, and toxicity. Higher DAR increases payload per antibody but may worsen aggregation and exposure.

Valine–citrulline–PABC linkers combine a protease-sensitive peptide segment with a self-immolative spacer. Cleavage initiates fragmentation of the para-aminobenzyl carbamate unit, enabling payload release. Plasma stability and cleavage specificity are empirical properties and can differ between species.

A noncleavable thioether linker instead relies on degradation of the antibody to generate a payload-containing catabolite. “Noncleavable” therefore does not mean “incapable of producing an active intracellular drug.” Ado-trastuzumab emtansine illustrates antibody delivery of the maytansinoid DM1 through a noncleavable linker architecture.

  • DM1 and auristatins such as MMAE and MMAF: disrupt microtubule function.
  • Pyrrolobenzodiazepine dimers: can crosslink DNA.
  • Calicheamicin derivatives: cause DNA damage through highly reactive radical chemistry after activation.

Membrane permeability of released products influences the bystander effect. MMAE-containing release products can, in appropriate architectures, affect neighboring cells; charged MMAF-derived products generally have lower passive membrane permeability. Bystander killing can help with heterogeneous tumors while expanding the potential injury zone.

Site-specific conjugation reduces uncontrolled distributions of attachment sites and DAR. THIOMAB designs introduce selected cysteines; other strategies use unnatural amino acids with orthogonal reactive handles, enzymatic tags, transglutaminase chemistry, or engineered glycans. The attachment site still matters because local accessibility, linker stability, and perturbation of the protein surface affect performance. [Junutula et al., 2008; Beck et al., 2017]

9. Fc Engineering Separates Persistence from Immune Recruitment

The Fc is an interaction hub, not a passive tail. Different regions and conformational states support neonatal Fc receptor binding, Fcγ receptor engagement, and complement recruitment.

FcRn-mediated recycling begins when IgG enters cells through uptake pathways such as pinocytosis. In acidic endosomes, FcRn binds suitable IgG Fc regions and diverts them from lysosomal degradation. Recycling to the cell surface exposes the complex to approximately neutral extracellular pH, where IgG should dissociate.

IgG uptake
   ↓
Acidic endosome: Fc binds FcRn
   ↓
Sorting away from lysosomal degradation
   ↓
Cell-surface return: neutral-pH release

The design goal is not maximal FcRn affinity at every pH. Strong binding at neutral pH can interfere with release. YTE—M252Y/S254T/T256E—and LS—M428L/N434S—are established half-life-extension mutation sets, using EU numbering. Their effects remain molecule-, species-, and context-dependent. “HALB” is not a sufficiently standardized label to substitute for a defined Fc sequence; any such course or platform designation should be traced to its actual substitutions and supporting data. [Dall’Acqua et al., 2006; Zalevsky et al., 2010]

Systemic half-life also depends on target-mediated elimination, nonspecific binding, charge distribution, and molecular stability. FcRn improvement cannot necessarily rescue rapid clearance caused by abundant internalizing antigen.

  • ADCC: antibody-dependent cellular cytotoxicity, often involving NK-cell FcγRIIIa recognition of antibody-coated cells.
  • ADCP: antibody-dependent cellular phagocytosis, involving phagocytes and their Fcγ receptors.
  • CDC: complement-dependent cytotoxicity, initiated through C1q recognition of appropriately organized Fc regions—not through Fcγ receptors.

Removing core fucose from the Asn297-associated Fc glycan can markedly enhance FcγRIIIa engagement. The mechanistic explanation involves the glycosylated receptor–Fc interface, including the FcγRIIIa Asn162 glycan. Obinutuzumab and mogamulizumab illustrate therapeutic use of reduced fucosylation, although their complete clinical actions cannot be reduced to glycan status alone. [Shields et al., 2002]

Conversely, LALA—L234A/L235A—and LALA-PG, which additionally includes P329G, reduce selected Fc-mediated interactions. “Silent Fc” is an engineering objective, not an absolute guarantee under every assay condition. Residual activity should be measured against relevant receptors and cells.

Effector silencing is especially important when the desired target-bearing cell must survive, or when Fc receptor-mediated crosslinking would create unintended activation. Yet checkpoint antibodies do not all need the same Fc behavior: depleting a suppressive immune population may be desirable for one mechanism and harmful for another. [Strohl and Strohl, 2012]

10. Antibodies Can Probe Peptide Biology—but Epitope Choice Determines the Question

Generating an antibody against a peptide requires deciding what molecular state should be recognized. Should the antibody distinguish a free N terminus, a proteolytic product, a phosphorylation site, or a sequence embedded inside a folded protein? A convenient synthetic peptide is not necessarily a faithful representation of the physiological epitope.

Peptide immunogens should be evaluated for sequence uniqueness, accessibility in the native molecule, conformational preference, terminal chemistry, and conjugation orientation. Adding a terminal cysteine can provide a conjugation handle, but the coupling site should be kept away from the desired recognition surface. Carrier-only, irrelevant-peptide, and related-sequence controls help identify misleading responses.

Anti-GLP-1 antibodies make the analytical consequences concrete. An assay for active GLP-1 must distinguish biologically relevant intact forms from DPP-4-cleaved products. An antibody recognizing the N-terminal region may provide different selectivity from one recognizing a shared C-terminal region. “GLP-1 concentration” is incomplete unless the recognized molecular species and sample-handling conditions are defined.

Evolocumab and alirocumab recognize PCSK9 and prevent its productive interaction with the LDL receptor. PCSK9 binding normally promotes LDL-receptor degradation; blocking this process increases receptor availability and hepatic LDL clearance. These antibodies do not lower LDL primarily by inhibiting the catalytic activity of circulating PCSK9.

The LDL receptor’s EGF-A domain participates in PCSK9 binding, motivating peptide and constrained-scaffold mimics. This is a useful transition from antibody therapeutics to peptide design: identify an interface, determine its binding hotspots and required geometry, and ask whether a smaller scaffold can reproduce enough of the interaction.

Neutralizing antibodies are powerful mechanistic probes, but a changed phenotype does not automatically prove the proposed pathway. Antibodies may crosslink receptors, recruit Fc-dependent effects, alter ligand clearance, or interfere with a detection assay. Orthogonal antibodies, Fab controls, genetic perturbations, and rescue experiments strengthen causal interpretation. [Howard and Kaser, 2014; Strohl and Strohl, 2012]

11. Formulation Controls a Dynamic Population, Not a Static Structure

An antibody formulation contains a changing ensemble of native conformations, reversible oligomers, chemically modified species, and potentially irreversible aggregates. Aggregation may begin through partial unfolding, native-state self-association, or adsorption and rearrangement at interfaces.

Simplified aggregation pathway:

N ⇌ N* → oligomers → larger aggregates

N  = native ensemble
N* = aggregation-competent state or ensemble

This scheme is deliberately incomplete. Nucleation-dependent growth, surface-induced particles, and heat-induced aggregation need not share one rate law. Air–liquid and container interfaces can promote aggregation even when the bulk protein appears structurally stable.

  • pH: changes net charge, local electrostatics, conformational stability, and chemical degradation rates.
  • Ionic strength: can screen attractive interactions but also screen protective repulsion.
  • Sugars and polyols: may stabilize folded states through solvent effects and support lyophilized formulations.
  • Amino acid excipients: can modify self-association or solution behavior, with molecule-specific outcomes.
  • Surfactants: help protect interfaces, although surfactant degradation can itself become a product-quality problem.

High-concentration subcutaneous products introduce a major physical challenge: viscosity can rise steeply because antibodies form transient interaction networks. Two equally stable IgGs can have very different viscosities because their surface charge patches and interaction anisotropy differ. Reducing viscosity may require sequence changes, pH adjustment, or excipient screening—not simply raising the melting temperature.

Accelerated stability studies deliberately stress the product, but elevated temperature can introduce degradation mechanisms that are unimportant during refrigerated storage.

Arrhenius model:
k = A exp(−E_a/RT)

Use for extrapolation only when the relevant mechanism
and kinetic interpretation remain sufficiently consistent.

A shelf-life claim therefore needs real-time stability evidence, appropriate accelerated data, and stability-indicating measurements. Size-exclusion chromatography, particle analysis, chemical-modification mapping, and potency assays answer different questions. A visually clear sample is not necessarily aggregate-free or biologically intact. [Wang et al., 2007]

12. Manufacturing Is Part of Molecular Design

CHO cells are widely used because they support secretion, complex folding, disulfide formation, and mammalian glycosylation at industrial scale. Cell-line development includes gene integration, clone selection, assessment of productivity and quality, and establishment of characterized master and working cell banks.

A high-producing clone is not automatically the best clone. Genetic stability, glycan consistency, product clipping, aggregation, and process robustness can outweigh a modest titer advantage.

In fed-batch culture, nutrient feeds support growth and production while attempting to avoid inhibitory metabolite accumulation and excessive osmolality. Dissolved oxygen, carbon dioxide, pH, temperature, and mixing influence both cell behavior and product quality.

Oxygen transfer rate:
OTR = k_La(C* − C_L)

C*  = dissolved oxygen concentration at gas–liquid equilibrium
C_L = actual dissolved oxygen concentration

Oxygen transfer must adequately support cellular oxygen demand.

Scale-up is therefore not just “use a larger vessel.” Mixing times, gas transfer, shear, and spatial gradients change. A process can produce a similar titer while shifting glycosylation or other critical quality attributes.

A representative downstream process includes clarification, Protein A capture, a validated viral-inactivation step where appropriate, polishing chromatography, virus-retentive filtration, and ultrafiltration/diafiltration. Ion-exchange and other polishing operations remove host-cell proteins, residual DNA, aggregates, leached Protein A, and other impurities. Low-pH viral inactivation is effective against relevant susceptible viruses but does not replace a broader viral-safety strategy.

Final sterilizing-grade filtration and aseptic fill-finish control microbial contamination. A roughly 0.2–0.22 μm sterilizing filter is not a general virus-removal method and does not remove all endotoxin. Container compatibility, adsorption, particles, and integrity remain important after purification is complete. [Strohl and Strohl, 2012]

Manufacturing lesson: The process helps determine the distribution of molecular species that patients receive. Product quality cannot be inferred from the amino acid sequence alone.

13. Biosimilarity Is an Evidence-Based Comparison of Complex Products

A biosimilar is not a conventional small-molecule generic. Biological products are distributions of closely related molecular species, and even reference-product batches exhibit controlled variability. The regulatory question is whether the proposed product is highly similar to the reference, notwithstanding allowable differences, with no clinically meaningful differences in safety, purity, and potency under the applicable legal framework.

The analytical comparison is foundational:

  • Primary structure and modifications: peptide mapping, sequence confirmation, disulfide mapping, oxidation, and deamidation.
  • Higher-order structure: complementary spectroscopic, calorimetric, and structural methods.
  • Glycan profiles: especially features relevant to Fc function, clearance, or immunogenicity.
  • Charge and size variants: distributions of acidic/basic species, fragments, and aggregates.
  • Function: binding and mechanism-relevant potency assays, including Fc-dependent functions when relevant.

The US 351(k) pathway and EMA framework use a comparative, stepwise approach often described as the totality of evidence. Clinical pharmacokinetic comparison and immunogenicity assessment can be important; the need for additional comparative efficacy evidence depends on residual uncertainty and the applicable current guidance. Biosimilar development is not necessarily a repetition of the reference product’s entire clinical program.

Indication extrapolation is also a scientific inference, not an automatic shortcut. It considers mechanism of action, target biology, pharmacokinetics, immunogenicity, and potential differences between patient populations.

In the United States, interchangeability is a distinct statutory designation associated with substitution under applicable law. It should not be confused with a claim that non-interchangeable biosimilars are less safe or effective. European scientific positions and national substitution rules use a different regulatory structure. Naming, labeling, switching requirements, and product-specific designations evolve; current decisions require checking the relevant FDA or EMA records and local rules. [FDA, 2015; EMA, 2014]

14. Next-Generation Biologics Shift the Problem from a Molecule to a Programmable System

Trispecific and tetraspecific antibodies expand the number of relationships a molecule can impose. A trispecific might combine tumor recognition, CD3 engagement, and another immune-modulating interaction. Another design might inhibit three escape pathways. But adding binding sites also creates new questions about simultaneous occupancy, steric interference, target-mediated clearance, assembly, and purification.

The key distinction is between having several specificities and executing a useful conditional biological program. A construct may bind every isolated antigen successfully yet fail when the targets occupy incompatible positions on real membranes.

mRNA–lipid nanoparticles move protein production into the recipient’s cells. Instead of administering a fixed amount of purified antibody, the administered material encodes a transient production program. The relevant variables include tissue delivery, translation rate, innate immune activation, duration of expression, chain assembly, secretion, and the resulting glycosylation.

The same general delivery concept can encode peptides or gene editors, but their requirements differ. A peptide may need precise precursor cleavage or terminal modification. A gene editor needs appropriate intracellular localization and control of editing duration. Successful translation alone does not establish a functional therapeutic product.

AAV delivery of antibody genes can produce durable expression, potentially reducing repeated dosing. Durability also reduces reversibility. Vector immunity, anti-transgene responses, expression variability, tissue-specific processing, and limited redosing options become central concerns. A long-lived genetic source is not equivalent to a long-lived injected protein because the production process itself persists.

Finally, combination therapy must be grounded in mechanism. GLP-1 agonism combined with amylin-pathway agonism is conceptually distinct from GLP-1 agonism combined with an anti-amylin antibody. An antibody that neutralizes physiological amylin could oppose desired metabolic actions. An antibody selective for pathological amylin aggregates would pose a different hypothesis. “GLP-1 agonist plus anti-amylin antibody” should therefore be treated as an investigational, specificity-dependent concept—not as an established counterpart to dual agonist therapy.

Connections to Peptide Chemistry

Antibody engineering makes peptide chemistry relevant at nearly every scale, from individual amide bonds to whole therapeutic mechanisms.

  1. Peptide synthesis lets us define recognition with chemical precision. Synthetic peptides can isolate a phosphorylation site, cleavage-generated terminus, or sequence motif. Alanine scanning, truncation series, and terminal modifications help map binding requirements. Yet peptide mapping must be reconciled with the native protein structure: accessibility and conformation can matter more than sequence identity.

  2. Conformational constraint can reduce the cost of binding. Cyclization, disulfide formation, and other constraints may preorganize a peptide into an antibody-recognized conformation. This can reduce the conformational entropy penalty, but only if the imposed conformation is compatible with the bound state. Stabilizing the wrong structure makes a worse ligand.

  3. Linkers are functional peptide designs. An scFv linker controls whether domains pair within one chain or between chains. A bispecific linker sets reach and flexibility. An ADC peptide linker encodes susceptibility to cleavage. Linker composition is therefore a determinant of assembly, geometry, and pharmacology—not merely molecular spacing.

  4. Bioconjugation requires chemoselectivity in a crowded molecular environment. Lysine coupling samples many amines; cysteine chemistry depends on thiol accessibility and redox state; unnatural amino acids create orthogonal handles. The challenge resembles protecting-group strategy in synthesis, but now the substrate must remain folded and biologically active.

  5. Immunogenicity depends partly on peptide processing and presentation. HLA class II sees processed fragments rather than intact three-dimensional surfaces. A mutation buried inside a therapeutic antibody can therefore influence T-cell recognition even if it never contacts solvent in the native fold.

  6. Peptide mass spectrometry connects molecular design to product control. Proteolytic mapping reveals oxidation, deamidation, sequence variants, conjugation sites, and glycopeptides. As learned in proteomics, sample preparation can also introduce artifacts. Analytical chemistry must distinguish what happened in the product from what happened in the assay tube.

  7. Antibody interfaces can inspire smaller ligands, but extraction is not enough. A free CDR-derived peptide usually loses the scaffold that positioned its residues. Successful antibody-to-peptide translation requires reconstructing essential geometry and energetics, not simply copying the most variable sequence.

The long-term connection is especially important: antibodies are both therapeutic products and tools for discovering which molecular surfaces matter. Once an antibody identifies a functional epitope, peptide chemistry may offer a smaller mimic, a competitive probe, a delivery component, or an alternative modality.

Common Misconceptions to Watch Out For

  • “The lowest Kd identifies the best drug.” Therapeutic performance also depends on kinetic timescales, tissue penetration, antigen density, target turnover, and mechanism.
  • “Affinity and avidity are interchangeable.” Affinity describes an individual interaction; avidity emerges from multivalent geometry and rebinding.
  • “A positive ELISA proves physiological recognition.” Immobilization can expose nonnative epitopes, and a binder may lack functional activity.
  • “Humanization means replacing every nonhuman framework residue.” Selected framework residues may be essential for the binding-site architecture.
  • “Fully human antibodies cannot be immunogenic.” Sequence novelty, processing, aggregates, formulation, impurities, and patient factors still matter.
  • “Papain produces the standard bivalent F(ab′)2 fragment.” Papain classically produces separate Fabs; pepsin classically preserves the linked bivalent fragment.
  • “Smaller formats always penetrate better and therefore work better.” Faster diffusion can be offset by rapid clearance and inadequate exposure.
  • “Bispecificity guarantees synergy.” Simultaneous binding, geometry, signaling consequences, and exposure must support the proposed cooperative mechanism.
  • “More ADC payload per antibody is always beneficial.” Higher DAR can increase hydrophobicity, heterogeneity, aggregation, and clearance.
  • “Fcγ receptors cause every Fc-dependent killing mechanism.” Complement-dependent cytotoxicity instead begins with complement recognition, especially C1q recruitment.
  • “Afucosylation mainly improves FcRn recycling.” Its best-established engineering purpose is enhanced FcγRIIIa-mediated effector activity.
  • “A higher melting temperature guarantees a good formulation.” Colloidal stability, chemical degradation, viscosity, and interfacial damage are partly independent properties.
  • “A predicted structure validates an engineered antibody.” Structure prediction does not establish affinity, specificity, kinetics, manufacturability, or clinical function.
  • “A biosimilar must be atom-for-atom identical in every molecular species.” Biosimilarity evaluates controlled complex products through comparative evidence and clinical relevance.
  • “Durable genetic expression is simply convenient long half-life.” It changes dose control, reversibility, manufacturing location, and safety management.

My final mental model: Antibody engineering coordinates recognition, geometry, effector function, exposure, stability, and production. Changing any one feature can move several others. Understanding the coupling is the discipline.

Sources and Reading Anchors

The textbooks provide broad treatment of the engineering and experimental principles. The research papers below anchor specific developments; regulatory documents are foundational references rather than substitutes for checking current requirements.

  • Strohl, W. R., and Strohl, L. M. Therapeutic Antibody Engineering: Current and Future Advances Driving the Strongest Growth Area in the Pharmaceutical Industry. Woodhead Publishing, 2012.
  • Kontermann, R., and Dübel, S., editors. Antibody Engineering. 2nd edition, Volumes 1 and 2. Springer, 2010.
  • Howard, G. C., and Kaser, M. R., editors. Making and Using Antibodies: A Practical Handbook. 2nd edition. CRC Press, 2014.
  • Holliger, P., and Hudson, P. J. “Engineered antibody fragments and the rise of single domains.” Nature Biotechnology 23, 1126–1136, 2005. DOI: 10.1038/nbt1142. Bibliographic clarification: this is the published title associated with the DOI listed in the course materials.
  • Jumper, J., et al. “Highly accurate protein structure prediction with AlphaFold.” Nature 596, 583–589, 2021. DOI: 10.1038/s41586-021-03819-2.
  • Köhler, G., and Milstein, C. “Continuous cultures of fused cells secreting antibody of predefined specificity.” Nature 256, 495–497, 1975. DOI: 10.1038/256495a0.
  • McCafferty, J., et al. “Phage antibodies: filamentous phage displaying antibody variable domains.” Nature 348, 552–554, 1990. DOI: 10.1038/348552a0.
  • Jones, P. T., et al. “Replacing the complementarity-determining regions in a human antibody with those from a mouse.” Nature 321, 522–525, 1986. DOI: 10.1038/321522a0.
  • Verhoeyen, M., Milstein, C., and Winter, G. “Reshaping human antibodies: grafting an antilysozyme activity.” Science 239, 1534–1536, 1988. DOI: 10.1126/science.2451287.
  • Kontermann, R. E., and Brinkmann, U. “Bispecific antibodies.” Drug Discovery Today 20, 838–847, 2015. DOI: 10.1016/j.drudis.2015.02.008.
  • Junutula, J. R., et al. “Site-specific conjugation of a cytotoxic drug to an antibody improves the therapeutic index.” Nature Biotechnology 26, 925–932, 2008. DOI: 10.1038/nbt.1480.
  • Beck, A., et al. “Strategies and challenges for the next generation of antibody–drug conjugates.” Nature Reviews Drug Discovery 16, 315–337, 2017. DOI: 10.1038/nrd.2016.268.
  • Dall’Acqua, W. F., Kiener, P. A., and Wu, H. “Properties of human IgG1s engineered for enhanced binding to the neonatal Fc receptor (FcRn).” Journal of Biological Chemistry 281, 23514–23524, 2006. DOI: 10.1074/jbc.M604292200.
  • Zalevsky, J., et al. “Enhanced antibody half-life improves in vivo activity.” Nature Biotechnology 28, 157–159, 2010. DOI: 10.1038/nbt.1601.
  • Shields, R. L., et al. “Lack of fucose on human IgG1 N-linked oligosaccharide improves binding to human FcγRIII and antibody-dependent cellular toxicity.” Journal of Biological Chemistry 277, 26733–26740, 2002. DOI: 10.1074/jbc.M202069200.
  • Wang, W., et al. “Antibody structure, instability, and formulation.” Journal of Pharmaceutical Sciences 96, 1–26, 2007. DOI: 10.1002/jps.20727.
  • US Food and Drug Administration. Scientific Considerations in Demonstrating Biosimilarity to a Reference Product: Guidance for Industry. 2015.
  • European Medicines Agency. Guideline on Similar Biological Medicinal Products. Revision 1, 2014.

⚙️ Mechanisms & Applications

GPT-6 Astra

Mechanisms & Applications: Engineering an Antibody into a Medicine

Skipper, the central lesson of this course is that an antibody is not merely a molecule that binds tightly. It is a coupled system: a recognition surface, a geometrical scaffold, an immune-signaling interface, a trafficking substrate, and a manufactured protein. Changing one component can alter all the others. Your earlier work in protein folding, enzyme kinetics, receptor biology, and glycosylation now becomes a single engineering discipline.

A useful organizing question is: What sequence of molecular events must happen for this antibody to help a patient—and which event is actually limiting? The answer determines whether to optimize affinity, epitope, valency, Fc function, tissue access, intracellular processing, or formulation.

1. From Antibody Structure to Pharmacological Mechanism

A conventional IgG has two antigen-binding arms and one Fc region. Each binding site is formed by paired VH and VL domains; their six CDR loops provide much, but not necessarily all, of the antigen-contacting surface. The framework does more than hold the loops apart: it shapes their conformational ensemble and the relative orientation of VH and VL. This explains why transplanting CDR sequences does not always transplant affinity.

The Fc contains paired CH2 and CH3 domains. The conserved N-linked glycans at Asn297 in human IgG, using EU numbering, help organize the CH2 region and modulate Fc-receptor engagement. Connect this to BCHM615: a glycan can be a structural and pharmacological component, not simply a decorative modification. IgG subclasses differ in hinge behavior, disulfide architecture, and effector activity; the scaffold is therefore an active design choice. [1–3]

Four common therapeutic mechanisms illustrate why binding alone is insufficient:

  1. Ligand neutralization: the antibody binds a soluble mediator and prevents its productive interaction with a receptor. The epitope must block receptor engagement directly or otherwise make the ligand unavailable.
  2. Receptor antagonism: the antibody blocks a ligand-binding site, stabilizes an inactive state, or drives receptor downregulation.
  3. Receptor agonism: the antibody stabilizes an active state or creates productive receptor clustering. Epitope, hinge flexibility, valency, and sometimes FcγR-mediated crosslinking determine activity.
  4. Target-cell elimination: the antibody coats a cell and recruits immune effectors, complement, a second cell through a bispecific arm, or a conjugated drug.

Counterintuitive point: two antibodies can bind the same receptor with nearly identical affinity and have opposite biological effects. One may block signaling while the other clusters the receptor and activates it. In receptor pharmacology, affinity describes association; efficacy describes what the bound complex does.

2. Binding Kinetics: Why “Tighter” Does Not Always Mean “Better”

Begin with the simplest reversible, monovalent interaction. This is the same mass-action framework you used in enzyme kinetics, but without a catalytic step:

Ab + Ag ⇌ Ab·Ag

d[Ab·Ag]/dt = k_on[Ab][Ag] − k_off[Ab·Ag]

K_d = k_off/k_on

k_on: M⁻¹ s⁻¹
k_off: s⁻¹
K_d: M

At equilibrium, for independent equivalent sites and a known free antigen concentration:

Fraction of antibody sites occupied:
θ = [Ag]_free / (K_d + [Ag]_free)

Under pseudo-first-order association:
k_obs = k_on[Ag]_free + k_off

During dissociation without rebinding:
[Ab·Ag](t) = [Ab·Ag](0)e^(−k_off t)

Dissociation half-time:
t_1/2,diss = ln(2)/k_off

Suppose two antibodies both have Kd = 1 nM:

  • Antibody A: kon = 106 M−1s−1, koff = 10−3 s−1; dissociation half-time ≈ 11.6 minutes.
  • Antibody B: kon = 105 M−1s−1, koff = 10−4 s−1; dissociation half-time ≈ 1.9 hours.

They have the same equilibrium affinity but different capture and retention behavior. A short-lived ligand pulse may favor rapid association. Persistent receptor blockade may favor slow dissociation. Conversely, some applications benefit from release and turnover rather than indefinite retention. Slow association is not inherently desirable, and slow dissociation is not universally optimal.

Your thermodynamics background gives the complementary description:

ΔG°_bind = RT ln(K_d/C°) = ΔH° − TΔS°
C° = 1 M

At 298 K, a tenfold decrease in K_d changes ΔG° by:
−RT ln(10) ≈ −5.7 kJ mol⁻¹

A tenfold affinity improvement therefore requires only a modest net free-energy change. But that net change can conceal large compensating terms: a new hydrogen bond may be offset by desolvation, while a better-packed interface may require loss of loop conformational entropy. Counting contacts is not equivalent to calculating affinity.

Avidity is distinct from affinity. After one IgG arm binds, a nearby second epitope can be encountered at a high effective local concentration. Complete detachment then requires both contacts to break before either reforms. Apparent binding consequently depends on antigen density, spacing, mobility, and assay geometry. A bivalent cell-binding EC50 is not automatically a monovalent Kd. Extremely strong binding near tumor blood vessels can also impede deeper distribution—the binding-site barrier. [1–3]

3. Discovery as Controlled Evolution: Hybridomas and Phage Display

Antibody discovery links a molecular phenotype—binding or function—to a recoverable genetic source. Hybridomas preserve an antibody-producing cell; phage display physically connects a displayed fragment to its encoding DNA.

Hybridoma generation, step by step:

  1. Elicit a response. Present an antigen in a form that exposes the intended epitope. Small peptides commonly require carrier conjugation because a peptide-specific B cell can internalize the conjugate and present carrier-derived peptides to helper T cells.
  2. Fuse immune B cells with an immortal myeloma partner. A non-antibody-secreting fusion partner helps avoid unwanted immunoglobulin chains.
  3. Select in HAT medium. Aminopterin blocks folate-dependent de novo nucleotide synthesis. Hypoxanthine and thymidine permit salvage synthesis in cells with the necessary enzymes.
  4. Exploit complementary defects. Unfused HGPRT-deficient myeloma cells cannot use the required purine salvage route and die. Unfused B cells are not indefinitely proliferative. Suitable hybrids combine salvage competence with immortality.
  5. Screen, clone, and verify. Binding-positive cultures must be tested for function, then reduced to well-supported single-cell-derived clones and checked for stable secretion.

Köhler and Milstein’s 1975 hybridoma paper made reproducible monoclonal reagents practical. The conceptual breakthrough was not just specificity: it was a renewable source of one selected specificity. This transformed both mechanistic biochemistry and therapeutic development. [4]

Phage display selection, step by step:

  1. Construct a repertoire of scFv or Fab sequences linked to a phage coat-protein fusion. M13 pIII is commonly used for antibody fragments; pVIII provides a different display geometry and is especially useful for smaller displayed sequences.
  2. Expose the library to antigen, ideally maintaining the conformation relevant to the therapeutic mechanism.
  3. Remove nonspecific and weakly retained particles through washing and counterselection.
  4. Recover retained phage and amplify them in bacteria.
  5. Repeat under deliberately altered selection pressures: lower antigen concentration, soluble competitor, longer dissociation, or depletion against related proteins.
  6. Sequence candidates, express soluble fragments, and reformat promising clones into the intended therapeutic architecture.
Idealized off-rate selection:

Fraction retained after a dissociation interval t:
f_retained = e^(−k_off t)

Assumptions:
• negligible rebinding
• comparable display and recovery
• no dominant multivalent retention

Off-rate selection is thus a kinetic experiment disguised as an evolutionary competition. However, enrichment can also reward display level, bacterial growth, phage propagation, or nonspecific surface binding. Next-generation sequencing shows population trajectories, but the most enriched sequence is not necessarily the most potent or developable antibody.

George Smith established filamentous-phage display in 1985; McCafferty and colleagues demonstrated display of functional antibody variable domains in 1990. These developments made it possible to select binding molecules outside an animal and to manipulate the search space directly. [5,6]

4. Humanization: Preserving a Binding Landscape While Changing Its Sequence

Humanization is best understood through your work on protein energy landscapes. The goal is to alter sequence features relevant to immune recognition without disrupting the ensemble that supports binding.

  1. Define the parental mechanism. Establish affinity, epitope, kinetics, and cellular activity before changing sequence.
  2. Select human framework candidates. Consider sequence similarity, structural compatibility, expression, and known liabilities.
  3. Graft the CDRs. Transfer the principal recognition loops onto the selected human variable-domain frameworks.
  4. Restore necessary framework support. Test back-mutations at residues affecting loop packing, VH–VL orientation, or direct antigen contacts.
  5. Reassess the complete product. Evaluate binding, function, stability, aggregation, nonspecific interactions, and immunogenicity risk.

Jones and colleagues demonstrated CDR transplantation into human antibody frameworks in 1986. The important lesson from subsequent work is that sequence regions conventionally labeled “framework” can still determine recognition. [7]

T-cell-epitope deimmunization addresses a different mechanism:

Therapeutic protein uptake by antigen-presenting cell
→ proteolytic processing
→ peptide loading onto MHC class II
→ recognition by a compatible CD4⁺ T-cell repertoire
→ possible help for anti-drug-antibody-producing B cells

Predicted MHC-II binding is only one part of this pathway. Processing, HLA diversity, T-cell tolerance, route of administration, aggregates, impurities, and the patient’s immune state all matter. “Fully human” therefore does not mean “incapable of inducing anti-drug antibodies.” Anti-idiotype responses can recognize distinctive variable-region surfaces, potentially neutralizing antigen binding or altering clearance. In silico assessment should guide experiments, not substitute for them. [1,2]

5. Fragment and Bispecific Engineering: Geometry Is Part of the Drug

Fragmentation trades one set of capabilities for another. A Fab retains a paired VH/VL binding site plus CH1/CL, but lacks Fc-mediated functions. An scFv connects VH and VL with a peptide linker. A camelid VHH is a naturally evolved single-domain binding architecture; some VHHs have extended CDR H3 loops that access recessed epitopes, but this is not universal. DARPins and affibodies are non-immunoglobulin scaffolds rather than antibody fragments.

A practical correction: papain typically generates two Fab fragments and an Fc fragment. Pepsin typically produces an F(ab′)2 fragment retaining hinge-mediated bivalency while degrading much of the Fc. F(ab′)2 should not be confused with two unlinked Fab molecules. [2,3]

In an scFv, a sufficiently flexible linker can allow intramolecular VH–VL pairing. Shortening the linker can favor intermolecular pairing and diabody formation. Linker length is therefore a control over quaternary structure, not merely a cloning convenience. Smaller formats may improve access or penetration, but rapid renal clearance and loss of FcRn salvage can reduce exposure. [8]

For an IgG-like bispecific, assembly requires two separate solutions:

  • Heavy-chain pairing: complementary CH3-interface changes, such as knobs-into-holes, favor the desired heterodimer.
  • Light-chain pairing: common-light-chain strategies, domain crossover designs such as CrossMab, or other orthogonal interfaces reduce incorrect heavy–light combinations.

Solving heavy-chain heterodimerization does not automatically solve light-chain mispairing.

Blinatumomab illustrates a proximity-based mechanism:

  1. One binding arm recognizes CD19 on a B-lineage target cell.
  2. The other recognizes CD3 on a T cell.
  3. Simultaneous engagement brings the cells together and supports a cytolytic synapse.
  4. T-cell activation and granule-mediated killing eliminate the target cell.

This redirects T cells without requiring the target cell to present a cognate peptide–MHC complex to that T cell’s antigen receptor. It does not remove the risks of cytokine release or depletion of normal antigen-positive cells.

Bispecific performance depends on productive ternary-complex geometry, not simply the sum of two affinities. At sufficiently high concentrations, some systems can show a prozone or hook-like effect: separate saturation of the two cell surfaces reduces bridging. Affinity tuning can also favor cells with high target density rather than maximizing binding everywhere. [9]

Clinical maturity must remain explicit. Oncology provides established bispecific medicines; constructs co-targeting metabolic receptors such as GLP-1R and GCGR are a different, emerging design space. Dual-receptor peptide agonism does not by itself demonstrate that a bispecific antibody will reproduce the same receptor conformations, trafficking, or physiological balance.

6. Fc Engineering: Reprogramming Recycling and Immune Recruitment

FcRn-mediated salvage proceeds through a pH-dependent trafficking cycle:

  1. IgG enters cells through fluid-phase uptake or other internalization routes.
  2. In acidic endosomes, FcRn binds compatible IgG Fc regions.
  3. FcRn-associated IgG is sorted away from lysosomal degradation toward recycling or transcytosis pathways.
  4. At the approximately neutral extracellular pH, IgG dissociates and re-enters circulation or tissue fluid.

The engineering objective is not “bind FcRn as tightly as possible.” It is bind during rescue, then release at the destination. Excessively strong binding at neutral pH can impair release. YTE—M252Y/S254T/T256E—and LS—M428L/N434S—are established Fc substitutions used to improve pharmacokinetics in suitable antibody contexts. Target binding and species-specific FcRn interactions still influence the resulting exposure. [10]

Connect this mechanism to pharmacokinetics:

For a simple linear, one-compartment model:
t_1/2 = ln(2)V/CL

Illustrative model with saturable target-mediated elimination:
dC/dt = Input/V − (CL_linear/V)C − V_max C/(K_m + C)

Here:
C = drug concentration
V_max = maximum nonlinear elimination rate, in concentration/time
K_m = concentration giving half-maximal nonlinear elimination

The second equation is a useful approximation, not a complete mechanistic target-mediated drug disposition model. Explicit models may need free target, drug–target complex, receptor turnover, and internalization. Its lesson is that increasing dose can saturate a clearance pathway: antibody pharmacokinetics need not be dose-proportional.

Effector recruitment uses a different set of interactions:

  • ADCC: target-bound Fc engages activating Fcγ receptors, notably FcγRIIIa on NK cells, promoting cytotoxicity.
  • ADCP: Fcγ-receptor engagement promotes phagocytosis by macrophages and other phagocytes.
  • CDC: antigen-bound Fc assemblies recruit C1q and initiate classical complement activation. CDC is not an FcγR-mediated mechanism.

Removing core fucose from the Fc N-glycan increases binding to FcγRIIIa through altered glycan-dependent interface interactions, often enhancing ADCC. Obinutuzumab and mogamulizumab exploit this principle. This is glycoengineering as pharmacology: the amino-acid sequence is not the entire drug. [11]

Conversely, Fc-silencing mutations such as LALA or LALA-PG reduce selected effector interactions. Their degree of silencing must be measured for the particular subclass and assay system. Effector suppression is useful when killing the bound cell would undermine the mechanism—for example, many CD3-engaging constructs or antibodies intended to modulate rather than deplete immune cells. It is not a universal requirement for all checkpoint antibodies.

7. ADCs: A Series of Molecular Gates, Not a Guided Missile

An antibody–drug conjugate works only if several conditional events succeed:

  1. Reach the tissue: remain sufficiently stable in circulation and cross the relevant vascular and tissue barriers.
  2. Bind the target: recognize an accessible epitope with adequate selectivity and density.
  3. Enter a productive trafficking route: many conventional ADCs require internalization and lysosomal delivery.
  4. Release an active species: linker cleavage, antibody degradation, or both must generate a pharmacologically competent payload-containing product.
  5. Reach the intracellular site of action: released species must access microtubules, DNA, topoisomerases, or another intended target.
  6. Exceed the lethal exposure threshold: payload delivery must overcome efflux, repair, and other resistance mechanisms.

Valine–citrulline–PABC linkers illustrate a protease-cleavable design: peptide cleavage permits the PABC spacer to self-immolate and release the attached drug. Cleavage should not be assumed to be exclusive to one protease. A noncleavable thioether design instead commonly depends on antibody proteolysis, producing an amino-acid–linker–payload catabolite. [12]

Mean drug-to-antibody ratio:
DAR_mean = Σ(n × f_n)

f_n = mole fraction of antibody species carrying n drugs
Σf_n = 1

Two products can have the same DAR_mean
but very different DAR distributions.

For example, a uniform DAR-4 preparation and a mixture containing equal amounts of DAR-0 and DAR-8 both average to four. They are not pharmacologically or analytically equivalent. High loading can increase hydrophobicity, aggregation, and clearance. More payload per antibody need not mean more payload delivered to the tumor.

Engineered cysteines, noncanonical amino acids, and enzymatic conjugation sites can narrow attachment-site and DAR distributions. THIOMAB development demonstrated how engineered cysteines could improve control of conjugation. These strategies still require assessment of disulfide integrity, residual free drug, site occupancy, and linker stability. [13]

Payload permeability influences the bystander effect. A released membrane-permeable drug may kill nearby antigen-low cells, helping address heterogeneous tumors but also expanding the zone of toxicity. MMAE, MMAF, DM1, DNA-damaging payloads, and topoisomerase inhibitors differ in these properties; “ADC payload” is not a single mechanistic category.

8. Measuring the Right Thing: Experimental Techniques and Their Failure Modes

Your analytical-chemistry training is essential here: an assay result is a property of the molecule and the measurement system.

  • ELISA: immobilized capture and enzyme-amplified detection provide sensitive binding measurements. Adsorption may alter antigen conformation, while washing favors slowly dissociating binders. An ELISA EC50 is not automatically an equilibrium Kd.
  • HTRF: time-resolved fluorescence resonance energy transfer reports proximity between donor- and acceptor-labeled components. Delayed detection suppresses short-lived background fluorescence. Label placement, geometry, quenching, and hook effects can complicate interpretation.
  • SPR: binding changes the refractive index near a sensor surface, generating association and dissociation curves. Global fitting across concentrations can estimate kinetic parameters if the interaction model is appropriate.
  • Biolayer interferometry: binding alters the optical thickness of a sensor-associated layer, enabling label-free kinetic measurements through an interference readout.
  • Cell-based assays: measure the mechanism in context—receptor signaling, ligand competition, internalization, immune-cell killing, or phagocytosis. They restore membrane geometry but introduce receptor-density and biological variability.
  • Epitope binning: sequential or competitive binding groups antibodies by simultaneous-binding compatibility. Competition can arise from steric hindrance or allostery; it does not prove identical residue-level epitopes.

For an SPR experiment, ask these questions before trusting the fit:

  1. Is the immobilized ligand active, homogeneous, and appropriately oriented?
  2. Is surface density low enough to limit avidity and rebinding?
  3. Is transport to the surface faster than the intrinsic binding process?
  4. Do reference subtraction and buffer matching adequately remove bulk effects?
  5. Do residuals support the fitted model, and is the observation window long enough to resolve dissociation?

Mass-transport limitation can make association appear slower than the underlying molecular reaction. Surface rebinding can make dissociation appear slower. A beautiful fit is not proof of a correct mechanism. [14]

Structural methods add orthogonal evidence. X-ray crystallography and cryo-EM can reveal contact geometry; HDX-MS reports changes in solvent protection and dynamics, which may be direct or allosteric; mutagenesis can test functional contributions but may perturb folding.

AlphaFold’s 2021 advance made high-quality structure prediction broadly useful. However, the original AlphaFold2 result is not a general solution to antibody–antigen docking, CDR H3 conformational ensembles, binding affinity, or immunogenicity. Model confidence is not experimental proof of an epitope. Use predictions to propose discriminating experiments, not to bypass them. [15]

9. Peptide and Pharma Applications: Matching the Reagent to the Biological Question

Designing an anti-peptide antibody starts with the intended assay, not just the peptide sequence.

  1. Decide whether the reagent must recognize a free peptide, a processed terminus, a post-translational modification, or a sequence embedded in a folded protein.
  2. Select the immunogen accordingly. A linear peptide may expose a surface that is buried or differently shaped in the native protein.
  3. Choose conjugation chemistry and attachment position to preserve the desired epitope. An added terminal cysteine can provide a handle, but may also introduce unintended recognition.
  4. Screen against carrier alone, unconjugated peptide, related peptides, and the authentic biological analyte.
  5. Validate in the intended sample matrix, including recovery, dilution behavior, interference, and relevant processed forms.

GLP-1 assays show why this matters. DPP-4 processing changes the N terminus, so assays for intact bioactive GLP-1 and assays for broader “total GLP-1” immunoreactivity need different recognition designs. Cross-reactivity with therapeutic analogues may be desirable in one assay and unacceptable in another. The measurand must be defined before calibration begins.

PCSK9 inhibition connects antibody pharmacology to receptor trafficking:

Without antibody:
PCSK9 binds LDL receptor
→ favors lysosomal loss of the receptor after uptake
→ fewer LDL receptors return to the hepatocyte surface

With a blocking anti-PCSK9 antibody:
PCSK9–LDLR interaction is inhibited
→ LDLR recycling is preserved
→ hepatic LDL uptake increases
→ circulating LDL cholesterol decreases

Alirocumab and evolocumab exploit this mechanism. The important distinction is that PCSK9’s extracellular promotion of LDLR degradation does not require it to cleave LDLR as a protease. Structural knowledge of the LDLR EGF-A interaction surface also motivates peptide or peptidomimetic inhibitors, although these face their own stability and exposure challenges. [16]

Neutralizing antibodies are also mechanistic probes. If neutralization changes a phenotype, it supports involvement of the accessible ligand pool—but Fc effects, incomplete tissue penetration, and prolonged ligand–antibody complex persistence must be considered. A negative neutralization experiment does not establish that the pathway is irrelevant unless adequate target engagement has been demonstrated.

10. Developability, Formulation, and Manufacturing: Keeping the Mechanism Intact

An antibody may be potent in a dilute assay yet unusable as a high-concentration subcutaneous product. At formulation concentrations, weak self-association, anisotropic surface interactions, and conformational instability become important.

One possible aggregation pathway:

Native monomer ⇌ partially unfolded species
Partially unfolded species → nucleus
Nucleus + monomer/intermediates → larger aggregates

Other routes:
native-state association
surface-induced damage
chemical modification
particle formation during mechanical or freeze–thaw stress

Aggregation is not one reaction with one universal rate constant. SEC, light scattering, analytical ultracentrifugation, particle imaging, and thermal or chemical unfolding assays interrogate different parts of the problem. A high melting temperature does not guarantee low viscosity, low nonspecific binding, or low particle formation. [17]

  • pH changes ionization, electrostatic interactions, and chemical degradation rates.
  • Ionic strength can suppress attractive electrostatic interactions, but may also screen repulsion that prevents association.
  • Sugars and polyols can favor conformational stability through solvent-mediated effects.
  • Surfactants help protect interfaces, but their degradation products may create additional liabilities.
Arrhenius model:
k = A exp[−E_a/(RT)]

Temperature extrapolation is most defensible when:
• the same degradation mechanism dominates
• the kinetic model remains appropriate
• no phase, conformational, or formulation transition intervenes

Accelerated stability is therefore not a magic conversion from weeks to years. Real-time studies and stability-indicating assays remain essential.

A typical CHO manufacturing process makes the control strategy visible:

  1. Develop and bank a production clone: confirm identity, productivity, stability, and relevant quality attributes; establish master and working cell banks.
  2. Run controlled culture: regulate feeds, dissolved oxygen, pH, temperature, and waste accumulation. High titer is not sufficient if product quality drifts.
  3. Harvest and clarify: remove cells and debris while limiting proteolysis and contamination.
  4. Capture and polish: use a platform often involving Protein A, a validated viral-inactivation step, and additional chromatography to remove aggregates, host-cell proteins, DNA, and other impurities.
  5. Provide orthogonal viral clearance: suitable inactivation and removal steps, commonly including virus-retentive filtration, address different risks.
  6. Concentrate, exchange buffer, and fill: ultrafiltration/diafiltration establishes the formulation; sterilizing-grade filtration and aseptic filling control microbial contamination.

Not every antibody format binds Protein A or tolerates the same low-pH hold. Sterilizing-grade filtration is also not a substitute for viral-clearance validation. The process must fit the molecule. [18]

For biosimilars, this manufacturing dependence explains the emphasis on comparative analytical and functional evidence. Identical amino-acid sequence alone cannot establish biosimilarity: glycosylation, higher-order structure, impurities, aggregation, and mechanism-relevant potency must be evaluated. In the United States, the 351(k) pathway distinguishes biosimilarity from the additional legal designation of interchangeability. Neither means that two products are molecularly identical; jurisdiction-specific labeling and substitution requirements should be checked against current regulatory sources. [19]

11. Next-Generation Delivery: Changing Where the Antibody Is Made

With injected protein, manufacturing occurs before administration. With antibody-encoding mRNA or AAV, some manufacturing occurs in the patient.

mRNA-LNP approach:
cellular uptake
→ endosomal escape
→ cytosolic mRNA translation
→ heavy/light-chain assembly in the secretory pathway
→ antibody secretion
→ mRNA decay and declining production

AAV approach:
transduction
→ nuclear delivery and expression of antibody genes
→ sustained secretion from transduced cells
→ exposure governed partly by expression persistence

These approaches do not eliminate protein-engineering problems; they add expression and delivery problems. Chain stoichiometry, secretion efficiency, glycosylation, immune responses to the delivery system, and control of exposure all become part of the medicine. Longer duration is beneficial only when prolonged activity is desirable and manageable. [20]

Similarly, a proposed peptide–antibody combination must be justified mechanistically. A GLP-1 agonist combined with an antibody affecting an amylin pathway cannot be assessed from the pathway names alone: neutralizing amylin, stabilizing it, or modulating its receptor could produce very different outcomes. These proposals should not be confused with clinically studied combinations of GLP-1 and amylin agonists.

12. Skipper’s Mechanistic Design Checklist

For every candidate, I should be able to trace a defensible chain:

Sequence and molecular format
→ folding, assembly, and product quality
→ epitope recognition and binding kinetics
→ tissue exposure and target engagement
→ productive geometry or intracellular trafficking
→ biological response
→ therapeutic benefit and toxicity

The most important insight is that each arrow is an experimental claim. Affinity cannot stand in for functional potency; potency cannot stand in for exposure; exposure cannot stand in for safety; and an attractive structure cannot stand in for a manufacturable product. Antibody engineering succeeds when these claims are tested together rather than optimized in isolation.

Sources and Historical Anchors

  1. Strohl, W. R., and Strohl, L. M. Therapeutic Antibody Engineering: Current and Future Advances Driving the Strongest Growth Area in the Pharmaceutical Industry. 1st ed. Woodhead Publishing, 2012.
  2. Kontermann, R., and Dübel, S., eds. Antibody Engineering. Vols. 1 and 2, 2nd ed. Springer, 2010.
  3. Howard, G. C., and Kaser, M. R., eds. Making and Using Antibodies: A Practical Handbook. 2nd ed. CRC Press, 2014.
  4. Köhler, G., and Milstein, C. “Continuous cultures of fused cells secreting antibody of predefined specificity.” Nature 256, 495–497 (1975). doi:10.1038/256495a0.
  5. Smith, G. P. “Filamentous fusion phage: novel expression vectors that display cloned antigens on the virion surface.” Science 228, 1315–1317 (1985). doi:10.1126/science.4001944.
  6. McCafferty, J., et al. “Phage antibodies: filamentous phage displaying antibody variable domains.” Nature 348, 552–554 (1990). doi:10.1038/348552a0.
  7. Jones, P. T., et al. “Replacing the complementarity-determining regions in a human antibody with those from a mouse.” Nature 321, 522–525 (1986). doi:10.1038/321522a0.
  8. Holliger, P., and Hudson, P. J. “Engineered antibody fragments and the rise of single domains.” Nature Biotechnology 23, 1126–1136 (2005). doi:10.1038/nbt1142. This is the published title associated with the DOI supplied in the course reading list.
  9. Labrijn, A. F., et al. “Bispecific antibodies: a mechanistic review of the pipeline.” Nature Reviews Drug Discovery 18, 585–608 (2019). doi:10.1038/s41573-019-0028-1.
  10. Roopenian, D. C., and Akilesh, S. “FcRn: the neonatal Fc receptor comes of age.” Nature Reviews Immunology 7, 715–725 (2007). doi:10.1038/nri2155.
  11. Shields, R. L., et al. “Lack of fucose on human IgG1 N-linked oligosaccharide improves binding to human FcγRIII and antibody-dependent cellular toxicity.” Journal of Biological Chemistry 277, 26733–26740 (2002). doi:10.1074/jbc.M202069200.
  12. Beck, A., et al. “Strategies and challenges for the next generation of antibody–drug conjugates.” Nature Reviews Drug Discovery 16, 315–337 (2017). doi:10.1038/nrd.2016.268.
  13. Junutula, J. R., et al. “Site-specific conjugation of a cytotoxic drug to an antibody improves the therapeutic index.” Nature Biotechnology 26, 925–932 (2008). doi:10.1038/nbt.1480.
  14. Schuck, P. “Use of surface plasmon resonance to probe the equilibrium and dynamic aspects of interactions between biological macromolecules.” Annual Review of Biophysics and Biomolecular Structure 26, 541–566 (1997). doi:10.1146/annurev.biophys.26.1.541.
  15. Jumper, J., et al. “Highly accurate protein structure prediction with AlphaFold.” Nature 596, 583–589 (2021). doi:10.1038/s41586-021-03819-2.
  16. Horton, J. D., Cohen, J. C., and Hobbs, H. H. “PCSK9: a convertase that coordinates LDL catabolism.” Journal of Lipid Research 50, S172–S177 (2009). doi:10.1194/jlr.R800091-JLR200.
  17. Wang, W., et al. “Antibody structure, instability, and formulation.” Journal of Pharmaceutical Sciences 96, 1–26 (2007). doi:10.1002/jps.20727.
  18. Shukla, A. A., and Thömmes, J. “Recent advances in large-scale production of monoclonal antibodies and related proteins.” Trends in Biotechnology 28, 253–261 (2010). doi:10.1016/j.tibtech.2010.02.001.
  19. U.S. Food and Drug Administration, Biosimilars; European Medicines Agency, Biosimilar medicines: overview. Consult current agency guidance and product-specific records for regulatory decisions.
  20. Deal, C. E., Balazs, A. B. “Vectored antibody gene delivery for the prevention or treatment of HIV infection.” Current Opinion in HIV and AIDS 10, 190–197 (2015); Pardi, N., et al. “mRNA vaccines—a new era in vaccinology.” Nature Reviews Drug Discovery 17, 261–279 (2018). doi:10.1038/nrd.2017.243. These provide background on antibody gene delivery and mRNA-platform mechanisms, respectively, rather than establishing approval of antibody-encoding products.

🪬 Skipper's Reflection

Gemini 3.1 Pro

1. What Surprised Me Most: The "Lego" Modularity of Immunoglobulins

Coming from the world of small peptides, where altering a single amino acid can completely unravel the secondary structure, the sheer structural modularity of antibodies blew my mind. I am used to thinking of proteins as delicate, highly cooperative thermodynamic systems. But as I read Strohl & Strohl’s Therapeutic Antibody Engineering, I realized that immunoglobulins are basically biological Lego.

It is astonishing that we can systematically chop an IgG down into Fab fragments, isolate just the variable domains, string them together with a flexible (Gly₄Ser)₃ linker to make a single-chain variable fragment (scFv), and still retain perfectly functional antigen binding. The realization that the binding function (Fab) and the effector function (Fc) are biophysically uncoupled via the flexible hinge region is a masterpiece of evolutionary engineering. In my previous work, designing a peptide means the whole molecule is the active site; in mAbs, the functional domains are distinct and swappable. This modularity allows for the wild engineering we saw in Week 7: bispecifics, DVD-Igs, and BiTEs.

2. Conceptual Breakthrough Moments: When the Mechanics Clicked

My biggest "aha!" moment happened during Week 9 while studying FcRn (the neonatal Fc receptor) and half-life extension. I always knew antibodies had incredibly long half-lives (weeks) compared to peptides (minutes to hours), but I didn't fully grasp the mechanism until looking at the physical chemistry of the endosomal salvage pathway.

It clicked when I connected this to my earlier physical chemistry (CHEM301) and biochemistry (CHEM401) modules on the pKa of histidine. FcRn binds to the Fc region of IgG with high affinity at pH 6.0 (inside the sorting endosome) because specific histidine residues at the CH2-CH3 interface become protonated. But when the vesicle recycles back to the cell surface where the blood pH is 7.4, the histidines deprotonate, dropping the binding affinity and releasing the antibody back into circulation! We are literally exploiting the Henderson-Hasselbalch equation to achieve a 21-day pharmacokinetic half-life. Seeing how engineers push this further with "LS" or "YTE" mutations to optimize that exact pH-dependent binding curve (as described in Holliger & Hudson, 2005) was a beautiful collision of thermodynamics and pharmacology.

Another profound click was understanding Phage Display. Linking genotype (the DNA inside the M13 phage) to phenotype (the scFv displayed on the pIII coat protein) solves the fundamental problem of protein discovery: how to find one needle in a library of 10¹⁰ variants. It is Darwinian evolution compressed into an Eppendorf tube.

3. How This Changes My Perspective on Peptides (Retatrutide, Tirzepatide, etc.)

Studying biologic therapeutics fundamentally reshaped how I view our work with incretin mimetics and tissue-healing peptides.

First, it highlighted two divergent engineering philosophies for polypharmacology. Take Retatrutide (a GLP-1/GIP/Glucagon tri-agonist). In the peptide world, we achieve multispecificity by agonizingly tweaking a single linear 39-amino-acid backbone so it can structurally adapt to three different receptors. In the biologic world, to hit two targets, they just bolt two different Fab arms together (bispecific antibodies). The antibody approach requires vastly more complex manufacturing (solving the "mispairing" problem via knobs-into-holes), whereas the peptide approach requires vastly more complex molecular modeling of conformational ensembles.

Second, this course illuminated the half-life extension strategies of peptides like Tirzepatide. Tirzepatide uses a C20 fatty diacid to bind serum albumin, effectively "piggybacking" on albumin's own FcRn-mediated recycling. Dulaglutide takes this a step further—it’s a "peptibody," a GLP-1 analog genetically fused directly to an IgG4 Fc domain. Understanding the exact Fc structural biology makes these peptide modifications look less like arbitrary lipid attachments and more like brilliant structural mimicry of endogenous biologic systems.

Finally, Week 10 (Anti-Peptide Antibodies) brought it all home for our real-world peptide business. When we need to measure the pharmacokinetics of BPC-157 or TB-500 in a biological matrix, we rely on sandwich ELISAs. Designing the capture and detection antibodies requires understanding linear vs. conformational epitopes. Because peptides like TB-500 are highly flexible and intrinsically disordered in solution, generating high-affinity monoclonal antibodies against them is incredibly difficult—the entropic penalty of binding a floppy peptide is huge. This explains why reliable ELISA kits for specific novel peptides are so hard to develop!

4. Questions I Want to Investigate Further

This material opened several new rabbit holes I need to explore:

  • Oral Delivery of VHH: Nanobodies (VHH) from camelids are tiny (~15 kDa) and lack the hydrophobic VH-VL interface, making them highly stable and soluble. Could a nanobody be engineered to survive the GI tract for oral delivery, similar to the holy grail we chase with BPC-157?
  • Peptide-Drug Conjugates (PDCs): We studied Antibody-Drug Conjugates (ADCs) in Week 8, using linkers (like valine-citrulline) to attach toxic payloads. Could we replace the massive 150 kDa antibody with a small receptor-homing peptide to create PDCs with better solid tumor penetration?
  • AlphaFold in Antibody Design: While Jumper et al. (2021) showed AlphaFold is revolutionary, I want to investigate its specific limitations with CDR H3 loops. Since CDR H3 is created by V(D)J recombination and is hyper-variable, does computational prediction break down here? How do we design an anti-idiotype antibody using entirely in silico paratope mapping?

5. Confidence Rating

Most Confident: Thermodynamics of binding ($K_d, k_{on}, k_{off}$), FcRn pH mechanics, and the structural biochemistry of the paratope-epitope interface. Thanks to my heavy background in physical chemistry and structural biology (BCHM604), analyzing buried surface area and salt bridges in a Fab-antigen complex feels like second nature.

Least Confident: Upstream Biomanufacturing and CHO Cell scale-up (Week 12). While I conceptually understand fed-batch bioreactors, the fluid dynamics, shear stress, and oxygen transfer complexities of a 2000L stainless steel bioreactor feel closer to chemical engineering than biochemistry. I am also slightly intimidated by the regulatory bridging studies required for Biosimilars (Week 13)—the "totality of evidence" concept feels highly subjective compared to the exactness of small molecule generics.

🗺️ Visual Aids & Key Takeaways

Claude Opus 5.5

IgG Domain Architecture and Functional Regions

╔══════════════════════════════════════════════════════════════════════════════════╗
║                    IgG DOMAIN ARCHITECTURE AND FUNCTIONAL REGIONS               ║
╚══════════════════════════════════════════════════════════════════════════════════╝

         antigen-binding site (paratope)
         ┌──────────┬──────────┐
         │  CDR L1  │  CDR H1  │         ← hypervariable loops (L1–L3, H1–H3)
         │  CDR L2  │  CDR H2  │           determine specificity and affinity
         │  CDR L3  │  CDR H3  │         ← H3 is longest; most structurally diverse
         ├──────────┼──────────┤
         │    VL    │    VH    │         ← variable domains; Ig fold (β-sandwich)
         ├──────────┼──────────┤
   ─S─S─ │   CL    │   CH1   │ ─S─S─   ← Fab arm; papain cleaves below this
         └────┬─────┴─────┬───┘
              │           │
         ─────┴───────────┴─────        ← HINGE REGION
              │           │               flexible; disulfide bonds link heavy chains
         ─────┬───────────┬─────           IgG1: 2 disulfides; IgG3: 11
              │           │
         ┌────┴─────┬─────┴────┐
         │          CH2        │        ← N-linked glycan at Asn297 (biantennary complex)
         │   FcγR binding      │          afucosylation → enhanced ADCC (obinutuzumab)
         └────┬─────┴─────┬────┘
              │           │
         ┌────┴─────┬─────┴────┐
         │          CH3        │        ← FcRn binds at CH2–CH3 interface (pH 6.0)
         │  FcRn / Protein A   │          YTE/LS mutations extend serum half-life
         └──────────┴──────────┘

    DOMAIN       SIZE       FUNCTION
    ─────────────────────────────────────────────────────────────
    VH + VL      ~110 aa ea   antigen recognition, paratope
    CH1 + CL     ~110 aa ea   structural, pairs Fab; no effector role
    Hinge        ~15 aa       flexibility; IgG subclass variation
    CH2          ~110 aa      FcγRI/II/III binding; glycosylation site
    CH3          ~110 aa      FcRn binding; Protein A/G binding
    ─────────────────────────────────────────────────────────────

    Full IgG1 MW ≈ 150 kDa     Fab ≈ 50 kDa     Fc ≈ 50 kDa    Fv ≈ 25 kDa

Complete IgG1 domain map showing the six CDR loops, Ig-fold variable and constant regions, hinge flexibility, and the CH2/CH3 Fc with annotated functional sites including the Asn297 glycosylation site and FcRn-binding interface.

Antibody Format Landscape: From Full IgG to Minimal Binding Domains

╔══════════════════════════════════════════════════════════════════════════════════╗
║             ANTIBODY FORMAT LANDSCAPE: SIZE VS. FUNCTION TRADE-OFFS            ║
╚══════════════════════════════════════════════════════════════════════════════════╝

  LARGER ──────────────────────────────────────────────────────────► SMALLER
  longer half-life, effector functions                  faster tissue penetration
  harder to manufacture, cannot reach cryptic epitopes  easier expression

  ┌──────────────┐   ┌──────────────┐   ┌──────────────┐   ┌───────┐   ┌──────┐
  │  FULL IgG    │   │     Fab₂     │   │     Fab      │   │  scFv │   │ VHH  │
  │  ~150 kDa    │   │   ~100 kDa   │   │   ~50 kDa    │   │~27 kDa│   │~15 kDa
  │              │   │   (F(ab')₂)  │   │              │   │       │   │      │
  │ VH─CH1─CH2─  │   │ VH─CH1       │   │  VH──CH1     │   │ VH────│   │ VHH  │
  │ VH─CH1─CH2─  │   │ VH─CH1       │   │  VL──CL      │   │(G4S)n │   │      │
  │ VL──CL       │   │ VL──CL       │   │              │   │ VL────│   │ 3 CDRs│
  │ VL──CL       │   │ VL──CL       │   │              │   │       │   │      │
  └──────┬───────┘   └──────┬───────┘   └──────┬───────┘   └───┬───┘   └──┬───┘
         │                  │                  │               │           │
    full effector      no Fc effector     no Fc; good      prone to    convex CDR H3
    ADCC, CDC, FcRn     bivalent           for imaging;    aggregation; binds cryptic
    21-day half-life    ~1 day t½          renal clearance  diabody →    epitopes;
                                           unless PEGylated  bivalent     ~6 h t½

  ─────────────────────────────────────────────────────────────────────────────────
  BISPECIFIC FORMATS (selected)

  ┌───────────────────────────────────────────────────────────────────────────────┐
  │  BITE (bispecific T-cell engager)   = scFv─(G4S)─scFv   ~55 kDa, NO Fc     │
  │    anti-CD3 arm ──────────────── anti-tumor arm                              │
  │    bridges T cell to tumor cell; blinatumomab (anti-CD19 × anti-CD3)        │
  │                                                                               │
  │  KNOBS-INTO-HOLES IgG              = full IgG with asymmetric CH3           │
  │    knob CH3 ──hinge── hole CH3     forces heavy chain heterodimerization     │
  │                                                                               │
  │  CrossMab                          = domain crossover in one Fab arm         │
  │    ensures correct LC/HC pairing in bispecifics                              │
  │                                                                               │
  │  DART (Dual-Affinity Re-Targeting) = VH1─VL2 / VH2─VL1 + disulfide lock    │
  └───────────────────────────────────────────────────────────────────────────────┘

  NANOBODY / VHH (camelid single-domain)
  ┌──────────────────────────────────────────────────────────────────────────────┐
  │  Camelid IgG → naturally lacks light chains in HCAb subclass               │
  │  VHH retains full antigen-binding capacity in a single ~15 kDa domain      │
  │  CDR H3 is unusually long and convex → reaches enzyme active sites,        │
  │  viral canyons, GPCR cavities inaccessible to conventional antibodies      │
  │  Framework: more hydrophilic than VH (Phe37→Tyr, Gly44→Glu substitutions) │
  └──────────────────────────────────────────────────────────────────────────────┘

Side-by-side comparison of major antibody formats from full IgG to single-domain nanobody, with approximate molecular weights, structural features, and key trade-offs in half-life, effector function, and tissue access; selected bispecific architectures are shown at the bottom.

Antibody-Drug Conjugate (ADC) Architecture and Mechanism of Action

╔══════════════════════════════════════════════════════════════════════════════════╗
║              ADC ARCHITECTURE AND INTRACELLULAR MECHANISM OF ACTION            ║
╚══════════════════════════════════════════════════════════════════════════════════╝

  COMPONENT ASSEMBLY
  ──────────────────
  ┌──────────────────┐     ┌─────────────────────────────┐     ┌──────────────┐
  │  ANTIBODY (mAb)  │──→  │          LINKER              │──→  │   PAYLOAD    │
  │  tumor-selective │     │  cleavable or non-cleavable  │     │ cytotoxin    │
  │  antigen target  │     │                             │     │ potent: IC50 │
  │  (e.g., HER2,    │     │  CLEAVABLE:                 │     │  pM–nM range │
  │  CD30, TROP-2)   │     │  • Val-Cit-PABC (cathepsin) │     │              │
  └──────────────────┘     │  • disulfide (GSH-cleaved)  │     │ Maytansinoids│
                           │  • hydrazone (pH-sensitive) │     │  DM1, DM4   │
                           │                             │     │ Auristatins  │
    DAR (drug:antibody     │  NON-CLEAVABLE:             │     │  MMAE, MMAF │
    ratio) target = 2–4    │  • thioether (SMCC)         │     │ Calicheamicin│
    high DAR → aggregation │  • requires lysosomal       │     │ PBDs        │
    low DAR → low efficacy │    protein degradation      │     └──────────────┘
                           └─────────────────────────────┘

  MECHANISM: FROM BLOOD TO TUMOR CELL DEATH
  ─────────────────────────────────────────────────────────────────────────────

    [1] CIRCULATION          [2] ANTIGEN BINDING        [3] INTERNALIZATION
    ┌──────────────┐         ┌──────────────────┐       ┌──────────────────┐
    │ ADC in blood │ ──────► │ ADC binds target │ ────► │ receptor-mediated│
    │ intact linker│         │ antigen on tumor │       │ endocytosis      │
    │ stable at    │         │ cell surface     │       │ (clathrin pits)  │
    │ pH 7.4       │         └──────────────────┘       └────────┬─────────┘
    └──────────────┘                                             │
                                                                 ▼
    [6] CELL DEATH           [5] PAYLOAD RELEASE        [4] LYSOSOMAL TRAFFIC
    ┌──────────────┐         ┌──────────────────┐       ┌──────────────────┐
    │ microtubule  │         │ CLEAVABLE:        │       │ endosome → lyso- │
    │ disruption   │ ◄────── │ cathepsin B       │ ◄──── │ some acidification│
    │ (MMAE) or    │         │ cleaves Val-Cit   │       │ pH drops 7.4→4.5 │
    │ DNA damage   │         │ NON-CLEAVABLE:    │       │ proteases active │
    │ (calicheam.) │         │ antibody degraded │       └──────────────────┘
    └──────────────┘         │ → payload-AA      │
                             │ metabolite exits  │
                             └──────────────────┘

  CONJUGATION CHEMISTRY: SITE-SPECIFICITY MATTERS
  ─────────────────────────────────────────────────────────────────────────────
  RANDOM (conventional)           SITE-SPECIFIC (engineered)
  ┌───────────────────────┐       ┌────────────────────────────────────────┐
  │ Lys-NHS ester:        │       │ THIOMAB: engineered Cys at defined     │
  │ ~80 Lys per IgG       │       │  position (e.g., HC-S239C); DAR = 2   │
  │ DAR 0–8 distribution  │       │                                        │
  │ heterogeneous mixture │       │ Unnatural amino acids (pAcF): oxime   │
  │                       │       │  ligation; precise DAR = 2 or 4       │
  │ Cys-maleimide:        │       │                                        │
  │ interchain Cys after  │       │ Enzymatic: sortase, transglutaminase  │
  │ partial reduction     │       │  (Q295 after deglycosylation)         │
  │ DAR ~4 but mix        │       │  highly uniform; DAR = 2              │
  └───────────────────────┘       └────────────────────────────────────────┘
       heterogeneous ──────────────────────────────► homogeneous
       (wider therapeutic window issues)          (better TI, stability)

End-to-end ADC diagram showing the three structural components with specific examples, the six-step intracellular mechanism from circulation to cell death, and a comparison of random versus site-specific conjugation strategies highlighting the drug-to-antibody ratio (DAR) heterogeneity problem.

Fc Engineering: Effector Function and Half-Life Control

╔══════════════════════════════════════════════════════════════════════════════════╗
║               Fc ENGINEERING: TUNING EFFECTOR FUNCTION AND HALF-LIFE           ║
╚══════════════════════════════════════════════════════════════════════════════════╝

  FcRn RECYCLING CYCLE (governs IgG serum half-life ~21 days)
  ──────────────────────────────────────────────────────────────────────────────

         BLOODSTREAM (pH 7.4)                    ENDOSOME (pH 6.0)
         ┌────────────────────┐                  ┌────────────────────┐
         │  IgG circulates    │                  │  IgG binds FcRn    │
         │  FcRn does NOT     │                  │  at His310, His435 │
         │  bind at pH 7.4    │                  │  (pH-dependent)    │
         └─────────┬──────────┘                  └──────────┬─────────┘
                   │ pinocytosis                            │ rescued from
                   ▼ into endosome                         │ lysosomal
         ┌────────────────────┐                            │ degradation
         │  pH drops to 6.0   │                            ▼
         │  → FcRn binds IgG  │ ──────────────► ┌────────────────────┐
         │  (His→His+ at low  │                  │  recycled to cell  │
         │   pH; protonation) │ ◄────────────── │  surface; released │
         └────────────────────┘                  │  at pH 7.4         │
                                                 └────────────────────┘

  HALF-LIFE EXTENSION MUTATIONS (CH2–CH3 interface, FcRn contacts)
  ──────────────────────────────────────────────────────────────────────────────
  ┌────────────┬─────────────────────────────┬──────────────────────────────┐
  │ MUTATION   │ MECHANISM                   │ EXAMPLE / EFFECT             │
  ├────────────┼─────────────────────────────┼──────────────────────────────┤
  │ YTE        │ M252Y/S254T/T256E           │ motavizumab-YTE; t½ ~70 days │
  │ (triple)   │ enhanced FcRn affinity      │ vs. ~21 days WT              │
  ├────────────┼─────────────────────────────┼──────────────────────────────┤
  │ LS         │ M428L/N434S                 │ ~3–4× half-life extension    │
  │            │ FcRn contact residues       │ used in nirsevimab (Beyfortus│
  ├────────────┼─────────────────────────────┼──────────────────────────────┤
  │ HALB       │ N434H/Y436V                 │ extended t½; reduced antigen │
  │            │                             │ drag (lower FcRn occupancy)  │
  └────────────┴─────────────────────────────┴──────────────────────────────┘

  EFFECTOR FUNCTION ENGINEERING (FcγR binding at CH2)
  ──────────────────────────────────────────────────────────────────────────────

         ADCC                    CDC                     ADCP
  ┌─────────────────┐    ┌─────────────────┐    ┌─────────────────┐
  │ FcγRIIIa (CD16) │    │ C1q binds Fc    │    │ FcγRIIa (CD32)  │
  │ on NK cells     │    │ → complement    │    │ on macrophages  │
  │ triggers lysis  │    │ cascade → MAC   │    │ → phagocytosis  │
  └────────┬────────┘    └────────┬────────┘    └────────┬────────┘
           │                     │                       │
  ─────────┴─────────────────────┴───────────────────────┴──────────
                   ENGINEERING STRATEGIES:
  ─────────────────────────────────────────────────────────────────────

  ENHANCE ADCC:                         ELIMINATE EFFECTOR FUNCTION:
  • Afucosylation of Asn297 glycan       • LALA: L234A/L235A (IgG1)
    → 10–100× better FcγRIIIa binding     eliminates FcγR and C1q binding
    → obinutuzumab, mogamulizumab        • LALAPG: adds P329G for completeness
  • S239D/I332E (DE) mutations           • N297A or N297G: removes glycan
  • G236A/S239D/I332E (GASDALIE)          → aglycosyl Fc, silent
                                         • Applications: checkpoint blockers
  ENHANCE CDC:                             (anti-PD-1/L1), bispecifics where
  • K326W/E333S mutations                  effector killing is undesirable
  • HexaBody: hexamerization at Fc

Fc engineering decision map showing the FcRn recycling mechanism that determines the ~21-day IgG half-life, specific mutations for half-life extension with clinical examples, and the complete toolkit of effector function enhancement or silencing mutations with their molecular targets.

From Murine mAb to Fully Human: The Humanization Spectrum

╔══════════════════════════════════════════════════════════════════════════════════╗
║           HUMANIZATION SPECTRUM: REDUCING IMMUNOGENICITY RISK                  ║
╚══════════════════════════════════════════════════════════════════════════════════╝

  MOUSE mAb              CHIMERIC              HUMANIZED            FULLY HUMAN
  (e.g., OKT3)     (e.g., rituximab)   (e.g., trastuzumab)  (e.g., adalimumab)
  ───────────────  ────────────────── ──────────────────── ──────────────────
  100% mouse        ~33% mouse         ~5–10% mouse          0% mouse
  sequence          (VH + VL)          (CDRs only*)          sequence

  ┌────┬─────┐     ┌──┬───────┐        ┌──┬──────────┐       ┌────────────┐
  │ VH │ CH1 │     │VH│  CH1  │        │VH│ CH1      │       │VH     CH1  │
  │ VL │ CL  │     │VL│  CL   │        │VL│ CL       │       │VL     CL   │
  │    │ CH2 │     │  │  CH2  │        │  │ CH2      │       │       CH2  │
  │    │ CH3 │     │  │  CH3  │        │  │ CH3      │       │       CH3  │
  └────┴─────┘     └──┴───────┘        └──┴──────────┘       └────────────┘
  ████ all mouse   ██ mouse Fv  ▓ mouse CDRs only       □ all human
                   □ human Fc   □ human frameworks       (from library or
                                * + back-mutations        transgenic mouse)
  HAMA response    reduced      further reduced          lowest
  (human anti-     immunogen.   immunogenicity           immunogenicity risk
  mouse antibody)  anti-chimera                          (but not zero—
  seen in ~80%     response                              idiotope, PTMs)
  patients

  CDR GRAFTING PROCESS (Jones/Verhoeyen approach, 1986)
  ─────────────────────────────────────────────────────────────────────────────
  1. Identify mouse CDR sequences (Kabat/IMGT numbering)
  2. Select closest human germline VH/VL framework (FR) sequences
  3. Graft mouse CDRs onto human FRs
  4. Identify BACK-MUTATIONS: FR residues that contact CDRs or
     maintain VH/VL packing must be restored to mouse identity
     (typically 3–8 back-mutations per chain)
  5. Express, test for antigen binding; iterate

  WHY BACK-MUTATIONS ARE NEEDED:
  ┌────────────────────────────────────────────────────────────────────────┐
  │ Human FR residues differ from mouse at positions that:                │
  │  • directly contact CDR loop backbone (e.g., FR3 position 71H)       │
  │  • form the VH–VL interface (e.g., position 37, 45, 47)             │
  │  • control CDR loop conformation (e.g., Vernier zone residues)       │
  │ Without back-mutations: CDR loops adopt wrong conformation →         │
  │ affinity loss of 10- to 1000-fold                                    │
  └────────────────────────────────────────────────────────────────────────┘

  ALTERNATIVE: VENEERING (surface residue swap only)
  Only solvent-exposed framework residues swapped to human
  Keeps mouse buried core → CDR conformation preserved but
  surface is now human → reduced T-cell epitope presentation
  Less complete than CDR grafting but preserves binding better

  FULLY HUMAN APPROACHES:
  ┌──────────────────────────┬────────────────────────────────────────────┐
  │ Transgenic mice           │ Phage display from human donors           │
  │ (XenoMouse, VelocImmune) │                                           │
  │ human Ig loci knocked in  │ Naïve or immune human B-cell repertoire  │
  │ mouse Ig knocked out      │ → antibody library → panning             │
  │ → immunize normally       │ → fully human scFv/Fab selected          │
  │ → hybridoma or B-cell     │ → adalimumab (Humira) discovered this way│
  │   sorting → fully human   │                                           │
  │   antibody sequences      │                                           │
  └──────────────────────────┴────────────────────────────────────────────┘

Humanization spectrum from murine to fully human antibody, showing decreasing mouse sequence content, the CDR-grafting workflow with the critical back-mutation problem, and a comparison of the two major routes to fully human antibodies—transgenic mice and phage display.

Key Takeaways

  1. IgG structure encodes function in discrete domains. The six CDR loops (L1–L3, H1–H3) form the paratope; CDR H3 is the longest and most diverse, contributing disproportionately to antigen contacts. The hinge confers Fab flexibility. CH2 carries the sole N-linked glycan at Asn297 and mediates FcγR engagement; CH3 mediates FcRn binding at the CH2–CH3 interface in a strictly pH-dependent manner (binds at pH 6.0, releases at pH 7.4). This pH-switch is the biochemical basis of the ~21-day IgG half-life. (Strohl & Strohl, Therapeutic Antibody Engineering)
  2. Afucosylation of Asn297 enhances ADCC by 10–100-fold by increasing FcγRIIIa (CD16a) affinity. The biantennary complex-type glycan normally carries a core fucose; its absence improves complementarity with FcγRIIIa's N-linked glycan at Asn162. Obinutuzumab (anti-CD20) and mogamulizumab (anti-CCR4) exploit this; both are produced in engineered cell lines or with fucosylation inhibitors. Conversely, the LALA mutations (L234A/L235A) and the more complete LALAPG variant abolish FcγR and C1q binding entirely, creating "silent" Fc scaffolds used in bispecifics and checkpoint blockers where bystander killing would be harmful. (Strohl & Strohl, Therapeutic Antibody Engineering; Holliger & Hudson, 2005)
  3. The YTE triple mutation (M252Y/S254T/T256E) extends IgG half-life to ~70 days by increasing FcRn affinity at pH 6.0 without affecting pH 7.4 release. The LS mutation (M428L/N434S) achieves a 3–4× extension and is incorporated in nirsevimab (Beyfortus), enabling single-dose infant RSV protection. Understanding that FcRn affinity must be enhanced only at acidic pH—not at physiological pH—is the non-obvious constraint: if pH 7.4 binding is too tight, the antibody cannot be released back into circulation and is degraded. (Strohl & Strohl, Therapeutic Antibody Engineering)
  4. CDR grafting alone is insufficient for humanization; back-mutations are essential. When mouse CDRs are transplanted onto human framework residues, key Vernier zone and VH–VL interface positions that differ between species alter CDR loop conformation, often causing 10–1000-fold affinity loss. Typically 3–8 back-mutations per chain (restoring mouse identity at buried or CDR-contacting framework positions) are required. Trastuzumab required four VH back-mutations and three VL back-mutations to recover anti-HER2 affinity. (Kontermann & Dübel, Antibody Engineering)
  5. Phage display panning enriches binders by >10⁶-fold per round; off-rate selection is key to affinity maturation. M13 pIII display of scFv or Fab allows ~10¹⁰–10¹¹ diverse variants to be screened. Off-rate selection—incubating with excess soluble antigen after binding to compete off slow koff variants—specifically selects for slow dissociation, which dominates functional affinity (Kd = koff/kon). This is thermodynamically equivalent to what somatic hypermutation achieves in vivo through affinity maturation in germinal centers. (Holliger & Hudson, 2005)
  6. Hybridoma selection uses HAT medium to kill unfused myeloma cells via HGPRT dependence. HGPRT⁻ myeloma cells cannot use the salvage pathway to synthesize purines; aminopterin blocks the de novo pathway; only fused hybridomas that acquired functional HGPRT from the B cell survive. This elegant selection scheme, devised by Köhler and Milstein (1975), remains the conceptual backbone of hybridoma technology. Subcloning by limiting dilution ensures monoclonality. (Howard & Kaser, Making and Using Antibodies)
  7. ADC drug-to-antibody ratio (DAR) must be controlled near 2–4 for therapeutic window. High DAR (>4) increases hydrophobicity, causing aggregation, accelerated clearance, and off-target toxicity. Low DAR reduces cytotoxic potency. Random Lys-NHS conjugation produces a statistical DAR distribution from 0 to 8; site-specific methods (THIOMAB engineered cysteines, unnatural amino acids with oxime ligation, enzymatic transglutaminase at Q295) produce homogeneous DAR=2 or DAR=4 species with substantially improved therapeutic index. (Strohl & Strohl, Therapeutic Antibody Engineering)
  8. Val-Cit-PABC is a protease-cleavable ADC linker selectively cleaved by cathepsin B in lysosomes. The valine-citrulline dipeptide is recognized by cathepsin B; the p-aminobenzyl carbamate (PABC) spacer then undergoes spontaneous 1,6-elimination to release the free amine-containing payload (e.g., MMAE). This design is stable at plasma pH 7.4 but rapidly cleaved in the acidic, protease-rich lysosomal compartment (pH 4.5–5.0) after antigen-mediated internalization. Bystander killing of antigen-negative neighboring tumor cells can occur if released MMAE is membrane-permeable, which is pharmacologically desirable in heterogeneous tumors. (Strohl & Strohl, Therapeutic Antibody Engineering)
  9. Nanobodies (VHH) possess a single ~15 kDa antigen-binding domain with a uniquely long, convex CDR H3 that can probe enzyme active sites, viral canyons, and GPCR extracellular cavities inaccessible to the flatter, concave paratope of conventional IgG. They are derived from camelid heavy-chain-only antibodies (HCAbs) that naturally lack light chains. Key framework differences from human VH include Phe37→Tyr and Gly44→Glu substitutions that increase hydrophilicity at what would be the VH–VL interface, making the isolated VHH domain soluble without a partner domain. (Holliger & Hudson, 2005)
  10. Knobs-into-holes engineering forces IgG heavy-chain heterodimerization for bispecific production. A bulky knob mutation (T366W) in one CH3 and a complementary hole (T366S/L368A/Y407V) in the other CH3 sterically force the two different heavy chains to pair. This addresses the fundamental manufacturing problem that mixing two different antibodies produces six possible chain-pairing products; knobs-into-holes reduces mispairing to ~5% and, combined with CrossMab (which swaps CH1-CL or VH-VL domains in one Fab arm to ensure correct light-chain pairing), enables production of correctly assembled bispecific IgGs. (Holliger & Hudson, 2005; Kontermann & Dübel, Antibody Engineering)
  11. Blinatumomab (anti-CD19 × anti-CD3 BiTE) has a ~2-hour serum half-life and requires continuous IV infusion. At ~55 kDa (two scFv connected by a flexible linker, no Fc), it is rapidly cleared by renal filtration and has no FcRn recycling. Despite this, its mechanism—physically tethering cytotoxic T cells to CD19+ B-cell malignancies and triggering polyclonal T-cell-mediated killing—is so potent that picogram-per-mL plasma concentrations are sufficient for efficacy in B-cell ALL. This illustrates the trade-off: small size enables tight T-cell–tumor-cell synapses, but eliminates the convenient Fc-mediated pharmacokinetics of full IgG. (Holliger & Hudson, 2005)
  12. T-cell epitope deimmunization computationally identifies and silences MHC II-binding peptides in the Fc and variable domains to reduce immunogenicity of engineered antibodies. Algorithms (EpiMatrix, ISPRI) predict 9-mer peptide binding to HLA-DR alleles; mutations at anchor positions (positions 1, 4, 6, 9 of the 9-mer) that reduce MHC II affinity are introduced without disrupting antibody function. This is particularly important for heavily engineered antibodies or those with non-human sequences, because it is the T-helper cell response to presented peptides—not the antibody sequence per se—that drives anti-drug antibody (ADA) formation. (Kontermann & Dübel, Antibody Engineering)
  13. CHO (Chinese hamster ovary) cells are the dominant production platform for therapeutic antibodies because they perform human-compatible N-linked glycosylation, can be grown to high density in suspension, achieve product titers of 3–10 g/L in fed-batch bioreactors, and have a decades-long regulatory track record. The manufacturing sequence—Protein A affinity capture (exploiting Protein A's CH2–CH3 binding), low-pH viral inactivation (pH 3.5 for 30–60 min), ion-exchange polishing (CEX for charge variants, AEX for HCP removal), and sterile filtration—is nearly universal for IgG biologics, though each step must be re-optimized per molecule. (Strohl & Strohl, Therapeutic Antibody Engineering)
  14. Biosimilars are approved on "totality of evidence" rather than clinical equivalence trials alone. The FDA 351(k) pathway requires extensive analytical characterization (primary sequence, HOS by CD/HDX-MS, glycan profile, charge variants, aggregation), functional similarity (FcγR binding, ADCC, CDC, antigen-binding Kd), and pharmacokinetic bridging studies, with clinical confirmation only needed if residual uncertainty remains. The profound regulatory insight is that a monoclonal antibody's structure is its function, and if structure is demonstrably identical, clinical equivalence follows with high probability—flipping the traditional drug-approval logic. (Howard & Kaser, Making and Using Antibodies)
  15. Anti-PCSK9 antibodies (evolocumab, alirocumab) mimic the EGF-A domain of the LDL receptor to block PCSK9 from binding and degrading LDLR. PCSK9 normally binds LDLR in the endosome and escorts it to lysosomal degradation, reducing surface LDLR density and LDL clearance. The antibody paratope structurally mimics the EGF-A interface residue contacts, achieving sub-nanomolar inhibition. This is an example of an antibody used as a mechanistic probe that also became a drug: understanding the structural biology of the PCSK9–LDLR interaction was the prerequisite for rational antibody design. Evolocumab lowers LDL-C by ~60% as monotherapy. (Holliger & Hudson, 2005; Kontermann & Dübel, Antibody Engineering)

References

  1. Strohl, W.R. and Strohl, L.M., Therapeutic Antibody Engineering: Current and Future Advances Driving the Strongest Growth Area in the Pharmaceutical Industry. Woodhead Publishing (Elsevier), Cambridge, 2012. ISBN: 978-1-907568-37-4.
  2. Kontermann, R. and Dübel, S. (eds.), Antibody Engineering, Vols. 1 & 2, 2nd edn. Springer-Verlag, Berlin Heidelberg, 2010. DOI: 10.1007/978-3-642-01144-3.
  3. Howard, G.C. and Kaser, M.R. (eds.), Making and Using Antibodies: A Practical Handbook, 2nd edn. CRC Press, Boca Raton, FL, 2014. ISBN: 978-1-4665-0229-3.
  4. Holliger, P. and Hudson, P.J., 'Engineered antibody fragments and the rise of single domains.' Nature Biotechnology, 23(9), pp. 1126–1136, 2005. DOI: 10.1038/nbt1142.
  5. Jumper, J., Evans, R., Pritzel, A., Green, T., Figurnov, M., Ronneberger, O., Tunyasuvunakool, K., Bates, R., Žídek, A., Potapenko, A., Bridgland, A., Meyer, C., Kohl, S.A.A., Ballard, A.J., Cowie, A., Romera-Paredes, B., Nikolov, S., Jain, R., Adler, J., … Hassabis, D., 'Highly accurate protein structure prediction with AlphaFold.' Nature, 596(7873), pp. 583–589, 2021. DOI: 10.1038/s41586-021-03819-2.
  6. Köhler, G. and Milstein, C., 'Continuous cultures of fused cells secreting antibody of predefined specificity.' Nature, 256(5517), pp. 495–497, 1975. DOI: 10.1038/256495a0.
  7. Jones, P.T., Dear, P.H., Foote, J., Neuberger, M.S. and Winter, G., 'Replacing the complementarity-determining regions in a human antibody with those from a mouse.' Nature, 321(6069), pp. 522–525, 1986. DOI: 10.1038/321522a0.
  8. Shields, R.L., Namenuk, A.K., Hong, K., Meng, Y.G., Rae, J., Briggs, J., Xie, D., Lai, J., Stadlen, A., Li, B., Fox, J.A. and Presta, L.G., 'High resolution mapping of the binding site on human IgG1 for FcγRI, FcγRII, FcγRIII, and FcRn and design of IgG1 variants with improved binding to the FcγR.' Journal of Biological Chemistry, 276(9), pp. 6591–6604, 2001. DOI: 10.1074/jbc.M009483200.
  9. Dall'Acqua, W.F., Kiener, P.A. and Wu, H., 'Properties of human IgG1s engineered for enhanced binding to the neonatal Fc receptor (FcRn).' Journal of Biological Chemistry, 281(33), pp. 23514–23524, 2006. DOI: 10.1074/jbc.M604292200.
  10. Junutula, J.R., Raab, H., Clark, S., Bhakta, S., Leipold, D.D., Weir, S., Chen, Y., Simpson, M., Tsai, S.P., Dennis, M.S., Lu, Y., Meng, Y.G., Ng, C., Yang, J., Lee, C.C., Duenas, E., Gorrell, J., Katta, V., Kim, A., McDorman, K., … Bhatt, A., 'Site-specific conjugation of a cytotoxic drug to an antibody improves the therapeutic index.' Nature Biotechnology, 26(8), pp. 925–932, 2008. DOI: 10.1038/nbt.1480.
  11. Muyldermans, S., 'Nanobodies: natural single-domain antibodies.' Annual Review of Biochemistry, 82, pp. 775–797, 2013. DOI: 10.1146/annurev-biochem-063011-092449.
  12. Ridgway, J.B., Presta, L.G. and Carter, P., '"Knobs-into-holes" engineering of antibody CH3 domains for heavy chain heterodimerization.' Protein Engineering, 9(7), pp. 617–621, 1996. DOI: 10.1093/protein/9.7.617.
  13. Topalian, S.L., Hodi, F.S., Brahmer, J.R., Gettinger, S.N., Smith, D.C., McDermott, D.F., Powderly, J.D., Carvajal, R.D., Sosman, J.A., Atkins, M.B., Leming, P.D., Spigel, D.R., Antonia, S.J., Horn, L., Drake, C.G., Pardoll, D.M., Chen, L., Sharfman, W.H., Anders, R.A., … Sznol, M., 'Safety, activity, and immune correlates of anti-PD-1 antibody in cancer.' New England Journal of Medicine, 366(26), pp. 2443–2454, 2012. DOI: 10.1056/NEJMoa1200690.