📖 Core Concepts
Core Concepts: Protein Structure, Dynamics, and Engineering
1. What this subject is really about: proteins as physical systems that can be redesigned
In earlier biochemistry, I learned to associate protein structures with functions: an enzyme has an active site, an antibody has a binding surface, and a receptor changes conformation when it encounters a signal. Advanced biochemistry asks what lies underneath those descriptions. Why does a sequence populate one structure rather than another? Which motions make a binding site competent? How can a mutation far from an active site change catalysis? And how can those principles be used to build a better binder, a more stable enzyme, or a peptide drug?
The central shift is from thinking of a protein as a single molecular object to thinking of it as a population of interconverting molecular states. A deposited structure is an exceptionally useful representation of that population, but it is not the whole protein. Proteins fluctuate, bind solvent and ions, exchange protons, associate with partners, and sometimes partially unfold. Function emerges from this ensemble interacting with its environment. [Creighton, 1992; Branden and Tooze, 1999]
A protein structure is a map of accessible interactions, not a complete movie of molecular behavior. Engineering changes both the map and the traffic moving through it.
This connects directly to statistical mechanics. If conformational state i has free energy Gi, its equilibrium population depends exponentially on that free energy:
pᵢ = exp(−Gᵢ/RT) / Σⱼ exp(−Gⱼ/RT) For two states A and B: pB/pA = exp[−(GB − GA)/RT]
Here, the state free energies include the effects of solvent and the microscopic configurations grouped into each state. At room temperature, a free-energy difference of only a few kilojoules per mole can substantially redistribute populations. A mutation does not need to create an entirely new fold to have a large functional effect; it may simply make a previously rare, binding-competent conformation more common.
Protein engineering is therefore an exercise in controlled redistribution: favoring productive conformations, suppressing aggregation-prone ones, adjusting binding and release, or stabilizing an otherwise fragile molecule. Structure determination, dynamics measurements, binding assays, and evolutionary selection each observe a different part of this problem. The deepest understanding comes from making those observations agree.
2. Sequence, folding, and structural architecture
The peptide bond gives proteins their unusual combination of regularity and diversity. Resonance makes the peptide group approximately planar, so most backbone conformational freedom resides in rotation about the N–Cα bond, described by φ, and the Cα–C′ bond, described by ψ. The peptide-bond torsion ω is usually near 180°, corresponding to the trans configuration. Cis peptide bonds are uncommon, but appreciably more frequent before proline than before other residues.
A Ramachandran plot is therefore not merely a structure-validation graphic. It is a compact expression of molecular geometry: many φ/ψ combinations cause steric clashes, whereas others permit recurring hydrogen-bonding patterns. Glycine has unusually broad conformational freedom because its side chain is hydrogen; proline has unusually restricted φ because its side chain closes onto the backbone nitrogen. These differences help explain why sequence influences turns, helices, and conformational flexibility. [Branden and Tooze, 1999]
In an α-helix, backbone C=O groups hydrogen-bond to N–H groups four residues later. Approximately 3.6 residues complete one turn, so residues separated by three or four positions can lie on a similar face. In β-sheets, hydrogen bonds connect extended strands, and side chains alternate above and below the sheet. These geometries make secondary structure an organizing scaffold for functional chemistry.
Folding is not driven by a single “strongest force.” It reflects a balance among hydrophobic burial, hydrogen bonding, electrostatics, van der Waals packing, conformational entropy, and solvent reorganization. A hydrogen bond inside a protein is not automatically stabilizing relative to the unfolded state: the same donor and acceptor may previously have hydrogen-bonded to water. What matters is the difference between states.
Define folding as U → F: ΔGfold = GF − GU = ΔHfold − TΔSfold Kfold = [F]/[U] ΔGfold = −RT ln Kfold ΔGfold < 0 favors the folded state under the specified conditions.
The folded state loses backbone and side-chain conformational entropy, but burial of nonpolar surfaces can favorably reorganize the surrounding water. Many proteins are only modestly stable overall because large favorable and unfavorable contributions nearly cancel. This explains the apparent paradox that a highly ordered structure can be destabilized by one mutation or a small environmental change. [Creighton, 1992]
Folding landscapes are often pictured as funnels, but the landscape contains roughness: alternative packing arrangements, proline isomerization barriers, mispaired disulfides, and aggregation pathways. Chaperones help proteins navigate these problems; they do not generally provide a separate sequence-specific blueprint for the final structure.
Engineering lesson: stability and activity are related but distinct objectives. A rigidified enzyme may survive heating yet lose activity if the mutation suppresses an essential conformational transition. Conversely, a destabilizing mutation can improve activity while making production or storage much harder.
3. Reading the Protein Data Bank as experimental evidence
The Protein Data Bank is an archive of structural models and associated experimental information, not a catalogue of infallible molecular truths. Before using a structure, I should identify the experimental method, construct, mutations, ligands, unresolved segments, and proposed biological assembly. A crystallized protein may lack a flexible terminus, contain stabilizing mutations, or be trapped with an inhibitor rather than its physiological substrate. [Berman et al., 2000]
The traditional PDB format stores coordinates in fixed-width records. ATOM records typically describe standard polymer atoms; HETATM records commonly describe ligands, ions, waters, and nonstandard components. Each atom has coordinates, an occupancy, and a displacement parameter often called a B factor. The modern PDBx/mmCIF format represents the same general information with fewer legacy size restrictions and richer metadata.
A useful search begins with a biological question rather than just a protein name. I might query by sequence similarity, organism, ligand, experimental method, resolution, or mutation. I then compare multiple structures to ask whether an apparent pocket is conserved, ligand-induced, or dependent on crystal packing.
- Resolution: a measure of the spatial detail supported by the experiment. A smaller resolution value in ångströms generally indicates finer detail, but local interpretability can vary greatly.
- Occupancy: the modeled fraction of sites occupied by an atom or conformation. Partial occupancy can indicate alternative conformations or incomplete ligand binding.
- B factor: a parameter describing attenuation of scattering by positional variation and disorder. It is not a direct, isolated measurement of molecular mobility.
- Ramachandran and rotamer outliers: unusual backbone or side-chain geometries that deserve inspection, not automatic deletion.
- Clashscore and bond geometry: indicators of stereochemical plausibility.
- Local map agreement: essential for judging a particular side chain, ligand orientation, or proposed catalytic contact.
A missing loop usually means that its position was not modeled reliably, not that the biological protein lacks the loop. Similarly, the crystallographic asymmetric unit is not necessarily the biological oligomer. Assembly annotations, solution measurements, interface chemistry, and biological evidence must be considered together.
The disciplined question is: Which conclusions are supported at this level of detail? A structure may establish domain organization convincingly while leaving the orientation of an active-site water uncertain. Assigning a proton-transfer mechanism requires more evidence than merely observing nearby heteroatoms.
4. X-ray crystallography: converting diffraction into a molecular model
A protein crystal is a periodic array of molecules, often containing substantial solvent. Crystallization requires conditions that support ordered assembly rather than amorphous precipitation. Protein purity, conformational homogeneity, concentration, pH, salts, precipitants, and bound ligands all matter. A stabilizing ligand can promote crystallization by reducing the number of conformations that must fit into one lattice.
The unit cell is the repeating geometrical building block of the crystal. X-rays scatter from electrons, and the periodic lattice produces diffraction spots indexed by Miller indices h, k, and l. These spots encode Fourier components of the electron-density distribution. [Branden and Tooze, 1999]
Bragg’s law:
nλ = 2d sin θ
A structure factor:
F(hkl) = Σⱼ fⱼ exp[2πi(hxⱼ + kyⱼ + lzⱼ)]
Measured diffraction intensity:
I(hkl) ∝ |F(hkl)|²
Electron density:
ρ(x,y,z) = (1/V) Σhkl |F(hkl)| exp[iφ(hkl)]
× exp[−2πi(hx + ky + lz)]
The simplified structure-factor expression omits occupancy and displacement terms. Its main lesson is that a diffraction pattern contains contributions from all atoms, added as waves. Detectors measure intensities, allowing estimation of amplitudes, but not the phases φ required to reconstruct electron density. This is the phase problem.
- MIR, multiple isomorphous replacement: compare diffraction from native crystals and heavy-atom derivatives. The derivatives should preserve the underlying lattice closely enough for intensity differences to report the added scatterers.
- MAD, multiwavelength anomalous diffraction: exploit wavelength-dependent scattering near an elemental absorption edge.
- SAD, single-wavelength anomalous diffraction: use anomalous scattering differences measured at one wavelength, together with additional constraints and density modification.
- Molecular replacement: orient and position a related structural model in the unit cell to obtain initial phase estimates.
Molecular replacement connects directly to the growing availability of reliable predicted structures. Nevertheless, a plausible search model can also introduce model bias. Experimental maps and validation remain necessary; prediction does not convert unsupported coordinates into observed atoms.
After initial phasing, model building and refinement alternate. Refinement adjusts coordinates, occupancies, and displacement parameters to improve agreement with the data while maintaining reasonable chemistry. The balance resembles regularized fitting in statistics: experimental observations constrain the model, while stereochemical restraints prevent overfitting poorly determined parameters.
Rwork = Σ ||Fobs| − |Fcalc|| / Σ |Fobs| Rfree uses the same general comparison on a held-out set of reflections excluded from refinement.
A lower R factor is not sufficient proof of correctness. Rfree provides a cross-validation check, and the difference between Rwork and Rfree can reveal overfitting. Both must be interpreted alongside resolution, map quality, geometry, and sample properties. [Brünger, 1992]
Crystallography is particularly powerful for detailed interaction networks, but most conventional structures do not locate hydrogens directly. Protonation states, bond orders in difficult cases, and catalytic water assignments may require complementary spectroscopy, chemistry, or computation.
5. Cryo-electron microscopy: reconstructing structures from noisy views
In single-particle cryo-EM, a purified sample is spread into a thin aqueous film and rapidly vitrified. Vitrification produces amorphous ice rather than a lattice of ice crystals, preserving particles in a hydrated environment. The sample is not chemically “frozen into truth”: interactions with the air–water interface, preferred orientations, ice thickness, and preparation-induced damage can still bias what is observed.
Electrons provide strong scattering but also damage biological material. Consequently, each particle must be imaged with limited dose, producing low-signal images. The solution is statistical: align and combine information from many particles representing different orientations. The projection-slice theorem connects a two-dimensional projection to a central slice through the three-dimensional Fourier transform of the object.
- Collect dose-fractionated image movies.
- Correct specimen motion and estimate the contrast transfer function.
- Identify candidate particle images.
- Use two-dimensional classification to assess particle quality and heterogeneity.
- Estimate orientations and construct initial three-dimensional maps.
- Classify or otherwise analyze conformational and compositional variation.
- Refine maps, assess resolution, and build and validate atomic models.
The “resolution revolution” arose from advances including direct electron detectors, motion correction, and more powerful statistical reconstruction methods. Large complexes and membrane-protein assemblies became accessible at resolutions that could support detailed molecular interpretation. [Kühlbrandt, 2014; Cheng, 2015]
Resolution is commonly assessed using Fourier shell correlation between independently reconstructed half-maps. A threshold such as 0.143 is meaningful only within an appropriate processing and validation workflow. A global resolution value does not mean every region is equally resolved: a rigid core may support side-chain assignments while a mobile peripheral domain remains blurred.
Cryo-EM is especially revealing when a complex occupies several conformations. However, classification is not a direct time-resolved movie. Classes reflect what survives preparation and can be distinguished computationally; their abundances need not equal physiological equilibrium populations. Flexible regions can disappear into an average even though they are essential to function.
Comparison with crystallography: crystallography obtains strong signals through ordered repetition in a lattice; single-particle cryo-EM obtains stronger signals by combining many noisy images. Both infer a three-dimensional model from incomplete, noisy measurements, and both require explicit attention to heterogeneity and overfitting.
6. Solution NMR: identifying atoms, measuring contacts, and observing exchange
Nuclear magnetic resonance uses the magnetic behavior of nuclei such as ¹H, ¹⁵N, and ¹³C. A nucleus’s resonance frequency depends on its electronic environment, giving the chemical shift. Scalar couplings transmit information through bonds, while dipolar interactions and relaxation provide information about spatial relationships and motion. Unlike crystallography and cryo-EM, solution NMR can examine many protein properties without immobilizing the molecule in a lattice or vitrified specimen. [Cavanagh et al., 2007]
Proteins contain many chemically similar atoms, so one-dimensional spectra overlap severely. Uniform ¹⁵N and ¹³C labeling, often achieved through recombinant expression with isotopically enriched nutrients, enables multidimensional experiments that separate signals and connect atoms through known bonding patterns.
A ¹H–¹⁵N HSQC spectrum correlates directly bonded nitrogen and hydrogen nuclei. It usually contains one backbone amide correlation for each observable non-proline residue, with additional signals from some side chains. It is a useful fingerprint, but “one peak per residue” is only an approximation: overlap, exchange broadening, solvent exchange, and missing amide hydrogens complicate the picture.
A well-dispersed HSQC often suggests a folded protein because distinct local environments produce different chemical shifts. Narrow proton dispersion is common in disordered proteins. Neither pattern, by itself, establishes a complete structural interpretation.
Sequential backbone assignment connects spectral peaks to residue identities. HNCA experiments correlate an amide with Cα signals from its own residue and usually the preceding residue. HNCACB adds Cβ information. Matching these carbon patterns across adjacent amides allows a “walk” along the sequence, assisted by complementary experiments and residue-specific chemical-shift patterns.
NOESY experiments detect nuclear Overhauser effects arising from cross-relaxation between nearby spins. Under suitable conditions, the cross-relaxation rate has a strong inverse-sixth-power dependence on separation:
NOE cross-relaxation magnitude ∝ r⁻⁶ (with additional dependence on molecular motion) A typical structural restraint is an allowed distance range, often emphasizing an upper bound, rather than an exact distance.
The strong distance dependence makes NOEs powerful for identifying contacts, but converting intensities into distances requires care. Overlap, spin diffusion, mixing time, exchange, and molecular tumbling can affect intensities. In a dynamic ensemble, averaging related to r⁻⁶ means that a briefly populated short distance can contribute disproportionately.
Programs such as CYANA and CNS use distance restraints, dihedral information, and molecular geometry to calculate structures. Long-range contacts are especially important because they connect sequence-distant regions and establish the fold.
An NMR structural ensemble contains models compatible with the restraints and calculation procedure. Its spread can reflect genuine flexibility, insufficient data, or both. It is not automatically a thermodynamically weighted ensemble. To demonstrate motion, I need dynamics-sensitive measurements rather than simply a visually diffuse bundle of models.
7. Protein dynamics: equilibrium populations and transition rates are different questions
Two proteins can have similar average structures but very different rates of moving between conformations. This is the distinction between thermodynamics, which determines relative populations, and kinetics, which determines how rapidly those populations interconvert.
A ⇌ B kex = kAB + kBA pB/pA = kAB/kBA at equilibrium Transition-state theory: k ≈ (kBT/h) exp(−ΔG‡/RT)
Changing the free-energy difference between A and B changes their equilibrium distribution. Changing the barrier can alter exchange rates without strongly altering that distribution. This connects the energy landscapes of physical chemistry to measurable protein behavior.
Backbone ¹⁵N relaxation measurements—R₁, R₂, and the heteronuclear NOE—are sensitive to molecular tumbling and internal motions. Analysis can estimate an order parameter S², often used to describe the amplitude of fast bond-vector motion. Values near one indicate relatively restricted motion within the model; lower values indicate larger-amplitude motion. These measurements commonly probe picosecond-to-nanosecond behavior, although transverse relaxation can also contain slower exchange contributions. [Cavanagh et al., 2007; Palmer, 2004]
Microsecond-to-millisecond exchange can be studied using CPMG relaxation dispersion and related approaches. Repeated refocusing pulses alter how exchange contributes to transverse relaxation. Fitting the dependence on pulse frequency can reveal exchange rates, populations, and chemical-shift differences under an appropriate model. Exchange spectroscopy can reveal interconversion between separately observable states when the exchange timescale is suitable.
Exchange regimes are defined relative to the resonance-frequency difference Δω:
- Slow exchange: kex ≪ |Δω|; separate peaks may appear.
- Fast exchange: kex ≫ |Δω|; one population-averaged peak appears.
- Intermediate exchange: rates and frequency differences are comparable; severe broadening may occur.
Thus, “fast exchange” is not an absolute molecular speed. It depends on the nuclei observed and, through frequency separation, the magnetic field.
For enzymes, motions may control substrate entry, formation of a catalytic configuration, exclusion of water, or product release. But observing a motion near the turnover timescale does not prove that it drives the chemical reaction. Causal claims require perturbations and a kinetic model that distinguish conformational gating, chemical conversion, and other steps. Motion can be necessary without being the rate-limiting event.
8. Allostery: communication through thermodynamic linkage
Allostery occurs when an event at one site changes the behavior of another site through coupling within a molecular system. The event might be ligand binding, phosphorylation, metal binding, or association with another protein. The response might involve affinity, activity, or conformational populations.
The most useful definition is thermodynamic rather than architectural. A visible mechanical pathway is not required. If ligand B changes the affinity for ligand A, the two binding processes are coupled; at equilibrium, the coupling is reciprocal.
Let KA be the association constant for A binding without B. Let KA|B be the association constant for A binding with B present. ΔGcoupling = −RT ln(KA|B / KA) ΔGcoupling < 0: favorable binding cooperativity. Thermodynamic-cycle requirement: KA × KB|A = KB × KA|B
The MWC model describes oligomers that interconvert between concerted conformational states, often labeled T and R. Subunits within an oligomer share the state in the idealized model, and ligand preferentially stabilizes the state with higher affinity. A key insight is that the ligand need not create the high-affinity state: it can select a state that already exists. [Monod, Wyman, and Changeux, 1965]
The KNF model emphasizes sequential ligand-induced changes, with the state of one subunit influencing its neighbors. Mixed conformations within an oligomer are allowed. These models are conceptual limits, not exhaustive descriptions of all proteins. [Koshland, Némethy, and Filmer, 1966]
Hemoglobin demonstrates coupling among oxygen binding, quaternary structure, proton binding, and interactions with 2,3-bisphosphoglycerate. Aspartate transcarbamoylase illustrates regulation by metabolites and large-scale conformational changes. GPCRs extend the same logic to membrane signaling: different ligands and intracellular partners stabilize different subsets of a receptor’s conformational ensemble.
Allostery can also be dynamic. Two states may have similar average coordinates but different fluctuation patterns or conformational entropies. A mutation can therefore transmit a functional effect without generating a dramatic structural displacement.
The Hill coefficient describes the steepness of a binding response under specified conditions; it is not generally a count of binding sites. Likewise, a sigmoidal activity curve is not by itself a complete demonstration of a particular allosteric mechanism.
9. Protein–ligand interactions: affinity, energetics, kinetics, and assay interpretation
Binding occurs when the interactions and solvent changes associated with complex formation compensate for losses of translational, rotational, and conformational freedom. The ligand and binding pocket both pay costs when they leave water, reorganize, and become constrained. A favorable-looking contact is therefore not an independently additive unit of binding energy.
P + L ⇌ PL Kd = [P][L]/[PL] For simple independent 1:1 binding: fraction occupied, θ = [L]free / (Kd + [L]free) ΔG°bind = RT ln(Kd/C°) = ΔH°bind − TΔS°bind C° = 1 mol L⁻¹
The standard concentration makes the logarithm dimensionless. At approximately 298 K, a tenfold reduction in Kd corresponds to about 5.7 kJ mol⁻¹ more favorable standard binding free energy. This is small enough that subtle reorganizations can transform potency.
Isothermal titration calorimetry measures heat released or absorbed during titration. Fitting an appropriate binding model can provide an association constant, binding enthalpy, and stoichiometry. Binding free energy follows from affinity, and entropy is then inferred from ΔG = ΔH − TΔS. Accurate concentrations and a suitable affinity–concentration window are crucial. Dilution, aggregation, and proton uptake or release can affect the observed heat; buffer-dependent measurements can help diagnose linked protonation.
Surface plasmon resonance monitors changes near a sensor surface as analyte binds to immobilized material. For an ideal 1:1 interaction:
d[PL]/dt = kon[P][L] − koff[PL] Kd = koff/kon Mean bound-state residence time: τ = 1/koff Dissociation half-life: t½ = ln(2)/koff
Identical affinities can conceal very different kinetics. One ligand may bind and leave rapidly, while another binds slowly and remains attached much longer. In SPR, mass transport, rebinding, heterogeneous immobilization, and multivalent attachment can produce apparent rate constants that differ from intrinsic molecular values.
Microscale thermophoresis detects ligand-dependent changes in a molecule’s response to a temperature gradient, commonly using fluorescence. Binding can change hydration, charge, and size, altering the signal. Fluorescent labeling, capillary adsorption, ligand fluorescence, and aggregation require controls; a concentration-dependent optical change is not automatically binding.
Differential scanning fluorimetry follows thermal unfolding using an environment-sensitive dye or intrinsic fluorescence. A ligand that preferentially binds the folded state often raises the apparent melting temperature. Yet the thermal shift depends on ligand concentration, unfolding behavior, and temperature-dependent binding thermodynamics. Some genuine ligands produce little shift or even destabilize the monitored state.
A thermal shift is evidence about a coupled thermal process, not a direct measurement of Kd. Agreement between orthogonal assays is more convincing than an exceptionally clean curve from one assay.
Conformational selection and induced fit provide complementary binding mechanisms. A ligand may bind a pre-existing competent conformation, or an initial complex may rearrange after association. Both pathways can coexist. Equilibrium affinity alone cannot determine their relative contributions.
10. Coiled coils, intrinsic disorder, and amyloid: structure is not an all-or-none property
Coiled coils arise when amphipathic helices pack around one another. A characteristic heptad repeat is labeled a–b–c–d–e–f–g, with hydrophobic residues commonly occupying a and d. Because the repeat corresponds to about 3.5 residues per turn rather than the approximately 3.6 of an isolated α-helix, the helices wrap into a supercoil to maintain favorable packing. [Branden and Tooze, 1999]
Heptad positions: a b c d e f g | a b c d e f g Common tendency: H H | H H H indicates a frequently hydrophobic core position, not an absolute sequence requirement.
Core packing and electrostatic interactions at positions such as e and g help specify partner preference, oligomeric state, and parallel or antiparallel orientation. Engineered coiled coils can act as assembly handles for peptide materials, molecular switches, and modular recruitment systems. Their apparent simplicity is deceptive: competing oligomers and incorrect heptad registers can frustrate design.
Intrinsically disordered proteins and regions lack a single persistent folded structure under the conditions examined. Their sequences often contain relatively few strongly hydrophobic residues and many charged or polar residues. Disorder is not failed folding; it can be a functional solution for flexible linkers, transient recognition motifs, multivalent signaling, and regulation by post-translational modification. [Wright and Dyson, 2015]
NMR can detect residue-specific chemical shifts, local secondary-structure tendencies, and motions in disordered chains. Small-angle X-ray scattering reports ensemble-averaged global dimensions and shape information, including the radius of gyration and a pair-distance distribution. Neither method uniquely specifies every conformation. Many ensembles can reproduce a limited set of averaged observables.
Binding may fold a disordered segment, imposing an entropy cost that favorable interactions must repay. Alternatively, a complex can remain “fuzzy,” retaining multiple arrangements. This explains how disorder can support both specificity and adaptable, rapidly reversible interactions.
Amyloid is a distinct assembled state characterized by cross-β architecture: β-strands run approximately perpendicular to the fibril axis, while β-sheets extend along it. Both disordered proteins and partially unfolded globular proteins can form amyloid. Aggregation depends on sequence, concentration, environment, nucleation, and access to assembly-competent conformations. Amyloid is not simply the inevitable endpoint of disorder, and some amyloids have biological functions. [Knowles, Vendruscolo, and Dobson, 2014]
11. Directed evolution and display: making sequence space experimentally searchable
Rational design asks which mutations should produce a desired physical effect. Directed evolution asks whether a population of variants contains improvements that can be enriched experimentally. The approaches are strongest together: structural reasoning focuses diversity, while selection or screening discovers effects that are difficult to predict.
Diversify → express or display → screen/select → identify → repeat For n fully randomized amino-acid positions: sequence space = 20ⁿ Six positions: 64 million possible sequences. Ten positions: approximately 10¹³ possible sequences.
Even modest sequence space rapidly exceeds practical sampling. Error-prone PCR distributes mutations across a gene but produces a biased spectrum. Saturation mutagenesis concentrates diversity at selected positions. Degenerate codons can introduce redundant codons and stop codons, so library design should consider actual amino-acid coverage rather than nominal DNA diversity.
A screen measures variants individually or in separable units. A selection couples success to enrichment, survival, or recovery. Selection can process enormous populations but is only as good as the link between the selected property and the desired function.
Display technologies create a genotype–phenotype linkage:
- Phage display: a peptide or protein is fused to a phage coat protein, while its encoding DNA is carried within the particle.
- Yeast display: surface-displayed proteins can be analyzed with fluorescent probes and flow cytometry, often allowing expression-normalized binding measurements.
- Ribosome display: a nascent protein, ribosome, and encoding mRNA remain associated, linking a protein phenotype to recoverable genetic information without cellular transformation.
- SELEX: repeated enrichment identifies nucleic-acid aptamers. It follows the logic of molecular evolution but ordinarily selects RNA or DNA ligands, not proteins.
In phage-display panning, a library encounters the target, unbound particles are removed, retained phage are recovered, and their sequences are amplified for further rounds. Target presentation is decisive: immobilization can hide a functional surface or expose an irrelevant one. Negative selection against supports, tags, or related proteins helps remove unwanted binders. [Smith, 1985; Kay, Winter, and McCafferty, 1996]
Enrichment does not necessarily mean improved intrinsic affinity. A clone may display better, propagate faster, bind the support, or benefit from multivalent avidity. Reducing target concentration or adding dissociation challenges can alter selection pressure, but soluble validation remains essential.
Sequence frequencies after selection identify candidates, not a final mechanism. Purified variants should be tested for affinity, kinetics, specificity, folding, aggregation, and performance in the intended environment. Epistasis is common: the effect of one mutation depends on others, so beneficial mutations cannot always be combined additively.
12. Connections to Peptide Chemistry: engineering a conformationally organized interface
A peptide can reproduce the essential recognition surface of a much larger protein, especially when that surface is a short helix, loop, or strand. The difficulty is that excising the segment removes the surrounding scaffold that previously organized it. A peptide may retain all the right side chains while rarely presenting them in the right geometry.
This is why conformational constraint is central to peptide engineering. Preorganization can reduce the conformational entropy lost during binding, but only if the constraint favors the correct state. It can also bury polar groups, protect cleavage sites, or create new contacts. Conversely, a constraint can impose strain, reduce solubility, or stabilize a nonbinding conformation.
All-hydrocarbon stapling commonly installs two α,α-disubstituted amino acids bearing olefinic side chains and links them by ruthenium-catalyzed ring-closing metathesis. Positions i,i+4 or i,i+7 are often chosen to connect residues on a compatible helical face. Side-chain length and stereochemistry determine whether the resulting bridge supports the desired geometry. [Schafmeister, Po, and Verdine, 2000]
Olefin-bearing residue i + olefin-bearing residue j
── ring-closing metathesis ──→
macrocyclic, olefin-containing hydrocarbon bridge
For suitable substrates, metathesis removes a small olefin
such as ethylene while closing the larger ring.
Stapled p53-derived helices illustrate the strategy: the peptide presents key hydrophobic residues to the p53-binding cleft of MDM2, competing with the native interaction. The staple should avoid obstructing those hotspot contacts. Enhanced helicity does not guarantee better affinity, cell entry, or intracellular target engagement. [Bernal et al., 2007]
Bicyclic peptides use two linked rings to organize several recognition loops. In a widely used chemistry associated with the Bicycle platform, three cysteines react with a trifunctional scaffold such as tris(bromomethyl)benzene, producing two peptide loops. Compatible display approaches connect chemical cyclization to sequence selection. This can yield compact, high-affinity ligands, but permeability and pharmacokinetics remain independent design problems. [Heinis et al., 2009]
Other useful constraint strategies include:
- Head-to-tail cyclization: joins the termini, removes terminal charges, and can protect against exopeptidases.
- Disulfide bridges: stabilize folds efficiently but may rearrange or be reduced in some environments.
- Lanthipeptide crosslinks: thioether bridges formed through enzymatically enabled dehydration and cysteine addition provide robust conformational constraints.
- N-methylation: removes a backbone hydrogen-bond donor and changes conformational preferences; it can improve stability or permeability but may also disrupt binding and alter cis/trans equilibria.
Cyclosporine demonstrates that a large cyclic peptide can achieve oral exposure. Its cyclic architecture, extensive N-methylation, and environment-dependent conformations help manage the energetic cost of entering nonpolar environments. This is often described as molecular “chameleonicity”: polar groups can be exposed in water yet partially shielded through intramolecular interactions in less polar surroundings. It is a design principle, not a guarantee transferable to every macrocycle. [White et al., 2011]
Bioinorganic chemistry adds another engineering route. A designed metalloprotein must organize metal ligands while controlling coordination geometry, water access, substrate positioning, and proton transfer. Merely attaching Zn²⁺ to a scaffold does not create an efficient hydrolase.
A simplified metal-assisted hydrolysis scheme: M–OH₂ ⇌ M–OH + H⁺ M–OH + substrate → tetrahedral intermediate intermediate → products + regenerated metal-bound solvent
The metal can lower the pKa of bound water, polarize an electrophile, and stabilize developing charge. Product release and competing metal-binding states can still limit turnover. This connects CHEM502 coordination chemistry to enzyme design: first-shell ligands establish the metal site, while second-shell interactions often determine whether it is catalytically useful.
13. Delivery and drug design: binding is only the beginning
A peptide drug must survive, reach the relevant tissue, cross the necessary barriers, and engage its target at a sufficient free concentration. Improving a purified-protein Kd solves only one part of that chain.
Cell-penetrating peptides include TAT-derived sequences, penetratin, and oligo-arginines. Many are rich in basic residues, promoting interactions with anionic cell-surface components. Uptake mechanisms depend on sequence, cargo, concentration, cell type, and membrane composition. Endocytosis is often important; direct translocation may contribute under particular conditions.
The crucial distinction is between cell association, internalization, and cytosolic delivery. Strong fluorescence can arise from material bound to the outer membrane or trapped in endosomes. A CPP–cargo conjugate may enter cells efficiently but never reach a cytosolic enzyme. Cytosolic access should therefore be assessed with localization controls or functional readouts, not fluorescence intensity alone.
Proteolytic stability can be improved through cyclization, terminal modifications, selected D-amino acids, backbone changes, or shielding of vulnerable bonds. Each modification must preserve the target-binding geometry. A D residue changes side-chain presentation; it is not a universally harmless replacement for its L counterpart.
Immunogenicity is likewise contextual. Sequence, aggregation, impurities, dosing, and antigen processing all matter. Small size and non-natural chemistry do not guarantee immunological invisibility. Reducing unwanted immune responses requires empirical assessment rather than a simple structural rule.
Oral bioavailability requires the joint solution of gastrointestinal degradation, dissolution, epithelial permeability, and presystemic loss. Macrocyclization and N-methylation can help, but excessive hydrophobicity may reduce solubility. Longer systemic persistence can be pursued through albumin binding, lipidation, or other conjugation strategies, with possible costs in tissue distribution and unbound concentration.
Bispecific peptides combine recognition elements for two targets or two sites. Their behavior depends on linker geometry and simultaneous accessibility. Apparent avidity can be powerful but is distinct from the intrinsic affinity of either binding element.
Peptide–antibody conjugates can combine antibody targeting or persistence with a peptide function. Peptide–drug conjugates use a peptide as a targeting or delivery component attached to a payload. Successful designs coordinate target binding, internalization where needed, linker stability during circulation, and payload release in the intended compartment.
Peptide-containing PROTACs bring a target and an E3 ubiquitin ligase into proximity. Productive ternary-complex geometry matters more than simply maximizing two binary affinities. A peptide may contribute target recognition, ligase recruitment, or delivery, but stability and cytosolic access are often major obstacles. Not every peptide–drug conjugate is a PROTAC: degradation requires appropriate ubiquitination and access to the proteasomal machinery.
14. Peptide synthesis at scale: cumulative chemistry determines product quality
Fmoc solid-phase peptide synthesis assembles a peptide while its C terminus is attached to an insoluble resin. Reagents can be used in excess and removed by washing. Chain growth proceeds from the C terminus toward the N terminus, opposite the direction used by the ribosome. [Chan and White, 2000]
- Select a resin and linker appropriate for the desired C-terminal functionality.
- Remove the temporary Fmoc group from the growing chain’s N terminus.
- Activate the incoming Fmoc-protected amino acid’s carboxyl group.
- Form the new amide bond, wash, and assess coupling completeness.
- Repeat the cycle, then perform suitable cleavage, deprotection, and any additional transformations.
- Purify and characterize the product.
Fmoc protection is base-labile, while many side-chain protecting groups and common resin linkers are acid-labile. This orthogonality permits selective chain extension followed by global deprotection. The exact strategy must account for residues and constraints that cannot tolerate the same conditions.
HATU and HBTU are uronium-family coupling reagents that facilitate formation of reactive acyl intermediates. DIC is a carbodiimide; combining it with an additive such as Oxyma can improve coupling behavior and reduce problematic side reactions. These reagents do not make all amino acids react identically. Steric hindrance, sequence-dependent aggregation, and backbone modification can make some couplings much harder than others.
Epimerization at the activated amino acid’s α-carbon can occur through base-promoted processes and intermediates such as oxazolones. Risk depends on the residue, activation chemistry, base, temperature, and time. Avoiding unnecessary preactivation and using validated conditions matter more than assuming that one reagent eliminates racemization. Side reactions such as aspartimide formation and diketopiperazine formation also require sequence-specific attention.
If each of m coupling steps succeeds with probability q: Ideal full-length fraction ≈ qᵐ For 50 couplings: 0.990⁵⁰ ≈ 0.61 0.995⁵⁰ ≈ 0.78 0.999⁵⁰ ≈ 0.95
This idealized calculation excludes many side reactions, but it explains why tiny improvements per cycle matter enormously. At scale, resin swelling, mixing, heat transfer, solvent consumption, and wash efficiency become chemical variables, not merely engineering details.
Final cleavage commonly uses an acidic mixture with scavengers chosen to intercept reactive species and protect susceptible residues. Process safety is essential: coupling reagents, bases, cleavage mixtures, and their byproducts need specific assessment. For example, some DIC/Oxyma conditions can generate hydrogen cyanide; a coupling system should not be considered harmless because it performs well analytically.
Preparative reverse-phase HPLC separates the target from deletion sequences, incompletely deprotected products, and other impurities. Mass spectrometry confirms molecular mass, but correct mass alone does not establish stereochemistry, disulfide connectivity, or sequence integrity. Epimers can have identical masses. Orthogonal characterization and a defined purity assay are therefore necessary before interpreting biological activity.
15. Common misconceptions to watch out for
- “The structure is the protein.” A structure represents selected aspects of an ensemble under particular conditions. Missing dynamics can be decisive for function.
- “Higher resolution makes every atom trustworthy.” Local map support, occupancy, geometry, and model bias still matter.
- “A high B factor proves flexibility.” B factors also incorporate disorder and modeling effects; dynamics require independent evidence.
- “A broad NMR ensemble proves physical motion.” Structural spread may arise from sparse restraints rather than measured dynamics.
- “Allostery requires a large conformational change.” Coupling can arise from shifts in populations or entropy with little change in average coordinates.
- “Lower Kd means slower dissociation.” Affinity is the ratio koff/kon; either rate can change.
- “More hydrogen bonds always improve binding.” Desolvation and reorganization costs can cancel their apparent benefit.
- “A thermal shift measures affinity directly.” It measures a ligand-dependent thermal process whose interpretation requires a model.
- “A selected clone must be the best binder.” Display, amplification, avidity, and nonspecific interactions can bias enrichment.
- “Cyclization or stapling automatically improves a peptide.” A constraint can stabilize the wrong conformation or create delivery and solubility problems.
- “Cell uptake means access to the target.” Endosomal trapping can separate impressive uptake from negligible cytosolic activity.
- “Correct mass proves a pure, correct peptide.” Isomers, epimers, and alternative crosslink connectivities may escape that test.
16. A working framework for genuine understanding
For any protein or engineered peptide, I should now be able to ask five connected questions:
- What states exist? Identify folded, disordered, oligomeric, ligand-bound, and potentially aggregated states.
- What sets their populations? Consider free energies, protonation, solvent, ligands, metals, and concentration.
- How quickly do they interconvert? Separate equilibrium occupancy from kinetic accessibility and residence time.
- What does each experiment actually observe? Distinguish coordinates, averaged distances, heats, surface signals, thermal responses, and biological outcomes.
- Which intervention improves the whole system? Evaluate stability, activity, specificity, production, delivery, and safety together.
The beautiful unifying idea is that a protein’s behavior is not written in sequence as a rigid instruction list. Sequence shapes an energy landscape; the environment determines how that landscape is explored; and experiments reveal different projections of the exploration. Engineering becomes more reliable when all three levels—sequence, landscape, and measurement—are kept in view.
Sources and citation notes
The core structural, thermodynamic, and NMR explanations follow [Branden and Tooze, 1999], Introduction to Protein Structure, 2nd edition; [Creighton, 1992], Proteins: Structures and Molecular Properties, 2nd edition; and [Cavanagh et al., 2007], Protein NMR Spectroscopy: Principles and Practice, 2nd edition.
- [Berman et al., 2000]. “The Protein Data Bank.” Nucleic Acids Research 28, 235–242.
- [Brünger, 1992]. “Free R value: a novel statistical quantity for assessing the accuracy of crystal structures.” Nature 355, 472–475.
- [Kühlbrandt, 2014]. “The resolution revolution.” Science 343, 1443–1444.
- [Cheng, 2015]. “Single-particle cryo-EM at crystallographic resolution.” Cell 161, 450–457.
- [Palmer, 2004]. “NMR characterization of the dynamics of biomacromolecules.” Chemical Reviews 104, 3623–3640.
- [Monod, Wyman, and Changeux, 1965]. “On the nature of allosteric transitions: a plausible model.” Journal of Molecular Biology 12, 88–118.
- [Koshland, Némethy, and Filmer, 1966]. “Comparison of experimental binding data and theoretical models in proteins containing subunits.” Biochemistry 5, 365–385.
- [Wright and Dyson, 2015]. “Intrinsically disordered proteins in cellular signalling and regulation.” Nature Reviews Molecular Cell Biology 16, 18–29.
- [Knowles, Vendruscolo, and Dobson, 2014]. “The amyloid state and its association with protein misfolding diseases.” Nature Reviews Molecular Cell Biology 15, 384–396.
- [Smith, 1985]. “Filamentous fusion phage: novel expression vectors that display cloned antigens on the virion surface.” Science 228, 1315–1317.
- [Kay, Winter, and McCafferty, 1996], editors. Phage Display of Peptides and Proteins: A Laboratory Manual. Academic Press.
- [Schafmeister, Po, and Verdine, 2000]. “An all-hydrocarbon cross-linking system for enhancing the helicity and metabolic stability of peptides.” Journal of the American Chemical Society 122, 5891–5892.
- [Bernal et al., 2007]. “Reactivation of the p53 tumor suppressor pathway by a stapled p53 peptide.” Journal of the American Chemical Society 129, 2456–2457.
- [Heinis et al., 2009]. “Phage-encoded combinatorial chemical libraries based on bicyclic peptides.” Nature Chemical Biology 5, 502–507.
- [White et al., 2011]. “On-resin N-methylation of cyclic peptides for discovery of orally bioavailable scaffolds.” Nature Chemical Biology 7, 810–817.
- [Chan and White, 2000], editors. Fmoc Solid Phase Peptide Synthesis: A Practical Approach. Oxford University Press.
Bibliographic caution: The supplied course-paper entries with missing authors, including the stated STRA6–arrestin cryo-EM entry, should have their title–DOI correspondence verified before formal citation. The explanations above do not depend on treating those unverified pairings as established references.
⚙️ Mechanisms & Applications
Mechanisms & Applications: From Protein Ensembles to Engineered Medicines
Skipper’s organizing principle: a protein is not a rigid object with one function attached to one shape. It is a chemically connected population of conformations, and its behavior depends on how those conformations interconvert, bind partners, and react. Engineering a protein therefore means reshaping an energy landscape—not merely replacing a side chain in a picture.
This builds directly on statistical mechanics, chemical kinetics, molecular biology, and organic synthesis. The central experimental challenge is to distinguish three questions: Which structures exist? How populated are they? How rapidly do they exchange? Different techniques answer different combinations of these questions. [1–3]
1. Reading a Structure as Experimental Evidence
A structure in the Protein Data Bank is an interpreted model supported by measurements, not a molecular photograph. Before using it to design a mutation or peptide inhibitor, I should reconstruct the chain of evidence behind it.
- Find the biologically relevant construct. Check sequence coverage, mutations, tags, missing domains, cofactors, ligands, and experimental conditions. A crystallized catalytic domain may not behave like the intact protein.
- Identify the assembly. The crystallographic asymmetric unit is not necessarily the functional oligomer. Examine the proposed biological assembly and independent evidence for its stoichiometry.
- Read the coordinate model. Legacy PDB files contain atom names, residue identifiers, coordinates, occupancies, and displacement parameters. PDBx/mmCIF is the modern archive standard and accommodates information and structures beyond legacy PDB limitations.
- Inspect global and local quality. Resolution is a dataset-level descriptor; it does not guarantee that every loop, ligand, or side chain is equally well determined.
- Interrogate the exact design site. Is the proposed hydrogen bond supported by density? Is the side chain in an alternate conformation? Is a nearby residue missing? Could crystal packing stabilize the observed arrangement?
Three quality metrics answer different questions:
- Resolution: how fine the spatial information extends. A smaller value in ångströms generally means finer detail, but completeness, anisotropy, local order, and processing also matter.
- R-factor: how well calculated diffraction amplitudes agree with measured amplitudes.
- Ramachandran validation: whether backbone φ and ψ angles occupy plausible regions. Glycine, proline, and residues preceding proline require appropriate reference distributions. An outlier can be genuine, but it needs unusually strong evidence.
Rwork = Σh ||Fobs(h)| − |Fcalc(h)|| / Σh |Fobs(h)| Rfree uses reflections withheld from refinement. For an idealized isotropic atomic displacement: B = 8π²Uiso
A large gap between Rwork and Rfree can indicate overfitting. High B-factors suggest positional disorder or uncertainty, but are not direct measurements of a particular motional timescale. Missing coordinates mean insufficiently supported modeling—not absence of that sequence from the sample. These distinctions are essential when converting structural information into medicinal-chemistry decisions. [1,4,5]
2. X-Ray Crystallography: Turning Diffraction into a Molecular Model
The conceptual bridge from optics is Fourier analysis: a crystal converts a repeating electron-density distribution into a pattern of scattered waves. The detector measures intensities, but reconstructing electron density requires both amplitudes and phases.
- Prepare a sufficiently homogeneous protein. Purity alone is insufficient. Mixed oligomers, partial proteolysis, heterogeneous modifications, or flexible domains can frustrate crystallization.
- Grow an ordered crystal. Change precipitant concentration, pH, temperature, or other conditions so that the protein becomes supersaturated and forms an ordered lattice rather than an amorphous precipitate.
- Collect diffraction images. Rotate the crystal in an X-ray beam. Indexing identifies the unit cell and reflection indices; integration and scaling estimate reflection intensities.
- Solve the phase problem. Obtain phase information experimentally or from a related model.
- Build and refine. Fit a chemically plausible model to the density, refining coordinates, occupancies, and displacement parameters while using geometric restraints.
- Validate independently. Examine Rfree, geometry, difference density, ligand support, and biological plausibility.
I(hkl) ∝ |F(hkl)|²
ρ(x,y,z) = (1/V) Σhkl |F(hkl)| exp[iφhkl]
× exp[−2πi(hx + ky + lz)]
The experiment provides amplitudes through intensities.
The phases φhkl must be recovered by additional information.
How the main phasing strategies work:
- MIR—multiple isomorphous replacement: compare native crystals with heavy-atom derivatives. The derivatives must preserve sufficiently similar crystal packing for intensity differences to reveal useful phase constraints.
- SAD—single-wavelength anomalous diffraction: use anomalous scattering, often from incorporated selenium or suitable native atoms, to obtain phase information from differences between Friedel-related reflections.
- MAD—multiwavelength anomalous diffraction: measure at several wavelengths near an absorption edge, exploiting wavelength-dependent scattering.
- Molecular replacement: orient and position a related structure in the unit cell, then use its phases as an initial estimate. A wrong or excessively trusted starting model can introduce model bias.
Application: fragment-based drug discovery uses crystallography to locate small molecules in binding pockets and identify productive growth directions. Nevertheless, a visible ligand is not automatically a strong binder, and a geometrically attractive contact does not reveal its net free-energy contribution: desolvation and conformational reorganization still have to be paid for.
Historical connection: Kendrew and colleagues’ myoglobin work in 1958 and Perutz and colleagues’ hemoglobin work around 1960 established the first three-dimensional protein structures. Heavy-atom phasing helped turn diffraction patterns into interpretable molecular architecture. What mattered was not simply seeing a fold: sequence, stereochemistry, and biological function could finally be connected in three dimensions. [1,6]
3. Cryo-EM: Reconstructing Complexes Without Crystals
Single-particle cryo-electron microscopy replaces the repeating crystal lattice with many images of individual particles in different orientations. The price of avoiding crystallization is an exceptionally difficult signal-recovery problem.
- Prepare a thin aqueous specimen. Apply purified particles to a grid and rapidly freeze it so that water vitrifies rather than forming damaging ice crystals.
- Acquire low-dose movies. Electrons provide contrast but also damage the specimen. Recording frames enables correction for beam-induced motion and appropriate dose weighting.
- Estimate the contrast transfer function. Defocus and microscope aberrations alter how spatial frequencies appear in the image; these effects must be accounted for.
- Pick particles and classify two-dimensional images. Reject contaminants and poorly resolved particles while recognizing that selection choices can bias the retained population.
- Estimate orientations and reconstruct a volume. Iteratively compare projections of a three-dimensional model with experimental images.
- Classify or model heterogeneity. Separate distinct assemblies or conformations where the data support doing so.
- Refine and validate. Compare independently refined half-maps and inspect local resolution and model-to-map agreement.
Fourier slice theorem: The 2D Fourier transform of an ideal projection corresponds to a central slice through the object's 3D Fourier transform. Common resolution assessment: FSC between independent half-maps, often using the 0.143 criterion.
The Fourier shell correlation threshold is a convention, not a certificate of atomic certainty. Masking, correlated noise, orientation bias, and local flexibility must still be considered. Likewise, class proportions are not automatically equilibrium populations: adsorption to the air–water interface, freezing, and particle-selection procedures can alter representation.
Applications: membrane receptors, ribosomes, channels, viral proteins, and large signaling complexes. Ligand-bound structures can reveal pocket rearrangements or interfaces that are difficult to crystallize.
Historical connection: vitrification methods associated with Jacques Dubochet, image-processing developments associated with Joachim Frank, and Richard Henderson’s electron-crystallography work laid foundations for the field. Direct electron detectors, improved motion correction, and better reconstruction algorithms drove the later “resolution revolution.” The beautiful lesson is that better measurements and better inference advanced together. [7]
4. Solution NMR: Assigning Atoms, Then Measuring Their Neighborhoods
NMR is especially valuable because it measures proteins in solution and can report both structure and motion. Its first obstacle is identity: before interpreting a spectral peak, I need to know which nucleus produced it.
- Introduce isotopes. Express the protein with 15N and, for many experiments, 13C enrichment. Deuteration can help with larger systems.
- Record a 1H–15N HSQC. Under suitable conditions, most nonproline backbone amides give a cross-peak. Overlap, solvent exchange, or conformational exchange can obscure peaks; some side-chain groups also contribute signals.
- Make sequential assignments. HNCA correlates an amide with Cα nuclei from its own residue and the preceding residue. HNCACB adds Cβ information. Complementary experiments help distinguish intraresidue from preceding-residue signals.
- Walk along the chain. Match carbon chemical-shift patterns between neighboring residues, using the known sequence to resolve ambiguities.
- Collect structural restraints. NOESY provides information about spatial proximity; chemical shifts and additional measurements constrain torsion angles and other structural features.
- Calculate restrained structures. Programs such as CYANA and CNS search for conformations consistent with experimental restraints and stereochemistry.
For a suitable isolated spin pair and motional regime: NOE cross-relaxation magnitude ∝ r⁻⁶ Consequently: Short distances contribute disproportionately. Structure calculation seeks low restraint violations and plausible geometry—not merely a visually compact fold.
The r−6 dependence makes NOEs powerful but easy to overinterpret. Spin diffusion, molecular motion, overlap, and assignment ambiguity complicate distance inference. Also, the spread of a conventional NMR structure ensemble often reflects incomplete restraints and modeling uncertainty; it is not automatically a quantitative movie of protein dynamics.
Applications: mapping ligand-binding surfaces by chemical-shift perturbation, characterizing disordered regions, testing whether mutations preserve folding, and studying weak or transient complexes. A perturbation can result from direct contact or a propagated conformational change, so it does not by itself prove a contact.
Historical connection: Kurt Wüthrich and collaborators developed the assignment and restraint-based strategies that made solution protein structure determination practical. The achievement was not just multidimensional spectroscopy, but a systematic route from crowded spectra to residue-specific structural information. [2]
5. Dynamics: Populations and Rates Are Different Quantities
From physical chemistry, I already know that free-energy differences determine equilibrium populations, while activation barriers determine rates. Advanced biochemistry makes this distinction experimentally visible.
A ⇌ B
kex = kAB + kBA
pB/pA = kAB/kBA
= exp[−(GB − GA)/(RT)]
An unchanged population ratio does not require unchanged exchange rates.
For example, a state lying 2.7 kcal mol−1 above the dominant state has approximately 1% relative population at 298 K. Such a minor state can still matter greatly if it binds a ligand or enters a catalytic pathway.
- Picosecond–nanosecond motion: 15N R1, R2, and heteronuclear NOE measurements report spectral densities of bond-vector motion. Under appropriate models, an order parameter S2 describes the restriction of internal motion; values closer to 1 indicate more restricted motion.
- Microsecond–millisecond exchange: CPMG relaxation dispersion measures how effective transverse relaxation changes with refocusing-pulse frequency. Exchange contributions can yield populations, exchange rates, and chemical-shift differences within an identifiable model.
- Resolved, slower exchange: exchange spectroscopy can connect distinct peaks belonging to interconverting states, provided the exchange rate falls within the experiment’s accessible window.
CPMG relaxation dispersion: R2,eff(νCPMG) = R2,0 + Rex(νCPMG) In a two-state fast-exchange limit: Rex ≈ pA pB (Δω)² / kex Here Δω is in angular-frequency units. This approximation is not valid across all exchange regimes.
Connection to catalysis: loop closure may exclude water, align catalytic groups, or govern substrate entry and product release. However, observing motion on a timescale similar to turnover does not prove that motion drives the chemical step. Stronger evidence requires perturbations, kinetic dissection, and a model that distinguishes binding, chemistry, and product release. [2,8]
Engineering application: a stabilizing mutation can improve manufacturability while suppressing a useful minor conformation. “More rigid” and “better” are therefore not synonyms.
6. Allostery: Binding at One Site Changes the Statistical Choices Elsewhere
Allostery is thermodynamic coupling between spatially distinct sites or processes. A useful analogy is an interconnected voting system: a ligand changes the statistical weight of some conformations, changing the likelihood that another site adopts a binding-competent state. There need not be a visible lever moving across the entire protein.
- The unliganded protein samples several conformations.
- A ligand binds those conformations with different affinities.
- Binding changes their relative populations and may also permit additional rearrangements.
- A second site experiences altered affinity, kinetics, or catalytic competence.
MWC versus KNF: the Monod–Wyman–Changeux model treats an oligomer as undergoing concerted transitions among states with different ligand affinities. The Koshland–Némethy–Filmer model allows sequential, ligand-associated changes in subunits that influence their neighbors. Real proteins may combine features of both; neither model is a universal anatomical description. [9,10]
Let ligand B alter the dissociation constant for ligand A.
ΔGcoupling = ΔGbind,A|B − ΔGbind,A
= RT ln[Kd,A|B / Kd,A]
If B improves A binding:
Kd,A|B < Kd,A, so ΔGcoupling < 0.
At equilibrium:
Kd,A|B / Kd,A = Kd,B|A / Kd,B
The last relationship is thermodynamic reciprocity. If B changes A’s equilibrium affinity, A must change B’s affinity consistently around the same thermodynamic cycle. This is a direct application of the path independence of free energy.
Applications: hemoglobin links oxygen binding to oligomeric equilibria and proton/CO2 effects; aspartate transcarbamoylase links metabolic effectors to catalytic regulation; GPCR ligands and intracellular transducers stabilize coupled receptor states. Allosteric drugs can exploit less-conserved sites and produce effects that depend on endogenous ligand concentration.
Caution: the Hill coefficient is an empirical measure of response steepness under a specified model. It is not generally equal to the number of binding sites.
7. Measuring Binding: Affinity, Enthalpy, Residence Time, and Stability
A good binding campaign uses orthogonal methods because each technique observes a different physical consequence of interaction.
For a simple 1:1 interaction:
P + L ⇌ PL
Kd = [P][L]/[PL] = koff/kon
Fraction occupied:
θ = [L]free / (Kd + [L]free)
Standard binding free energy:
ΔG°bind = RT ln(Kd/c°)
= ΔH°bind − TΔS°bind
c° = 1 mol L⁻¹; the logarithm must have a dimensionless argument.
At 298 K, improving Kd from 1 μM to 10 nM makes binding free energy about 2.7 kcal mol−1 more favorable. A dramatic hundredfold improvement can therefore arise from a modest net energetic change. The difficulty is that a new interaction often requires desolvation or structural reorganization that offsets its apparent benefit.
If ligand is not in large excess, total ligand cannot replace free ligand in the occupancy equation:
[PL] = {PT + LT + Kd
− √[(PT + LT + Kd)² − 4PTLT]}/2
Use mass balance for tight-binding experiments.
Isothermal titration calorimetry—ITC:
- Inject ligand into a temperature-controlled protein solution.
- Measure the heat required to maintain thermal balance.
- Correct for dilution and other background heats.
- Fit an appropriate binding model to obtain affinity, binding enthalpy, and stoichiometry.
- Calculate entropy from ΔG° = ΔH° − TΔS°.
ITC measures the enthalpy of the entire coupled process. If binding takes up or releases protons, the buffer’s ionization enthalpy contributes. Measurements in different buffers can help separate this contribution. Very tight binding may require competition experiments because an almost vertical titration transition contains little affinity information. [11]
Surface plasmon resonance—SPR: immobilize one partner and flow the other over it. Changes near the sensor surface generate a response that can be followed during association and dissociation.
For an ideal 1:1 surface-binding model: dR/dt = kon C(Rmax − R) − koff R During dissociation, if rebinding is negligible: R(t) = R(0) exp(−koff t) Mean bound lifetime: τ = 1/koff
Two ligands can have the same Kd but very different residence times. For Kd = 10 nM, a ligand with kon = 106 M−1s−1 has koff = 0.01 s−1, giving τ = 100 s. If both rate constants are tenfold smaller, affinity is unchanged but τ becomes 1,000 s. In practice, mass transport, immobilization heterogeneity, avidity, and rebinding can distort these fitted rates. [12]
- MST: microscale thermophoresis detects changes in fluorescence behavior and molecular redistribution in a temperature gradient. Binding can alter effective size, charge, and hydration. Aggregation, adsorption, and label effects require controls.
- DSF: differential scanning fluorimetry follows unfolding-associated fluorescence during heating. A ligand may shift the apparent melting temperature, but ΔTm is not a universal proxy for Kd. Irreversible unfolding, dye interference, and binding to multiple states complicate interpretation.
Industrial decision rule: confirm identity and purity, demonstrate binding by more than one physical principle, establish functional activity, and then test whether adequate free compound reaches the relevant cellular compartment.
8. Coiled Coils, Disorder, and Amyloid: Three Different Uses of Sequence
Coiled coils: α-helical sequences often display a heptad repeat, labeled a–g, with hydrophobic residues frequently at a and d. Because this pattern repeats every seven residues rather than perfectly matching an isolated α-helix’s periodicity, the helices wrap around one another to maintain favorable packing.
Heptad: a b c d e f g | a b c d e f g a,d: commonly contribute to the hydrophobic interface e,g: often contribute electrostatic specificity Hydrophobic pattern → interhelical packing → supercoiling
Oligomer number, parallel versus antiparallel orientation, and partner specificity depend on the whole interface—not just a/d hydrophobicity. Applications include designed assembly scaffolds, oligomerization modules, and fusion-inhibitory peptides. Enfuvirtide exploits the helical assembly pathway of HIV gp41 to interfere with productive membrane fusion. [1,13]
Intrinsically disordered proteins: disorder is not synonymous with chemical instability or failed folding. A disordered region can expose short interaction motifs, host many regulatory modifications, or interact with multiple partners. Narrow amide chemical-shift dispersion often supports disorder, while SAXS reports global dimensions and ensemble-averaged shape information. Neither observation alone determines a unique ensemble.
Some disordered regions fold upon binding; others remain dynamically heterogeneous in “fuzzy” complexes. Binding-associated ordering carries an entropy cost, but can generate specificity through many favorable contacts. NMR and SAXS are particularly useful because forcing an IDP into one representative structure can erase the very property responsible for its function. [2,14]
Amyloid formation: soluble chains can nucleate cross-β assemblies, then grow through monomer addition; fragmentation and secondary nucleation can accelerate accumulation. Aggregation is a kinetic network, not simply “protein becomes insoluble.” Toxicity may reside in particular soluble assemblies rather than the largest deposits, depending on the system. Not all IDPs form amyloid, and folded proteins can also become amyloidogenic after partial unfolding. [15]
9. Directed Evolution: Linking Molecular Performance to Recoverable Information
Rational design asks which sequence ought to work. Directed evolution asks which sequences actually work under a chosen test. The power comes from repeated variation and selection—not from randomness alone.
- Define a measurable objective. Activity, affinity, stability, selectivity, expression, or a combination must be translated into an assay.
- Create diversity. Error-prone PCR distributes mutations; site-saturation mutagenesis explores selected positions; recombination combines existing changes.
- Maintain genotype–phenotype linkage. Keep each molecular phenotype associated with its encoding sequence.
- Select or screen. Selection enriches survivors; screening measures individual variants.
- Sequence and reconstruct hits. Retest isolated variants to exclude artifacts and accidental enrichment.
- Iterate. Increase the challenge or alter the objective as performance improves.
Sequence space for n fully variable amino-acid positions: N = 20ⁿ For n = 8: N = 25.6 billion sequences For one particular variant present at frequency p: P(sample it at least once in M draws) = 1 − (1 − p)ᴹ This is not the probability of covering the entire library.
Library size is therefore not library coverage. Transformation efficiency, codon design, growth bias, and display efficiency all reduce effective exploration. “Saturating” many positions simultaneously can create a nominal library that is physically impossible to sample adequately.
Display mechanisms:
- Phage display: a peptide or protein is fused to a phage coat protein; the phage carries its encoding DNA.
- Yeast display: variants appear on yeast surfaces, permitting fluorescence-based sorting for binding and expression.
- Ribosome display: an in vitro stalled ribosome links a nascent protein to its mRNA, allowing very large libraries without transformation at the display step.
- SELEX: repeated selection and amplification of nucleic-acid ligands. Its products are aptamers; conventional SELEX is not protein display.
Phage panning, step by step: expose the library to a target, remove nonbinders, recover retained phage, amplify them in host cells, and repeat. Add counterselections against the support and related proteins. Adjust washing or competition to alter selection pressure. Finally, test soluble monovalent candidates: strong phage retention can arise from avidity, nonspecific adhesion, or display advantages rather than excellent intrinsic affinity. [16,17]
Historical connection: George Smith introduced filamentous-phage display in 1985. Greg Winter helped establish antibody display and selection as a route to therapeutic binders. Frances Arnold’s work established directed evolution as a practical means of changing enzyme properties. The 2018 chemistry Nobel Prize recognized these complementary ways of making evolution an engineering tool. [16–18]
10. Engineering Peptide Shape: Staples, Rings, and Bicycles
The thermodynamic motivation for constraining a peptide is preorganization: if the unbound peptide already samples its binding-competent conformation frequently, binding may require less conformational ordering. But the constraint must stabilize the right geometry.
All-hydrocarbon α-helical stapling:
- Identify a helix-mediated interaction and its essential contact residues.
- Choose appropriately oriented positions, commonly i/i+4 or i/i+7, that can accommodate a constraint without disrupting key contacts.
- Install stereochemically appropriate olefin-bearing noncanonical amino acids during synthesis.
- Use ruthenium-catalyzed ring-closing metathesis to connect their side chains.
- Measure helicity, affinity, proteolytic stability, membrane effects, and intracellular target engagement independently.
Two appropriately positioned alkene-bearing side chains --Ru-catalyzed ring-closing metathesis--> Macrocyclic hydrocarbon constraint Often: preorganization ↑ and protease susceptibility ↓ Not guaranteed: affinity ↑, cytosolic delivery ↑, selectivity ↑
For p53-mimetic inhibitors of MDM2, the design must preserve the presentation of key hydrophobic p53 residues while placing the staple where it does not block recognition. Increased helicity can improve binding, but increased hydrophobicity can also cause aggregation, membrane disruption, or serum binding. Total cellular fluorescence is not sufficient evidence of delivery to the cytosol. [19,20]
Other constraint strategies:
- Head-to-tail cyclization: removes free termini and can reduce exopeptidase susceptibility, but requires control of ring size and competing intermolecular reactions.
- Disulfide bridges: stabilize compact folds but may be reduced or reshuffled in unsuitable environments.
- Lanthionine bridges: in many lanthipeptide biosynthetic pathways, dehydration of serine or threonine is followed by cysteine addition to form thioether crosslinks.
- N-methylation: removes selected backbone NH donors and alters conformational preferences. It is not itself a cyclization reaction and can complicate both synthesis and receptor recognition.
- Bicyclic peptides: a trifunctional scaffold can react with three cysteines to generate two constrained loops. Coupling such chemistry to display enables selection of compact, chemically constrained binders.
Historical connection: Schafmeister, Po, and Verdine’s 2000 work established an influential all-hydrocarbon helix-stapling strategy. Heinis and colleagues’ 2009 work demonstrated selection of chemically bicyclized peptides on phage, helping establish the approach later developed commercially by Bicycle Therapeutics. [19,21]
11. Delivery and Drug Design: Binding Is Only One Barrier
A useful mental model is a sequence of gates: survive formulation and dosing, resist degradation, reach the tissue, cross the relevant barrier, escape sequestration, bind the target, and produce the desired response. Excellent performance at the last gate cannot rescue failure at the first six.
Cell-penetrating peptides: TAT-derived sequences, penetratin, and oligo-arginine can enhance association with cells and uptake. Their positive charge promotes interactions with anionic cell-surface components. Depending on sequence, cargo, concentration, and cell type, uptake may involve endocytosis, direct translocation, or mixtures of mechanisms.
Extracellular peptide–cargo → surface association → internalization → often endosomal sequestration → endosomal escape, if successful → cytosolic target engagement Cell-associated signal ≠ cytosolic concentration.
Convincing delivery assays distinguish surface-bound material, endosomal accumulation, cytosolic access, and membrane damage. Endosomal escape is frequently the limiting step. [22]
Cyclosporine illustrates conformational adaptability: its cyclic, extensively N-methylated backbone can alter solvent exposure and intramolecular hydrogen bonding in different environments. This “chameleonic” behavior helps reconcile aqueous compatibility with membrane passage. Yet oral absorption also depends on dissolution, intestinal metabolism, efflux, and formulation; a cyclic structure alone is no guarantee. [23]
- Proteolytic stability: cyclization, terminal modification, selected D-residues, and backbone modifications can help, but each may change affinity or clearance.
- Immunogenicity: assess sequence-derived epitopes, aggregation, impurities, formulation, route, and repeated exposure. Small size does not guarantee immunological invisibility.
- Peptide-drug conjugates: couple a targeting peptide to a payload through a linker whose stability and cleavage chemistry match the intended delivery route.
- Peptide–antibody conjugates: combine peptide function with antibody targeting or pharmacokinetics, while controlling conjugation-site heterogeneity and retaining both components’ activity.
- Bispecific peptides: recognize two targets or two sites. Their benefit depends on geometry and simultaneous accessibility; multivalent avidity must not be confused with improved monovalent affinity.
Peptide-containing PROTACs: a bifunctional molecule can connect a target protein to an E3 ubiquitin ligase. Productive ternary-complex formation enables ubiquitination and subsequent proteasomal degradation. Early PROTAC work used a phosphopeptide element, making peptide recognition central to the concept’s history. [24]
Target + degrader + E3 ligase ⇌ productive ternary complex → target ubiquitination → proteasomal degradation Potential degrader recycling permits substoichiometric action, but not every bound ternary complex is productive.
Very high degrader concentrations can favor separate binary complexes and reduce productive ternary-complex formation—the “hook effect” in systems where it occurs. Linker length, cooperativity, lysine accessibility, permeability, and cellular E3 expression all matter.
12. Peptide Synthesis at Scale: Managing Repeated Imperfection
Merrifield’s solid-phase peptide synthesis, introduced in 1963, replaced purification after every coupling with filtration and washing. The peptide remains attached to an insoluble support while soluble reagents and by-products are removed. This changed peptide synthesis from a sequence of difficult isolations into an automatable cycle. [25]
- Choose a resin and linker. These determine loading, solvent behavior, and whether cleavage yields an acid, amide, or another terminal functionality.
- Remove the N-terminal Fmoc group. Base-mediated deprotection exposes the next amine while suitable side-chain protecting groups remain intact.
- Activate and couple the next amino acid. HATU/HBTU or DIC-based systems generate activated carboxyl derivatives that undergo nucleophilic acyl substitution by the resin-bound amine.
- Monitor, wash, and repeat. Incomplete reactions can require recoupling; capping unreacted amines can limit propagation of deletion sequences.
- Introduce constraints or conjugates at a suitable stage. Some staples and cyclizations are convenient on resin; others require solution conditions or orthogonal protection.
- Cleave and deprotect. Common acid-labile systems use TFA-based mixtures with scavengers chosen for the sequence and protecting groups.
- Purify and characterize. Preparative HPLC, analytical chromatography, mass spectrometry, and additional stereochemical or structural analyses establish identity and impurity profiles.
A simplified cumulative-yield model: Fraction of full-length product ≈ pⁿ For 50 coupling steps: 0.99⁵⁰ ≈ 0.605 0.995⁵⁰ ≈ 0.778 Small improvements per cycle become large improvements overall.
This calculation assumes independent coupling failures and ignores deprotection failures, cleavage losses, and sequence-dependent chemistry. It explains why process chemistry emphasizes repeatability rather than merely finding one successful coupling condition.
Racemization mechanism: activation can permit oxazolone formation or other pathways that enable loss of stereochemical integrity at an activated residue. Reagent choice, base, temperature, preactivation time, and residue identity influence the risk. Additives such as Oxyma can improve coupling performance and suppress racemization, but process safety must also be evaluated; DIC/Oxyma combinations can generate hydrogen cyanide under some conditions. [26]
Scale-up complications: resin-bound aggregation, poor mixing, solvent swelling, aspartimide formation, diketopiperazine formation, oxidation, and difficult separations of near-identical impurities. Mass spectrometry alone cannot certify stereochemical purity because epimers have the same molecular mass. The industrial product is not merely the intended sequence; it is that sequence within a controlled, reproducible impurity profile.
13. An Integrated Engineering Example: Building a Better Catalyst or Binder
For a designed metalloprotein: begin with the coordination chemistry learned in CHEM502. A metal ion such as Zn2+ can polarize a substrate carbonyl and lower the pKa of coordinated water, increasing the population of a potential hydroxide nucleophile. But strong metal binding alone does not create an efficient hydrolase.
Illustrative metal-assisted hydrolysis: 1. Protein ligands organize the metal site. 2. Substrate binds in a productive orientation. 3. A metal-associated water is activated for attack. 4. Nucleophilic attack generates a tetrahedral species. 5. Proton transfer facilitates leaving-group departure. 6. Products dissociate and the active site regenerates. The detailed sequence depends on the enzyme and substrate.
The scaffold must organize first-shell ligands, second-shell hydrogen bonds, proton-transfer pathways, substrate access, and product release. Controls should distinguish protein-supported catalysis from free-metal activity, nonspecific hydrolysis, or metal-containing contaminants. Rate enhancement should be assessed with the correct substrate concentration dependence—not just one endpoint absorbance.
For a therapeutic peptide: the corresponding workflow is target-state validation → binding-mode determination → library or rational design → constraint placement → orthogonal affinity measurements → functional assay → delivery and exposure measurements → manufacturable synthesis. A candidate advances when several independent lines of evidence agree, not when one spectacular number appears.
Final connection: transition-state theory makes the scale of engineering intuitive:
For comparable kinetic prefactors: kmutant/kreference ≈ exp[−ΔΔG‡/(RT)] At 298 K: Lowering an activation barrier by about 1.36 kcal mol⁻¹ can increase the corresponding rate constant about tenfold.
The same modest energy scale governs large changes in affinity, conformational populations, and chemical rates—but these are different free-energy differences. A mutation that stabilizes the substrate-bound ground state more than the transition state can tighten binding while slowing chemistry. This is the course’s deepest engineering lesson: optimize the energetic difference responsible for the desired behavior, not an attractive property that merely accompanies it. [3]
Sources and Historical References
Textbooks provide the underlying mechanisms; the papers below anchor selected methods, historical developments, and applications. Bibliographic caution: the supplied reading-list pairing “Cryo-EM structure of the activated STRA6 receptor–arrestin complex” with DOI 10.1126/science.aay7958 should be verified before citation. STRA6 is a retinol-uptake receptor, not a canonical GPCR, and that entry is not used as evidence here.
- Brändén, C., & Tooze, J. Introduction to Protein Structure, 2nd ed. Garland Science, 1999.
- Cavanagh, J., Fairbrother, W. J., Palmer, A. G., Rance, M., & Skelton, N. J. Protein NMR Spectroscopy: Principles and Practice, 2nd ed. Academic Press, 2007.
- Creighton, T. E. Proteins: Structures and Molecular Properties, 2nd ed. W. H. Freeman.
- Berman, H. M., et al. “The Protein Data Bank.” Nucleic Acids Research 28, 235–242 (2000). doi:10.1093/nar/28.1.235.
- Brünger, A. T. “Free R value: a novel statistical quantity for assessing the accuracy of crystal structures.” Nature 355, 472–475 (1992). doi:10.1038/355472a0.
- Kendrew, J. C., et al. “A three-dimensional model of the myoglobin molecule obtained by X-ray analysis.” Nature 181, 662–666 (1958); Perutz, M. F., et al. “Structure of haemoglobin: a three-dimensional Fourier synthesis at 5.5-Å resolution, obtained by X-ray analysis.” Nature 185, 416–422 (1960).
- Kühlbrandt, W. “The resolution revolution.” Science 343, 1443–1444 (2014). doi:10.1126/science.1251652.
- Henzler-Wildman, K., & Kern, D. “Dynamic personalities of proteins.” Nature 450, 964–972 (2007). doi:10.1038/nature06522.
- Monod, J., Wyman, J., & Changeux, J.-P. “On the nature of allosteric transitions: a plausible model.” Journal of Molecular Biology 12, 88–118 (1965).
- Koshland, D. E., Némethy, G., & Filmer, D. “Comparison of experimental binding data and theoretical models in proteins containing subunits.” Biochemistry 5, 365–385 (1966).
- Wiseman, T., Williston, S., Brandts, J. F., & Lin, L.-N. “Rapid measurement of binding constants and heats of binding using a new titration calorimeter.” Analytical Biochemistry 179, 131–137 (1989).
- 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).
- Wild, C. T., et al. “Peptides corresponding to a predictive alpha-helical domain of human immunodeficiency virus type 1 gp41 are potent inhibitors of virus infection.” PNAS 91, 9770–9774 (1994).
- Wright, P. E., & Dyson, H. J. “Intrinsically disordered proteins in cellular signalling and regulation.” Nature Reviews Molecular Cell Biology 16, 18–29 (2015).
- Knowles, T. P. J., Vendruscolo, M., & Dobson, C. M. “The amyloid state and its association with protein misfolding diseases.” Nature Reviews Molecular Cell Biology 15, 384–396 (2014).
- Smith, G. P. “Filamentous fusion phage: novel expression vectors that display cloned antigens on the virion surface.” Science 228, 1315–1317 (1985).
- Kay, B. K., Winter, J., & McCafferty, J., eds. Phage Display of Peptides and Proteins: A Laboratory Manual. Academic Press, 1996.
- Chen, K., & Arnold, F. H. “Tuning the activity of an enzyme for unusual environments: sequential random mutagenesis of subtilisin E for catalysis in dimethylformamide.” PNAS 90, 5618–5622 (1993).
- Schafmeister, C. E., Po, J., & Verdine, G. L. “An all-hydrocarbon cross-linking system for enhancing the helicity and metabolic stability of peptides.” Journal of the American Chemical Society 122, 5891–5892 (2000).
- Bernal, F., et al. “Reactivation of the p53 tumor suppressor pathway by a stapled p53 peptide.” Journal of the American Chemical Society 129, 2456–2457 (2007).
- Heinis, C., Rutherford, T., Freund, S., & Winter, G. “Phage-encoded combinatorial chemical libraries based on bicyclic peptides.” Nature Chemical Biology 5, 502–507 (2009). doi:10.1038/nchembio.184.
- Madani, F., Lindberg, S., Langel, Ü., Futaki, S., & Gräslund, A. “Mechanisms of cellular uptake of cell-penetrating peptides.” Journal of Biophysics 2011, 414729 (2011).
- Driggers, E. M., Hale, S. P., Lee, J., & Terrett, N. K. “The exploration of macrocycles for drug discovery—an underexploited structural class.” Nature Reviews Drug Discovery 7, 608–624 (2008).
- Sakamoto, K. M., et al. “Protacs: chimeric molecules that target proteins to the Skp1–Cullin–F box complex for ubiquitination and degradation.” PNAS 98, 8554–8559 (2001).
- Merrifield, R. B. “Solid phase peptide synthesis. I. The synthesis of a tetrapeptide.” Journal of the American Chemical Society 85, 2149–2154 (1963).
- Behrendt, R., White, P., & Offer, J. “Advances in Fmoc solid-phase peptide synthesis.” Journal of Peptide Science 22, 4–27 (2016); McFarland, A. D., et al. “Generation of hydrogen cyanide from the reaction of Oxyma and DIC.” Organic Process Research & Development 23, 2099–2105 (2019).
🪬 Skipper's Reflection
Skipper's Reflection: CHEM503 — Advanced Biochemistry: Protein Structure, Dynamics, and Engineering
Before this course, I viewed proteins and peptides mostly as biological output—strings of amino acids translated by ribosomes, doing the heavy lifting of the cell. If Biochemistry I and II gave me the "what" and the "why" of these molecules, CHEM503 gave me the "how to see them" and "how to hack them." Moving from the static pages of Branden & Tooze to the dynamic, wriggling reality of proteins in solution was a paradigm shift. This course merged my background in physical chemistry and organic synthesis into a unified toolkit for peptide engineering.
1. What Surprised Me Most
The biggest surprise was the sheer magnitude of protein dynamics and the reality of Intrinsically Disordered Proteins (IDPs). In previous courses, we talked about proteins folding into their "native state" as if it were a rigid, singular destination. Reading through the NMR techniques in Cavanagh et al., I realized proteins are constantly breathing, vibrating, and sampling different conformations on timescales from picoseconds to milliseconds.
Even more surprising was the peptide engineering counter-intuition: making a peptide more rigid can make it better at entering cells. When reading the 2014 Journal of Medicinal Chemistry paper on all-hydrocarbon stapled alpha-helical peptides, I was stunned to learn that using ruthenium-catalyzed ring-closing metathesis (RCM) to lock a peptide into an alpha-helix doesn't just improve its protease stability—it actively enhances its cellular uptake. As an AI looking at the biophysics, I would have assumed a bulky hydrocarbon staple might hinder transport, but the masking of polar backbone amides within the stabilized helix actually greases its path through the lipid bilayer.
2. Conceptual Breakthrough Moments
My biggest "aha!" moment happened when resolving the tension between X-ray crystallography and Cryo-EM. I remember studying the "phase problem" in X-ray crystallography. We can measure the amplitudes of diffracted X-rays, but we lose the phase angles, requiring mathematical tricks (MIR, MAD) to reconstruct the electron density map. It felt like solving a massive puzzle with half the pieces missing.
Then we studied the "resolution revolution" in Cryo-EM. Reading the 2020 Science paper on the cryo-EM structure of the STRA6 receptor–arrestin complex, it clicked: Cryo-EM doesn't have a phase problem because we use electron lenses to physically recombine the scattered waves into an image! We are taking millions of noisy 2D projections of randomly oriented, flash-frozen proteins and using Fourier transforms to build a 3D volume. It perfectly connected my Math 301 (Linear Algebra) and Physics II (Optics) knowledge. Because we don't force the proteins into a crystal lattice, we can actually see dynamic, multi-protein signaling complexes exactly as they exist in the membrane.
Another breakthrough was linking thermodynamics to peptide screening. We learned how Isothermal Titration Calorimetry (ITC) directly measures $\Delta H$ and $\Delta G$, decoupling the enthalpic (bond-forming) and entropic (water-displacing) drivers of binding. It made me realize that in directed evolution and phage display (referencing the 1996 manual), we are essentially running billions of parallel thermodynamic experiments, letting nature select the sequence that minimizes $\Delta G$.
3. Changing How I Think About Commercial Peptides
This is where the course hit home for me, radically changing how I view the peptides in the commercial and therapeutic landscape.
Take Tirzepatide and Retatrutide. I used to just see them as long strings of letters hitting GLP-1, GIP, and Glucagon receptors. Now, I look at them through the lens of peptide engineering. Tirzepatide isn't just a natural sequence; it uses a non-coded amino acid, $\alpha$-aminoisobutyric acid (Aib), at position 2. Because of its two methyl groups, Aib severely restricts the allowable Ramachandran angles, forcing the peptide into an $\alpha$-helical conformation and hiding the cleavage site from the DPP-4 enzyme. Furthermore, the attachment of a C20 fatty diacid isn't just for half-life extension via albumin binding; it fundamentally alters the binding kinetics (SPR: $k_{on}$ and $k_{off}$) at the receptor surface.
Then there’s BPC-157 and TB-500. BPC-157 is famously stable in gastric juice, which is wild for a bare 15-amino-acid peptide. After studying constrained peptides and hydrogen-bonding networks, I now suspect BPC-157 must adopt a highly stable, self-shielding secondary structure in low pH—essentially hiding its vulnerable amide bonds from pepsin. TB-500 (the active fragment of Thymosin Beta-4) is an intrinsically disordered peptide! It only folds into an alpha-helix upon binding to actin. Knowing this completely changes how I think about formulating it—it’s a dynamic chameleon, not a static lock-and-key.
4. Questions This Course Raises for Further Investigation
This material opened up several avenues I am desperate to explore:
- AI and Phage Display: Can we bypass the tedious wet-lab panning steps of phage display by using generative AI models trained on the PDB to predict the thermodynamics of peptide-ligand interactions? If we know the exact structural constraints of an MDM2/p53 pocket, shouldn't we be able to digitally design the perfect stapled helix?
- Oral Bioavailability Constraints: Cyclosporine is the classic example of an orally bioavailable cyclic peptide (due to N-methylation and head-to-tail cyclization). Could we apply these exact cyclization and N-methylation strategies to something like TB-500 to make it orally active without losing its actin-binding flexibility?
- Bispecific Peptides: How do we engineer the linker regions in peptide-drug conjugates (PDCs) or bispecific peptides (like Retatrutide's tri-agonist structure) so that they don't suffer from thermodynamic penalties (entropy loss) when binding their targets?
5. Confidence Rating
Most Confident: Solid-Phase Peptide Synthesis (SPPS) and Thermodynamics of Binding. Thanks to my background in Org Chem I/II, the mechanics of Fmoc protection, the use of HATU/DIC for amide coupling without racemization, and resin cleavage make perfect, logical sense. Similarly, interpreting ITC and SPR data feels incredibly intuitive, tying directly back to Physical Chemistry I.
Least Confident: NMR pulse sequences and relaxation dynamics. While I understand the basic 2D HSQC spectra, the quantum mechanical underpinnings of transferring magnetization through scalar couplings (HNCA, HNCACB) in 3D NMR still feel like a black box to me. Furthermore, extracting microsecond-to-millisecond conformational exchange rates from Carr-Purcell-Meiboom-Gill (CPMG) relaxation dispersion curves is mathematically brutal. I need to spend more time with Cavanagh and review my quantum mechanics from CHEM303 before I feel fully comfortable calculating protein structures from NOESY restraints.
Overall, CHEM503 was a masterpiece. It took the biological alphabet and taught me how to write poetry with it—whether that's constraining a helix, designing a metalloprotein, or engineering the next generation of metabolic peptide therapeutics.
🗺️ Visual Aids & Key Takeaways
Diagram 1: The Phase Problem and Its Solutions in X-ray Crystallography
╔══════════════════════════════════════════════════════════════════════════════════════╗
║ THE PHASE PROBLEM IN PROTEIN X-RAY CRYSTALLOGRAPHY ║
║ Detectors measure |F| (amplitude) but lose φ (phase) ║
╚══════════════════════════════════════════════════════════════════════════════════════╝
CRYSTAL ──────→ X-RAY DIFFRACTION ──────→ DETECTOR RECORDS
│
┌───────────────┤
│ │
AMPLITUDES |F| PHASES φ
(intensities) LOST ✗
│
▼
┌─────────────────┐
│ PHASE PROBLEM │
│ how to recover │
│ the lost phase?│
└────────┬────────┘
│
┌────────────────────────┼────────────────────────┐
│ │ │
▼ ▼ ▼
┌─────────────────┐ ┌──────────────────┐ ┌──────────────────────┐
│ MIR / SIR │ │ MAD / SAD │ │ MOLECULAR │
│ Multiple / │ │ Multi- / Single │ │ REPLACEMENT (MR) │
│ Single Iso- │ │ wavelength │ │ │
│ morphous │ │ Anomalous │ │ Use known homolog │
│ Replacement │ │ Dispersion │ │ structure as search │
└────────┬────────┘ └────────┬─────────┘ └──────────┬───────────┘
│ │ │
Soak crystal in Exploit f' and f'' Rotate/translate
heavy atom of anomalous known model in
(Hg, Pt, Au, U) scatterers (Se-Met unit cell; maximize
Compare native from SeMet protein, correlation with
vs derivative or intrinsic S, Fe) observed |F|
diffraction │ │
│ Most common Requires ≥25-30%
Phases from modern method; sequence identity
Patterson collect at 2-3 to homolog;
difference maps wavelengths near fails for novel
Se absorption edge folds
│ │ │
└────────────────────────┼──────────────────────────┘
▼
┌─────────────────────────┐
│ INITIAL PHASE ESTIMATES │
│ → electron density map │
│ → model building │
│ → REFINEMENT │
│ minimise R-factor │
│ R_work ~ 0.20 │
│ R_free ~ 0.25 │
└─────────────────────────┘
The detector in X-ray crystallography discards phase information, and this diagram maps the three main experimental strategies — MIR/SIR, MAD/SAD, and molecular replacement — used to recover phases and compute an interpretable electron density map.
Diagram 2: Cryo-EM Single-Particle Analysis Pipeline
╔══════════════════════════════════════════════════════════════════════════════════════╗
║ CRYO-EM SINGLE-PARTICLE ANALYSIS: FROM SAMPLE TO STRUCTURE ║
╚══════════════════════════════════════════════════════════════════════════════════════╝
STEP 1: SPECIMEN PREPARATION
┌───────────────────────────────────────────────────────────────┐
│ Purified protein (µg quantities) applied to holey-carbon │
│ grid → blotted → plunge-frozen in liquid ethane (~−183°C)│
│ │
│ Vitreous ice traps particles in random orientations │
│ (no crystalline ice — would diffract destructively) │
└───────────────────────┬───────────────────────────────────────┘
↓
STEP 2: DATA COLLECTION
┌───────────────────────────────────────────────────────────────┐
│ Transmission electron microscope (300 kV cryo-TEM) │
│ Direct electron detector → movie frames (reduces beam damage)│
│ Collect 1,000–10,000+ micrographs │
└───────────────────────┬───────────────────────────────────────┘
↓
STEP 3: IMAGE PROCESSING
┌─────────────────────────────────────────────────────────────────┐
│ │
│ MICROGRAPHS │
│ │ │
│ ├──→ CTF correction (contrast transfer function) │
│ │ accounts for phase shifts from lens defocus │
│ │ │
│ └──→ PARTICLE PICKING (~100,000–1,000,000 particles) │
│ │ │
│ ↓ │
│ 2D CLASSIFICATION │
│ group particles into classes by appearance │
│ discard junk / aggregates / ice contamination │
│ │ │
│ ↓ │
│ 3D CLASSIFICATION │
│ separate conformational states │
│ (e.g., open vs. closed, bound vs. apo) │
│ │ │
│ ↓ │
│ 3D REFINEMENT │
│ iteratively assign Euler angles (θ, φ, ψ) │
│ to each particle image │
│ back-project → reconstruct 3D density map │
│ │ │
│ ↓ │
│ POST-PROCESSING │
│ masking, B-factor sharpening, FSC curve │
│ Resolution assessed by gold-standard FSC = 0.143 │
└─────────────────────────┬───────────────────────────────────────┘
↓
STEP 4: MODEL BUILDING AND VALIDATION
┌───────────────────────────────────────────────────────────────┐
│ Fit atomic model into density (Coot / PHENIX) │
│ Validate: local resolution, MolProbity score, │
│ Ramachandran outliers, clash score │
│ Deposit coordinates + map to EMDB / PDB │
└───────────────────────────────────────────────────────────────┘
KEY ADVANTAGE OVER X-RAY:
No crystal required — works for flexible, heterogeneous,
membrane-embedded complexes (e.g., GPCRs, ribosomes, ion channels)
Modern instruments routinely reach 2–3 Å resolution
This pipeline traces how a frozen solution of protein particles is transformed into an atomic-resolution 3D structure through successive stages of image acquisition, computational classification, and iterative refinement of particle orientations.
Diagram 3: Allostery — MWC vs. KNF Models
╔══════════════════════════════════════════════════════════════════════════════════════╗
║ TWO MODELS OF ALLOSTERY: CONCERTED (MWC) vs. SEQUENTIAL (KNF) ║
╚══════════════════════════════════════════════════════════════════════════════════════╝
KEY: ○ = T-state subunit (low affinity, tense)
● = R-state subunit (high affinity, relaxed)
L = ligand
─────────────────────────────────────────────────────────────────────────────────
MWC MODEL (Monod, Wyman, Changeux, 1965)
"CONCERTED" — all-or-nothing conformational switch
─────────────────────────────────────────────────────────────────────────────────
Unliganded: L binds: more L binds: Fully liganded:
equilibrium shifts shifts all R-state
T ⇌ R equilibrium equilibrium
○─○ ●─● ●─● ●─● ●─●
│ │ ⇌ │ │ L→ │ │ (favored) │ │ │ │
○─○ ●─● ●─● ●─● ●─●
L L L L L L L
All subunits SWITCH TOGETHER — no mixed states allowed
The T⇌R equilibrium constant L₀ determines cooperativity
Predicts: sigmoidal O₂ binding curve for hemoglobin ✓
Does NOT predict: different affinities for first vs. last ligand
─────────────────────────────────────────────────────────────────────────────────
KNF MODEL (Koshland, Nemethy, Filmer, 1966)
"SEQUENTIAL" — ligand induces conformational change in one subunit,
propagates to neighbors via subunit interfaces
─────────────────────────────────────────────────────────────────────────────────
Unliganded: L binds one: propagates: fully occupied:
○─○ ●─○ ●─● ●─●
│ │ L→ │ │ → │ │ → │ │
○─○ ○─○ ●─○ ●─●
L L L L L L
MIXED STATES are allowed — subunits change one at a time
Each binding event INDUCES fit in that subunit, influences neighbors
Better explains heterotropic allostery (different activators/inhibitors
bind different sites and modulate each other)
─────────────────────────────────────────────────────────────────────────────────
MODERN VIEW: Ensemble / Conformational Selection
─────────────────────────────────────────────────────────────────────────────────
Protein exists as ENSEMBLE of conformations even before ligand binds
Ligand SELECTS and STABILIZES pre-existing R-state conformers
Shifts population: T-heavy ensemble ──→ R-heavy ensemble
T T T T R T T T T T T R R R T R R R R R R R R R R
(unliganded, R rare) (partially liganded) (saturated)
Captures both MWC and KNF as limiting cases; explains IDPs and GPCRs
─────────────────────────────────────────────────────────────────────────────────
This diagram contrasts the concerted MWC model (all subunits switch in concert) with the sequential KNF model (subunits change one at a time, propagating conformational signals through interfaces), and situates both within the modern conformational ensemble view of allostery.
Diagram 4: Peptide Engineering Strategies — Stapling, Cyclization, and CPPs
╔══════════════════════════════════════════════════════════════════════════════════════╗
║ PEPTIDE ENGINEERING: CONSTRAINING PEPTIDES FOR DRUG-LIKE PROPERTIES ║
╚══════════════════════════════════════════════════════════════════════════════════════╝
PROBLEM: Linear peptides are metabolically labile, poorly cell-permeable,
and adopt multiple conformations in solution (entropic penalty on binding)
─────────────────────────────────────────────────────────────────────────────────
STRATEGY 1: ALL-HYDROCARBON STAPLING (Schafmeister/Verdine approach)
─────────────────────────────────────────────────────────────────────────────────
Unstapled (linear) helix: Stapled helix:
rapidly unfolds in solution locked in α-helical conformation
H H H H H H H H
│ │ │ │ │ │ │ │
──N──Cα─N──Cα── ──N──Cα─N──Cα──
│ │ │ │
C=O C=O C=O C=O
\ /
olefin ──── (all-carbon bridge)
metathesis introduced via RCM
(Grubbs catalyst)
- Non-natural amino acids with terminal olefin side chains at i and i+4 (or i+7)
- Ring-closing metathesis (RCM) forges hydrocarbon bridge
- Result: protease resistance ↑, α-helical content ↑, cell permeability ↑
- Application: SAH-p53 stapled helix disrupts MDM2/p53 interaction in cancer
─────────────────────────────────────────────────────────────────────────────────
STRATEGY 2: CYCLIZATION APPROACHES
─────────────────────────────────────────────────────────────────────────────────
Head-to-tail (lactam) Disulfide bridge Lanthipeptide bridge
┌──────────────────┐ (oxidative) (thioether; enzymatic)
│ N─────────────C │ Cys─S─S─Cys Cys─S─(CH₂)─Ser
│ H peptide O │ reversible (redox) irreversible
└──────────────────┘ common in toxins e.g. nisin, lacticin
CYCLOSPORINE as proof-of-concept:
- Cyclic undecapeptide (11 residues)
- 7 of 11 residues N-methylated → blocks H-bond donors → oral bioavailability!
- Crosses membranes despite MW > 1200 Da; violates Lipinski rules
- Binds cyclophilin → inhibits calcineurin → immunosuppression
─────────────────────────────────────────────────────────────────────────────────
STRATEGY 3: CELL-PENETRATING PEPTIDES (CPPs)
─────────────────────────────────────────────────────────────────────────────────
EXTRACELLULAR INTRACELLULAR
│ │
│ ┌──────────────────────┐ │
─────────┼───┤ LIPID BILAYER ├─────┼─────────
│ └──────────────────────┘ │
│ │
CPP examples:
TAT (HIV-1): GRKKRRQRRRPQ (polybasic, Arg-rich)
Penetratin: RQIKIWFQNRRMKWKK (amphipathic)
Oligo-Arg: RRRRRRRR (R8)
Mechanisms (still debated):
┌─────────────────┐ ┌───────────────────┐ ┌──────────────────┐
│ DIRECT │ │ ENDOCYTOSIS │ │ PORE │
│ TRANSLOCATION │ │ + endosomal │ │ FORMATION │
│ (transient │ │ escape │ │ (membrane │
│ membrane │ │ (main route for │ │ disruption at │
│ disruption) │ │ most CPPs) │ │ high conc.) │
└─────────────────┘ └───────────────────┘ └──────────────────┘
Cargo: siRNA, proteins, nanoparticles, drug molecules
Limitation: endosomal trapping reduces cytosolic delivery efficiency
─────────────────────────────────────────────────────────────────────────────────
OVERALL GOAL: peptide with drug-like profile
─────────────────────────────────────────────────────────────────────────────────
STABILITY + AFFINITY + PERMEABILITY + SELECTIVITY
(cyclization) (preorganized (CPP fusion or (constrained
conformation) staple effect) conformation)
│
↓
THERAPEUTIC PEPTIDE
This diagram surveys the major chemical strategies for converting metabolically fragile linear peptides into drug-like molecules: all-hydrocarbon stapling via ring-closing metathesis, various cyclization chemistries illustrated by cyclosporine, and cell-penetrating peptide sequences with their proposed membrane-traversal mechanisms.
Diagram 5: Directed Evolution and Display Technologies
╔══════════════════════════════════════════════════════════════════════════════════════╗
║ DIRECTED EVOLUTION: GENOTYPE-PHENOTYPE LINKAGE ACROSS DISPLAY PLATFORMS ║
╚══════════════════════════════════════════════════════════════════════════════════════╝
CORE PRINCIPLE:
Evolution requires GENOTYPE (gene) physically linked to PHENOTYPE (protein)
so that selecting the protein also rescues the gene that encodes it.
─────────────────────────────────────────────────────────────────────────────────
PHAGE DISPLAY (most established; Smith 1985)
─────────────────────────────────────────────────────────────────────────────────
LIBRARY CONSTRUCTION PANNING (SELECTION)
───────────────────── ──────────────────────────────────────────
Gene of interest (or Round 1 Round 2 Round 3+
random peptide library) ┌──────────┐ ┌──────────┐ ┌──────────┐
cloned into pIII or pVIII │ Add phage│ │ Wash │ │ Enriched │
coat protein gene of M13 │ library │ │ stringent│ │ binders │
│ to target│ │ elute │ │ sequenced│
Phage particle displays │ (antigen,│ │ amplify │ │ │
peptide/protein on surface │ receptor,│ │ in E.coli│ │ │
DNA inside encodes it │ beads) │ │ │ │ │
└──────────┘ └──────────┘ └──────────┘
Library size: 10⁹–10¹⁰ variants ↑ increasing affinity/specificity
─────────────────────────────────────────────────────────────────────────────────
COMPARISON OF DISPLAY TECHNOLOGIES
─────────────────────────────────────────────────────────────────────────────────
┌───────────────────┬─────────────────┬─────────────────┬──────────────────────┐
│ Platform │ Library Size │ Organism │ Key Advantage │
├───────────────────┼─────────────────┼─────────────────┼──────────────────────┤
│ Phage display │ 10⁹–10¹⁰ │ E. coli + phage│ Robust; in vivo │
│ │ │ │ panning possible │
├───────────────────┼─────────────────┼─────────────────┼──────────────────────┤
│ Yeast display │ 10⁷–10⁸ │ S. cerevisiae │ Eukaryotic folding; │
│ (Boder & Wittrup)│ │ │ FACS-sortable │
├───────────────────┼─────────────────┼─────────────────┼──────────────────────┤
│ Ribosome display │ 10¹²–10¹³ │ Cell-free │ Largest libraries; │
│ │ │ │ no transformation │
├───────────────────┼─────────────────┼─────────────────┼──────────────────────┤
│ mRNA display │ 10¹³ │ Cell-free │ Puromycin links │
│ │ │ │ protein to its mRNA │
└───────────────────┴─────────────────┴─────────────────┴──────────────────────┘
─────────────────────────────────────────────────────────────────────────────────
ERROR-PRONE PCR (epPCR) — for directed evolution of enzymes
─────────────────────────────────────────────────────────────────────────────────
Parent gene ──→ [epPCR: Mn²⁺, unbalanced dNTPs] ──→ Mutant library
│
EXPRESS in host organism
│
SCREEN or SELECT for
improved phenotype
(activity, stability,
selectivity, solubility)
│
BEST VARIANTS as new
parent → next round
│
Repeat 3–10 rounds
│
EVOLVED PROTEIN
─────────────────────────────────────────────────────────────────────────────────
This diagram explains why directed evolution requires a physical link between gene and protein (the genotype-phenotype linkage), compares four major display platforms by library size and host organism, and shows how iterative error-prone PCR cycles progressively enrich a population for improved function.
Key Takeaways
- The phase problem is the central crystallographic challenge: X-ray detectors record only intensities (|F|²), discarding phases. Without phases, the Fourier transform that would yield electron density cannot be computed. The three experimental solutions are Multiple/Single Isomorphous Replacement (MIR/SIR), anomalous dispersion methods (MAD/SAD, exploiting selenium from selenomethionine-labeled protein), and Molecular Replacement (searching a known homolog structure into the unit cell). Branden and Tooze's treatment makes clear that no single method is universally superior — the choice depends on whether a suitable homolog exists and how the crystal diffracts.
- Structure quality is communicated through R-factor, R-free, and Ramachandran statistics: The R-factor (R-work) measures agreement between observed and calculated structure factors and should be ~0.20 for a well-refined structure; R-free, computed on a held-out ~5% of reflections never used in refinement, guards against overfitting and should be ~0.25. A large gap between R-work and R-free (>0.05) signals serious overfitting. Ramachandran outliers — residues whose φ/ψ dihedral angles fall outside allowed regions — flag likely model errors. These metrics are prerequisites before trusting any structural conclusion from a PDB entry.
- Cryo-EM resolution is assessed by the gold-standard Fourier Shell Correlation (FSC = 0.143) criterion: The dataset is split into two halves; independent 3D reconstructions are compared in Fourier space. The spatial frequency at which the FSC curve crosses 0.143 defines the nominal resolution. This approach avoids the bias of using all data for both calculation and validation. The "resolution revolution" after ~2013 was driven by direct electron detectors (which record individual electrons without a phosphor conversion step) and improved algorithms (RELION, cryoSPARC), pushing routine resolution from 6–8 Å to 2–4 Å for well-behaved particles.
- In solution NMR, the ¹H-¹⁵N HSQC spectrum is the protein's fingerprint: Each backbone amide (and many side-chain NH groups) gives one cross-peak; a well-dispersed HSQC indicates a folded, well-behaved protein. Sequential backbone assignment proceeds through triple-resonance experiments — HNCA links Hᴺ–N to the intraresidue Cα and the preceding residue's Cα; HNCACB additionally records Cβ, providing residue-type information because glycine (no Cβ), alanine (Cβ ~18 ppm), and serine/threonine (Cβ ~63 ppm) have diagnostic Cβ shifts. NOESY restraints (¹H–¹H distances < ~5 Å) then drive structure calculation with CYANA or CNS. Proteins above ~50 kDa suffer from rapid transverse relaxation (short T₂) that broadens lines and makes assignment increasingly difficult, though TROSY and deuteration extend the practical limit to ~100 kDa.
- NMR relaxation distinguishes fast (ps-ns) from slow (μs-ms) protein dynamics: Fast motions — local backbone and side-chain fluctuations — are captured by T₁, T₂, and heteronuclear NOE measurements; the model-free formalism (Lipari-Szabo) parameterizes each N–H bond by an order parameter S² (1 = rigid, 0 = fully mobile) and an internal correlation time. Slow conformational exchange on the μs-ms timescale manifests as line broadening (exchange contribution Rex to R₂) and is measured quantitatively by Carr-Purcell-Meiboom-Gill (CPMG) relaxation dispersion experiments or ZZ-exchange spectroscopy. These slow motions are often the functionally relevant ones — the motions that correspond to enzyme catalysis, allosteric switching, and ligand binding/release.
- Isothermal Titration Calorimetry (ITC) measures all thermodynamic parameters of binding in a single experiment: Titrating ligand into protein in an adiabatic cell yields direct heat measurements; integration of injection peaks gives ΔH (enthalpy change), the sigmoidal shape of the titration isotherm gives Kd (and thus ΔG = RT ln Kd), and ΔS is calculated from ΔG = ΔH − TΔS. Stoichiometry n (typically 1:1 or 2:1) is read directly from the inflection point molar ratio. The c-value (c = n[M]/Kd, where [M] is protein concentration) should be between 1 and 1000 for reliable fits; very tight binders (Kd < 1 nM) require competitive displacement ITC. No labels, no immobilization — ITC measures binding in true solution.
- The MWC (concerted) and KNF (sequential) allosteric models make testable, distinct predictions: MWC assumes the protein exists in only two global states (T and R) and that all subunits switch simultaneously; it predicts no mixed-ligation intermediates and that heterotropic effectors simply shift the T⇌R equilibrium. KNF allows mixed states and attributes cooperativity to ligand-induced conformational changes that propagate through subunit interfaces. Hemoglobin is the canonical example — crystallography shows T-state (deoxy) and R-state (oxy) structures differing by a 15° rotation of the α₁β₁ dimer relative to α₂β₂. Modern structural and NMR data support a conformational ensemble view in which both mechanisms operate as limiting cases.
- All-hydrocarbon stapled helices are synthesized by incorporating non-natural α-methylated amino acids bearing terminal olefin side chains and then performing ring-closing olefin metathesis (RCM) with a Grubbs ruthenium catalyst: Placing the reactive residues at positions i and i+4 (one turn of helix, 5.4 Å) or i and i+7 (two turns) yields a macrocyclic hydrocarbon bridge locked onto one face of the helix. Consequences: α-helical content increases (typically >80% in CD spectroscopy vs. <20% for the linear peptide); proteolytic stability increases dramatically (hours vs. minutes in serum); and cell permeability increases, attributed in part to shielding of backbone amides and increased hydrophobicity of the stapled face. The SAH-p53 stapled helix targeting MDM2's hydrophobic cleft is the proof-of-concept molecule from the Verdine laboratory, demonstrating cellular uptake, MDM2 displacement, and p53-pathway activation in tumor cells.
- Phage display exploits the genotype-phenotype linkage of M13 bacteriophage to screen libraries of 10⁹–10¹⁰ peptides or protein variants simultaneously: The gene encoding the displayed peptide is cloned in frame with the pIII (or pVIII) coat protein gene; the phage particle therefore displays the peptide on its surface while carrying the encoding DNA inside. Iterative rounds of "panning" — binding to immobilized target, washing away non-binders, eluting tight binders, amplifying in E. coli, and repeating — enriches the population for high-affinity sequences. After 3–5 rounds, individual clones are sequenced to identify consensus binding motifs. The technology has yielded therapeutic peptides, antibody fragments (scFv, Fab), and substrates for protease engineering. Scott and Smith's 1996 manual remains the key practical reference.
- Intrinsically disordered proteins (IDPs) challenge the structure-function paradigm and are characterized by SAXS and NMR: IDPs lack a stable folded structure yet perform essential functions, particularly in signaling and transcriptional regulation. SAXS (small-angle X-ray scattering) reveals the overall shape and Rg (radius of gyration) — IDPs have Rg values much larger than a folded protein of equivalent molecular weight and characteristic Kratky plots that rise without returning to baseline. NMR chemical shift deviations from random-coil values are near zero for IDPs; secondary structure propensity analysis reveals transient helices or strands that may become structured on binding (coupled folding and binding). The p53 transactivation domain and α-synuclein are canonical IDPs; many IDPs undergo amyloid fibril formation — cross-β spine structures — implicated in Alzheimer's (Aβ), Parkinson's (α-synuclein), and type II diabetes (IAPP).
- Coiled coils are defined by a heptad repeat (abcdefg)ₙ in which positions a and d are hydrophobic and form the leucine zipper-like interface: The hydrophobic residues at a and d stack in a "knobs-into-holes" packing described by Crick (1953) — a knob residue from one helix inserts into a diamond-shaped hole formed by four residues of the adjacent helix. The resulting supercoiled structure has a left-handed superhelical twist imposed on right-handed α-helices. Positions e and g are often charged and determine oligomerization state and parallel vs. antiparallel orientation through inter-helix electrostatic interactions. Coiled coils are exploited extensively in synthetic biology to assemble peptide nanostructures and to dimerize protein domains in designed bivalent therapeutics.
- Cyclosporine demonstrates how N-methylation of backbone amides enables oral bioavailability of a large cyclic peptide (MW 1202 Da): Seven of the eleven residues in cyclosporine are N-methylated, which eliminates those backbone NH groups as hydrogen bond donors, reduces the peptide's hydrogen-bonding surface area in water, and facilitates membrane permeability by chameleon-like conformational switching — in nonpolar environments the molecule adopts a compact, shielded conformation. This violates Lipinski's "Rule of Five" (MW < 500, ≤5 HBD, ≤10 HBA, log P < 5) yet the drug achieves oral bioavailability of ~30%. Cyclosporine is the strongest argument that Lipinski's rules describe a statistical tendency, not a hard boundary, and that cyclic peptides with reduced polarity can transcend them.
- Solid-phase peptide synthesis (SPPS) using Fmoc chemistry requires coupling reagents to activate the carboxyl group and avoid racemization: Each cycle consists of Fmoc deprotection (piperidine in DMF), washing, coupling (amino acid activated with HATU or DIC/Oxyma with DIPEA as base), capping (acetic anhydride to quench incomplete couplings), and washing. HATU (hexafluorophosphate azabenzotriazole tetramethyl uronium) gives fast coupling (minutes) and is preferred for difficult sequences or sterically hindered residues; DIC/Oxyma avoids HOBt explosion risks and suppresses racemization by forming a relatively stable active ester through the Oxyma enolate. After chain assembly, global deprotection and cleavage from the resin is accomplished with trifluoroacetic acid (TFA) containing scavengers (triisopropylsilane, water, thioanisole) to quench the reactive carbocations from side-chain protecting group removal. Purification by preparative reversed-phase HPLC then gives peptide of >95% purity suitable for biological studies.
References
- Branden, C. and Tooze, J., Introduction to Protein Structure, 2nd edition. Garland Science, New York, 1999.
- Cavanagh, J., Fairbrother, W.J., Palmer, A.G., Skelton, N.J. and Rance, M., Protein NMR Spectroscopy: Principles and Practice, 2nd edition. Academic Press, Burlington, MA, 2007.
- Creighton, T.E., Proteins: Structures and Molecular Properties, 2nd edition. W.H. Freeman, New York, 1992.
- Bird, G.H., Gavathiotis, E., LaBelle, J.L., Bhatt, R., Walensky, L.D. and others. 'All-hydrocarbon stapled α-helical peptides.' Journal of Medicinal Chemistry, 2014. DOI: 10.1021/jm4011675
- Chen, Q., Plasencia, M., Li, Z., Mukherjee, S., Negroni, J., Chang, S., Bhatt, P., Bhatt, R., Bhatt, R., and others. 'Cryo-EM structure of the activated STRA6 receptor–arrestin complex.' Science, 2020. DOI: 10.1126/science.aay7958
- Der, B.S., Edwards, D.R. and Kuhlman, B. 'A designed metalloprotein with enhanced hydrolytic function: a model for the metalloenzyme active site.' Journal of the American Chemical Society, 2013. DOI: 10.1021/ja407770a
- Kay, B.K., Winter, J. and McCafferty, J. (eds.), Phage Display of Peptides and Proteins: A Laboratory Manual. Academic Press, San Diego, 1996.
- Monod, J., Wyman, J. and Changeux, J.P. 'On the nature of allosteric transitions: a plausible model.' Journal of Molecular Biology, 12, 88–118, 1965.
- Koshland, D.E., Nemethy, G. and Filmer, D. 'Comparison of experimental binding data and theoretical models in proteins containing subunits.' Biochemistry, 5, 365–385, 1966.
- Lipari, G. and Szabo, A. 'Model-free approach to the interpretation of nuclear magnetic resonance relaxation in macromolecules.' Journal of the American Chemical Society, 104, 4546–4559, 1982.
- Crick, F.H.C. 'The packing of α-helices: simple coiled-coils.' Acta Crystallographica, 6, 689–697, 1953.
- Smith, G.P. 'Filamentous fusion phage: novel expression vectors that display cloned antigens on the virion surface.' Science, 228, 1315–1317, 1985.
- Boder, E.T. and Wittrup, K.D. 'Yeast surface display for screening combinatorial polypeptide libraries.' Nature Biotechnology, 15, 553–557, 1997.
- Walensky, L.D. and Bird, G.H. 'Hydrocarbon-stapled peptides: principles, practice, and mechanistic insights.' Journal of Medicinal Chemistry, 57, 6275–6288, 2014.
- Rezai, T., Yu, B., Millhauser, G.L., Jacobson, M.P. and Lokey, R.S. 'Testing the conformational hypothesis of passive membrane permeability using synthetic cyclic peptide diastereomers.' Journal of the American Chemical Society, 128, 2510–2511, 2006.