Foldon Model: Modular Protein Folding

• Foldon Model is a hierarchical framework in which proteins fold via orderly formation of discrete, cooperative units (foldons) comprising about 20-30 residues.
• The model contrasts with funnel-based descriptions by proposing a single, well-defined folding pathway characterized by reproducible intermediates.
• Experimental evidence from hydrogen-deuterium exchange and single-molecule force spectroscopy supports the presence of distinct foldons guiding proper protein assembly.

Searching arXiv for recent and foundational papers on the protein foldon model and related analyses. The foldon model is a hierarchical framework for protein folding in which proteins assemble through a defined sequence of discrete, cooperative intermediates built from small structural units termed foldons. In this formulation, foldons are groups of typically about 20 to 30 residues that acquire secondary and partial tertiary structure in a strict order, producing a unique or strongly channeled pathway to the native state. Recent syntheses present the model as a competing paradigm to funnel-based energy landscape descriptions, while frustration-based analyses reinterpret foldons as quasi-independent modules whose foldability can sometimes be related to conserved exons and recurrent architectural partitions (Bustamante et al., 31 Oct 2025, Galpern et al., 2024).

1. Definition, origin, and basic assumptions

The concept was introduced and elaborated by Englander and collaborators to rationalize hydrogen-deuterium exchange observations in cytochrome c, where the same intermediates appeared in the same order on every folding trajectory. Subsequent hydrogen-deuterium exchange studies on ribonuclease H similarly resolved sequentially forming foldons. In the model’s canonical form, proteins contain discrete folding units of a few tens of residues; each unit forms cooperatively in an all-or-none manner on the timescale of the experiment; early foldons stabilize later ones; and the amino-acid sequence encodes not only the native structure but also the order and coupling among foldons, so that proteins behave less like polymers of residues and more like “oligomers of foldons” (Bustamante et al., 31 Oct 2025).

These assumptions give the foldon model a sharply modular interpretation of folding. A foldon is not merely a secondary-structure fragment, nor simply any stable local motif. It is a cooperative unit whose internal formation and coupling to previously formed units generate a hierarchy of obligatory intermediates. Primary foldons initiate folding, whereas later foldons require both intra-foldon interactions and inter-foldon coupling with already formed units.

A related but not identical usage appears in energy-landscape work on modularity. There, foldons are described as quasi-independent cooperative units that fold in one step when their internal native interactions overcome the entropy lost upon ordering. Panchenko and collaborators formalized foldons as independent modules with strong internal contacts that can fold autonomously, and later work used mutational frustration analysis to evaluate such independence quantitatively across protein families (Galpern et al., 2024).

2. Relation to funnel landscapes and the status of the controversy

The principal conceptual contrast is with the funnel model, which represents folding as diffusion down a rugged, multidimensional free-energy landscape with many parallel routes to the native basin. In that picture, folding information is distributed across the sequence, initiation can occur in many locations so long as native bias increases, and route multiplicity provides robustness. By contrast, the foldon model asserts a channeled pathway composed of ordered intermediates rather than broadly distributed initiation and many alternative trajectories (Bustamante et al., 31 Oct 2025).

The tension between the two frameworks is both kinetic and interpretive. In funnel landscapes, lowering free energy does not necessarily increase the fraction of native contacts, and stable intermediates with few native contacts can become kinetic traps. The multiplicity of routes therefore implies possible frustration and trapping. The foldon model instead interprets reproducible, ordered protection patterns as evidence for obligatory intermediates whose appearance is more naturally explained by hierarchical assembly than by a broad ensemble of parallel routes.

As presented in recent reviews, this disagreement remains unresolved. The foldon model is not treated as a settled replacement for the funnel model, and the funnel model is not treated as having absorbed the foldon picture. Rather, the current position is that decisive experiments are needed to determine whether proteins traverse different routes on different folding events or whether they follow a single ordered sequence of intermediates. At the same time, the literature notes larger-scale hierarchy, including domains as “super-foldons,” which suggests that modularity and channeling may coexist with landscape concepts at an appropriate coarse-graining level. This suggests that the sharpest disagreement may concern the mechanistic granularity at which folding pathways are best described, rather than the existence of free-energy landscapes as such.

3. Experimental evidence for hierarchical assembly

Hydrogen/deuterium exchange provides the most direct classical evidence for foldons. In HDX-MS, backbone amide hydrogens exchange with deuterium during a labeling phase; exchange is then quenched by lowering pH, followed by pepsin digestion and mass spectrometry. Regions that form structure earlier become protected from exchange sooner, whereas late-forming regions exchange more deuterium. In cytochrome c, the same intermediates appeared in the same order across folding trajectories, defining a unique pathway. In ribonuclease H, time-resolved HDX resolved multiple foldons that structure sequentially, and the associated schematic free-energy representation displayed successive basins corresponding to ordered foldon formation. Typical protein-folding timescales are milliseconds to seconds, while HDX labeling windows can span seconds to hours, allowing both rapid and slower protection events to be captured (Bustamante et al., 31 Oct 2025).

Single-molecule force spectroscopy supplies an orthogonal line of evidence. Optical tweezers can tether a single protein between beads and monitor extension or force changes as the molecule unfolds and refolds in real time. In calmodulin, force spectroscopy resolved discrete transitions among states labeled $U$, $F12$, $F34$, $F123$, and $F1234$, together with an off-pathway misfolded state $F23$. These transitions occur at distinct forces and on millisecond-to-second timescales; some transitions are resolved within approximately $10\ \mathrm{ms}$, whereas native assembly is visible on approximately $1\ \mathrm{s}$ timescales. The modular EF-hand motifs behave as cooperative units whose formation or rupture produces quantized mechanical signatures, consistent with foldon-like modules and hierarchical pathways.

These results support two distinct but related claims. First, productive folding can proceed through reproducible intermediates rather than through a continuum of weakly distinguished states. Second, the productive pathway can coexist with nonproductive detours. The calmodulin data are therefore important because they counter a common oversimplification: the foldon model does not require the total absence of off-pathway states. Rather, the evidence cited for foldons concerns the ordering and reproducibility of the productive route, while optional or misfolded states may still be observed under force or kinetic bias.

4. Kinetic, thermodynamic, and spectroscopic formulation

The foldon model is often interpreted through standard HDX/HX kinetics and multistate folding formalisms rather than through a single dedicated equation set. In the EX2 regime, where opening and closing are at equilibrium and intrinsic exchange is slower than reclosing, the observed rate is written as

$$k_{\mathrm{obs}} = k_{\mathrm{int}} \times K_{\mathrm{op}},$$

with

$$K_{\mathrm{op}} = \exp\left(-\frac{\Delta G_{\mathrm{open}}}{k_B T}\right) = \frac{k_{\mathrm{op}}}{k_{\mathrm{cl}}}.$$

In the EX1 regime, where exchange is faster than reclosing, $F12$0, one has

$F12$1

The protection factor is

$F12$2

Larger $F12$3 implies stronger protection and is typically interpreted as evidence for foldon formation. In time-resolved HDX, stepwise increases in $F12$4 over labeling time are taken to reveal the appearance of discrete foldons, because formation of a foldon markedly reduces the opening probability for its amides and often for neighboring segments through inter-foldon coupling (Bustamante et al., 31 Oct 2025).

For kinetics and thermodynamics more generally, the standard barrier-crossing relation

$F12$5

is used to describe transitions between intermediates, with $F12$6 the activation free energy. The equilibrium free energy

$F12$7

provides the usual bookkeeping for stabilization through enthalpic packing and entropic cost. Sequential intermediates can be written in multistate form as

$F12$8

where $F12$9 is the population of intermediate $F34$0. In a strictly hierarchical sequence, the dominant rates connect adjacent states in one direction under folding conditions, producing ordered population flow. A schematic free-energy profile $F34$1, with $F34$2 a reaction coordinate such as the fraction of native contacts, is then expected to show distinct basins at increasing $F34$3 as foldons form in sequence.

Single-molecule observables are interpreted analogously. Discrete changes in force or extension are taken to signal cooperative formation or rupture of structural units. In the calmodulin example, each mechanical step corresponds to a foldon-like event, such as formation of an EF-hand motif or inter-motif coupling. The underlying review does not specify a polymer-elasticity model such as a worm-like chain; its emphasis is instead on the use of high-resolution trajectories to determine whether the same ordered sequence of intermediates recurs repeatedly, which is a central empirical prediction of the foldon framework.

5. Modularity, exons, and frustration analysis

Energy-landscape theory supplies a second route to the foldon concept by quantifying the independent foldability of segments. In this formulation, if $F34$4 is the energy gap between the native configuration and an ensemble of compact non-native molten-globule decoys, and $F34$5 is the energy variance of those decoys, the total frustration index is defined as

$F34$6

The folding and glass temperatures scale as

$F34$7

and a foldability measure can be written as

$F34$8

A segment with large negative $F34$9 and hence larger $F123$0 is more likely to fold cooperatively and independently. In this language, foldons are quasi-independent cooperative units that can assemble into the full structure (Galpern et al., 2024).

A systematic reassessment of the exon-foldon correspondence examined 38 abundant and conserved protein families by mapping genomic exon boundaries onto protein sequence and structure alignments and then evaluating whether exonic segments behave like foldons according to frustration metrics. Segment energetics were computed with AWSEM using two Hamiltonians: an “Independent” form $F123$1, which retains only burial and contact terms internal to a segment, and a “Context” form $F123$2, which also includes contacts between the segment and the rest of the protein. To quantify foldon-like independence, the study defined

$F123$3

Small $F123$4 indicates that a segment remains minimally frustrated in isolation, whereas large positive $F123$5 indicates that substantial stabilization arises only through inter-segment contacts.

Several quantitative findings emerged. Considering all families together and exons with abundance at least $F123$6, the median $F123$7 decreases with exon abundance. Below $F123$8 abundance, natural exon $F123$9 distributions are indistinguishable from size-matched controls; above $F1234$0, the trend decreases smoothly, indicating that more conserved exons are more foldon-like. In dihydrofolate reductase, exon 1 had $F1234$1, $F1234$2, and $F1234$3, consistent with near-independent foldability, whereas exon 4 had $F1234$4, $F1234$5, and $F1234$6, indicating strong stabilization by context and poor independent foldability. Across families, actual exons generally showed greater independent foldability than random segments of similar length, but only about one third of families had minimal common exons that were significantly more independently foldable than controls (Galpern et al., 2024).

The same study introduced exon boundary hot spots and minimal common exons. Boundary hot spots were defined as local maxima in exon-boundary histograms along a multiple-sequence alignment, using a 10-residue neighborhood on each side, excluding peaks with density below $F1234$7 and maxima too close to one another or to alignment ends. Tiling the alignment between successive hot spots produced minimal common exons, which represent a family’s consensus modular partition. In most families these minimal common exons coincide with uninterrupted secondary-structure elements or simple combinations such as an entire helix, an entire strand, or a helix-turn-helix segment, and their boundaries tend not to interrupt $F1234$8-helices or $F1234$9-strands. Positive $F23$0 and $F23$1 values in the majority of families show that these boundaries occur more often than neutral alternatives in coil-like regions and at positions of higher local frustration, although notable exceptions exist.

These results refine the relation among motifs, foldons, exons, and domains. Domains are larger architectural units that are often autonomously foldable; motifs are smaller recurrent patterns that are often not autonomously stable; foldons occupy an intermediate scale. The frustration analysis suggests that the most conserved exons in many families behave like foldons, whereas minimal common exons are often more motif-like and may need to combine into larger segments to achieve independent foldability. This suggests that exon organization can align with foldable units, but the alignment is neither exact nor universal.

6. Computational implications, biomedical relevance, and open problems

Recent computational advances sharpen the stakes of the foldon model without yet resolving it. AlphaFold has transformed static structure prediction but remains agnostic about dynamics and folding pathways. The current view is that extending machine learning to dynamics, using evolutionary covariation, structural databases, HDX, and single-molecule data, could begin to infer pathway information, including candidate foldon units and their coupling. Coarse-grained and enhanced-sampling methods such as Markov state models and replica exchange can bridge inaccessible timescales but sacrifice some resolution, making close integration between simulation and experiment essential (Bustamante et al., 31 Oct 2025).

The foldon framework also offers a mechanistic view of misfolding. If productive folding proceeds through defined intermediates, then off-pathway states and kinetic traps arise when one or more foldons fail to assemble in the proper order or misassemble. In that setting, the hierarchy of foldon formation identifies where the process goes awry. The literature therefore points to several practical implications: stabilizing early foldons to promote correct pathway channeling, destabilizing misfold-prone intermediates, modulating inter-foldon coupling, redesigning sequences to strengthen primary foldons or reorganize interfaces among “super-foldons,” and understanding why some proteins require chaperones because particular foldons are slow or vulnerable (Bustamante et al., 31 Oct 2025).

At the same time, the limitations are substantial. HDX is ensemble-averaged, and synchronizing populations through dynamic processes is challenging; observed intermediates may reflect redistribution within an unfolded ensemble rather than discrete thermodynamic states. Single-molecule trajectories reveal intermediates directly, but structural annotation of mechanical steps and demonstration that the same ordered sequence recurs on every trajectory remain difficult. Evidence is strongest for a limited set of model proteins and modular systems, especially cytochrome c, ribonuclease H, and calmodulin, so generality across broader protein classes, including multi-domain and membrane proteins, remains open. Frustration-based exon analyses likewise show heterogeneity: not all families display strong exon-foldon correspondence, and factors such as intrinsically disordered regions, multi-domain scaffolds, binding-induced folding, and allostery may decouple genomic segmentation from folding modularity (Bustamante et al., 31 Oct 2025, Galpern et al., 2024).

The current synthesis therefore presents the foldon model as a rigorous but still contested framework. Its strongest empirical claim is that at least some proteins fold through discrete, cooperative units that appear in a reproducible hierarchy. Its strongest theoretical extension is that these units can be characterized by near-independent foldability and by nonrandom alignment with conserved exon architecture in many, though not all, protein families. Whether this hierarchical organization is a special mechanism for certain proteins or a broadly general principle of folding remains an open question.