Fold-switching Proteins

Devlina Chakravarty; Lauren L. Porter

REVIEW 2 major objections 4 minor 9 references

Fold-switching Proteins

T0 review · 2 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This review argues that fold-switching proteins are widespread, biologically functional, and shaped by natural selection, not rare evolutionary accidents.

desk verdict Solid review of fold switching with a central selection claim that is stronger than its coevolutionary evidence warrants. read the letter →

arxiv 2507.10839 v1 pith:2IN5VRPN submitted 2025-07-14 q-bio.BM

Devlina Chakravarty , Lauren L. Porter This is my paper · ORCID

classification q-bio.BM

keywords fold-switchingproteinsmetamorphicdual-foldcoevolutionproteinenergylandscapesonesequencestructurepredictionalternativeconformationscircadianclock

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper sets out to establish that some proteins with a single amino-acid sequence reliably adopt two different folded structures, and that this dual-folding behavior is a selected, functional trait rather than a mistake. It assembles evidence that fold-switching proteins appear across all kingdoms of life, that both folds of many such proteins carry evolutionary coevolutionary signals, and that their alternative conformations are slow, structured transitions on multi-well energy landscapes. If the review is right, the textbook rule that one sequence encodes one structure is a special case, and current structure-prediction tools trained on that rule will continue to miss a meaningful slice of the proteome. The review also lays out how switches happen, how new folds evolve through stepwise mutation, and which computational and experimental methods are beginning to find new switchers.

What carries the argument

The load-bearing object is the multi-well energy landscape: fold-switching proteins occupy at least two structurally defined, native-like minima rather than a single deep funnel, and transitions between minima are slow, on the order of seconds to hours, and often pass through partially disordered or hybrid intermediates. The second pillar is dual-fold coevolution, the presence of evolutionary couplings uniquely corresponding to each of the two conformations, which is used as evidence that both folds are functionally selected. Supporting machinery includes NMR exchange techniques such as CEST, zz-exchange, and CPMG for observing slow interconversion, structure-based models with dual basins for simulating transitions, and sequence-based predictors such as CF-random that generate alternative conformations and feed the coevolutionary filter.

What would settle it

A decisive test would be an independent, blinded survey: take a large set of predicted fold-switching E. coli proteins, express and purify them, and use NMR or hydrogen-deuterium exchange to ask whether a substantial fraction actually populate two distinct folded states under physiological conditions; if the confirmation rate falls far below the predicted 5%, the prevalence and selection claims would need revision.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central claim is that fold-switching proteins challenge the classical expectation that a globular protein's sequence encodes a single fixed fold. The review reports roughly one hundred experimentally characterized fold switchers, estimates that up to 4% of proteins in the Protein Data Bank and up to 5% of E. coli proteins may switch folds, and argues that dual-fold coevolutionary signals show both conformations of many switchers are under selection. It also argues that fold switching can be an evolutionary end point, as in XCL1 and RfaH, or an evolutionary intermediate, as in the stepwise helix-turn-helix to winged-helix transition seen in bacterial response regulators. The review treats the emergence of these proteins as evidence that fold space is more fluid than the one-sequence-one-structure doctrine implies, and that AI-based structure predictors often fail on fold switchers because they memorize training-set structures rather than inferring alternative folds from coevolution.

Load-bearing premise

The central wager is that the structural screens and coevolutionary signals used to count fold switchers really identify proteins whose two folds are both biologically functional; if those signals largely reflect database bias or correlated noise, the widespread-and-selected story weakens.

Editorial extensions

If this is right

If both folds of many fold-switching proteins are under selection, then genome-wide searches should expect to find functional fold switchers in diverse biological processes, including transcription regulation, circadian timing, cell division, and immune defense.
If up to 5% of E. coli proteins switch folds, then current structural and functional annotations that assume a single conformation are incomplete for a substantial fraction of proteins in that organism and possibly across bacteria.
If AlphaFold-family predictors rely on memorized training structures, then reliable structure prediction for fold-switching proteins will require new training schemes or new inference methods that explicitly model alternative states.
If stepwise mutation can convert one fold family into another, then protein fold evolution is better described as a connected, fluid landscape than as isolated discrete folds.
If temperature, pH, binding, or post-translational modification can trigger switching, then disease mutations and alternative splicing variants that alter folding equilibria become plausible mechanisms of dysfunction.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Beyond the paper, the estimated 4-5% prevalence is likely a lower bound, because structural databases and current screens are biased toward stable, highly populated conformations and will miss switchers whose alternative folds are transient or condition-dependent.
Beyond the paper, if dual-fold coevolution is a reliable signature, then scanning sequence databases alone, before any structure is solved, could nominate fold-switching candidates across entire proteomes, extending the approach well beyond E. coli.
Beyond the paper, the claim that AlphaFold2 memorizes fold-switched structures implies a concrete test: retraining predictors on datasets deliberately stripped of all known fold-switching homologs should eliminate confident predictions of alternative folds if the memorization hypothesis is correct.
Beyond the paper, linking temperature-sensitive switching to circadian timescales suggests that other biological clocks may tune their periods through slow fold-switching equilibria rather than through faster allosteric changes.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

You should know this paper is a review, not primary research. What it does well: it is a clear, well-organized synthesis of the fold-switching field, with good concrete examples (RfaH, KaiB, XCL1), a fair summary of the biophysics (energy landscapes, NMR timescales) and a genuinely useful roundup of prediction and design attempts. It also includes appropriate caveats: CF-random only hits 35% of known switchers, and no computational method has yet predicted a genuinely new fold switcher that was later confirmed. The authors do not oversell the experimental validation of predicted E. coli hits.

The soft spot is the central claim in the Abstract: 'evolution has selected for their dual-folding behavior.' That is a strong statement about selection. The evidence in Section 4.2 is evolutionary-coupling analysis (ref 95), which finds contacts unique to each fold. The review mentions that the signal comes from subfamilies with fold-switchers rather than superfamilies of single folders, which is a reasonable control, but it does not demonstrate that a family composed entirely of single-fold members with different folds (like the HTH-to-wH response regulators described in Section 4.1) would not produce the same dual-fold coupling signal. Until that negative control is published, the inference that both folds of a single sequence are under selection is not fully secured. The stress-test concern lands. It is a review, so the claim is secondhand, but the Abstract states it as fact. That should be softened or flagged as a hypothesis.

Also, the prevalence estimates (up to 4% of PDB, up to 5% of E. coli) come from the authors' own methods and have not been independently replicated. The review notes this implicitly, but could be more explicit.

The AlphaFold 'memorization' interpretation is presented as settled; it is actually contested. That section is a bit one-sided, though the authors have published supporting analyses.

Nothing here is a fatal flaw. The paper is a competent review that will be useful to anyone entering the field. It deserves a serious referee, and with minor revisions to the Abstract and more hedging in the evolution section, it would be a solid accept. If it were primary research, I'd want the negative control; as a review, it is acceptable but should mark the selection claim as an interpretation.

Recommendation: engage, accept after minor revision.

Referee Report

2 major / 4 minor

Summary. This is an invited review of the fold-switching protein field, written by two leading contributors. The paper surveys the biophysical basis of fold switching (energy landscapes, NMR-detected interconversion mechanisms, molecular dynamics simulations), the evolutionary evidence for fold switching as both an intermediate and an adaptive end point, computational approaches to predict and design fold switchers, and open questions. The central thesis, stated in the abstract and reiterated in Section 4.2 and Summary Point 4, is that evolution has selected for the dual-folding behavior of many fold-switching proteins, so that both conformations are functional and adaptive rather than accidental.

Significance. The review is timely, authoritative, and generally well balanced. It synthesizes the recent explosion of work on fold switching, including the authors' own contributions such as the ACE coevolutionary analysis (ref 95), the CF-random prediction approach (ref 59), and the AlphaFold2 memorization studies (refs 94, 96). It is commendably honest in stating limitations: CF-random's 35% success rate is acknowledged, the lack of experimentally validated de novo predictions is stated explicitly in Section 6, and the caveat that 'dual-fold coevolutionary signals originate from the sequences of protein subfamilies populated by fold-switching proteins, rather than from superfamilies typically dominated by single-fold proteins' is included. If published, this would be a useful reference for both specialists and newcomers. However, the strength of the abstract's claim that evolution has selected for dual-folding behavior is somewhat ahead of the evidence as presented in the review, which is the main substantive concern.

major comments (2)

[Abstract and Section 4.2] The abstract's claim that 'evolution has selected for their dual-folding behavior' is presented as a settled conclusion, and Section 4.2 asserts that 'dual-fold selection can be identified by evolutionary couplings uniquely corresponding to the different conformations of fold-switching proteins (95).' The review does not, however, address the alternative interpretation that contacts unique to each fold could arise from phylogenetic mixtures of single-fold homologs within the analyzed subfamilies, the very situation described for HTH/wH response regulators in Section 4.1. The statement that these signals 'originate from the sequences of protein subfamilies populated by fold-switching proteins, rather than from superfamilies typically dominated by single-fold proteins' is an assertion about the control, not a description of it; a reader cannot determine whether the negative control (showing that ACE does not fire on mixed single-fold families) was performed. Please either (i) briefly describe the controls in ref 95 that exclude this confound, or (ii) temper the abstract and Summary Point 4 to say that 'many' or 'most' known fold-switching proteins show evidence of dual-fold selection, reserving the universal phrasing for the subset of families where the phylogenetic-mixture confound has been excluded.
[Section 5.1] The conclusion that 'AF-cluster works by associating memorized structures or substructures with MSA sequences, not coevolutionary inference of input MSAs as originally claimed (113)' is stated as a settled fact. The cited evidence (refs 13, 59, 94, 96) is consistent with this interpretation, but the matter is an active debate in the literature, and the original authors (ref 113) have not conceded the point. As a review, this passage would be more appropriate phrased as a strong inference ('these observations suggest that AF-cluster's predictions are driven by training-set memorization rather than by coevolutionary inference from the input MSA') rather than a definitive refutation. This is a secondary issue relative to the central selection claim, but it bears on the review's authority in a contentious area.

minor comments (4)

[Introduction] The phrase 'For over 50 years, globular proteins been expected to assume single folded structures' is missing the auxiliary verb 'have' and should read 'have been expected.'
[Abstract] The symbols for alpha-helices and beta-sheets appear as 'a-helices' and 'b-sheets' in several places (including the abstract and Section 3); consistent use of Greek letters (α, β) would improve readability.
[Section 5.1] Several cited sources are bioRxiv preprints (refs 27, 36, 92, 102) and are treated on par with peer-reviewed literature in the text; consider flagging their preprint status when first mentioned so readers can weigh them appropriately.
[Figure 3 caption] The caption for Figure 3 is very long and combines the legend for three different panels (a, b, c) with extensive interpretation of phylogenetic results. Splitting the legend into a per-panel format would aid the reader.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular reduction demonstrated; self-cited evolutionary and prevalence analyses are falsifiable and anchored by independent experimental examples.

full rationale

This is a review, so its derivation chain is mostly citation-based rather than a fitted loop. I found no step where a fitted parameter is renamed as a prediction and no definition that forces the conclusion by construction. The abstract's claim that evolution has selected both folds relies heavily on the authors' own ACE study (ref 95), stated in Section 4.2 as: 'Dual-fold selection can be identified by evolutionary couplings uniquely corresponding to the different conformations of fold-switching proteins (95).' Similarly, the 4-5% prevalence estimates rest on the authors' own screens (refs 59, 83). These are self-citations, but they are not circular: the ACE and CF-random analyses are falsifiable computational procedures applied to independent sequence and structure data, and the review supplements them with independent experimental work on XCL1, RfaH, KaiB, and designed systems. The review also explicitly concedes limitations, noting that reliable de novo prediction of new fold switchers is not yet achieved and that CF-random 'achieves only a 35% success rate.' The phylogenetic-mixture confound raised by skeptics is a threat to the inference from coevolutionary signal to selection, but that is a correctness risk rather than a circular reduction of the conclusion to its input. Overall, the paper is self-contained against external benchmarks for its core phenomenology, so no specific circular step is established. The score of 2 reflects the density of author self-citations in load-bearing claims rather than a demonstrated circularity.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

As a review, this paper introduces no new free parameters or entities. Its claims rest on the validity of prior experimental and computational studies, including several from the authors' own laboratory, as listed above.

assumptions (4)

domain assumption The PDB contains enough near-identical sequence pairs with different folds to support prevalence estimates.
Section 1 and ref 83 infer that up to 4% of PDB proteins switch folds based on a screen for identical sequences with different secondary structures; this assumes the PDB is representative and the screen has low false positives.
domain assumption Dual-fold coevolutionary signals indicate that both folds are evolutionarily selected.
Section 4.2 and ref 95 interpret fold-specific evolutionary couplings as evidence of selection; this assumes the signals are not artifacts of alignment or phylogeny.
domain assumption NMR-observed slow exchange (seconds to hours) generalizes to most fold switchers.
Section 3.2 generalizes from a handful of systems (RfaH, XCL1, KaiB, PimA, Sa1) that interconversion is slow; other mechanisms may exist.
domain assumption AlphaFold's failures on fold switchers are due to structure memorization of training data.
Section 5.1 states this conclusion based on refs 16, 17, 94, 96, almost all from the authors' lab; the memorization explanation is not yet independently established.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Fold-switching Proteins." pith.science (2026). https://pith.science/paper/2IN5VRPN

@misc{pith2026250710839,
  author       = {Pith},
  title        = {Pith review of: Fold-switching Proteins},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2IN5VRPN}},
  note         = {Machine review of arXiv:2507.10839}
}

read the original abstract

Globular proteins are expected to assume folds with fixed secondary structures, alpha-helices and beta-sheets. Fold-switching proteins challenge this expectation by remodeling their secondary and/or tertiary structures in response to cellular stimuli. Though these shapeshifting proteins were once thought to be haphazard evolutionary byproducts with little intrinsic biological relevance, recent work has shown that evolution has selected for their dual-folding behavior, which plays critical roles in biological processes across all kingdoms of life. The widening scope of fold switching draws attention to the ways it challenges conventional wisdom, raising fundamental unanswered questions about protein structure, biophysics, and evolution. Here we discuss the progress being made to answer these questions and suggest future directions for the field.

Figures

Figures reproduced from arXiv: 2507.10839 by the authors.

**Figure 1.** Three well-characterized foldswitching proteins. (a) The C-terminal domain (CTD, teal) of E. coli RfaH reversibly switches from an α-helical to β-sheet fold upon binding RNA polymerase and operon polarity suppressor DNA. Adapted from (82). (b) Under physiological conditions, the human chemokine XCL1 (mustard) reversibly interconverts between a chemokine fold involved in signaling and a dimeric β-sheet fold that bin… view at source ↗

**Figure 4.** Predicting E. coli proteins that may switch folds. (A). 2126 E. coli and phage proteins were run through CF-random to test whether they switch folds (59). Seashell-like image represents these proteins by length; the inner circle represents 1111 candidates for which sufficiently deep MSAs could not be generated, and outer, the 2126 proteins that were then run through CF-random. If two or more distinct conformations w… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

9 extracted references · 8 canonical work pages

[1]

one sequence–one structure

Introduction For over 50 years, globular proteins been expected to assume single folded structures fostering their biological functions (7). Here we describe an emerging class of proteins that defy this expectation. Fold-switching proteins, once thought to be transitory evolutionary intermediates with little intrinsic biological relevance (105, 116), have...

work page
[3]

Fold-switching transitions are slow (on the order of seconds or slower) and can occur through intermediates

Fold-switching proteins have multi-well energy landscapes that are impacted by environmental factors such as temperature. Fold-switching transitions are slow (on the order of seconds or slower) and can occur through intermediates. 4. Evolution has selected for both folds of many fold-switching proteins, and new folds can emerge in protein families mediate...

work page 2026
[6]

Future opportunities Fold-switching proteins challenge conventional wisdom and raise fundamental unanswered questions about protein structure, biophysics, and evolution. Recent work has begun to address these questions by characterizing the stabilities and interconversions of fold switchers, finding coevolutionary signals unique to both folds, identifying...

work page
[14]

Cai M, Ying J, Lopez JM, Huang Y , Clore GM. 2025. Unraveling structural transitions and kinetics along the fold-switching pathway of the RfaH C-terminal domain using exchange-based NMR. Proc Natl Acad Sci U S A 122: e2506441122 15. Chakravarty D, Lee M, Porter LL. 2025. Proteins with alternative folds reveal blind spots in AlphaFold-based protein structu...

work page 2025
[34]

Drozdetskiy A, Cole C, Procter J, Barton GJ. 2015. JPred4: a protein secondary structure prediction server. Nucleic acids research 43: W389-W94 35. Farrow NA, Zhang O, Forman-Kay JD, Kay LE. 1995. Comparison of the backbone dynamics of a folded and an unfolded SH3 domain existing in equilibrium in aqueous buffer. Biochemistry 34: 868-78 36. Ferrari AJR, D...

arXiv 2015
[53]

Khatua P, Ray AJ, Hansmann UHE. 2020. Bifurcated Hydrogen Bonds and the Fold Switching of Lymphotactin. J Phys Chem B 124: 6555-64 54. Kim AK, Looger LL, Porter LL. 2021. A high-throughput predictive method for sequence-similar fold switchers. Biopolymers 112: e23416 55. Kim AK, Porter LL. 2021. Functional and Regulatory Roles of Fold-Switching Proteins. ...

work page 2020
[71]

Markin CJ, Mokhtari DA, Sunden F, Appel MJ, Akiva E, et al. 2021. Revealing enzyme functional architecture via high-throughput microfluidic enzyme kinetics. Science 373 72. Marks DS, Colwell LJ, Sheridan R, Hopf TA, Pagnani A, et al. 2011. Protein 3D structure computed from evolutionary sequence variation. PLoS One 6: e28766 73. Mishra S, Looger LL, Porte...

work page 2021
[93]

invisible

Sala D, Engelberger F, McHaourab HS, Meiler J. 2023. Modeling conformational states of proteins with AlphaFold. Curr Opin Struct Biol 81: 102645 94. Schafer JW, Lee M, Chakravarty D, Thole JF, Chen EA, Porter LL. 2025. Sequence clustering confounds AlphaFold2. Nature 638: E8-E12 95. Schafer JW, Porter LL. 2023. Evolutionary selection of proteins with two ...

work page 2023

Show all 9 references

[114]

Wayment-Steele HK, Otten R, Pitsawong W, Ojoawo AM, Glaser A, et al. 2024. The conformational landscape of fold-switcher KaiB is tuned to the circadian rhythm timescale. Proc Natl Acad Sci U S A 121: e2412293121 115. Wei KY , Moschidi D, Bick MJ, Nerli S, McShan AC, et al. 202...

2024

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.

[1] [1]

one sequence–one structure

Introduction For over 50 years, globular proteins been expected to assume single folded structures fostering their biological functions (7). Here we describe an emerging class of proteins that defy this expectation. Fold-switching proteins, once thought to be transitory evolutionary intermediates with little intrinsic biological relevance (105, 116), have...

work page

[2] [3]

Fold-switching transitions are slow (on the order of seconds or slower) and can occur through intermediates

Fold-switching proteins have multi-well energy landscapes that are impacted by environmental factors such as temperature. Fold-switching transitions are slow (on the order of seconds or slower) and can occur through intermediates. 4. Evolution has selected for both folds of many fold-switching proteins, and new folds can emerge in protein families mediate...

work page 2026

[3] [6]

Future opportunities Fold-switching proteins challenge conventional wisdom and raise fundamental unanswered questions about protein structure, biophysics, and evolution. Recent work has begun to address these questions by characterizing the stabilities and interconversions of fold switchers, finding coevolutionary signals unique to both folds, identifying...

work page

[4] [14]

Cai M, Ying J, Lopez JM, Huang Y , Clore GM. 2025. Unraveling structural transitions and kinetics along the fold-switching pathway of the RfaH C-terminal domain using exchange-based NMR. Proc Natl Acad Sci U S A 122: e2506441122 15. Chakravarty D, Lee M, Porter LL. 2025. Proteins with alternative folds reveal blind spots in AlphaFold-based protein structu...

work page 2025

[5] [34]

Drozdetskiy A, Cole C, Procter J, Barton GJ. 2015. JPred4: a protein secondary structure prediction server. Nucleic acids research 43: W389-W94 35. Farrow NA, Zhang O, Forman-Kay JD, Kay LE. 1995. Comparison of the backbone dynamics of a folded and an unfolded SH3 domain existing in equilibrium in aqueous buffer. Biochemistry 34: 868-78 36. Ferrari AJR, D...

arXiv 2015

[6] [53]

Khatua P, Ray AJ, Hansmann UHE. 2020. Bifurcated Hydrogen Bonds and the Fold Switching of Lymphotactin. J Phys Chem B 124: 6555-64 54. Kim AK, Looger LL, Porter LL. 2021. A high-throughput predictive method for sequence-similar fold switchers. Biopolymers 112: e23416 55. Kim AK, Porter LL. 2021. Functional and Regulatory Roles of Fold-Switching Proteins. ...

work page 2020

[7] [71]

Markin CJ, Mokhtari DA, Sunden F, Appel MJ, Akiva E, et al. 2021. Revealing enzyme functional architecture via high-throughput microfluidic enzyme kinetics. Science 373 72. Marks DS, Colwell LJ, Sheridan R, Hopf TA, Pagnani A, et al. 2011. Protein 3D structure computed from evolutionary sequence variation. PLoS One 6: e28766 73. Mishra S, Looger LL, Porte...

work page 2021

[8] [93]

invisible

Sala D, Engelberger F, McHaourab HS, Meiler J. 2023. Modeling conformational states of proteins with AlphaFold. Curr Opin Struct Biol 81: 102645 94. Schafer JW, Lee M, Chakravarty D, Thole JF, Chen EA, Porter LL. 2025. Sequence clustering confounds AlphaFold2. Nature 638: E8-E12 95. Schafer JW, Porter LL. 2023. Evolutionary selection of proteins with two ...

work page 2023